Current Issues In Linguistic Theory 314-harry.pdf

ENGLISH HISTORICAL LINGUISTICS 2008

AMSTERDAM STUDIES IN THE THEORY AND HISTORY OF LINGUISTIC SCIENCE General Editor E.F.K. KOERNER Zentrum für Allgemeine Sprachwissenschaft, Typologie und Universalienforschung, Berlin [email protected] Series IV – CURRENT ISSUES IN LINGUISTIC THEORY Advisory Editorial Board Lyle Campbell (Salt Lake City) Sheila Embleton (Toronto) Elly van Gelderen (Tempe, Ariz.) John E. Joseph (Edinburgh) Manfred Krifka (Berlin) Martin Maiden (Oxford) Martha Ratliff (Detroit, Mich.) E. Wyn Roberts (Vancouver, B.C.) Joseph C. Salmons (Madison, Wis.)

Volume 314

Ursula Lenker, Judith Huber and Robert Mailhammer (eds.) English Historical Linguistics 2008. Selected papers from the fifteenth International Conference on English Historical Linguistics (ICEHL 15), Munich, 24–30 August 2008. Volume I: The history of English verbal and nominal constructions

ENGLISH HISTORICAL LINGUISTICS 2008 SELECTED PAPERS FROM THE FIFTEENTH INTERNATIONAL CONFERENCE ON ENGLISH HISTORICAL LINGUISTICS (ICEHL 15), MUNICH, 24–30 AUGUST 2008 VOLUME I: THE HISTORY OF ENGLISH VERBAL AND NOMINAL CONSTRUCTIONS Edited by

URSULA LENKER JUDITH HUBER ROBERT MAILHAMMER Catholic University of Eichstätt-Ingolstadt

JOHN BENJAMINS PUBLISHING COMPANY AMSTERDAM/PHILADELPHIA

8

TM

The paper used in this publication meets the minimum requirements of American National Standard for Information Sciences – Permanence of Paper for Printed Library Materials, ansi z39.48-1984.

Library of Congress Cataloging-in-Publication Data International Conference on English Historical Linguistics (15th : 2008 : University of Munich) English historical linguistics 2008 : selected papers from the fifteenth International Conference on English historical linguistics (ICEHL 15), Munich, 24-30 August 2008. p. cm. (Amsterdam studies in the theory and history of linguistic science. Series IV, Current Issues in Linguistic Theory, issn 0304-0763 ; v. 314) The fourteen contributions to this volume, which were selected from papers delivered at the 15th International Conference on the History of the English Language (ICEHL) held at the University of Munich (24-30 August 2008). Includes bibliographical references and index. 1. English language--Grammar, Historical--Congresses. 2. English language--History-Congresses. I. Lenker, Ursula, 1963-. II. Huber, Judith, 1981-. III. Mailhammer, Robert, 1975-. IV. Title. PE1075.I57   2010 425--dc22 2010025973 isbn 978 90 272 4832 9 (Hb ; alk. paper) isbn 978 90 272 8779 3 (Eb)

© 2010 – John Benjamins B.V. No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the publisher. John Benjamins Publishing Co. · P.O. Box 36224 · 1020 me Amsterdam · The Netherlands John Benjamins North America · P.O. Box 27519 · Philadelphia pa 19118-0519 · usa

Table of contents Acknowledgements Introduction: Capturing and explaining syntactic change in the history of English Ursula Lenker, Judith Huber, Robert Mailhammer

vii

1

part i.  Verbal constructions “Þonne hate we hine morgensteorra”: On verb complementation in Old English Nils-Lennart Johannesson

11

Tracking and explaining variation and change in the grammar of American English: A case study, with evidence from the TIME Corpus 29 Juhani Rudanko Prevent and the battle of the ‑ing clauses: Semantic divergence? Elina Sellgren Prescription or practice? Be/have variation with past participles of mutative intransitive verbs in the letters of Joseph Priestley Robin Straaijer On the idiomatization of “give + O + to” constructions Minoji Akimoto The clausal complementation of good in extraposition constructions: The emergence of partially filled constructions An Van linden

45

63

79

95

part ii.  Modality and (marginal) modals The ‘ fail to’ construction in Late Modern and Present-Day English Thomas Egan

123



English Historical Linguistics 2008

The interplay of modal verbs and adverbs: A history of mæg eaþe Jerzy Nykiel Current change in the modal system of English: A case study of must, have to and have got to Joanne Close & Bas Aarts

143

165

part iii.  Developments in the English noun phrase Discontinuous quantificational structures in Old English Artur Bartnik Genitive variation in letters, history writing and sermons in Late Middle and Early Modern English Teo Juvonen

185

197

part iv.  Syntactic variation and change through contact On the use of beon and wesan in Old English Ilse Wischer

217

The reflexes of OE beon as a marker of futurity in early Middle English Margaret Laing

237

Stylistic fronting in the history of English Masayuki Ohkado

255

Subject and Word index

279

Acknowledgements The fourteen contributions to this volume, which were selected from papers delivered at the 15th International Conference on the History of the English Language (ICEHL) held at the University of Munich (24–30 August 2008), focus on the genesis and variation of nominal and verbal constructions in the history of English. At ICEHL 15, about 170 papers were given, of which fifty were handed in for consideration in the proceedings. The editorial board agreed on compiling three volumes, which were to be organized thematically. For the present volume, the editorial team from the Catholic University of Eichstätt-Ingolstadt picked those papers which were related enough to form a fairly coherent volume dealing with syntactic and morpho‑syntactic topics and ultimately, after a thorough inspection by the editors and reviewers, the fourteen papers here united were chosen for the present publication. The editing was completed in time for the subsequent ICEHL conference in Pécs, Hungary, in August 2010. In addition to this book, an editorial team based at the University of Munich (editor‑in‑chief: Hans Sauer) will compile at least one more volume, to be published with John Benjamins, comprising selections from the remaining papers carrying a different thematic focus. The editors wish to thank the organizing committee of the 15th International Conference on the History of the English Language (ICEHL) at the University of Munich for organizing and supporting such a huge conference logistically and academically. Our heartfelt gratitude also goes out to all the colleagues who have been involved in the selection process for this volume as referees, readers and advisors. Specifically, we would like to thank Daniel Hole, Ulrike Krischke, Angelika Lutz, Wolfgang Mager, Hans Sauer, Gaby Waxenberger and those reviewers who wish to remain anonymous for their quick and thorough review of the contributions, substantially improving sections of this book with their pertinent comments. We are also very grateful to the series’ editor, E.F.K. Koerner, and the staff of John Benjamins for their support and, of course, to the authors for their cooperation and patience. Last but not least, thanks are due to Veronika Bischofberger, Jonas Bodensohn, Andrea Graf, Andrea Halbich, Christoph Hauf, Barbara Hefele, Jutta Öhlschläger and Stefan Stenke for their supporting work during the production process of the manuscripts and to Anke de Looper and Martine van Marsbergen at John Benjamins, for their help in the final stages of guiding the volume through the press. Eichstätt, May 2010

Ursula Lenker, Judith Huber, Robert Mailhammer

Introduction Capturing and explaining syntactic change in the history of English Ursula Lenker, Judith Huber, Robert Mailhammer Catholic University of Eichstätt-Ingolstadt

1. New paths of investigation: Corpus and constructionist approaches In recent years, the study of historical syntax has in particular benefited from a methodological trend towards usage-based or functional approaches. This is largely due to the intrinsically compositional nature of syntax. Unlike morpho­ logy and phonology, which describe the formation and use of single speech signs, syntax is about stringing items together in order to form more complex functions and meanings. As a result, syntax forms a larger context, which it is, however, at the same time dependent on, since the co‑ and contextual parameters feed back on how words can be put together. In a synchronic approach, the contextual information can be backgrounded in favour of formal issues, but from a diachronic perspective it is exactly the changes in collocations, constructions (in a general sense) and sentence types that form the object of investigation. These changes occur in a complex interplay of conventionalization and variation, which is why it is important to pay attention not only to standardized or widely-used varieties of a language, but also to its dialectal varieties. Consequently, studies in historical syntax, which either – explicitly or implicitly – contrast historical stages of a language with its later stages, or which describe a diachronic development, have increasingly made use of corpus studies and of a range of what can be labelled usage-based approaches, such as frequency-based accounts and, more recently, various kinds of constructional grammar (see Croft 2001: 14–21 for an overview of different approaches, with references), investigating not only the standard language but also various kinds of diatopic, diastratic and diaphasic varieties. Such functional investigations combine description with explanation, in the sense that changes in syntactic conventions are



Ursula Lenker, Judith Huber, Robert Mailhammer

documented, which may then be explained with, for instance, the help of a formal theory. The key significance of such fine-grained studies of micro-variation for a theory of language change has been increasingly recognized in recent years (see, e.g., Enfield 2005). The problem of examining language change has always been to “catch it red-handed”, mainly because the perspective was often too wide, the method too crude and the resolution of the study too low. The gradual nature of change and the increasing entrenchment of new forms and meanings make it necessary to look more closely and to single out phenomena which can then be studied more carefully and more comprehensively. Traditional philology has always inherently assumed such a viewpoint, and this is why the editions of the nineteenth and early twentieth century are so valuable. In this sense, corpus linguistics takes philology to the next level, employing essentially the same perspective, but with much richer datasets. Ever more often, linguists also examine ongoing language change and use data from comparatively large contemporary corpora such as the Brown quartet (Brown/LOB for 1961 and Frown/F‑LOB for 1991/92), the TIME Corpus or the BNC (see the contributions by Close & Aarts, Rudanko, Sellgren and Van linden, all in this volume) not only to study variation and change in Present-Day English, but to employ a wider perspective aimed at retrieving regular patterns of linguistic change (for a general discussion of the value of contemporary data, see Mair 2006:  12–35 and Leech, Hundt, Mair & Smith 2009: 1–50). And this is exactly what is needed to capture, and ultimately, explain language change: a wealth of data on one well-defined, clearly bounded phenomenon. In syntax, these smallest observable phenomena are constructions in a broad sense, collocations which first become conventionalized and may then be grammaticalized. Therefore, the investigation of syntactic change starts with an analysis of the elements that are strung together on a regular basis and an examination of their behaviour through time (see, e.g., Ferraresi & Goldbach 2008). Recent studies in English historical linguistics have frequently followed this trend. The use of corpora has by now become part and parcel of the standard methodology analyzing the emergence, variation and developments of syntactic units, collocations, idioms and constructions in all historical stages of English and across varieties, both in a general and in a theory-determined sense. However, there is another dimension to language change, and this is the fact that language change is a behavioural change of speakers, since changes of speech patterns require a change in behaviour. This has become especially clear through the network-based sociolinguistic studies of language variation (see, for instance, Milroy & Milroy 1992). In identifying reasons for such behavioural changes, recent theories on language change have made great progress (see, e.g., Labov 2001



Introduction

and Enfield 2003). One such reason is external influence. The investigation of externally motivated language change depends on the availability of a broad range of linguistic and extra‑linguistic data and the application of a strict methodology (see Thomason & Kaufman 1988), but the explanatory potential of change through language contact has been demonstrated time and time again. This is especially valid for the history of English with its massive impact – not only in the lexicon – from, in particular, Norse (see the discussions in Laing and Ohkado, both in this volume), Latin and French (see, for example, Akimoto and Egan, both in this volume). In recent years, the significance of Celtic influence on the structure of English has become increasingly apparent after it had been underestimated in earlier scholarship (see Filppula & Klemola 2009 and also Wischer, this volume). 2. Functional and distributional approaches to the history of nominal and verbal constructions 2.1

Approaches

A considerable proportion of the articles assembled here apply a usage‑based approach, a majority of them investigating questions of constructions and complementation in the history of English. The general idea of usage-based approaches is to focus on the way certain forms and syntactic units are employed, discussing, for example, the frequencies and kinds of formal properties these constructions display and which functions they perform. Based on such a comprehensive design, the different approaches also emphasize different aspects, such as frequency, pragmatics or the degree of fixed contextual embedding and constructional entrenchment. In a sense, this could be termed a neo-philological or post-philological approach, a move ‘back to the roots’ of linguistics, in which the data are of primary concern (for a discussion of the importance of early Middle English data, see, for instance, Laing, this volume). Interestingly, this parallels the recent comeback of documentary linguistics as a linguistic subfield in its own right, whose primary aim is to document language and to make it accessible to outsiders (see Himmelmann 1998, for an overview of the field). In most of the papers brought together in the present volume, however, the comprehensive and detailed empirical base serves as a foundation for more general and also theoretical inferences, when, for instance, data on the conventionalization of constructions (in a general sense as well as in the sense of construction grammar) are linked up further with theories on grammaticalization (see Egan, this volume) and language change (see, for instance, Sellgren and Rudanko, both in this volume).





Ursula Lenker, Judith Huber, Robert Mailhammer

2.2

Verbal constructions

Many studies in the volume set out to capture data on constructions – whether ad hoc, conventionalized or grammaticalized – and document variation and stability. In a considerable number of the contributions “constructions” such as verb complementation patterns, phrasal units or idioms are investigated without explicit reference to construction grammar as a theory (see Goldberg 1995). The papers by Rudanko, Sellgren and Straaijer, for instance, examine questions of verb complementation from a basically distributional viewpoint, whereas Johannesson’s paper on verb complementation in Old English assumes a distinctly theoretical perspective. Johannesson develops a three-partite system for Old English verbs – according to the dimensions of copularity, transitivity and status (personal vs. impersonal) – and uses it to account for the fact that Old English verbs of naming can take an object complement either marked for nominative or accusative. In one of the distributional analyses of a large corpus (the TIME Corpus), Rudanko shows that the patterns of complementation of accustomed have radically changed in the course of the twentieth century. In the 1920s, the older pattern with to‑infinitive is roughly as frequent as the newer one with to ‑ing. Yet, Rudanko finds that this was not a matter of free variation, but that the new complementation pattern emerged in combination with non‑volitional agents as well as in contexts without complement or adjunct extraction. The older to‑infinitive complementation remained more widespread with volitional agents and in contexts involving complement or adjunct extraction. By the 1960s, however, the infinitival complementation was largely replaced by to ‑ing. In a similar vein, Sellgren (using data from the BNC) compares different complementation patterns of prevent in Present-Day English, the older (prevent) somebody from doing something, and the more recent (prevent) somebody doing something. She proposes that one of the factors that govern the choice of complement is a semantic distinction: the from‑pattern is preferred if the prevented action is a hypothetical one, while the other pattern is chosen more often in less hypothetical cases, where, for instance, the prevented action has been realized before. In an even more fine-grained micro-study, Straaijer focuses on the usage of one individual author. He compares the usage of the auxiliaries be and have with mutative intransitive verbs (such as begin or arrive) in the private writings of the eighteenth-century grammarian Joseph Priestley with a) the rules of his grammar and b) a reference corpus of contemporary letters (for variation in the forms of be in Old and early Middle English and their functions, see also Wischer and Laing, both in this volume). In his Rudiments of English Grammar, Priestley suggests that the choice of auxiliary with these verbs depends on the context, have



Introduction

being more appropriate with reference to past or completed events, be in other cases. Since this distribution is also found in his own usage and is furthermore in line with the eighteenth century’s general usage, Straaijer adds further evidence to the claim that Priestley is a de‑ rather than a pre‑scriptivist. While most of the studies thus deal with constructions in a more general sense, the contributions by Akimoto and Van linden make explicit use of construction grammar in the shape of the approach developed by Goldberg (1995). Akimoto compares the distribution of give + Oi + Od and give + Od + to through the history of English and traces the development of idiomatic expressions such as give rise to from the latter. He finds that the direct object slot in these expressions has ever more frequently been filled with short, eventive nouns from Early Modern English onwards. In a functional synchronic-diachronic description of the clausal complement patterns found with good in extraposition constructions, Van linden shows that PDE good differs from other deontic-evaluative adjectives, such as important or convenient, in that it preferably combines with propositional to‑infinitive clauses (as in it is good to hear that) instead of mandative that‑clauses (as in it is important that you go there). The frequency of propositional to‑clauses with good is attributed to the existence of two partially filled constructions in the sense of Goldberg (1995): the locative pattern and the knowledge/acquisitionof-knowledge pattern. Both of these papers using large corpora illustrate that, as Croft (2001: 11) asserts, one of the fundamental methods of construction grammar is a quantitative and distributional analysis of different diachronic occurrences of a given construction. From this basis, further inferences are possible via frequency‑based models of language change, which ties in further with cognitive theories (see Bybee 2003) and grammaticalization (see Egan, this volume). 2.3

Modality and (marginal) modals

Three contributions in this volume address constructional issues of modality and (marginal) modals in the history of English. Egan’s paper is a good example for linking distributional findings to issues of grammaticalization. He shows that the employment of fail to has changed radically from being almost always negated in the eighteenth century to being hardly ever used with negation in Present-Day English. He discusses its semantic development and argues that fail to is to some extent grammaticalized as a negative marker in Present-Day English. The developmental ancestor of many constructions in a general as well as a theory-specific sense is the collocation, the topic of Nykiel’s paper. He traces the development of the Old English collocation mæg eaþe ‘may easily’, which is on its way to lexicalization in Old English, to its disappearance in Middle English. Nykiel finds that OE eaþe may collocate with mæg in all the verb’s meanings





Ursula Lenker, Judith Huber, Robert Mailhammer

except that of “permission” and suggests that the collocation mæg eaþe reinforced the deontic and possibility‑interpretation of mæg. In their contribution on current changes in the system of English deontic/ epistemic modals, Close & Aarts look at the uses of must and have to/have got to in the Diachronic Corpus of Present-Day Spoken English. In environments were both must and have (got) to are possible, the deontic as well as the epistemic use of must is shown to have declined from the 1960s to the 1990s, while deontic have to, already more frequent than must in the 1960s, keeps rising. Contrary to what one might expect, however, have to is not increasing in its epistemic use, and also have got to, the most recent of the three, is in decline. 2.4

Developments in the English noun phrase

Two papers in this volume focus on features in the diachrony of noun phrases. Bartnik discusses Old English discontinuous quantificational structures, i.e., quantifiers that are not adjacent to the noun or pronoun they quantify. He shows that the group is heterogeneous and offers a classification according to whether or not they show morphological agreement. While the first case involves movement such as topicalization or scrambling, the quantifier is base-generated in the other case. Juvonen’s analysis of s- and of-genitives in late Middle and Early Modern English challenges the claim that the s-genitive rises from being a mere relic in 1400 to becoming a normal genitive variant from the fifteenth to the seventeenth century, arguing that the s‑genitive was the unmarked choice for human possessors all along (particularly widespread in the informal genre of letters), while the of-genitive is more frequent in formal contexts like sermons. Thus genre is seen as the most important factor in genitive variation, with factors such as weight, topicality, givenness and genitive function also playing a role. 3. Syntactic variation and change through contact The second strain of investigation present in this volume focuses on language variation and, in a further extension, on explanation operating with language contact. The papers by Wischer and Laing complement each other in this successive application of research methodology, examining the forms of the copula verbs OE beon/wesan in Old and early Middle English. In her distributional investigation of the double paradigm of beon and wesan in Old English, Wischer finds that the difference between the two verbs, one of future/habitual present vs. (current) present meaning, is linked to a similar one in Celtic. Wischer argues that it is the prolonged contact with Celtic after the



Introduction

Anglo-Saxon settlement that the spread of the b-forms to the subjunctive, imperative and infinitive, and the development of the Old English full double paradigm, an exception among the West‑Germanic languages, is to be attributed to. In a fine-grained study, Laing investigates how the Old English distribution of b‑ and s‑roots of the BE-paradigm, with b‑ more often referring to future time, develops in regional varieties of Middle English. She identifies four different types, some of which continue the Old English distinction in the present indicative singular only, while others – with ar-/er- instead of b-forms in the plural – also maintain the distinction to the plural. While the impact of the Celtic languages (as discussed by Wischer, this volume, in particular) is a topic which has been revived in recent years (see now Filppula & Klemola 2009), the linguistic impact of Old Norse has been a long‑standing and more stable area of research. In his paper, Ohkado re‑examines the evidence for the Norse origins of a particular word order pattern in English. He shows that in Old and Middle English clauses with pronominal or empty subjects, the finite verb tends to be preceded by adjectives, participles, infinitives, negatives and other adverbs more frequently than in clauses with a full NP subject, where the verb tends to be followed by these elements. Ohkado attributes this difference in word order to stylistic fronting of these elements into the subject gap, a phenomenon known, for instance, from Old Norse. On the basis that stylistic fronting can be observed in all Middle English dialects, but even more so in non‑Northern ones, the paper argues (against Trips 2002 and also Ohkado 2006) that it is not only due to Scandinavian contact but has to be a genuine property of Old and Middle English. This volume thus brings together two avenues of research which have been extremely fruitful in the investigation of English historical syntax, namely, functional and data-oriented as well as contact‑linguistic approaches. The key strength of the former perspective is its fine-grained examination of micro‑variation and micro‑change, whereas motivating language change is at the heart of the latter. In recent years, it has become clear that both of these perspectives are indispensable parts of a theory of language change. In particular, the study of constructional change and language variation, which the majority of contributions to this volume are concerned with, has shown to provide crucial data for the description and explanation of syntactic change in the history of English. References Bybee, Joan. 2003. “Mechanisms of Change in Grammaticization: The Role of Frequency”. The Handbook of Historical Linguistics ed. by Brian D. Joseph & Richard D. Janda, 602–623. Oxford: Blackwell.





Ursula Lenker, Judith Huber, Robert Mailhammer Croft, William. 2001. Radical Construction Grammar: Syntactic Theory in Typological Perspective. Oxford: Oxford University Press. Enfield, Nicholas. 2003. Linguistic Epidemiology: Semantics and Grammar of Language Contact in Mainland Southeast Asia. London: Routledge/Curzon. Enfield, Nicholas. 2005. “Areal Linguistics and Mainland Southeast Asia”. Annual Review of Anthropology 34.181–206. Ferraresi, Gisella & Maria Goldbach, eds. 2008. Principles of Syntactic Reconstruction. (= Current Issues in Linguistic Theory, 302.) Amsterdam & Philadelphia: John Benjamins. Filppula, Markku & Juhani Klemola, eds. 2009. Re-Evaluating the Celtic Hypothesis. (= Special issue of English Language and Linguistics 13:2.) Cambridge: Cambridge University Press. Goldberg, Adele E. 1995. Constructions: A Construction Grammar Approach to Argument Structure. Chicago: University of Chicago Press. Himmelmann, Nikolaus. 1998. “Documentary and Descriptive Linguistics”. Linguistics 36.161–195. Labov, William. 2001. Principles of Linguistic Change. Vol. II: Social Factors. Malden, Mass. & Oxford: Blackwell. Leech, Geoffrey, Marianne Hundt, Christian Mair & Nicholas Smith. 2009. Change in Contemporary English: A Grammatical Study. Cambridge: Cambridge University Press. Mair, Christian. 2006. Twentieth-Century English: History, Variation, and Standardization. Cambridge: Cambridge University Press. Milroy, Lesley & James Milroy. 1992. “Social Network and Social Class: Toward an Integrated Sociolinguistic Model”. Language in Society 21.1–26. Ohkado, Masayuki. 2001. Old English Constructions with Multiple Predicates. Tokyo: Hituzi Syobo. Thomason, Sarah & Terrence Kaufman. 1988. Language Contact, Creolization and Genetic Linguistics. Berkeley: University of California Press. Trips, Carola. 2002. From OV to VO in Early Middle English. Amsterdam & Philadelphia: John Benjamins.

part i

Verbal constructions

“Þonne hate we hine morgensteorra” On verb complementation in Old English Nils-Lennart Johannesson Stockholm University

Traditional works on Old English syntax offer no explanation why verbs of naming could combine accusative and nominative complements. This paper offers such an explanation, based on an account of Old English verb complementation in terms of transitivity, copularity and status. transitivity is defined in terms of the number of internal arguments marked for oblique case (none, one or two), copularity in terms of the presence or absence of an internal argument in the nominative case, and status in terms of the presence or absence of an external argument of the verb marked for nominative case. By treating these features as dimensions rather than mutually exclusive categories (in particular transitivity and copularity), it becomes possible to account for combinations of internal arguments marked for nominative case and oblique case, such as þas þing þincað þam arasedum clericum unweorðlice (‘these things seem trivial to experienced clerks’) and þonne hate we hine morgensteorra (‘then we call it the Morning Star’), as well as for verbs of naming showing alternation between, on the one hand, a combination of accusative object and nominative complement, as in the previous example, and on the other hand two accusative objects, as in se steorra ... þone sume menn hatað þone fexedan steorran (‘the star ... which some people call the long-haired star’).

1. Introduction In Old English, we find examples of transitive verbs followed by a combination of an oblique‑case complement and a nominative complement. The combination of dative + nominative is illustrated here by the clause swa þas þing þincað þam



Nils-Lennart Johannesson

arasedum clericum unweorðlice in (1); the combination of accusative + nominative by the clause þonne hate we hine morgensteorra in (2).1

(1) We witon þæt þære næddran attor ys þæs mannes unhælo;

swa [þas þing]N þincað [þam so these.nom things.nom seem.prs.ind.pl the.pl.dat arasedum clericum]D [unweorðlice]N, ac þam þe þæt experienced.pl.dat clerks.dat trivial.pl.nom Lyden ne understandað, hig magon þe leohtlicor witan hwæt (ByrM 1 2.1.419) gerimcræft forstande.2 ‘We know that the serpent’s poison is a man’s undoing; in the same way, these things seem trivial to experienced clerks, but those who do not understand Latin may the more easily know what the calendar signifies.’ (2) þonne hate we [hine]A [morgensteorra]N, forþam then call.prs.ind.pl we him.acc morningstar.nom he cymð eastan up; bodað þære sunnan cyme. (Bo 39.136.4) ‘Then we call it the Morning Star, because it rises in the east; [it] heralds the rising of the sun.’ To the best of my knowledge, the construction with dative + nominative has never disturbed anyone. Visser (1963: 212) merely states, “[i]n Old English þyncan (= to seem) was often used as a copula, both with a personal pronoun in the dative ..., and without such a pronoun”. He never seems to reflect on the status of the dative constituent (which does not have to be a pronoun, as (1) above shows). Mitchell discusses the passive use of þyncan (1985: 436–438) but seems to have nothing to say about the use of a dative NP after þyncan. On one occasion he mentions þyncan in passing as a copula: “copula verbs such as beon/wesan, weorþan, þyncan and others, ...” (1985: 636), without indicating, however, that þyncan is in any way different from the other copulas. The combination accusative + nominative, by contrast, seems to have caused a raised eyebrow or two. Visser comments, “A remarkable feature is the occasional 1. All Old English examples in this paper are quoted from the 2004 release of the Dictionary of Old English Corpus. Texts are identified by means of their Short Short Title, line references are given as stated in the Corpus files. Angle brackets are used in examples, as in the Corpus, to mark words modified by the editor of the text. 2. In the numbered examples, a subscript upper-case letter indicates the case of the preceding noun phrase, adjective phrase or small clause (A = accusative, D = dative, G = genitive, N = nominative); a finite clause used as the complement of a verb is marked Cl. The letter t indicates the trace of a moved constituent; a subscript lower-case letter links the trace to the moved constituent. The relevant verb is bolded. In the modern English translation, lower-case subscript letters are used to indicate identity of reference, [ ] enclose additions, // enclose explanations.



Verb complementation in Old English

use of the nominative after the verb hatan” (1963: 553), and Mitchell acknowledges the existence of the nominative variant, but has no explanation to offer: §1476. When the person or thing named is the object of a verb of naming, the name can be nominative or accusative ... Both usages are idiomatic. I must leave it at that. §1477. ... the naming element ... can be an adjective, which may also be nominative ... or accusative ... My collections are not complete and it may be coincidental that my examples with the nominative are early and those with the accusative late. The reverse would have been less surprising in the light of the loss of inflexions. (Mitchell 1985: 618)

This paper will outline a system for the categorization of Old English verb complementation within which examples such as these can be easily accounted for. Since the verb complementation system is partly dependent on the general syntactic framework employed, a certain amount of ground will need to be covered before we can come back to examples like (1) and (2): in Section 2 below, a brief introduction to the framework used to describe Old English clause structure will be outlined, in Section  3 the verb complementation system will be presented, with some comments on impersonal verbs like þyncan, and in Section 4 the discussion will finally home in on the properties of naming verbs like hatan. 2. Descriptive framework The description of Old English clause structure in this paper will be couched in the framework of Government and Binding Theory (Haegeman 1994). It is not possible in the space available here to present the overall treatment of Old English clause structure within the chosen framework; for an account of the general framework the interested reader is referred to Haegeman (1994), for its application to Old English clauses to Johannesson (2010), in particular Chapter 2. An easily accessible overview of Old English syntax which is generally compatible with that in Johannesson (2010) can be found in Traugott (1992). A few comments about the status of subjects in the current framework may be called for, however, since it differs in certain respects from that of Traugott (1992: 213). A prototypical subject (in the following referred to as a personal subject) of an active clause is base-generated as a Specifier in the verb phrase (VP). This can be demonstrated when a subject noun phrase moves from that position leaving behind a quantifier such as ealle. The personal subject is always marked for nominative case, and is always raised from [Spec, VP] to [Spec, IP], the specifier position in the inflectional phrase. From this position the subject controls Subject-Verb agreement by transferring its number and person (in the singular only) features to the cluster of inflectional features under the I-node, the Head of the inflection phrase. An





Nils-Lennart Johannesson

example of this type of subject is shown in (3), and the derivation of the clause, including the raising of the subject from [Spec, VP] to [Spec, IP], is shown in (4).3 (3) Heora weorc beoð æfre untodæledlice, and hi they.pl.nom habbað ealle ane godcundnysse, have.pl all.pl.nom one.f.acc divine-nature.f.acc and ealle an gecynd and ænne mægenþrymm. (ÆHom 6 247) ‘Their works are ever indivisible, and they have all one divine nature, and all one kind, and one majesty.’

(4) hi habbað ealle ane godcundnysse



CP Spec

C'

hi

C

IP

habbað Spec

I'

hi

3

VP

I

Spec 4

NP NP 1

V' V

Q

hi ealle

NPA

habb- ane godcundnysse

+T pres +A[pl] habbað 2

3. The diagram in (4) also shows the movement of the verb stem habb‑ from V to I to pick up its inflectional features and emerge as the form habbað, the topicalization of the pronoun hi (movement to [Spec, CP]), and finally the verb movement to C, triggered by the topicalization and serving to place the verb in second position in the clause. A moving constituent leaves a trace behind. To simplify the reading of the tree diagrams used here, such a trace is shown as an italicized version of the moved constituent, and an arrow is used to represent the movement. The curved arrows in these diagrams represent the transfer of agreement features from the subject in [Spec, IP] to the I-node.



Verb complementation in Old English

But not every clause has a subject base-generated in [Spec, VP]. In a passive clause the [Spec, VP] is empty, and a subject is provided by raising an object noun phrase to [Spec, IP]. If the raised object was marked for accusative or genitive case in the object position, its case changes to nominative upon moving to [Spec, IP], and it controls Subject-Verb agreement.4 An object marked for dative case, however, retains its case marking and does not control Subject-Verb agreement; the verb assumes its default form, which is identical with the third person singular form. A subject of this kind will in the current framework be called a derived subject. Derived subjects are not only used in passive clauses, but also constitute one option for providing a surface subject in impersonal constructions (see further Section 3 below). Derived subjects in passive clauses are illustrated in (5) to (7); the derivation of the final clause in (7) is illustrated in (8).

(5) Þa cwæð se ealda symeon. To þære eadigan marian. His swurd sceal þurhgan þine saule. þæt swurd getacnade cristes þrowunge.

Næs [seo eadige neg.was the.sg.f.nom blessed.sg.f.nom na ofslegen ne gemartyrod lichomlice: maria]A→N Mary.sg.f.nom not killed.sg (ÆCHom I, 9 254.169) Ac gastlice.5 ‘Then the old Simeon said to the Blessed Mary, “His sword will pierce your soul.” The sword signified Christ’s Passion. The Blessed Mary was not killed or martyred bodily, but spiritually.’ wæs, (6) And þa englas dydon swa [heom]D beboden as them.dat commanded.sg was.3.sg and he astah on heofonas. (LS 1.1 (AndrewBright) 100) ‘And the angels did as they had been commanded, and He /Christ/ ascended to Heaven.’

(7) Helias se witega wæs us mannum gelic. ðrowiendlic swa swa

wæs we. and he swa ðeah abæd. þæt [ren]G→N so-that rain.m.nom was.3.sg 4. Traugott (1992: 213) claims that “only NPs that take accusative in active constructions may be passivised”. The counterpart in the current framework of “passivised” in this quotation would be “raised to [Spec, IP] and changed to nominative case”. However, in Old English clauses with ditransitive verbs which combine dative and genitive objects, the genitive object was raised to become the passive subject, as illustrated in (7). 5. The notation A→N here and in later examples marks an accusative object which has been raised to become a passive subject, thereby shifting case from accusative to nominative.





Nils-Lennart Johannesson

forwyrned. [ðam wiðerweardum folce]D denied.sg the.sg.dat contrary.sg.dat people.sg.dat to ðreora geara fyrste. and syx monða fæce;6 (ÆCHom II, 21 189.277) ‘The prophet Elias was like us men, capable of suffering like us, and he nevertheless prayed so that rain was denied the contrary people for a period of three years and six months.’

(8) þæt ren wæs forwyrned. ðam wiðerweardum folce. to ðreora geara fyrste. and syx monða fæce; C' C þæt

IP Spec

I'

NPN

VP

ren Spec

I V' IP

V

3

+T pret +A[3sg]

bēo-

Spec

wæs

NPN

I' VP

I

ren Spec

V' V'

V 1 4

NPD

–T pp PP +A[masc,sg]

NPG to ðreora geara fyrste. and syx renes monða fæce

forwyrn- ðam wiðerforwyrned weardum folce

2

6. The notation G→N marks a genitive object which has been raised to become a passive subject, thereby shifting case from genitive to nominative.



Verb complementation in Old English

For subjects in impersonal constructions, see Section 3 below. 3. Dimensions of complementation Old English verbs can be categorized with respect to their syntactic properties in terms of three separate dimensions: copularity, transitivity, and status. The first two dimensions are defined on the basis of the case of the complement(s) of a verb (if any), the third dimension is defined on the basis of the kind of subject the verb takes. The definition of the dimension of copularity is based on the presence or absence of a complement marked for nominative case. Thus a copular verb has one complement marked for nominative case, whereas a non-copular verb has no such complement. The most common copular verbs in Old English are, of course, the standard copulas beon/wesan ‘to be’ and weorþan ‘to become’. As we will see, there are also other kinds of copular verbs, even some which simultaneously are transitive: such verbs take both a subject complement (nominative case) and an object (marked for accusative or dative case). The definition of the dimension of transitivity is based on the presence or absence of a complement marked for an oblique (i.e., non-nominative) case. For the purposes of verb complementation, the accusative, dative and genitive cases make up the group of oblique cases in Old English (the remnants of the instrumental case occur only in adverbial expressions). An intransitive verb is one which takes no oblique-case complement, a monotransitive verb takes one oblique-case complement (object), and a ditransitive verb takes two. All three oblique cases can be found after different monotransitive verbs, and all possible combinations of two different oblique cases can be found after different ditransitive verbs. The definition of the dimension of status, finally, is based on the kind of subject that the verb combines with. A personal verb (or, more precisely, a verb which enters into a personal construction) is one which has a subject which originates within the VP (it is base-generated in [Spec, VP]). Such a subject is always marked for nominative case. An impersonal verb (or a verb which enters into an impersonal construction) is one which does not have a subject base-generated in [Spec, VP]. When a clause with an impersonal verb nevertheless typically has a surface subject, that subject is then of a different type: Inserted subject: A “dummy” subject (hit or þæt) may be inserted in the course of the derivation of the sentence, if, for example, the second complement of þyncan is realized by a finite clause, as in (9). There, hit has been inserted in [Spec, IP].





Nils-Lennart Johannesson

(9) Hit þuhte [þam dysegum hæþenum]D It.nom seemed.3.sg the.pl.dat foolish.pl.dat pagans.pl.dat [þæt seo dæd waclic wære]Cl, þæt he wolde sylfwilles on rode [that ... was].clause beon gefæstnod; ... . (ÆHomM 12 125) ‘It seemed to the foolish pagans that the action was despicable, that He would voluntarily be fastened on the cross; ... .’ Raised subject: A second possibility is that the subject of a small clause which functions as the nominative complement of the impersonal verb is raised to [Spec, IP] in the higher clause and thus controls the number and person agreement of the impersonal verb, as shown in (10). (10) We witon þæt þære næddran attor ys þæs mannes unhælo; swa þincað [þam [þas þing]i,N these.nom things.nom seem.prs.ind.pl the.pl.dat arasedum clericum]D [ti unweorðlice]N, ... experienced.pl.dat clerks.dat trivial.pl.nom (ByrM 1 2.1.419) ‘We know that the serpent’s poison is a man’s undoing; in the same way, these things seem trivial to experienced clerks, ... .’ Derived subject: A third possibility for impersonal verbs is that the surface subject is derived from an oblique-case complement (object) of the verb which is raised to [Spec, IP]. As in passive constructions, an accusative or genitive complement switches to nominative case when it moves to [Spec, IP], and the construction takes on a “pseudo-personal” look, as shown in (11). A dative complement, by contrast, remains unchanged, as shown in (12). “Dative subjects” cannot affect the agreement form of the verb; when a dative complement is raised to [Spec, IP], the verb assumes its default form, which is identical with the third person singular form. (11) Almachius se arleasa het þa ardlice gefeccan þa eadigan Cecilian, and hi axode sona hwylcere mægðe heo wære, and hi motodon lange, oþþæt [þam deman]D until the.sg.m.dat judge.sg.m.dat ofþuhte [hyre drystignyss]G→N, and cwæð orhlice eft to annoyed.3.sg her boldness.f.nom þam mædene, Nast þu mine mihte? (ÆLS (Cecilia) 308) ‘The wicked Almachius then ordered [his men] to quickly fetch the blessed Cecilia, and [he] straightway asked her of what family she came, and they argued for a long time, until the judge was annoyed by her boldness, and [he] replied to the maiden scornfully, “Don’t you know my power?”’



Verb complementation in Old English 

(12) For ði [him]D ofhreow [þæs mannes]G: Therefore him.dat rued.3.sg the.sg.m.gen man.sg.m.gen for ðan þe he wæs bepæht mid þæs deofles searocræftum.  (ÆCHom I, 13 281.12) ‘Therefore Hei felt sorry for Manj, because hej had been deceived by the devil’s treachery.’ 4. Analysis The three dimensions of syntactic verb categorization in Old English are independent of one another and must all be specified in the description of a given clause. There are two values in the dimension of copularity, three in the dimension of transitivity, and a further two in the dimension of status. Thus, twelve different combinations of values are theoretically possible. In actual fact, however, only ten of these combinations have been attested: ditransitive copular verbs, whether personal or impersonal, do not seem to have existed in Old English. The ten attested combinations are presented in Tables 1 and 2, where a few brief examples of each type are given; subscript capital letters are used in the table examples to indicate the case of complements. For a comprehensive discussion of all the different personal and impersonal complementation patterns in Old English the reader is referred to Johannesson (2010). The remainder of this paper will be devoted to a discussion of monotransitive copular personal verbs, more specifically naming verbs such as cigan, hatan and nemnan. Monotransitive copular personal verbs take, by definition, one oblique-case complement and one nominative complement; they have a subject base-generated in [Spec, VP]. The oblique-case complement is always marked for accusative case. A few examples of constructions with this type of verb, in addition to (2), are given in (13) to (15). (13) Witodlice he ongan þisne licetere næmnian & mid ceorliscum forþon worde [hine]A cigan [æswica]N, him.acc call.inf hypocrite.sg.m.nom þe he eowde hine sylfne beforan manna eagum III dagas & III niht on gebede standan. (GDPref and 3 (C) 14.200.11) ‘Indeed, hei began to name the otherj a deceiver and with churlish words to call himj a hypocrite, because hej showed himself j before the eyes of men standing in prayer for three days and three nights.’



Nils-Lennart Johannesson

Table 1.  Complementation types, personal verbs (A = accusative, D = dative, G = genitive, N = nominative) Personal constructions (subject base-generated in [Spec, VP]) Non-copular verbs Copular verbs Intransitive

Monotransitive

Ditransitive

... & we forðferdon. (Alex 16.1) ‘... and we marched off.’ ... ðær sind weallas ymbe þone flod ... (HomU 12.2 30) ‘There are embankments on either side of the river.’ & se husbunda ofsloh [ðone man]A. (ChronF 1050.13) ‘and the owner of the house killed the man.’ Beorgað [steopcildum]D ... . (WHom 11 197) ‘Defend orphans.’ ... & þa stowa gelicodon [his eagum]D ... (GD 2 (H) 13.129.14) ‘... and those places pleased his eyes ...’ ... he fægnode [þæs]G ... (ÆLS (Chrysanthus) 24) ‘... he rejoiced at that ...’

[Ðisne anweald]A forgeaf crist [þam apostolum]D. (ÆCHom I, 16 309.75) (D + A) ‘Christ gave this power to the apostles.’ & ic geann [Siferðe]D ... [anes swurdes]G ... . (Ch 1503 60) (D + G) ‘And I bequeath to Siferð ... a sword.’ ... ær þan þe ge [hine]A [æniges þinges]G biddon. (ÆCHom I, 10 261.99) (A + G) ‘... before you ask him for anything.’ ... þu ... [hine]A [þeodscipe ðinne]A lærest ... . (PPs 93.11) (A + A) ‘... you ... teach him your law.’

Ic eom [fiscere]N (ÆColl 87) ‘I am a fisherman.’ ... þa wearð he [dreorig]N (ÆCHom I, 13 284.87) ‘... then he became sorrowful.’ he ongan ... [hine]A cigan [æswica]N (GDPref and 3 (C) 14.200.11) (A + N) ‘he began ... to call him a hypocrite’ þonne hate we [hine]A [morgensteorra]N ... (Bo 39.136.4) (A + N) ‘Then we call it the Morning Star.’ [Þone monað]A we nemnað on ure geðeode [se æftera lyða]N. (Mart 2.1 Jy 0, A.5) (A + N) ‘That month /July/ we call in our language “After-Lithe”.’ NB: cigan, hatan and nemnan are occasionally constructed like læran (non-copular, ditransitive, A + A).



Verb complementation in Old English 

Table 2.  Complementation types, impersonal verbs (A = accusative, D = dative, G = genitive, N = nominative; A→N, G→N: case shift from accusative/genitive to nominative when a complement is raised to subject position in [Spec, VP]; t = trace of moved constituent) Impersonal constructions (no subject base-generated in [Spec, VP]) Non-copular verbs Copular verbs Intransitive

& hit rinde ða ... (Gen 7.12) ‘And then it rained ...’ ... hit sniwð ... . (ÆGram 128.14) ‘... it is snowing ...’ ... & swa miclum Δ sniwde ... (Alex 30.11) ‘... and [it] snowed so much ...’

Monotransitive

... ac syþðan [him]D hingrode. (ÆCHom I, 11 266.9) (D) ‘... but after that he became hungry.’ He læsode his swyn & geseah, þæt hy æton beanbylgas, & [hine]A hingrode. (GD 2 (H) 3.106.28) (A) ‘He fed his pigs and saw that they ate bean-pods, and he became hungry.’ For ði [him]D ofhreow [þæs mannes]G ... (ÆCHom I, 13 281.12) (D + G) ‘Therefore He felt sorry for Man.’ ... forðon þe [þone welan]A lyst [anwaldes]G, & [þone anwald]A lyst [weorðscipes]G, & [þone weorðscipe]A lyst [mærða]G. (CP 33.76.33) (A + G) ‘... because wealth desires power, power desires honour, and honour desires fame.’ ... [þam deman]D ofþuhte [hyre drystignyss]G→N ... (ÆLS (Cecilia) 308) (D + G) ‘... her boldness annoyed the judge ...’ ... þa [se mæssepreost]A→N [þæs mannes]G ofhreow ... (ÆLS (Oswald) 262) (A + G) ‘then the priest felt sorry for the man’

Ditransitive

... ic hit eom. (GDPref and 3 (C) 25.228.21) ‘It is me.’ [Ic Δ eom] [þe ðær syngode] ... (ÆLS (Pr Moses) 257) ‘It was I who sinned there’ ...sei þincð nu [ti wærrest & geapest], þe oðerne mæig beswican ... (LS 28 (Neot) 167) ‘he seems now most prudent and shrewdest, who can deceive another man ...’ ... [þas þing]i,N þincað [þam arasedum clericum]D [ti unweorðlice]N, ... (ByrM 1 2.1.419) (D + N) ‘... these things seem trivial to experienced clerks, ... .’



Nils-Lennart Johannesson

(14) Ða nolde he gecigan god mildheortne. ac het [hine]A but called.3.sg Him.acc [his mildheortnyss]N. þus cweðende. min god and min his mercy.sg.f.nom mildheortnyss; (ÆCHom II, 5 49.219) ‘Then he would not call God merciful, but called Him his mercy, saying thus, “My God and my mercy”.’ we nemnað on ure geðeode (15) [Þone monað]A that.sg.m.acc month.sg.m.acc we call.pl [se æftera lyða]N. the.sg.m.nom second.sg.m.nom Lithe.sg.m.nom (Mart 2.1 Jy 0, A.5) ‘That month /July/ we call in our language “After-Lithe”.’ As both Visser (1970: 553) and Mitchell (1985: 618) point out, an alternative construction with the NP expressing the “name” marked for accusative case also occurs, as shown in (16) to (20). (16) & God gecygde [ða drignysse]A God.m.nom called.3.sg the.sg.f.acc dryness.f.acc [eorðan]A earth.f.acc

& ðæra wætera gegaderunga he het sæ; God geseah

ða ðæt hit god wæs. (Gen 1.10) ‘And God called the dry place “earth” and the gatherings of water He called “sea”; God saw then that it was good.’ [mildheortne]A. (17) Ða nolde he gecigan [god]A then neg-would.3.sg he call.inf God.m.acc merciful.sg.m.acc ac het hine his mildheortnyss. þus cweðende. min god and min mildheortnyss; (ÆCHom II, 5 49.219) ‘Then he would not call God merciful, but called Him his mercy, saying thus, “My God and my mercy”.’ (18) Þa wearð geond eall Engla land swylc tacen on heofenum gesewen swylce nan mann ær ne geseh. Sume menn cwædon þæt hyt cometa se steorra wære [þone]A which.sg.m.acc sume menn hatað [þone some.pl.m.nom people.pl.m.nom call.pl the.sg.m.acc



Verb complementation in Old English 

fexedan <steorran>]A, & he æteowde ærest on long-haired.sg.m.acc star.sg.m.acc þone æfen Letania Maiora, þæt ys VIII Kł. Mai., & swa scean ealle þa VII niht. (ChronC 1066.4–6) ‘Then throughout England there was such a sign seen in the sky as no man had ever seen. Some men said that it was the star “comet”, which some people call “the long-haired star”, and it appeared first on the eve of the Greater Litanies, that is 24 April, and so it was shining for seven nights.’ (19) Sume næmmeð [þone cæstel]A some.pl.m.nom call.pl that.m.acc fortress.m.acc [Magdalum]A, þe Maria wæs of Magdalenisc geclypod, & Magdalus.m.acc þæt becumð wel to þyssere trahtnunge. For Magdalus is stepel geclypod, & betacneð eadmodnysse. (InFestisSMarie 56) ‘Some call that fortress “Magdalus”, from which Mary was called “Magdalen”, and that befits this exposition well. For Magdalus means “steeple”, and signifies humility.’ (20) Sume eac sæcgeað þæt heo stelan mid twigum hysopan gelicne hæbbe & leaf beanum gelice, þanon [hy]A sume whence it.f.acc some.pl.m.nom men þam sylfum naman nemnað [hysopan]A. people.pl.m.nom call.pl hyssop.sg.f.acc  (Lch I (Herb) 57.1) ‘Some people also say that it /a plant/ has a stalk with branches similar to hyssop and leaves similar to beans, whence some people call it hyssop by that very name.’ It will be proposed here that the verbs of naming in this kind of construction have the same complementation pattern as læran ‘teach’, as illustrated in (21). In other words, the verbs of naming are used as ditransitive non-copular personal verbs with two accusative objects. (21) Ond þa sona het se biscop ðone cniht him to gelædan ond [gastlicne wisdom]A, [hine]A lærde him.acc taught.3.sg spiritual.sg.m.acc learning.sg.m.acc ond he wæs eall mid godes snyttro gefylled. (Mart 5 My 2, A.11) ‘And then straightway the bishopi commanded that the boy j should be led to himi, and [hei] taught himj spiritual learning, and hej was completely filled with God’s wisdom.’



Nils-Lennart Johannesson

When a verb of naming used as a monotransitive copular personal verb is passivized, the accusative object is raised to become the passive subject, marked for nominative case, as shown in (22) to (26). The nominative “name” complement remains in the nominative case. (22) Him he getimbrade in Suðrigna lande bi Temese streame in is geceged [Ceortes eig]N ... stowe, [seo]N which.3.sg.f.nom is called.sg Ceort’s island.sg.f.nom  (Bede 4 7.282.12) ‘For himself he built [a monastery] in Surrey by the River Thames in a place which is called Chertsey ...’ cnapa byst [þæs (23) And [þu]N and you.sg.nom boy.m.nom are.2.sg the.sg.m.gen genemned, þu gæst hehstan witega]N Highest.sg.m.gen prophet.sg.m.nom called.sg beforan Drihtnes ansyne his wegas gearwian ... (Lk (WSCp) 1.76) ‘And you, boy, will be called the prophet of the Highest; you will go before the face of the Lord to prepare His ways.’ (24) Hælendes þegna sum þa wæs swyþe gebolgen,

[se]N wæs he.nom was

haten [Iudas se Scariothisca]N; called Judas.nom the.sg.m.nom man-of-Scariot.sg.m.nom forþon he com of þæm tune þe Scariot hatte; he cwæþ, To hwon sceolde þeos smyrenes þus beon to lore gedon? (HomS 21 45) ‘One of Jesus’ disciples was then greatly annoyed, he was called Judas Iscariot; because he came from the town that was called Scariot/Kerioth/; he said, “Why should this ointment be wasted thus?”’ hatte (25) [Se oðer monð]N The.sg.m.nom second.sg.m.nom month.sg.m.nom is-called [Februarius]N.  February.sg.m.nom ‘The second month is called February.’

(ByrM 1 2.1.17)

(26) Saga me hwilc wyrt ys betst and selust. Ic þe secge, hatte [seo wyrt]N [lilige]N lily.sg.f.nom is-called the.sg.f.nom plant.sg.f.nom for þon þe heo getacnað Crist. (Sol I 28.1–2) ‘Tell me which plant is best and most excellent. I tell you, “lily” is the name of the plant, because it signifies Christ.’



Verb complementation in Old English 

Passivization of Old English verbs with two accusative objects seems to have been avoided (cf. Traugott 1992: 213; Mitchell 1985: §835 on the lack of Old English passive constructions of the type ‘he was taught singing’). The verb læran could alternatively be constructed with the person-denoting object marked for dative case (27); this construction can occasionally be found in the passive voice (28). Alternatively, the “thing taught”’ could be marked for dative case, with the person-denoting object marked for accusative case, which was changed to the nominative when the clause was passivized (29). The most common passive use of læran, however, involves a person-denoting passive subject in the nominative case (derived from an accusative object) and a prepositional phrase expressing the “thing taught” (29). (27) Hæfde he se ilca cyning þæs biscopes broðor mid hine, se wæs Cælin haten; se wæs mæssepreost & <efenlice> Godes mon til: [godcunde se [him & his hiwum]D he.nom him.dat and his household.pl.dat divine.sg.f.acc lærde, & þa geryno þegnade þæs halgan lare]A lore.sg.f.acc taught.3.sg geleafan. (Bede 3 17.230.12) ‘That same kingi had the bishop’s brotherj, who was called Cælin, with himi; hej was a priest and equally a good man of God: hej taught himi and hisi household the divine lore, and served the mysteries of the holy faith.’ (28) Wæs in ða tid ðeaw Ongolcynnes folcum þæt, þonne mæssepreost oðþe oðer in tun com, ðæt hie ealle to his bebode gesomnedon Godes word to gehyrenne; & fuslice gehyrdon, ða ðe [him]D gelærde wæron, ond eac [pro]A→N [which.pl.nom] part them.dat taught.pl were.pl swylce neodlice mid dædum læstton, ða ðe hie ongeotan (Bede 4 28.362.23) mehton.7 7. My interpretation of the two occurrences of the pronoun ða in the last two lines of (28) differs from what is suggested by the editor’s punctuation. I take ða to be a plural accusative form, the object of gehyrdon and læstton, respectively. The first ocurrence of ða is the antecedent of the relative pronoun in the relative clause ðe him gelærde wæron. The relative pronoun in that clause is a phonologically empty (and hence invisible) pronoun, which is fronted to [Spec, CP] in the relative clause. This pronoun will here be labelled PRO; whether this is the best label or not (cf. the discussion in Haegeman 1994: 422–426) is irrelevant to our present concerns. In this clause PRO starts out as the accusative object (‘the thing taught’) of læran, which is first raised to [Spec, IP] to become the passive subject (shifting case in the process from accusative to nominative), and is later raised to [Spec, CP]. Its presence and its number (plural) and case (nominative) are shown by the fact that from its temporary position in



Nils-Lennart Johannesson

‘At that time it was a custom among the people in England that when a priest or someone else came to a village, they all gathered on his command to hear the word of God; and they listened eagerly to those [things] that they were taught, and also diligently carried out in their actions those which they could understand.’ (29) Þa wæs se Godes mon bebyrged in Sancte Paules cirican þæs [he]A→N apostoles, [þæs monungum]D whose.sg.m.gen admonitions.f.dat he.nom

gelæred wæs & geleornade, þæt he ða heofonlecan eadignesse taught.sg was.3.sg gehyhte. (Bede 4 14.296.32) ‘Then the man of God was buried in the church of St. Paul the Apostle, whose admonitions he had been taught and [had] learnt, so that he trusted in the heavenly bliss.’

(30) Eac swylce he sumne æðelne cyricsangere begeat, se wæs Mafa on Cent [on sangcræft]PP gelæred haten, [se]A→N wæs he.nom was.3.sg in Kent in singing.m.acc taught.sg fram æfterfyligendum þara discipula þæs eadigan papan Sancte Gregorii; & he þone XII winter mid him hæfde. (Bede 5 18.466.16) ‘Similarly hei /Acca, bishop of Hexham/ received a famous cantorj who was called Mafa; hej had been instructed in singing in Kent from successors of the disciples of the blessed pope St Gregory; and hei had himj with himi for twelve years.’ Some verbs of naming, by contrast, do occur in passive constructions with the object referring to the “thing/person named” raised to become a nominative passive subject, and with the “name” object retaining its accusative case marking, as shown in (31) to (35). Admittedly, these examples are found in contexts where Latin influence will have been strong, and the “name” NPs are mostly Latin names of various kinds. Example (31) is the only one I have been able to find with unambiguous Old English accusative forms. But to the extent scribes recognized the Latin accusatives as accusatives, these constructions do fit the general pattern of passivization: only the object that is raised to the position of the passive subject has its case changed to the nominative.

[Spec, IP] it controls subject-verb agreement, resulting in the plural forms of the past participle (gelærde) and the finite verb (wæron).



Verb complementation in Old English 

(31) Ond æfter Cristes upastigennysse he gecyrde twa mægða to godes geleafan, [þa]A→N wæron þus genemned, those.nom were.pl thus called ond [Achaiam [Scyððiam þa mægðe]A Scythia.f.acc the.sg.f.acc nation.f.acc and Achaia.f.acc þa mægðe]A. the.sg.f.acc nation.f.acc (Mart 5 No 30, A.6) ‘And after Christ’s Ascension he /St Andrew/ converted two nations to the faith of God, they were called thus, the nation of Scythia and the nation of Achaia.’ (32) And hi þa ongunnon ealle heora godas, & þær wæs [sum]A→N genemned [ypochratin]A & there was.3.sg one.m.nom called.sg Hippocrates.m.acc [oðer]A→N [mercurium]A & [se þridda]A→N [galligenum]A.  (LS 30 (Pantaleon) 197) ‘And they then began to invoke all their gods, and there was one called Hippocrates and another [was called] Mercury and the third [was called] Galligenus.’ (33) And hi þa drihtene lof sungon & sægdon, & ferdon in þa dune þe is nemned [Montem Olifete]A. [pro]A→N [which.sg.f.nom] part is called.sg Mount.m.acc Olivet  (HomS 19 47) ‘And they then sang and spoke in praise of the Lord, and went to the hill that is called the Mount of Olives.’ (34) Þa sende se cyning sumne heretogan, [Holofernem]A gehaten, mid mycelre fyrdinge and bebead Holofernes.m.acc called.sg him þas word: Ne wanda þu nan ðing, ne ara ðu nanum rice, ne ænigre burhscire, ac ælce getrymmede burh þu gebige to me.8  (ÆHomM 15 45) ‘Then the king sent a certain general, [who was] called Holofernes, with a great army and commanded him [with] these words: “Do not desist from anything, do not spare any country, nor any city, but subject each fortified city to me.”’ 8. The past participle clause Holofernem gehaten can be regarded as an abbreviated version of a passive relative clause, se wæs Holofernem gehaten.



Nils-Lennart Johannesson

(35) [Se stan]A→N is haten [piriten]A. That.m.nom stone.m.nom is called.sg pyrite.m.acc ‘That stone is called “pyrite”.’

(Lap 24)

5. Conclusion The system for categorization of Old English verb complementation outlined in this paper was originally developed in an attempt to come to terms with the kind of syntactic variation found in clauses with impersonal verbs, as illustrated in examples (9) to (11) above. The three-dimensional verb complementation model together with the recognition of similarities between passive uses of personal verbs and the provision of different types of subjects (inserted, raised, derived) in clauses with impersonal verbs made it possible to distinguish “pseudo-personal” uses of impersonal verbs (as shown in (11), as well as in Table 2) and constructions with personal verbs proper, which have a subject base-generated in [Spec, VP]. The analysis of the use of naming verbs in clauses like þonne hate we hine morgensteorra and se steorra ... þone sume menn hatað þone fexedan steorran emerged as a spin-off effect of the adoption of the three-dimensional verb complementation model. In other words, the system turned out to be capable of accounting for cases it had not primarily been developed to handle. Mitchell (1985: 453) warns against “elaborate and rigid schemes of classification”. It has hopefully emerged that the classification system proposed here is, by contrast, simple and flexible, while at the same time powerful enough to handle all types of Old English verb complementation patterns, including those used by verbs of naming. References Aarts, Bas. 2008. English Syntax and Argumentation. 3rd ed. Oxford: Blackwell. Haegeman, Liliane. 1994. Introduction to Government & Binding Theory. 2nd ed. Oxford: Blackwell. Johannesson, Nils-Lennart. 2010. Stæfcræft. An Introduction to Old English Syntax. 3rd ed. Stockholm: Dept. of English, Stockholm University. Mitchell, Bruce. 1985. Old English Syntax. Volume I. Concord, the Parts of Speech, and the Sentence. Oxford: Oxford University Press. Traugott, Elizabeth Closs. 1992. “Syntax.” The Cambridge History of the English Language. Vol. 1. The Beginnings to 1066 ed. by Richard M. Hogg, 168–289. Cambridge: Cambridge University Press. Visser, Fredericus Theodorus. 1963. An Historical Syntax of the English Language. 1, Syntactical Units with One Verb. Leiden: Brill.

Tracking and explaining variation and change in the grammar of American English A case study, with evidence from the TIME Corpus Juhani Rudanko

Department of English, University of Tampere* Both to infinitive and to ‑ing complement clauses selected by adjectival heads may involve subject control in English, but there are syntactic differences between them. In spite of the differences, some matrix predicates have shown variation and change between the two types of sentential complement in recent times. The article examines the adjective accustomed from this point of view, with evidence from the TIME Corpus. It is observed that in the first decade covered by the corpus, from 1923 through 1932, to infinitives were slightly more frequent than to ‑ing complements, but that in the 1960s the to ‑ing pattern had become predominant in comparison with the to infinitive pattern. The article points to such variation and change affecting the sentential complements of accustomed in one particular text type of American English, and identifies two explanatory principles to account for the variation observed. The first is the Extraction Principle, which is defined more broadly than in some earlier work, and the second is the postulation of a semantic contrast between the two complementation patterns. It is argued that both the Extraction Principle and the semantic distinction are statistically significant in explaining variation at one stage in the overall process of change that has been called the Great Complement Shift.

* I am grateful to the participants of the ICEHL conference, the anonymous referees and the editors for their comments and suggestions. I am also grateful to Ian Gurney, Tiina Harjula, Veera Peteri, and Elina Sellgren, all of the University of Tampere, for helping me finalize the article. All remaining shortcomings are my responsibility.



Juhani Rudanko

1. Introduction This study examines variation and change between two patterns of sentential complementation selected by the adjective accustomed in recent American English. To introduce the patterns independently of this adjective, consider (1a–b): (1) a. John is eager to change the agreement. b. John is averse to changing the agreement. The pattern of (1a) may be termed the to infinitive pattern, and the pattern of (1b) the to ‑ing pattern. In the latter pattern the ‑ing form is a gerund. (1a) and (1b) are similar in that each contains a matrix adjective that selects a complement. It is assumed here that the complement is sentential, with its own subject even though the subject is covert or implicit. This assumption was for instance made by traditional grammarians, including Jespersen, who writes: Very often a gerund stands alone without any subject, but as in other nexuses (nexus substantives, infinitives, etc.) the connexion of a subject with the verbal idea is always implied. (Jespersen [1940] 1961: 140)

It is convenient here to follow this traditional approach, in order to represent the subject arguments of the lower predicates in (1a–b). Another similarity between the sentences of (1a–b) is that each higher predicate, be eager and be averse, assigns a semantic role to its subject. The sentences with their sentential complements are therefore control constructions, rather than NP Movement constructions. Both involve subject control, with PRO as the lower subject, to use a term from current work. However, there are also important differences between (1a) and (1b). First, the complements are not interchangeable: (2) a. *John is eager to changing the agreement. b. *John is averse to change the agreement. Second, the string that follows to may be replaced by the pronoun it or that in the case of (1a), but not in the case of (1a):

(3) a. *John is eager to change the agreement, but I am not eager to that. b. John is averse to changing the agreement, but I am not averse to that.

Further, the string that follows to is subject to ellipsis in (1a), but not in (1b): (4) a. John is eager to change the agreement, but I am not eager to. b. *John is averse to changing the agreement, but I am not averse to.



Variation and change in American English

One way of accounting for the differences observed is to make the traditional assumption that the to of (1a) should be analyzed differently from the to of (1b). In particular, the to of (1a) may be termed the infinitival marker,1 to adopt a term from Quirk et al. (1985: 1178, note a). For its part, the to of (1b) is a preposition. In syntactic terms, the infinitival marker to is under the Infl node (Chomsky 1981: 18–19), or the Aux node, to use more traditional terminology. By contrast, the prepositional to is under a P node, introducing a prepositional phrase, with the prepositional phrase consisting of the preposition and what may be termed a nominal clause, to adopt a term from traditional grammar. A nominal clause may be viewed as a sentence dominated by a NP node. Analyzed in the way described, the sentences of (1a–b) may be represented as in (5a´–b´):

(5) a.´ [John is [eager]Adj [[pro]NP [to]Aux [change the agreement]VP]S2]S1

b.´ [John is [averse]Adj [to]Prep [[[pro]NP [changing the agreement]VP]S2]NP]S1 The ellipsis constructions of (4a–b) involve the ellipsis of a VP (cf. Huddleston & Pullum 2002: 1526), and the result is well-formed in (4a) because the string that follows to is a VP in (5a´), whereas this is not the case in (5b´), and (4b) is therefore ill-formed. As for the pronoun usage of (3a–b), since the string that follows the word to is a NP in (5b´), it may be replaced by a pronoun of the type of it or that, that is, of the type of pronoun that may generally replace NPs. In spite of robust differences between the two types of complement, it has been observed in the literature (for instance, Kjellmer 1980; Denison 1998: 266–267; Rudanko 1998: Chapter 3; Rudanko 1999: Chapter 2; Rudanko 2000: Chapter 2, 3 and 4; Vosberg 2003a; 2003b; Vosberg 2006) that there are matrix predicates, including verbs, adjectives, and nouns, that have shown variation and change between the two constructions in recent centuries. In broad terms, it is the purpose of this article to examine one such predicate, the adjective accustomed. The general story of accustomed is well known (Kjellmer 1980; Vosberg 2003a: 314–315; Rudanko 1999:  9–10; 2006; 2007; Leech et al.  2009:  183–184). In the eighteenth and nineteenth centuries it was frequent with sentential complements involving subject control, and those complements were overwhelmingly of the to infinitive type. Some tokens of to ‑ing complements have come to light in the literature, but the frequency of such tokens in these two centuries is very low (Rudanko 2006). The predominance of the to infinitive pattern is also reflected in the OED treatment of the adjective (s.v. accustom, vb., 3.b). There are altogether four 1. Even though the term “infinitive marker” is used here, it is important to point out that this does not mean that I hold the view that the infinitival to is necessarily without any meaning or without any semantic contribution. For some discussion, see Rudanko (1989: 34–35).





Juhani Rudanko

illustrations of sentential complements involving subject control given, all of the to infinitive variety and none of the to ‑ing type. However, in Present-Day English it is in fact the to ‑ing pattern that predominates in both British and American English. The predominance is less pronounced in a more conservative text type as represented by the “Books” segment of the Bank of English Corpus, but is nearly total in the more agile text types of newspaper English and of spoken English (see Rudanko 2007). Overall, the competition of the to ­‑ing pattern with the to infinitive pattern and the spread of the former at the expense of the latter illustrate one aspect of a set of changes in the system of English predicate complementation that Rohdenburg (2006) has recently termed the Great Complement Shift. The present article aims to shed further light on principles that may explain variation in the argument structure properties of accustomed and on the process of change that has resulted in the present situation. Most of the earlier work on the adjective has been on British English, but the focus here is on a text type of American English. The data come from the new TIME Corpus, compiled by Mark Davies at Brigham Young University. The corpus comprises some one hundred million words in total, and two decades were chosen for the present investigation, namely, 1923–1932 and the 1960s. The corpus starts from the beginning of 1923, and in order to reach as far back as possible, it seems appropriate to select the first decade of the corpus, from 1923 to the end of December 1932, as the first segment to be examined. As regards the choice of the 1960s, it may be noted that present-day usage of accustomed was for instance investigated in Rudanko (2006: 39–45), and the 1960s is a decade that is approximately in the middle of the period that extends from the first decade of the TIME Corpus to the present day. In the case of each decade the first task is to ascertain the facts regarding the use of the two types of sentential complements involving subject control. Going beyond descriptive findings, the study also identifies and discusses two principles that may be invoked to explain change and variation in the system of English predicate complementation. 2. The adjective accustomed in the first decade of the TIME corpus 2.1

Variation in the data and the extraction principle

The search string adopted is simply accustomed. This is ideal from the point of view of recall, and is possible from the point of view of precision because the number of irrelevant tokens is not excessively high.



Variation and change in American English 

For the first decade of the TIME Corpus the search string retrieves 209 tokens. The total number of words for this period in the corpus is 11.3 million. The raw figure thus represents a frequency of 18.5 per million words. The search string also retrieves a number of tokens that are irrelevant. Sometimes the word form accustomed is a form of the verb accustom, as in (6), rather than of the adjective.

(6) As the eye accustomed itself to the religious dim, it caught the gleam of many a fine garment. (1925)

Usages where accustomed is a premodifier inside an NP, as in (7), can be set aside in the study of the complementation patterns of the adjective:

(7) ... he played in his accustomed effective style. (1923)

The adjective accustomed is also frequently found with to NP complements, as in (8):

(8) For people who are not accustomed to it office is a very heady wine. (1924)

As argued in Section 1, a to ‑ing complement involves a nominal clause, and it is therefore of interest to note the presence of such to NP complements in the corpus, but the focus here is on sentential complements involving subject control. With these exclusions made, I counted 95 relevant tokens in the material. This represents a frequency of 8.4 per million words. Table 1 gives information on the two types of sentential complement. Table 1.  To Infinitive and to ‑ing Complements of Accustomed in the First Decade of the TIME Corpus

to infinitive to ‑ing

raw frequency

normalized frequency

53 42

4.7 3.7

A first conclusion is therefore that both types of complement were found in robust numbers in the text type of American news magazine English during the first decade of the TIME Corpus. Here is an initial illustration of each type: (9) a. ... Captain N. E. Cousins of the liner Ruth Alexander, accustomed to navigating Pacific coastal waters for 30 years. (1923) b. ... I want them to become accustomed to respect the popular voice of the people ... (1924)





Juhani Rudanko

Given the amount of variation in the material, it is of interest to attempt to identify explanatory principles to account for the variation. Two such principles are examined here. The first is the non-semantic factor of extraction. To appreciate the potential impact of extraction, it is appropriate to quote Vosberg’s original definition of the Extraction Principle: In the case of infinitival or gerundial complement options, the infinitive will tend to be favoured in environments where a complement of the subordinate clause is extracted (by topicalization, relativization, comparativization, or interrogation, etc.) from its original position and crosses clause boundaries. (Vosberg 2003a: 308)

The Extraction Principle is an important principle explaining variation between different complement options. However, extractions in English are not limited to the extraction of complements.2 Adjuncts may also often be extracted, and while it is worth keeping the distinction between complements and adjuncts in mind as an important theoretical distinction, a broader definition of the principle is applied here, and the word “complement” of Vosberg’s definition is replaced with the word “constituent”. This is done in order to take the extraction of adjuncts into account. The reason underlying the Extraction Principle has to do with two interconnected considerations. First, it is unquestionable, in view of John Robert Ross’s work in the 1970s, that to infinitive complements are more explicitly sentential than ‑ing complements (Ross 1973). The latter are sentential complements, but there is a gradient of sententiality among sentential complements, and ‑ing complements are at the nouny end of that scale. Here it is again also possible to recall the grammatical differences between the two patterns discussed in Section  1, where it was argued that a to ‑ing complement involves a nominal clause and a to infinitive is sentential without being a NP. Second, it is also unquestionable that the extraction of a constituent, whether of a complement or an adjunct, produces a more complex sentence compared to the canonical, non-extraction version of the sentence. Given these two considerations, it is possible to view the Extraction Principle as a manifestation of Rohdenburg’s Complexity Principle: In the case of more or less explicit grammatical options the more explicit one(s) will tend to be favored in cognitively more complex environments. (Rohdenburg 1996: 151) 2. The possibility of a broader interpretation of extraction is also raised in Vosberg (2006: 69) and Rohdenburg (2006: 151). The latter source thus refers to “extracting postverbal elements out of complement clauses”, without restricting such elements to complements.



Variation and change in American English 

In other words, since extraction produces a complex environment, a more explicitly sentential complement – the to infinitive – tends to be preferred over a less explicitly sentential complement – the to ‑ing complement. Applying the modified version of the Extraction Principle to the present material, the results given in Table 2 are obtained. Table 2.  To infinitive and to ‑ing Complements in Canonical and Extraction Contexts in the First Decade of the Corpus to infinitive raw normalized frequency frequency canonical extraction

41 12

3.6 1.1

to ‑ing raw normalized frequency frequency 40  2

3.5 0.2

Applying the Chi square test to these findings, with the Yates correction factor, the Chi square value is 4.62, and the results are significant at the .05 level. Here are the two illustrations of extractions out of to ‑ing complements and some illustrations of extractions out of to infinitive complements, with the extraction rule in question spelled out in brackets after each illustration: (10) a. ... the emotional appeal of the case is emphasized much more than we are accustomed to seeing. (1924) [Comparativization] b. Think of all the adjectives I can now employ! Where I have been accustomed to using “great,” “magnificent,” “heart-rending,” I ... (1927)  [Wh-Movement] (11) a. The States of Europe descended from the mountains of animosity from which they have long been accustomed to glare at one another, went down the rugged mountain paths of doubt, ... (1924) [Relativization] b. ... a notebook of 90 chaotic pages in which Coleridge was accustomed to scrawl the names of books which he had read or intended to read, ... (1927) [Relativization] c. We sit in our library, in a room where we are accustomed to study and reflect, where all the surroundings are natural. (1928)[Relativization] d. Before long Hudkins’ coarse face, misshapen by the beatings he is accustomed to take even when he wins a fight, was made even more than normally ferocious ... (1929) [Relativization] e. Such advertisements socialite Japanese matrons have long been accustomed to read in magazines of the highest class ... (1930)[Topicalization]



Juhani Rudanko

The illustrations show that extractions of complements and extractions of adjuncts are found in the material and that more than one kind of extraction rule is encountered. For instance, in (10a) the extraction rule operative is Comparativization, and the constituent extracted is a complement out of a sentential complement of the adjective accustomed. On the other hand, while (11d) and (11e) also involve the extraction of a complement, the extraction rules involved are Relativization and Topicalization, respectively. Extractions of adjuncts appear often to involve adjuncts of place, as in (10b) and in (11a–c). It is also clear in the light of these illustrations that the extraction rule that is most commonly found with extractions of adjuncts is Relativization. 2.2

A semantic distinction

The other explanatory principle is semantic. In an earlier publication I observed that with a to ‑ing complement the adjective has a sense close to that of used to, with the “complement of the adjective expressing a regular situation” (Rudanko 2006: 39). As regards to infinitive complements the sense of the adjective may be close to that of ‘tend’, with the complement of the adjective expressing a regular practice. There may thus be more of a sense of choice on the part of the referent of the matrix subject in the case of the to infinitive complement than in the case of the to ‑ing complement. (I am grateful to Ian Gurney, personal communication, for commenting on the distinction.)  (Rudanko 2006: 39–40)

Both the construction accustomed + to infinitive and the construction accustomed + to ‑ing are subject control constructions, and this means that the higher subject and the lower subject are coreferential. It is then possible to take up the suggestion made in Rudanko (2006) that accustomed with the to infinitive pattern is associated with a sense of choice and that accustomed with the to ‑ing pattern is associated with lack of choice and to subject it to a systematic investigation in the present material, where, as we have seen, there is abundant variation between the two patterns. Since the extraction factor was found to be significant above, the sentences involving extraction are excluded from this meaning‑based investigation (see also note 4 below). Lower subjects expressing or encoding a sense of choice on the part of the lower subject may be designated as +Choice, and those that encode lack of a sense of choice may be designated as –Choice. The same terms are used to characterize the predicates in question. The notion of choice is here interpreted on the basis of the semantic role of Agent. This association is appropriate because choosing implies a volitional act and intentional engagement, and because the concept of volitional engagement in



Variation and change in American English 

the event or state is an important feature of what Dowty (1991: 572) has termed the “Agent Proto-Role”. For his part, Jackendoff (1990:  129) identifies three “semi-autonomous parts” when summing up his view of an Agent: “doer of action”, “volitional Actor”, and “extrinsic instigator”. In a similar vein, the notions of “volition, control, and responsibility” are prominent in Hundt’s (2004: 49) discussion of agentivity. A +Choice lower subject is therefore viewed here as one that is agentive, and a –Choice lower subject as one that is ‑agentive. Some illustrations of +Choice lower subjects are given in (12a–d), and of –Choice lower subjects in (13a–d). (12) a. Speculators are now accustomed to buy White Rock on margin or–if they dare–sell Canada Dry Ginger ... (1926) b. To all appearances the excessively freedom-loving Riffi, each of whom is accustomed to guard his personal freedom with his own rifle, can not be made ... (1926) c. Anselmo Bonin, accustomed to boasting that he had a great fortune in gold and who was always ready to lend to his neighbors, was disturbed by a visit from four masked bandits. (1923) d. Washington political experts are accustomed to writing of Borah as the one man in the Senate who “goes his own gait” – regardless of what people say; ... (1923) (13) a. The hearts of real Presbyterians had rather grown accustomed to be saddened by the utterances of Dr. S. Parkes Cadman, ... (1929) b. The child becomes accustomed to see the law disobeyed ... (1930) c. ... for the Anglicans are thoroughly accustomed to being a minority bloc in a Church which ... (1928) d. Harvardmen recall that last spring, long accustomed to being driven by a chauffeur. (1931) Predicates of the type of buy White Rock, as in (12a), guard one’s freedom, as in (12b), boast that he had a fortune in gold, as in (12c), and write of Borah as the one man ..., as in (12d), all encode an event or a situation that involves an agentive or +Choice subject. By contrast, predicates of the type of be saddened by the utterances, as in (12a), see a law disobeyed, as in (13b),3 be a minority bloc, as in (13c), and be driven by a chauffeur, as in (13d), all encode an event or a situation that involves a non-agentive or –Choice subject. 3. On viewing see as a non-agentive verb, in uses that are similar to those discussed in the text, cf. Gruber (1967: 943), who contrasts it with the verb look, which is typically agentive. Compare also Taylor’s (2003: 30–31) comments on see and look from the point of view of the notion of choice.



Juhani Rudanko

It is possible to strengthen such intuitions with tests. One such test that is salient in the present context is the imperative construction, for an imperative tends to be more easily compatible with an agentive predicate than with a non-agentive predicate. Here it is appropriate to quote a remark on imperatives and choice by Taylor: Prototypically, an imperative instructs a person to do something, and is therefore only acceptable if a person has a choice between carrying out the instruction or not. (Taylor 2003: 31)

Applying this test in the present context, we might for instance compare (12a) and (13b). Thus Buy White Rock! is readily conceivable as an imperative, but ?Be saddened by the utterances! is less so, motivating the distinction made between the two predicates here. The quotation from Taylor also links the imperative to the notion of choice. It may be added that the lower sentence may be in the passive, as for instance in (13a), and in that case the subject represents the object argument of the verb in terms of argument structure. Such a subject is prototypically non-agentive and –Choice because it prototypically carries the Patient or Undergoer role. It is clear from the illustrations given in (12a–d) and (13a–d), that the semantic principle that the accustomed + to infinitive pattern is linked to a +Choice lower subject and the accustomed + to ‑ing pattern is linked to a –Choice pattern cannot be maintained as an absolute rule. Thus (12c–d) illustrate to ‑ing complements with +Choice subjects, and (13a–b) illustrate to infinitive complements with –Choice subjects. However, while the principle is not always adhered to, it is still appropriate to inquire whether it may have validity as a tendency. When all the relevant sentences are examined in the present material, the results given in Table 3 are obtained. The Chi-square value is 6.42 and the results are statistically significant at the .05 level of significance (df = 1).4 The present discussion suggests, therefore, that lower predicates with –Choice subjects offered a particularly fertile ground for the spread of the to ‑ing pattern with the adjective accustomed. By contrast, the to infinitive complement was linked to scenarios where the lower clause encoded the lower subject as expressing a degree of agentivity. 4. As noted, the 14 tokens involving extraction were excluded from the present figures, since extraction was found to be a significant factor bearing on variation. However, it may be observed that as regards the two to ‑ing tokens, one of them has a +Choice and the other a –Choice lower subject, but little can be concluded from them because of their low number. As for the 12 to infinitives with extraction, it turns out that as many as 11 of the 12 have +Choice subjects. They therefore strengthen the tendency observed in the text that is based on the tokens in canonical contexts.



Variation and change in American English 

Table 3.  To infinitive and to ‑ing Complements with +Choice and –Choice Lower Predicates in the First Decade, Excluding Extraction Contexts to infinitive raw normalized frequency frequency +Choice –Choice

35  6

to ‑ing

3.1 0.5

raw frequency

normalized frequency

23 17

2.0 1.5

It may be added that when the lower clause is in the passive voice, the lower subject is prototypically –Choice. Unfortunately, there are only five such sentences in the material. No claim of statistical significance can be based on such a small number of tokens. However, it is still of interest to observe that among the five tokens, there are four to ‑ing complements, of which one is given as (13d), and only one to infinitive complement, which is given as (13a). This is in line with the tendency identified. 3. Accustomed with sentential complements involving subject control in the 1960s Turning to the 1960s, the size of the subcorpus is 16.1 million words. The search term used was again simply the form accustomed. The search string accustomed retrieves 243 tokens of the word. This represents a frequency of 15.1 per million for all uses of the word, as compared with 18.5 in the decade of 1923 to 1932. Most of the 243 tokens are irrelevant, for similar reasons as in the case of the earlier corpus, but 64 of them are of the salient type, with the adjective accustomed selecting a sentential complement involving subject control. This represents a frequency of 4.0 per million, which is considerably lower than the corresponding figure of 8.4 per million for the first decade. The picture that the 64 tokens offer is very different from that of the first decade. The frequencies of the two types of complement are given in Table 4 below. Table 4.  To Infinitive and to ‑ing Complements of Accustomed in the 1960s

to infinitive to ‑ing

raw frequency

normalized frequency

 3 61

0.2 3.8

Here are the three to infinitive complements and four of the 61 to ‑ing complements:



Juhani Rudanko

(18) a. Algiers has lived so intimately with violence that well-dressed women are accustomed to step daintily over the bodies of ... (1962) b. We are shaken in a strange manner by the rapid friction of beings and things, and we become accustomed to perceive the forces of beings through the very forces enslaved by them. (1966) c. When her fearless father Macnamara led her across the peat bogs, he was accustomed to throw her across the wider draining ditches. (1967) (19) a. In no time at all Premier Menderes was such a frequent caller at the singer’s apartment that other tenants grew grumpily accustomed to being stopped and searched by bodyguards. (1960) b. The nation is becoming accustomed to waving farewell to Mr. Wilson ... (1966) c. After all, they had grown accustomed to opening to the middle of the paper ... (1966) d. ... modern audiences, who are accustomed to seeing females in female roles. (1967) The quantitative findings show that the to ‑ing pattern was overwhelmingly predominant over the to infinitive pattern in complements of the adjective accustomed in the text type of news magazines in American English in the 1960s. A major grammatical change had thus been almost fully completed within some four decades. At the same time, there are still questions of a qualitative nature to investigate, especially in the light of the discussion of the first decade. Regarding the nature of the three sentences with to infinitives, (18a) and (18c) are sentences where the situation encoded by the lower clause is of the +Choice type, with an agentive lower subject. In both (18a) and (18c) it is possible to conceive of turning the complements of accustomed into imperative sentences, of the type Throw her across the wider ditches!, applying the test to sentence (18c). In view of the discussion of ±Choice contexts above, it is not surprising that isolated remnants of to infinitive complements should have survived in such +Choice contexts. By contrast, (18b) stands out, for its lower subject is an Experiencer and the situation encoded by the lower clause is –Choice in its flavor. However, the survival of the to infinitive complement in (18b) is easily explained, for an examination of the larger context of the token revealed that the sentence is originally from 1913. The surviving tokens thus show that there may still be some lingering connection between +Choice contexts and to infinitive complements, but the numbers are of course too low for such a tendency to have any statistical significance. The surviving tokens of to infinitive complements are also of interest from the point of view of what they do not contain. In particular, there is no token involving extraction among them. This raises the question of whether extractions can



Variation and change in American English 

be found among the tokens of to ‑ing complements or whether extractions have more or less disappeared with the disappearance of to infinitive complements. The answer is that extractions have not disappeared, though their proportion is slightly lower than in the earlier corpus. In the 1923–1932 corpus there were 14 extractions among 95 tokens. Now there are six extractions among the 64 tokens, but all of them are found among the 61 to ‑ing complements. Here are some illustrations: (20) a. In This Side of Paradise, F. Scott Fitzgerald alarmed mothers by telling them “how casually their daughters were accustomed to being [Wh Movement] kissed”; (1964)5 b. “Any new reporter in Washington, fresh from the city hall beat where he was accustomed to putting nothing in the paper without identifying the source, will find that if he tries that here, his sources will dry up on him.” (1965) [Relativization] c. In her four months as Prime Minister, Indira Gandhi has been largely spared the biting public criticism that Indian politicians are accustomed to meting out to their leaders. (1966) [Relativization] d. The successful businessman or professional entering Government service, however, may draw an official salary that is far less than the sum he is accustomed to paying in taxes. (1969) [Relativization] The extractions found testify to the entrenchment of the to ‑ing pattern, but the slight decrease in their proportion may suggest that the triumph of the to ‑ing pattern is not yet quite complete in this respect. It may also be noted that both extractions of adjuncts and extractions of complements out of complements of accustomed are found in the material. (20a–b) involve the extraction of adjuncts and (20c–d) involve the extraction of complements. The other factor to consider here concerns the question of +Choice contexts. It was observed in the data from the first decade that to ‑ing complements tended to be more common in –Choice contexts than in +Choice contexts, and it is of interest to examine whether they have now spread beyond such contexts to be embraced by +Choice contexts or whether there is a residual reluctance to use them in such contexts. Of the four initial illustrations in (19a–d), (19a) and (19d) involve –Choice contexts, while (19b–c) involve +Choice contexts. When all 61 tokens are examined, it is observed that the to ‑ing pattern is indeed readily found in +Choice contexts. The six extraction contexts, with one exception, are of the +Choice type, as for instance 5. Accustomed in (20a) may at first sight look verbal in nature, but on reflection it seems clear that the word is adjectival, since the sentence is not about the process of accustoming the daughters to being kissed casually but about the daughters being accustomed to being kissed casually.



Juhani Rudanko

in (20b–c), and when the extraction contexts are set aside, there are 29 tokens of the +Choice type, and 26 of the –Choice type. Here are two more tokens of both types: (21) a. Long accustomed to dictating his own terms, Hughes refused. (1960) b. Accustomed to interviewing McGeorge Bundy regularly when he was dean of Harvard’s Faculty of Arts and Sciences, the editors of the Harvard Crimson telephoned him at the White House to ask if he had anything to do with eliminating Latin from diplomas ... (1961) (22) a. His long-suffering friends are even accustomed to having him turn up with his equipment to record their squalling, hiccuping children. (1961) b. Doctors are so accustomed to hearing about bad effects of smoking ... (1966) Overall, the ease with which to ‑ing complements can now occur in +Choice contexts may be regarded as another indication, of a qualitative nature, of the predominance of this pattern over the to infinitive pattern with accustomed in American English variety of the text type of news magazine English by the 1960s. 4. Concluding observations It is argued in the present investigation that there are robust grammatical differences between the to infinitive and to ‑ing patterns in English. However, some higher predicates have shown variation and change between the two patterns. The adjective accustomed offers a particularly striking example of a matrix adjective of this type. The investigation of the sentential complements of accustomed involving subject control in the first decade of the TIME Corpus shows that the decade was characterized by a considerable amount of variation between the two types of complement. This study offers two explanatory principles to account for the variation. The first principle was the Extraction Principle, which was broadened to include the extraction of adjuncts. It was shown that the Extraction Principle played a significant role in preserving to infinitives. The second principle was semantic. It was observed that when the lower predicate selected a +Choice subject, a to infinitive was the likely type of complement. On the other hand, when the lower subject was of the –Choice type, the complement tended to be of the to ‑ing type. Since the to ‑ing pattern was the new and emerging one with accustomed in the overall context of the Great Complement Shift (Rohdenburg 2006), it is possible to say that non-agentive lower predicates provided a semantic niche for its spread.



Variation and change in American English 

The establishment of the two principles means that what may at first sight look like free variation between two complementation patterns that are different from a grammatical point of view is in fact rule governed, at least up to a significant extent. Turning to the 1960s, the investigation revealed a radical change in the argument structure of accustomed. Within the space of some 30 or 40 years the to infinitive pattern had become very rare, with only one or two tokens. To ‑ing complements were now also found in extraction contexts, though perhaps not quite as freely as to infinitives used to be found in such contexts in the first decade. To ‑ing complements were also readily used in +Choice contexts in the material from the 1960s, which further testifies to their entrenchment from a qualitative point of view. In other words, neither extraction nor the ±Choice factor can be used to explain what variation there is in the 1960s. The present study is focused on accustomed, and it raises the question of whether other adjectives in English selecting sentential complements involving subject control may have followed a similar trajectory of change in recent English, and whether the explanatory principles employed here can be applied in their study. It is only the advent of large electronic corpora of the type of the TIME Corpus that makes it possible to raise and to explore such questions in a meaningful and systematic way. The scholarly community owes a debt of gratitude to Mark Davies and others who have compiled such large corpora and made them publicly available. It may be hoped that the present article demonstrates some of the possibilities that are offered by such corpora in the study of the core grammar of English and of the recent history of English. Sources Davies (2007) = Davies, Mark. TIME Corpus (100 million words, 1920s–2000s). Available online at http://corpus.byu.edu/time.

References Bolinger, Dwight. 1968. “Entailment and the Meaning of Structures”. Glossa 2.119–127. Chomsky, Noam. 1981. Lectures on Government and Binding. Dordrecht: Foris. Denison, David. 1998. “Syntax”. The Cambridge History of the English Language, vol. IV: 1776– 1997 ed. by Suzanne Romaine, 92–329. Cambridge: Cambridge University Press. Dowty, David. 1991. “Thematic Proto-Roles and Argument Selection”. Language 67.547–619. Gruber, Jeffrey S. 1967. “Look and See”. Language 43.937–947. Huddleston, Rodney & Geoffrey K. Pullum. 2002. The Cambridge Grammar of the English Language. Cambridge: Cambridge University Press.



Juhani Rudanko Hundt, Marianne. 2004. “Animacy, Agentivity, and the Spread of the Progressive in Modern English”. English Language and Linguistics 8.47–69. Jackendoff, Ray. 1990. Semantic Structures. Cambridge, Mass.: MIT Press. Jespersen, Otto. [1940] 1961. A Modern English Grammar on Historical Principles. Part V: Syntax (Volume IV). London and Copenhagen: Allen and Unwin and Ejnar Munksgaard. Kjellmer, Göran. 1980. “Accustomed to Swim; Accustomed to Swimming: On Verbal Forms after TO”. A Linguistically Varied Assortment of Readings. Studies Presented to Alvar Ellegård on the Occasion of his 60th Birthday ed. by Jens Allwood & Magnus Ljung, 75–90. Stockholm: Stockholm Papers in English Linguistics and Literature 1. Stockholm: Almqvist and Wiksell. Leech, Geoffrey, Marianne Hundt, Christian Mair & Nicholas Smith. 2009. Change in Contemporary English. Cambridge: Cambridge University Press. OED = The Oxford English Dictionary. Second edition. 1989. Prepared by J. Simpson & E. Weiner. Oxford: Clarendon Press. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech & Jan Svartvik. 1985. A Comprehensive Grammar of the English Language. London: Longman. Rohdenburg, Günter. 1996. “Cognitive Complexity and Increased Grammatical Explicitness in English”. Cognitive Linguistics 7:2.149–182. Rohdenburg, Günter. 2006. “The Role of Functional Constraints in the Evolution of the English Complementation System”. Syntax, Style and Grammatical Norms ed. by Christiane DaltonPuffer, Dieter Kastovsky, Nikolaus Ritt & Herbert Schendl, 143–166. Bern: Peter Lang. Ross, John R. 1973. “Nouniness”. Three Dimensions of Linguistic Research ed. by Osamu Fujimura, 137–258. Tokyo: TEC Company. Rudanko, Juhani. 1989. Complementation and Case Grammar. Albany, New York: The State University of New York Press. Rudanko, Juhani. 1998. Change and Continuity in the English Language. Lanham, Maryland: University Press of America. Rudanko, Juhani. 1999. Diachronic Studies of English Complementation Patterns. Lanham, Maryland: University Press of America. Rudanko, Juhani. 2000. Corpora and Complementation. Lanham, Maryland: University Press of America. Rudanko, Juhani. 2006. “Watching English Grammar Change: A Case Study on Complement Selection in British and American English”. English Language and Linguistics 10.31–48. Rudanko, Juhani. 2007. “Text Type and Current Grammatical Change in British and American English: A Case Study with Evidence from the Bank of English Corpus”. English Studies 88.465–483. Taylor, John R. 2003. “Meaning and Context”. Motivation in Language. Studies in Honor of Günter Radden ed. by Hubert Cuyckens, Thomas Berg, René Dirven & Klaus-Uwe Panther, 27–48. Amsterdam & Philadelphia: John Benjamins. Vosberg, Uwe. 2003a. “The Role of Extractions and Horror Aequi in the Evolution of ‑ing Complements in Modern English.” Determinants of Grammatical Variation in English ed. by Günter Rohdenburg & Britta Mondorf, 305–327. Berlin & New York: Mouton de Gruyter. Vosberg, Uwe. 2003b. ”Cognitive Complexity and the Establishment of ‑ing Constructions with Retrospective Verbs in Modern English.” Insights into Late Modern English ed. by Marina Dossena & Charles Jones, 197–220. Bern: Peter Lang. Vosberg, Uwe. 2006. Die Grosse Komplementverschiebung. Tübingen: Gunter Narr Verlag.

Prevent and the battle of the ‑ing clauses Semantic divergence? Elina Sellgren

University of Tampere The article discusses the variation between the two most common sentential complements of the verb prevent, as in prevent me from going and prevent me going, from a semantic point of view. The variant me going became significantly more common in British English in the twentieth century, competing with the variant with from. Mair (2002) has suggested that a similar phenomenon may be incipient with semantically similar verbs like hinder and stop, signalling a more general grammatical change that is restricted to British English. With data from the British National Corpus, the article proposes a semantic distinction, a consequence of the recent competition, in order to partially explain the variation. The distinction links the notion of hypotheticality to the ‑ing clause in the prepositional variant, whereas the ‑ing clause without from expresses a realized event, or an existing property of the object NP of prevent.

1. Introduction In Present-Day British English (BrE), prevent can choose two nearly identical sentential complements. The historically more common prepositional variant with from (from ‑ing), exemplified in (1), is nowadays used as often as the variant without the preposition from (0 ‑ing), as exemplified in (2) (Mair 2002).1

(1) I prevented him from robbing the bank. (2) I prevented him robbing the bank.

1. The corpora used by Mair (2002) were the following: Brown, Frown, LOB and FLOB, 1 million words each. Brown and Frown represent American English and LOB and FLOB British English. The pairs Brown-LOB and Frown-FLOB represent the same decades respectively (1961 and 1991).



Elina Sellgren

The gerund with the subject of the ‑ing clause in the genitive or possessive case (poss ‑ing) can also be found today, exemplified in (3) below, but it is referred to as being “archaic” by Mair (2002: 112).

(3) I prevented his robbing the bank.

Due to its rarity today, poss ‑ing does not really compete with either of the aforementioned variants.2 In order to get a picture of just how archaic and rare this variant is, one can run a tag sequence search in the BNC Web, a search facility for the 100 million word British National Corpus. By defining the positions after the key term (i.e. any verb form of prevent) as “DPS + VVG”, the resulting example sentences should be of the form “prevent + possessive determiner + ‑ing form”. This procedure yields only 62 examples altogether. Instances where the object NP is a common noun in the genitive can be assumed to be equally rare. Because this complement type is most likely to follow the matrix verb prevent immediately, this figure of 62 cases of poss ‑ing in the 100 million word BNC can be taken to be representative. A study by Heyvaert et al. (2005) provides further evidence that poss ‑ing is rare in twentieth century British English with prevent: the Times and UK Spoken subcorpora in the Collins COBUILD corpus together contain only 32 such instances. From a diachronic point of view, poss ‑ing may have had a bigger part to play in the battle of the ‑ing complements selected by prevent. In the present paper, however, the focus is on synchronic variation. It is notable that in Present-Day American English (AmE), only the prepositional variant from ‑ing as in (1) is used (e.g. Rohdenburg 1995: 87, Dixon 1995: 217 and Mair 2002). The archaic variant poss ‑ing as in (3) is rare, just like in British English. In earlier American English the variant without from as in (2), 0 ‑ing, was apparently used to some extent (Mair 2002). According to Dixon (1995), the 0 ‑ing variant is found also in Australian English. The OED seems to suggest that 0 ‑ing appeared later with prevent than poss ‑ing or from ‑ing, even though examples are given from the seventeenth century: “prevent me going appears to be short for prevent me from going, perh[aps] influenced by prevent my going”. Unlike Mair (2002), the OED sees 0 ‑ing as a reduced form of from ‑ing. As for poss ‑ing, Visser (1973: 2363) mentions that this variant has declined dramatically in frequency with prevent since the nineteenth century, whereas the other two variants from ‑ing and 0 ‑ing are simply noted to coexist. The roughly equal frequency between from ‑ing and 0 ‑ing in British English is very recent. Mair (2002) observes that in the structurally and diachronically parallel ICAME corpora of LOB and FLOB (British English) and Brown and Frown (American English), the variant without from increased rapidly in 2. Cf. also Mair (2002).

Prevent and the battle of the ‑ing clauses 



frequency from 1961 to 1991. In the LOB corpus, containing British English texts from 1961, from ‑ing is dominant with 37 examples following the base form of prevent, against only seven cases of the 0 ‑ing variant. Three decades later, both variants are attested in FLOB with 24 examples, whereas in Brown and Frown (American English) 0 ‑ing was found only once, namely in the 1961 corpus of Brown. While the number of examples in these one million word corpora is not particularly high, they can be taken as suggestive of a notable shift in the frequencies of these complementation variants over the latter half of the twentieth century. Such a rapid change invites attempts to explain the progressively more even distribution. When and why can the preposition from be left out of the complement clause? Mair proposes that in the future, 0 ‑ing may become the sole sentential variant with prevent in British English, whereas in American English the prepositional variant will remain dominant. This process of divergence between two language variants would be part of a grammatical change where in British English semantically similar verbs, for example hinder, block, and stop, will also increasingly gear towards 0 ‑ing at the expense of from ‑ing, and vice versa in American English. According to Mair (2006: 133), 0 ‑ing with stop never had a foothold in older American English and is non-existent today. Rohdenburg (2009: 211) posits that the negative feature associated with from may “play a very special role” in American English. In American English, verbs of negative causation (e.g. prevent, stop, save) have followed a trend where they mark negative orientation by the use of from without exception, whereas in British English the verbs of negative causation have followed suit only in the passive. Indeed, in any variety of Present-Day English only the from ‑ing variant can be used when prevent is in the passive voice (Huddleston & Pullum 2002: 1238):

(4) I was prevented from robbing the bank. (5) *I was prevented robbing the bank.

One instance of passivization with 0 ‑ing is mentioned in Poutsma (1926: 649– 650), but it comes from one of Jane Austen’s novels, and thus obviously does not represent present-day usage norms. In the BNC, only one instance of 0 ‑ing with prevent in the passive can be found by running a tag sequence search with VVG in position +1 after prevent:

(6) If one person is prevented getting AIDS from this officer, we’ve paid for that person for two years ... (KRL, 4350)

In this connection, Mair (2006: 131) makes another interesting point: the passivization of the complement verb also has an effect on the choice of the complement type. The variant 0 ‑ing seems to be encouraged when the complement verb



Elina Sellgren

is in the passive voice. In Mair’s (2006) OED Baseline Corpus, there are five instances of the type “prevent the free copies being sold” and only one of the type “prevent free copies from being sold”. In the active voice, however, at least some verbs of negative causation can leave out from in British English: Mair (2002) reported the increasing use of ‑ing without from with prevent and stop over the twentieth century in British English. This is a complete reversal of an earlier trend where verbs of negative causation increasingly tended to mark negative orientation by means of from, a trend that has reached completion in American English (Rohdenburg 2009: 211). The reversal in British English of this historical trend of marking negative orientation with from is most visible with the highest‑frequency verbs prevent and stop, but could possibly be present, or incipient, with other verbs of negative causation. But even in the 100 million word BNC, verbs like hinder, deter, debar and dissuade, all selecting from ‑ing, are only represented by a couple of hundred examples, the majority of which show nominal complementation. Thus, the lemma search for hinder gives 594 matches, 882 for deter, 212 for dissuade and 120 matches for debar. Mair (2006) found, however, individual instances of 0 ‑ing with discourage, block and ban in the BNC and certain newspapers (examples quoted from Mair 2006: 133): (7) ... aim to discourage Britain returning to the exchange rate mechanism.  (CEK 2355) (8) ... it became possible that the relatives of the dead [...] would seek a judicial review to block the findings being published ...(Private Eye 1046: 13) (9) ... the editorial I wrote for Palatinate over the banning of our ice hockey team visiting Communist East Berlin ... (Private Eye 1053: 3) Another verb that also seems to be gearing towards using 0 ‑ing is save. Mair (2006: 134) found 13 instances of 0 ‑ing with save in the BNC, as against only three of the from ‑ing type. (10) It saves you from becoming under-insured as a result of inflation ...  (AYP 1603) (11) Oh well, it saves you penning. (GYT 157) Though they are not very numerous, these examples could support Mair’s prediction of 0 ‑ing becoming a syntactical Briticism with verbs of negative causation. Even though the variation between the nearly identical complements from ‑ing and 0 ‑ing with prevent has its roots in historical developments, synchronic factors either promoting or hindering the further advance of 0 ‑ing with prevent and perhaps other verbs as well cannot be ruled out. If 0 ‑ing is about to defeat from ‑ing, it is likely to follow a particular pattern.



Prevent and the battle of the ‑ing clauses 

2. Explaining the complementation of prevent In Mair (2002), the two most common variants, from ‑ing and 0 ‑ing, were found to have become serious competitors only in recent decades. The phenomenon where the preposition from is increasingly absent from a complement clause is certainly intriguing, especially as this change happened so rapidly. Wherever the competition will lead in the future, it is of interest to examine how it has developed, and what kind of factors direct the choice of the different complements synchronically. Apart from diachronic change, one can postulate syntactic and semantic explanations for the variation. Semantic distinctions between the variants, though unsubstantiated by systematic study, have been suggested before. Dixon (1991: 236) proposes that 0 ‑ing is used when the agent denoted by the subject in the sentence employs direct means to prevent something, and from ‑ing is used when indirect means are employed. Rudanko (2002: 58) offers the explanation that from ‑ing may involve indeterminacy, while 0 ‑ing has a sense of immediacy and external observability. Moreover, he suggests two different readings of prevent which entail two different underlying structures: the ‘bring about’ reading, which is connected to an NP Movement structure and the ‘act on’ reading, which is connected to a control structure. The problem with any semantic description, such as the suggestion by Dixon (1991) that from ‑ing involves indirect means of prevention on the part of the agent and 0 ‑ing direct means, lies with sentences with underspecified contexts. The notion of direct or indirect means in the act of prevention can only apply to animate and non-abstract object noun phrases or “preventees”. The distinction can therefore only apply to part of the data. In addition to semantic explanations, the functional approach of Rohdenburg’s complexity principle (e.g. 1995, 1996 and 2000) is useful: In the case of more or less explicit grammatical options, the more explicit one(s) will tend to be favored in cognitively more complex environments. (Rohdenburg 1996: 151)

This principle is valuable as an explanative factor from a functional point of view. The principle predicts that the more explicit variant will be found more often in complex syntactical environments. In general terms, speakers resort to the more explicit variant in environments that are “for some reason more difficult, more complex, less entrenched, less frequent, less accessible or in any other way cognitively more complex” (Rohdenburg 2009: 88). In the case of prevent, such complex environments can be passivized sentences, relativizations, extractions or sentences with long and complex object noun phrases. With prevent, from ‑ing is the structurally most explicit variant,



Elina Sellgren

whereas 0 ‑ing and poss ‑ing are less explicit. In Rohdenburg’s analysis, the archaic variant (poss ‑ing) is the structurally least explicit option, due to its ambiguity as a nominal-verbal hybrid: the same structure can sometimes be analyzed as more nominal (I prevented his going) and sometimes as more verbal, when an object is present (I prevented his stealing the book). As was observed, the structurally more explicit from ‑ing complement is indeed dominant when prevent is passivized. The principle rests on the hypothesis that out of several different variants, the structurally most explicit one will be favoured in complex environments. This is because explicit constructions are easier to process cognitively, especially in sentences with complex structures, because they decrease the processing load created by structural complexity. To exemplify this in the case of prevent, consider example (12) below where the author has chosen to include from in the complement clause. Not only is the object NP of prevent very long (the fundamental beliefs in the nation and “the historic integrity of the island of Ireland”), but also it is followed by a finite subordinate clause, before the from ‑ing complement finally completes the message of the sentence. (12) But the fight did not prevent the fundamental beliefs in the nation and “the historic integrity of the island of Ireland”, as nationalist parties described it in their New Ireland Forum (1983–1984: i. 28), from remaining basic to the perceptions of both parties. (BNC, A07, 317) Apparently, the syntactic complexity of a sentence can to some extent protect the historically more common and structurally more explicit variant from ‑ing against the advancing 0 ‑ing variant. Using data from Collins COBUILD corpora, in a recent study on prevent (Babováková 2005) it was found that the increasing length and complexity of insertions as well as object noun phrases between the matrix verb prevent and the ‑ing participle tend to favour the explicit variant, from ‑ing. However, when the object noun phrase of prevent consists of only one word, from ‑ing dominates, regardless of the head of the noun phrase (pronoun or common noun). Theoretically, short object NPs after prevent could support the use of the 0 ‑ing variant because they represent a “cognitively simple” environment. The predominance of from ‑ing following both short and long object noun phrases may simply reflect the overall distribution of the competitors in the corpora used. Roughly 57% to 70% of the sentential examples were used with from. It is only in the case of passivization that the complexity principle is firm – but otherwise there appear to be no hard and fast rules to the variation from the point of view of syntactical complexity. The question remains, what kind of factors promote the use of the relatively new big player in the field, 0 ‑ing? What governs the choice of either variant when the sentences examined are not particularly complex? This is the starting point of

Prevent and the battle of the ‑ing clauses



the present study which seeks to find a possible semantic distinction between the variants, working systematically through two small samples. While it is acknowledged that there are probably many factors at play simultaneously, the semantic distinction proposed here appears to be valid enough to be seriously considered as an important factor in the variation between the competing complements. This study is inspired by Bolinger’s (1968) generalization that a difference in syntactic form entails a difference in meaning. It is proposed here that the complementation variants from ‑ing and 0 ‑ing may be developing slightly different meanings due to the increasing competition in British English. A preliminary look at data from earlier British English3 seems to suggest that we cannot find this semantic distinction any earlier than the twentieth century. This implies that the semantic distinction in question arose due to the increasing competition between the two variants, a competition which itself may have been caused by other than semantic motivations. 3. Data The data for the study was collected from the British National Corpus (BNC) by using the BNC Web search facility. The study was initially prompted by the observation during pilot searches that in many individual texts with several tokens of prevent, the verb was found with both near-identical sentential variants in a seemingly random fashion. When obtaining several different random sets of 100 examples of prevent from the BNC, it soon becomes obvious that the competition is most palpable between the from ‑ing and 0 ‑ing variants. For every set of a hundred tokens of prevent, only one or two instances of poss ‑ing are likely to be found (cf. Section 1). It is also clear that different authors use the more common variants to varying degrees, some use only one of the variants, some use both variants but with a pronounced preference one way or the other. Two data sets were compiled for analysis. Data set 1 consisted of texts that each had several occurrences of prevent with a sentential complement. This set was compiled for exploring whether the varying usage of the complement patterns of prevent by individual authors reflects a subtle semantic distinction between its complement variants. By collecting a selection of texts that each have several tokens of the verb prevent (data set 1), it is possible to compare the usage behaviour of individual authors. When both complement variants are found in one text, it is interesting to examine whether one of the variants is clearly preferred and whether 3. Data extracted from the Corpus of Late Modern English Texts, a corpus compiled by Hendrik de Smet at the University of Leuven.





Elina Sellgren

the choice of either variant follows some kind of a rule. Data set 1 represents the 14 texts in the BNC with the biggest number of tokens of prevent (381 tokens altogether; see Table 1), and altogether they contain about 1.9 million words. Data set 2 was a random sample of 100 tokens of prevent, comprising 44 instances of sentential complementation. This set was collected in order to discover whether any semantic distinctions could be found when comparing different authors. In other words, data set 1 is oriented towards the authors, whereas data set 2 compares the variants that are in competition with each other in general, disregarding idiosyncratic preferences for either of the variants. The examination of the use of prevent by individual authors is justified, because emerging semantic distinctions cannot be assumed to be found in the language use of all speakers. One can hypothesize that an incipient semantic distinction between the competing complementation variants will wiggle its way through into mainstream usage by the innovativeness of individual speakers and authors, for instance newspaper journalists or internet writers. Whether the semantic distinction explored in the next sections has solidified its position to any degree can be examined by comparing individual instances of the variants from different texts (data set 2), rather than within individual texts (data set 1). 4. Discussion 4.1

Data set 1: Variation within individual texts

In data set 1, there is a remarkable amount of variation among the authors with respect to preference for either sentential complement of prevent (see Table 1). The texts chosen contained from 8 to 42 tokens of the verb prevent. The number of tokens with sentential complements ranged from 5 to 37. The author of the text ARF4 (extract from an academic book on medicine), for example, uses from ‑ing exclusively, while the rest of the authors use 0 ‑ing in at least one sentence. In some texts, like J7B and CJE, both variants are found without a clear preference for either of them. In J7B (extract from an academic book on law practices in commerce), from ‑ing is found 23 times and 0 ‑ing 14 times. In CJE (a training guide for dog owners), from ‑ing is found 10 times and 0 ‑ing six times.

4. The combinations of three letters and/or numbers refer to the codename for the file in the BNC.

Prevent and the battle of the ‑ing clauses 



Table 1.  Data set 1 Text

No. of tokens of prevent

from -ing

0 -ing

A0J ARF ASB CCX CJE EDC ECJ EV5 GVR HH6 HHV HHX J7B K5M Total

  14   38   31   36   27   24   21    8   12   28   42   30   42   17 381

   6    5    1    3   10    5    1    2   11    1   30   19   23   11 128

   5    0    4   15    6    1   19    6    1   27    6    4   14    5 113

In most cases, however, authors prefer one of the variants very strongly. The author of HHV, for example, uses from ‑ing 30 times, but 0 ‑ing only six times. (13) It is none other than the Prime Minister who, as Minister for Social Security, introduced measures to prevent people from receiving benefit in those circumstances. (HHV 3204) (14) Does he agree that if – in his admirable phrase – we are to prevent the Commission interfering in the nooks and crannies of everyday life.  (HHV 23850) The author of ECJ, on the other hand, uses 0 ‑ing 19 times, but from ‑ing only once. (15) Other safety factors to consider are the provision of side restraints in the form of a low wall or balustrade, to prevent anyone from falling over the side edges of the steps. (ECJ 91) (16) Lift up the ball valve arm as high as it will go, and tie up to prevent the cistern refilling. (ECJ 521) Such huge differences in the preference for either complement demonstrate that the nearly equal overall distribution of the variants in general (cf. Section 1; Mair 2002) does not reflect the situation accurately. Obviously, many authors do not see the variants as interchangeable and grammatically unmarked options, nor do they use them to an equal degree. Especially cases where one of the variants is



Elina Sellgren

found only once or twice are interesting from the point of view of semantic distinctions. Even though the choice of using a preposition or leaving it out cannot be unwitting on the part of the author, it is likely that in less formal types of writing the level of linguistic self-correction and monitoring is lower. It is of course possible that the “exceptions” in the language use of a certain author are caused by the syntactical complexity of the given sentence, thus following the complexity principle mentioned earlier. More often than not, however, the sentences in data set 1 were not that different as regards syntactical complexity. There are undoubtedly many cases where hardly any distinction can be seen between the sentences using the different variants – they appear to be in free variation. Consider the following examples: (17) Father Morrow said he had acted “in an attempt to prevent future comatose, disabled and elderly patients from sharing the fate of Tony Bland”  (CFB, 1201) (18) He said he had acted “in an attempt to prevent future coma, disabled and elderly patients suffering the same fate as Tony Bland”. (CFB, 1227) Here the writer repeats a quotation with nearly identical sentences, but with different complementation variants. There are minor differences: comatose has been abbreviated into coma, and the lower clause ‑ing participle has changed from sharing to suffering. To summon up even a far-fetched explanation, one could make the case that sharing does not indicate as much passiveness on the part of the patients as suffering does. Hence the patients, denoted by the object noun phrase, may be able to keep themselves at a distance from the event or situation of sharing the fate of Tony Bland, which is the prevented event denoted by the ‑ing clause. Suffering implies no volition on the part of the subject of the ‑ing clause (patients), and the patients are thus much closer to that fate. Here we may consider Huddleston & Pullum’s (2002: 657) description of the semantic function of the preposition from as giving the following ‑ing participle the role of an “intended action as a spatial goal”. The absence of the preposition may then be seen to collate the semantic roles of the ‑ing form and the object noun phrase of prevent, denoting the prevented event together. This notion may be related better to verbs implying no volition or agentiveness on the part of their subject noun phrase. There are many other instances where the similarity of the sentences is rather striking. In (19) and (20) below, the syntactical environments of the sentences are again nearly identical.



Prevent and the battle of the ‑ing clauses

(19) Care must be taken not to keep the stick forward for too long, as too much forward movement during the take-off run will prevent the glider from leaving the ground until a much higher speed is reached. (A0H, 637) (20) Opening the airbrakes at that moment will prevent the glider leaving the ground again. (A0H, 691) The ‑ing clauses do not differ much in complexity: in (19) (the glider) from leaving the ground is followed by a separate subordinate clause, in (20) (the glider) leaving the ground is followed by an adverb. It is unlikely that anything outside the complementation sequence would have an effect on the choice of a complement variant. Example (21) demonstrates a further instance of such striking similarity from the same text: (21) The launch should always be stopped if an overrun occurs, and if the launch continues in spite of pulling the release the pilot should make every effort to prevent the glider from leaving the ground by keeping the stick right forward. (A0H, 759) We can here recall Dixon’s (1995) idea that the from‑less variant involves a more direct, physical interaction between the actor and the goal of action: it is easy to see the difficulty in formulating this kind of semantic distinctions. In the examples above, the referents of both the subject and the object are inanimate, and the semantic role of actor does not apply in the context. As mentioned also before, Rudanko (2002: 58) suggests that NP from ‑ing may be more indeterminate and less specific and NP ‑ing more immediate, involving a sense of external observability. Starting with this distinction in mind, (22) and (23) below with from ‑ing may sound rather hypothetical, as in giving instructions beforehand: (22) The wing unstalls immediately and prevents the spin from developing.  (A0H, 1023) (23) Any forward movement will prevent a spin from developing, whereas failure to move forwards can result either in the spin continuing or, if the opposite rudder has been applied, in the direction of the spin suddenly reversing. (A0H, 1032) Then again, (22) above is part of a description of a demonstration by an instructor, and the verb unstalls in the third person present tense gives the sentence a very immediate feel. With 0 ‑ing in (24) below, on the other hand, the tone is very certain: a spin will not develop. (24) This shows conclusively that any movement forward will prevent a spin developing even if the rudder is forgotten altogether. (A0H, 1025)





Elina Sellgren

Reconsidering now the examples (19) to (21) above, repeated below, there may be some differences after all between the different variants. (19’) Care must be taken not to keep the stick forward for too long, as too much forward movement during the take-off run will prevent the glider from leaving the ground until a much higher speed is reached. (A0H, 637) (20’) Opening the airbrakes at that moment will prevent the glider leaving the ground again. (A0H, 691) (21’) The launch should always be stopped if an overrun occurs, and if the launch continues in spite of pulling the release the pilot should make every effort to prevent the glider from leaving the ground by keeping the stick right forward. (A0H, 759) Examples (19’) and (21’) with from ‑ing sound much more hypothetical than (20’) with 0 ‑ing, just as examples (22) to (23) with from ‑ing. “Too much forward movement during the take-off run will prevent the glider from leaving the ground” is something one would say when imagining a hypothetical scenario. The crucial difference between the two very similar sentences may be the adverb again in (20’): it implies that the glider is commonly known to leave the ground unless the airbrakes are opened at the right moment. This more concrete sentence discusses something that has been known to happen before. There is a sense of the writer imagining the scenario in action in his or her mind’s eye in (20’), but this is probably due to the adverbial at that moment. Many other cases of apparently free variation in the data are more easily explained. The examples (25) to (28) below come from an extract from the book Train your dog, classified in the domain “Leisure” in the BNC, text type “Written books and periodicals”. In examples (25) to (27), the from ‑ing variant is used – but in (28), the preposition from is left out. Such a sudden change of practice points to some special motivation behind the choice of a different complement variant. (25) Regular exercise off the leash will help to prevent a young dog from becoming destructive around the home. (CJE, 1106) (26) Yet if it is not alarmed by this procedure, you may well be able to relieve the obstruction and so prevent the dog from choking at a time when rapid action is required. (CJE, 771) (27) In an emergency this may prevent a dog from straying into a potentially dangerous situation, for example if you should suddenly encounter riders on horseback when you are out for a walk along a narrow path.(CJE, 569) (28) Desperate owners sometimes ask their vet if it is possible to prevent a dog barking by surgical means. (CJE, 1038)



Prevent and the battle of the ‑ing clauses 

One can again look at the meaning of the syntactically lower clause: in the first three examples, it describes potentially harmful scenarios to the dog (e.g. choking), whereas barking in (28) simply denotes a permanent quality or property of the dog. Judging by these examples, it could be said that from ‑ing is used when talking about hypothetical, unrealized situations. Accordingly, 0 ‑ing is used when talking about an event or a property of the object noun phrase that already exists, or an event that happened before respectively. The notion of from giving the following participle the role of an intended action as a spatial goal is particularly clear here. In (25) to (27) above, the unwanted, hypothetical scenarios denoted by the ‑ing participles effectively have the role of “non-intended” actions as spatial goals. The preposition from puts the preventee, the dog, at a conceptual spatial distance from those spatial goals. Without from indicating spatial distance between the roles of the preventee and the prevented action or event in (28), the object noun phrase a dog and the ‑ing form barking consequently seem to form a more closely knit semantic unit. To extend this idea further, the following examples may also serve to demonstrate hypotheticality versus existing or realized situations or properties. (29) ...By including express restrictive covenants in an employment contract an employer will seek to achieve three goals once employment is over: (a) to prevent the ex-employee canvassing orders from the employer’s customers; (b) to prevent the ex-employee competing with his business (usually within a defined geographical area); (c) to prevent the ex-employee from using/disclosing any legitimate business secrets. (J7B, 1309) (30) Note that (a) and (c) alone are not enough: neither would prevent an employee from serving those customers of his ex-employer who approached him. (J7B, 1310) In (29), 0 ‑ing is used with the first two instances of prevent, whereas from ‑ing is used with the last one, within the same sentence. An ex-employee canvassing orders from their ex‑employer’s customers (see 29 a) and competing with their ex‑employer’s business (see 29 b) are both realistic scenarios. They are not strictly unethical actions, in fact they are a natural part of making business. Using or disclosing business secrets of an ex‑employer, on the other hand, may be seen as dishonest. Perhaps this scenario (see 29 c) is therefore seen as more unlikely than the other two, and therefore perceived as hypothetical. Example (30), elaborating on the previous sentence, may also discuss a hypothetical scenario, in the sense that the ex‑employer’s customers approaching the ex‑employee are perhaps the minority among the ex‑employee’s customers in general.



Elina Sellgren

4.2

Data set 2: Variation between different texts

The data from the random sample, data set 2, was examined firstly with regard to the idea of hypotheticality related to the from ‑ing complement. Secondly, the idea was explored that the 0 ‑ing complement implies a closer connection between the object noun phrase of prevent and the ‑ing form, and that together they denote a realized event or a property of the object noun phrase. After discarding examples of nominal complementation, the random sample of 100 tokens of prevent produced 25 examples of the from ‑ing variant and 19 examples of the 0 ‑ing variant. The distribution follows quite closely that observed in other studies: the variant from ‑ing is represented by 57% of the examples and 0 ‑ing by 43% (cf. Mair 2002 who finds a fifty-fifty ratio between these variants). The sample is modest in size, but adequate for an exploratory semantic analysis. The majority of the examples seem to corroborate the ideas proposed above in Section 3. Examples (31) to (32) are prime candidates for the hypotheses, the first with from ‑ing and the second with 0 ‑ing. Example (31) below demonstrates the idea of hypotheticality: (31) Indeed the only reason that modern living things are able to survive in the presence of oxygen, is that they contain a variety of compounds that prevent it from reacting with materials such as fats: compounds that include vitamins C and E, and uric acid. (AMS, 224) If there were no compounds to prevent oxygen from reacting with materials inside living beings, there would be no “modern living things”. Any sense of the compounds merely stopping something already happening, i.e. the oxygen reacting with the materials, is not possible: the act denoted by the participle has never come to pass in a living creature, or it would not be alive. Example (32), on the other hand, clearly discusses an event that some people wanted to prevent, but which did actually happen (i.e. the commemoration of the 1916 Easter Rising). (32) A classic clash of symbols occurred in April when the government took no action to prevent the 1916 Easter Rising being commemorated in Belfast. (AD2, 199) To take another pair of contrasting examples, example (33) with from ‑ing discusses something hypothetical: the trader commits an offence unless he has tried to prevent consumers from relying on misleading information.



Prevent and the battle of the ‑ing clauses 

(33) If in that situation some consumers might reasonably be expected to rely on the misleading information, the trader commits an offence unless he has taken all reasonable steps to prevent them from doing so. Example (34) below with 0 ‑ing, on the other hand, is concerned with an existing situation, one where hens lay outdoors and roost in the trees if they are allowed to. The intention of the parties concerned is to prevent the “event” of these hens laying outdoors and roosting in the trees at the moment of speaking, with prevent having a meaning closer to stop than hinder. The preventive measures stop an activity that is already taking place, rather than pre-empting a hypothetical activity that has not taken place yet. (34) To prevent hens laying outdoors and roosting in the trees, the report suggests that the birds be kept in their housing until mid-morning, and be presented with a main feed in-house in the evening before being shut in for the night. In (35) with from ‑ing, it obviously happened that the speaker was prevented from going to the woman, hence it is a hypothetical scenario. (35) Vividly could he still recall his nurses preventing him from going to her there, and he knew he had stood at a window just like this, gazing with longing through the trees to the building where his mother was imprisoned. (HGV, 4757) In some cases, however, the semantic distinction proposed here is not quite so clear. In seven cases it was not possible to maintain the idea of hypotheticality with from ‑ing, or the idea of a realized event/property with 0 ‑ing, or the information provided by the context was insufficient. For instance, in (36) the discussion of a depreciation of the exchange rate seems hypothetical, and apparently it has not happened due to the preventive measures by the government. (36) Ultimately there would be a depreciation of the exchange rate under a free-floating regime, but the policy response of the government has been to prevent this occurring by sustaining a high level of domestic interest rates. (EC3, 658) All in all, there are occasional cases where the semantic distinction does not apply, or it is simply impossible to determine due to lack of context. In such cases, there are certainly other factors at work that can explain the choice of one variant over the other. The level of formality of the text or the nature of the text type may affect the choice of the complement if from ‑ing is felt to be the “default” or more



Elina Sellgren

correct variant in a prescriptive sense. Considering Rohdenburg’s complexity principle discussed above (Section 2), the length of the object noun phrase of prevent, the length of the ‑ing participle, or even the inflectional form of prevent may all guide the choice of the complement. On a more general level, it is possible that through its historical spread, the 0 ‑ing variant is simply more and more often seen as a neutral choice from a prescriptive point of view, in the sense of the authors’ intuition about normative rules of language use. 5. Conclusion This paper has offered a new, semantic perspective on the variation between the two most common sentential complements of prevent, prevent me from going (from ‑ing) and prevent me going (0 ‑ing). The variant from ‑ing was linked to the idea of hypotheticality denoted by the ‑ing form, and 0 ‑ing was connected to the idea of a realized event or a permanent or existing property of the object NP of prevent, denoted by the object NP and the ‑ing form together. A surprisingly large proportion of the examples in the data samples appeared to follow this semantic distinction. In subsequent studies, statistical methods for validation may be helpful in determining how viable the distinction is in more numerical terms. There have been several attempts at explaining the close competition, ranging from Rohdenburg’s (e.g.  1996) complexity principle to Dixon’s (1991) and Rudanko’s (2002) proposals regarding semantic and underlying structural and syntactic distinctions, but they have never been systematically examined with more data than a couple of examples. As was mentioned in the introduction, Mair (2002) has suggested that prevent is spearheading a grammatical change whereby semantically similar verbs in British English (e.g. hinder, block, stop) are increasingly favouring the variant without from, perhaps eventually ousting the prepositional complement completely. So far, it seems unlikely that such a process could be carried to completion, considering the strong preference for the prepositional variant in passivized sentences with prevent. On the other hand, the distribution of from ‑ing and 0 ‑ing in relation to each other may not have stabilized yet, following a very rapid change over the twentieth century. If both variants continue to be used on a par in British English as regards their relative frequencies, it is possible that they select their own niches of use, semantically or otherwise determined.

Prevent and the battle of the ‑ing clauses 



Sources BNC = The British National Corpus: BROWN = A Standard Corpus of Present-Day Edited American English, for Use with Digital Computers (Brown) compiled by W. N. Francis & H. Kučera. Brown University. Providence, Rhode Island. 1964, 1971, 1979. CLMET = The Corpus of Late Modern English Texts compiled by Hendrik De Smet, University of Leuven. F-LOB = The Freiburg-LOB Corpus of British English compiled by Christian Mair, AlbertLudwigs-Universität Freiburg. 1999. FROWN = The Freiburg-Brown Corpus of American English compiled by Christian Mair, Albert-Ludwigs-Universität Freiburg. 1999. LOB = The Lancaster-Oslo-Bergen Corpus compiled by Geoffrey Leech, Lancaster University, Stig Johansson, University of Oslo & Knut Hofland, University of Bergen. 1976.

References Babováková, Petra. 2005. The Complements of Prevent. A Master’s Thesis. Tampere: Tampere University Press. Bolinger, Dwight. 1968. “Entailment and the Meaning of Structures.” Glossa 2.119–127. Dixon, Robert M.W. 1991. A New Approach to English Grammar, on Semantic Principles. Oxford: Clarendon Press. Dixon, Robert M.W. 1995. “Complement Clauses and Complementation Strategies”. Grammar and Meaning: Essays in Honour of Sir John Lyons ed. by Frank R. Palmer, 175–220. Cambridge: Cambridge University Press. Heyvaert, Liesbet, Hella Rogiers & Nadine Vermeylen. 2005. “Pronominal Determiners in Gerundive Nominalization: A ‘Case’ Study”. English Studies 86.71–88. Huddleston, Rodney & Geoffrey K. Pullum. 2002. The Cambridge Grammar of the English Language. Cambridge: Cambridge University Press. Hundt, Marianne, Geoffrey Leech, Christian Mair & Nicholas Smith. 2009. Change in Contemporary English: A Grammatical Study. Cambridge: Cambridge University Press. Mair, Christian. 2002. “Three Changing Patterns of Verb Complementation in Late Modern English: A Real-Time Study Based on Matching Text Corpora”. English Language and Linguistics 6.105–131. Mair, Christian. 2006. Twentieth Century English. History, Variation and Standardization. Cambridge: Cambridge University Press. Poutsma, Hendrik. 1926. A Grammar of Late Modern English: For the Use of Continental, especially Dutch, Students. Part 2, Section 2: The Parts of Speech: The Verb and the Particles. Groningen: Noordhoff. Rohdenburg, Günter. 1995. “On the Replacement of Finite Complement Clauses by Infinitives in English”. English Studies 76:4.367–388. Rohdenburg, Günter. 1996. “Cognitive Complexity and Increased Grammatical Explicitness in English”. Cognitive Linguistics 7:2.149–182. Rohdenburg, Günter. 2000. “The Complexity Principle as a Factor Determining Grammatical Variation and Change in English”. Language Use, Language Acquisition and Language



Elina Sellgren History: (Mostly) Empirical Studies in Honour of Rüdiger Zimmermann ed. by Ingo Plag & Klaus Peter Schneider, 25–44. Trier: Wissenschaftlicher Verlag. Rohdenburg, Günter. 2009. “Nominal Complements”. One Language, Two Grammars? Differences between British and American English ed. by Günter Rohdenburg & Julia Schlüter, 194–211. Cambridge: Cambridge University Press. Rudanko, Juhani. 2002. Complements and Constructions. Corpus-based Studies on Sentential Complements in English in Recent Centuries. Lanham, Maryland: University Press of America. Visser, F. Theodorus. 1973. An Historical Syntax of the English Language, Part Three. Leiden: E. J. Brill.

Prescription or practice? Be/have variation with past participles of mutative intransitive verbs in the letters of Joseph Priestley Robin Straaijer* Leiden University

This paper investigates variation in the usage of be and have with participles of mutative intransitive verbs by the Late Modern English grammarian Joseph Priestley. His usage, as exemplified by a corpus compiled from his private manuscript letters, is compared with a reference corpus and with the “rules” in his own grammar. The study shows that Priestley’s usage reflects the general change for this linguistic feature from a predominant use of be to a greater occurrence of have in the late eighteenth century, as discussed by Rydén & Brorström (1987). It is also shown that when compared to his grammar, Priestley’s usage upholds his reputation as an early descriptivist.

1. Introduction This paper explores the usage of the auxiliaries be and have with the participles of mutative intransitive verbs by Joseph Priestley (1733–1804)1, one of the most important eighteenth-century English grammarians. Mutative intransitive verbs are verbs that denote some form of change in place or condition, such as the verbs go or become respectively. In Old English, the original function of be or have with a past participle was to denote a state. Originally in Old English, have was only used with transitive verbs, but this was changing to include intransitive and * The research for this paper was carried out as part of the research project “The Codifiers and the English Language: Tracing the Norms of Standard English” at Leiden University, sponsored by the Netherlands Organization for Scientific Research (NWO). I would like to thank Ingrid Tieken-Boon van Ostade, the editors of this volume and the two anonymous reviewers for their helpful comments on earlier versions of this paper. 1. Life dates throughout this paper were taken from the online Oxford Dictionary of National Biography (ODNB).



Robin Straaijer

mutative verbs. During the Middle English period, be was already the regular marker of perfectivity, without referring to aspect, but the use of have in the same context was slowly rising (Rydén & Brorström 1987:  16–18). In Late Modern English we also find an aspectual distinction between contexts where be is used to indicate a state and have to indicate an action or process (Rydén & Brorström 1987: 26–27). In general though, be remained dominant throughout the seventeenth century. The passages from Priestley’s letters given as examples (1) and (2) illustrate the auxiliary variation in his own language with the mutative intransitive verb begin (despite the fact that the tenses are not the same). (1) The last time that I heard from Mr Belsham the war was not begun, but apprehended. (letter to Theophilus Lindsey, 27 August 1803) (2) I shall, however, hardly undertake more than completing, in the best manner that I can, what I have begun.  (letter to Theophilus Lindsey, 16 June 1798)

The change from be to have as the dominant auxiliary with mutative intransitive verbs in Late Modern English has been extensively studied in a diachronic context by Rydén & Brorström (1987), who found that the occurrence of have rose from 20 percent around 1700 to 40 percent at the beginning of the nineteenth century, and that have reached a “paradigmatic majority” in the first decades of that century (Rydén & Brorström 1987: 196). In Present Day English, the auxiliary have is used almost exclusively with mutative intransitive verbs. Priestley is an interesting subject for a study of this kind, since he is considered to be one of the few descriptivists in the predominantly prescriptive eighteenth century. The following remark by Leonard is illustrative of this: “Priestley is undoubtedly the first writer in English, and apparently the only one in the eighteenth century, to take a reasonably consistent view of usage” (1929:  142). Similarly, Baugh & Cable say that of “all the grammarians of this period only Priestley seems to have doubted the propriety of ex cathedra utterances and to have been truly humble before the facts of usage” (2002:  278). Paradoxically, Priestley is also counted among the founders of English prescriptivism (Bryan 1923). Recent studies such as Hodson (2006) and Straaijer (2009) have endeavoured to show a more balanced view to Priestley’s position with regard to these two extremes of descriptivist and prescriptivist. In that light, this paper is an addition to this. His choice of auxiliary with mutative intransitives compared to his prescription – or rather description? – in his grammar will lead to more nuance in Priestley’s reputation as a descriptivist. The discussions involving participles in Late Modern English grammars appear to be more concerned with the passive voice of transitive verbs than with



Be/have variation with mutative intransitive verbs 

anything regarding intransitives. An example is James Greenwood (1683?–1737), in his Essay Towards a Practical English Grammar (1711): 2 The Participle which ends in ed is called the Passive Participle, because we having in English no Passive Voice, that is, no distinct Ending to distinguish a Verb that signifies doing, from a Verb that signifies suffering, we make up this Want by the Help of the Verb Am, and this Participle, as, I am loved, I am burned (Greenwood 1711: 124)

It seems therefore as though the grammars do not say much about the choice of be or have with mutative intransitives. However, the terminology of the time must also be taken into account. The term “verbs neuter” that we find in these grammars appears to be roughly equivalent to what we now call mutative intransitive verbs, yet neuter verbs could also become transitive, so the match is not exactly one-to-one. In his Short Introduction to English Grammar (1762), the grammarian Robert Lowth (1710–1787) defines the verb neuter in terms of transitivity: the Verb Neuter is called Intransitive, because the effect is confined within the Agent and doth not pass over to any Object (Lowth 1762: 45)

On the other hand, Greenwood defines the verb neuter in terms of active and passive use: The Signification of Verbs Absolute (or Neuter, which signify Action) is in a manner Passive: and therefore Verbs Absolute and Passive are frequently us’d for each other (Greenwood 1711: 155)

Rydén & Brorström also note Greenwood’s remarks (1987: 208–209) and take it as a statement of his preference for have with these intransitives. James Buchanan (fl. 1753–1773) seems to have pleaded for the use of be with mutative intransitives in his British Grammar (1762) although we cannot rule out the possibility that he was merely recording such usage: Come, gone, run, set, risen, fallen, grown, withered, decayed, arrived, and such like intransitive Verbs, which have frequently the passive Signs, am, art, &c. instead of have in the Perfect Time; and was, wast, &c. instead of had in the Pluperfect; as, I am come, for I have come; I was come, for I had come, &c. (Buchanan 1762: 142)

In a footnote in the Short Introduction to English Grammar Lowth appears to proscribe the use of be with mutative intransitives (see also Tieken-Boon van Ostade 2006: 553): 2. Full-text editions used in this study were taken from Eighteenth Century Collections Online (ECCO).



Robin Straaijer

I doubt very much of the propriety of the following examples: “The rules of our holy Religion, from which we are infinitely swerved.” [...] “The whole obligation of that law and covenant, which God made with the Jews, was also ceased.” [...] “Whose number was now amounted to three hundred.” (Lowth 1762: 63)

It seems from this that Lowth was in favour of the use of have with these verbs, which would be a progressive stance in light of the change in progress. However, this footnote refers to the following passage in the main text in which Lowth discusses neuter verbs. The Neuter Verb is varied like the Active; but, having somewhat of the Nature of the Passive form, retaining still the Neuter signification; chiefly in such Verbs as signify some sort of motion, or change of place or condition: as, I am come; I was gone; I am grown; I was fallen. The Verb am in this case precisely defines the Time of the action or event, but does not change the nature of it; the Passive form still expressing, not properly a Passion, but only a state or condition of Being. (Lowth 1762: 61–63)

It seems that by the “Passive form” Lowth means be + participle, which makes his position on the subject less clear. According to Rydén & Brorström, Lowth’s proscription may have made him a linguistic innovator for the change from be to have “[s]ome late eighteenth and early nineteenth century grammarians, starting with Lowth 1762, express doubts about the use of be with (certain uses of) some verbs” (Rydén & Brorström 1987:  209). However, Tieken-Boon van Ostade (2002: 163) hedges this position and intimates that more research on actual usage is required in order to determine this conclusively: Lowth may indeed have been the first to discuss the subject [...] but whether [...] his grammar would also have acted as a linguistic innovator from the perspective of the social network model is a question that can only be answered when the expansion of the [Corpus of Early English Correspondence] will have had its effect (2002: 163)

Whatever his stance, Lowth shows an awareness of a difference in meaning between the uses of one auxiliary or the other with mutative intransitives. Priestley mentions the subject for the first time in the second edition of his Rudiments of English Grammar (1768), where he does his reputation as a descriptivist credit. He makes no pronouncement of preference of neither be nor have, but argues that though strictly speaking be may be more appropriate, the choice of auxiliary should be determined by the context: It seems not to have been determined by the English grammarians, whether the passive participles of verbs neuter require the auxiliary am or have before them. [...] I think we have an advantage in the choice of these forms of expression,

Be/have variation with mutative intransitive verbs 



as it appears to me, that we use them to express different modifications of the sense. When I say, I am fallen, I mean at this present instant; whereas, if I say, I have fallen, my meaning comprehends, indeed, the foregoing; but has likewise, a secret reference to some period of time past, as some time in this day, or in this hour, I have fallen; implying some continuance of time, which the other form of expression does not. (Priestley 1768: 127–128)

From the quotation above, it appears that Priestley advocates the use of be in a context of immediacy or “recentness” and the use of have in a context of duration or “pastness”. From this admittedly rather limited survey of Late Modern English grammars it seems as if Lowth was indeed one of the first grammarians to note the existence of a variation in choice of auxiliary with mutative verbs. Priestley did not discuss it until the second edition of his grammar and we know that he was familiar with Lowth’s, so it is possible that his remarks were at least partly inspired by that grammar. There are two main questions that this study will try to answer. What is the position of Priestley’s usage in light of the shift from be to have with participles of mutative intransitives? And how does Priestley’s practice compare to his prescription? In other words, is Priestley’s usage conservative or progressive and does he practise what he preaches? 2. Method In order to discover Priestley’s actual usage, I shall look at the distribution of be and have with intransitive mutatives in his personal correspondence. This study uses autograph letters, rather than published letters, as primary data. Manuscripts show the actual usage of their author more faithfully than edited correspondence, because editing practices are often not transparent (see also Dury 2006). I transcribed 419 of Priestley’s private letters and collected them as a machine-readable idiolect corpus. At present, the Joseph Priestley Letter Corpus (JPLC) has a size of 197,794 words of running text and is still under construction.3 Using this corpus I will examine whether there are diachronic or stylistic (in terms of register) variations in the choice of auxiliary. I will also investigate the syntactic-semantic environments in which variation with be and have occurs. The JPLC covers the latter half of the eighteenth century, and I divided the letters into three roughly equal periods, 1766–1779, 1780–1793 and 1794–1804. Rather than using an arbitrary division into decades, I chose to use three periods 3. When completed the JPLC will contain 431 letters with a total size of 209,736 words running text.



Robin Straaijer

that were delimited by changes in Priestley’s personal situation,4 as such changes may influence his usage. In order to analyze possible register variation, the corpus was also divided into formal and informal letters. The level of formality of a letter was based on the form of its salutation and end formula. For instance, an end formula such as “yours sincerely” usually indicates an informal letter, while the type “your humble servant” is usually indicative of a formal letter (see also Tieken-Boon van Ostade 1999). I will compare the findings for the JPLC to one part of the corpus of texts compiled by Rydén & Brorström, namely the part that consists of letters from the period 1750 to 1800. In order to be able to compare the results from my searches with theirs, I use the same list of 161 mutative intransitive verbs (Rydén & Brorström 1987: 228–230),5 which is too long to repeat here. No comparison between their corpus and the JPLC could be made for the verbs finish, meet, miscarry and (re)move because Rydén & Brorström give no data for these verbs for the letters of the second half of the eighteenth century. Another, more recent corpus-based investigation of the variation of be and have with the participles of intransitive mutative verbs is Kytö (1997). It is a long diachronic study covering the period from 1350 to 1990, which uses a large multigenre corpus of nearly 2.8 million words. Because it reports the sizes of the corpora used, Kytö’s study would appear to be more suited for a comparison with this study than Rydén & Brorström’s. However, the much longer period and the large number of text types under investigation has the consequence that the data for the relatively short period and the specific text type in which I am interested is too small to allow a good comparison to be made. For the period from 1750 to 1800 Kytö’s corpus has no more than a total of 30 occurrences of be and have for the genre of letters, which is too small a number to perform a quantitative analysis of, for instance, syntacticsemantic environments with any useful results. Hence, although the later study is in some way more rigorously performed, I will be using the older study by Rydén & Brorström as a reference corpus with which to compare my data. It should be noted that, since the total size of Rydén & Brorström’s corpus is not known, the statistical analysis of a comparison between the JPLC and Rydén & Brorström’s corpus – or parts of it – has to be based on the total number of verb forms rather than the total number of words in the corpus. Because of the low 4. In 1780 Priestley quit the employ of the Earl of Shelburne in Wiltshire and moved to Birmingham, where he became a member of the Lunar Society. This had a considerable effect on his social situation and his connections. In 1794 he emigrated to the United States of America to escape religious and political persecution in Britain. The emigration itself and his subsequent settlement in the back country of Philadelphia radically altered his personal and social circumstances. 5. WordSmith Tools 5.0 (Scott 2008) was used for the concordance searches.

Be/have variation with mutative intransitive verbs 



frequencies in the JPLC, the log-likelihood (G2) statistic rather than the chisquared (X 2) was used to calculate statistical significance.6 3. Results The most frequently occurring mutative intransitive verbs in the JPLC are in increasing order of frequency: return, miscarry, become, recover, go, arrive and come. Each of these occurs at least ten times in the whole corpus. Table 1 shows their frequencies in alphabetical order compared to Rydén & Brorström’s corpus. For easier comparison with the corpus of Rydén & Brorström, I have normalized frequencies of occurrences to percentages rather than to the number of occurrences per ten thousand words because of the relatively low raw scores (given in parentheses). As can be seen from Table 1, the use of have in the JPLC in general is close to that of Rydén & Brorström’s letters for the second half of the eighteenth century (1987: 232). There is no statistically significant difference between these two percentages (39 percent in the JPLC versus 33 percent in Rydén & Brorström) which suggests that in general Priestley’s usage was neither conservative nor very progressive. Diachronic variation can be indicative of a conscious effort by the author to change his usage, driven by changing contemporary norms for instance. There is only a little evidence for diachronic variation in this feature within the JPLC. Between the second and the third period there is a weakly significant increase in Table 1.  Be/have variation with mutative intransitive verbs in the JPLC compared to Rydén & Brorström (1987) verb

JPLC be

arrive become come go miscarry recover return all verbs

89% 83% 47% 57%   0% 88% 83% 61%

R & B 1987 have

(31) (10) (21) (13) (0) (14) (10) (187)

  11%   17%   53%   43% 100%   13%   17%   39%

be (4) (2) (25) (10) (12) (2) (2) (120)

85% 75% 72% 77% – 81% 71% 67%

have (68) (12) (93) (169) – (30) (27) (544)

15% 25% 28% 23% – 19% 29% 33%

(12) (4) (37) (50) – (7) (11) (270)

6. The difference was considered not statistically significant when a G2-score corresponded to p > 0.05, weakly significant when 0.01 < p < 0.05, significant when 0.001 < p < 0.01, and highly significant when p < 0.001.



Robin Straaijer

the use of have in general, which is consistent with the direction of the change in the English language. In the formal register, be and have both occur equally often, while the distribution between be and have within the informal register is close to that of the entire corpus (60 percent be and 40 percent have). However, the difference between the two registers is not statistically significant. Since both time and register are not significant variables, I have not given a tabular view of them and they will not be included in the following discussion. Rydén & Brorström mention several syntactic-semantic contexts that heavily promote the use of have: past perfect, perfect infinitive, ing-constructions, and conditional, concessive, optative, iterative, durative and negative contexts (Rydén & Brorström 1987: 184–193). It appears that Priestley’s usage is governed by simpler criteria, since only contexts of ‘pastness’ and conditional contexts significantly favour the use of have. Conversely, only contexts of ‘recentness’ and the present perfect significantly favour the use of be. These contexts of ‘pastness’ and ‘recentness’ are understood as being signalled primarily by the use of temporal adverbials denoting past or present. Examples of this are given in the passage from Priestley’s letters quoted in the following discussion. 4. Discussion In this section I will discuss the most frequently occurring mutative intransitive verbs in the JPLC in more detail. In order to be able to say anything meaningful about the distribution of be and have with these verbs, I will limit myself to those verbs which occur at least ten times in the whole corpus, as shown in Table 1. I will briefly discuss the semantic and/or syntactic constraints of the construction of these verbs’ participles with the auxiliaries be and have. I will discuss these verbs in increasing order of frequency: return, miscarry, become, recover, go, arrive and come. As have is the rising variant in the paradigm, I will discuss them primarily in relation to this auxiliary. 4.1

Return

There is no significant difference in the use of either auxiliary with return between the JPLC and Rydén & Brorström. Priestley appears to favour have when the sense is decidedly past, as indicated by “the day before yesterday” in (3), although this could not be verified statistically.

(3) the day before yesterday, she coughing violently at ten times, in consequence of taking something into the wind pipe, she burst a blood vessel

Be/have variation with mutative intransitive verbs 



and continued to spit blood perhaps two hours, but in no great quantity, and it has not returned since (letter to Theophilus Lindsey, 2 July 1790) In a context of recentness, i.e. when a relatively recent event is referred to, as in (4), Priestley has a significant preference for be as an auxiliary. (4) William is just returned to us from Boston, and better than I expected.  (letter to Theophilus Lindsey, 12 July 1795) 4.2

Miscarry

Priestley uses miscarry only with the sense of letters going missing or being delivered to the wrong place and indeed half of the occurrences of miscarry refer explicitly to letters, of which (5) and (6) are typical examples. The preference for the use of have as the auxiliary with miscarry is highly significant. Indeed, the use of be with this verb is not attested at all in the JPLC. (5) As several of Mr Lindsay’s letters to me have miscarried, I fear it may be the same with mine to you, or yours to me.  (letter to John Wilkinson, 30 November 1797) (6) Mrs Rayner’s letter must have miscarried, as I have never received it.  (letter to Theophilus Lindsey, 23 December 1798)

Note that according to Rydén & Brorström the use of have is dominant in this construction from the time it emerged around 1600 (1987: 132). The possible alternative interpretation of this construction as a transitive use would promote the use of be. The fact that have is nevertheless preferred is indicative of the general trend towards the use of this auxiliary. In the JPLC there are no other syntacticsemantic environments in which miscarry occurs. 4.3

Become

There is no significant difference in the use of either auxiliary with become between the JPLC and Rydén & Brorström. Priestley uses have as the auxiliary when the sense is more clearly past, as in (7).

(7) I think I told you that the air which I extracted from nitre, and which was, at first, wholesome and good, had, in the course of a year, become highly noxious. (letter to Richard Price, 26 November 1772)

The only two occurrences of have with become are of this type. When the sense is obviously recent or present, as in example (8), Priestley has a significant preference for the auxiliary be.



Robin Straaijer



(8) I have just found, to my great surprise, that a quantity of air, which I got about a year ago from salt petre, and which was then quite wholesome, and in which candle burned perfectly well, is now become in the highest degree noxious. (letter to Richard Price, 11 November 1772)

The tendency towards “free variation” between be and have that Rydén & Brorström observe towards the end of the eighteenth century (Rydén & Brorström 1987: 50) is not corroborated in Priestley’s letters. 4.4 Recover The distribution between be and have with recover in Rydén & Brorström is similar to that in the JPLC, where there is a significant general preference for be. As shown in (9) and (10), Priestley uses both be and have with recover in the sense of ‘to improve in health’. (9) The pope has recovered contrary to our expectations and wishes.  (letter to Theophilus Lindsey, 27 August 1803) (10) I am now returned from Philadelphia, and tho I had a long relapse into a state of very indifferent health after I informed Mr Lindsey that I was perfectly recovered, I am now again much better than I lately expected I ever should be (letter to Thomas Belsham, 2 April 1803) In contexts of recentness there is a significant preference for be, as exemplified by (11). Indeed, there is only evidence of the use of be as the auxiliary in Priestley’s letters. (11) Mr Russell, who is just recovered from a feverish complaint that has held him a long time, is affected more than I am.  (letter to John Wilkinson, 4 June 1793) In a context of “regular” definite pastness, as in (12), have seems preferred, though the scores were too low to support this statistically. (12) He had had an ague, which was very generally prevalent in this country, the last summer, but had recovered very well, and even a second time after a relapse; but after those he had frequent colds and indispositions, which affected his stomach and bowels. (letter to John Wilkinson, 17 December 1795) 4.5

Go

Priestley does not generally favour either be or have with go, whilst he uses have significantly more and be significantly less than in the reference corpus. According to Rydén & Brorström, have occurs mostly with go in conditional and iterative/

Be/have variation with mutative intransitive verbs 



durative contexts (1987: 105). In the case of conditional contexts the results from the JPLC would seem to agree, as the largest proportion of uses of have are in a conditional context, such as in (13). (13) If I had gone on a mistake of your [intention], it would have been kind to have [???] apprized me of it after the first, or second time of drawing, before I had involved myself so far as I have done.  (letter to John Wilkinson 3 November 1796) However, in the other six instances in which Priestley uses have with this verb, it is not in iterative/durative contexts, but rather in contexts in which there is a definite sense of pastness and/or completion. For Priestley, the use of have appears to signal the perfective better than be, as in (14): (14) Yourself and Mrs Lindsey as well as some others, have gone far beyond you[r] proportion. (letter to Theophilus Lindsey, 12 July 1795) Rydén & Brorström found that be is used normally in the go + to-infinitive construction (1987: 105), as in example (15). This is also the case in the JPLC, where these constructions show a significant preference for be. (15) In the winter we were very fortunate in having a man and his wife with us from England, who are now gone to settle on a farm.  (letter to Theophilus Lindsey, 12 July 1795) As expected, this is also a context of recentness, as shown by the occurrence of now in example (15). The go + to-infinitive construction does not occur with have as its auxiliary. 4.6 Arrive There is no significant difference in the use of either auxiliary with arrive between the JPLC and Rydén & Brorström. In general, Priestley strongly favours the use of be, the greater occurrence of which is highly significant. This corroborates Rydén & Brorström’s findings that “[i]n the eighteenth century be is still heavily predominant” (1987:  44). The frequently recurring construction arrive + locP/ dirP (locative or directional prepositional phrase) occurs exclusively with be. This type, shown in (16), is the most frequent construction with arrive in the JPLC. (16) I write to inform you that I and my wife are just safely arrived at Sandy Hook, and expect to be at New York before Midnight.  (letter to John Vaughan, 3 June 1794)



Robin Straaijer

There are no occurrences of be + arrive +temP (temporal prepositional phrase) in the JPLC. Rather, temporal relations are mostly conveyed by adverbials of time. Have does not occur frequently enough with arrive in any one context to say anything meaningful about its distribution, although there might be a preference for have in negative contexts. 4.7

Come

In general, Priestley uses be significantly less and have significantly more with come than is found in Rydén & Brorström. These authors report that be is the dominant alternative (1987: 62). However, this is not reflected in the JPLC, where there is no significant preference for either auxiliary. Rydén & Brorström also found that be predominates in negative contexts (1987: 62). Again, this is not reflected in the JPLC, where there is no significant difference between the use of be and have, each occurring about equally often. Rydén & Brorström note a predominance of be in the collocation come home (1987: 62–63). The JPLC has eight instances of the semantically related collocation come hither, of which (17) is the only attestation with be. (17) But being come hither, and having, at a great risk and expence, brought my library and Apparatus hither, tho that scheme has failed, I cannot remove any more. (letter to John Adams, 13 November 1794) However, it should be noted that the collocation come hither in the sense of arrive, as illustrated in (18), seems more strongly perfective than come and thus would promote the use of have. (18) I would send you one of me, taken since I have come hither, but I am so different from what I was, chiefly in consequence of wearing my hair, that you would hardly know me. (letter to Theophilus Lindsey, 29 May 1797) Indeed, the preference of have in this construction (come hither) in the JPLC is weakly significant. Incidentally, this is also an example of the construction come + locP/dirP, which is the single most frequently occurring construction with come in the JPLC. The preference for the use of have in this construction is weakly significant. An example of this construction with be is given in (19). (19) Almost all the ships expected to arrive before winter are come in.  (letter to Theophilus Lindsey, 7 November 1801) To find a broader pattern of auxiliary selection, I investigated Priestley’s use of be and have across all mutative intransitive verbs in the following contexts: negative, conditional, pastness, recentness, present perfect and V + locP/dirP. There are

Be/have variation with mutative intransitive verbs 



significant differences between the use of be and have in the conditional, pastness and recentness contexts, and in the present perfect. In conditional contexts and in contexts denoting pastness, have is used significantly more often, whilst be is used more often in contexts denoting recentness and in the present perfect. Some of these results corroborate the findings of the earlier studies. Both Rydén & Brorström, and Kytö find a greater use of have in conditional contexts (Rydén & Brorström 1987: 186; Kytö 1997: 58). Kytö reports more be in the present perfect, which also indicates recentness, and more have in the past perfect, which indicates pastness (Kytö 1997: 52–53). Rydén & Brorström find a preponderance of be in the past perfect as well as in the present perfect (1987: 63), but this is not observed in the JPLC, which has only two instances of the past perfect construction, one with be and one with have. 5. Conclusions Rydén & Brorström note that it is often difficult to distinguish between the function of be + past participle as copula + adjective and that of perfective auxiliary with intransitive verb (1987: 24, 27), which they illustrate as follows: The functional range of be + PP covers [...] be doing duty as (1) copula (with adjectival statal PP): he is changed ‘he is different’ (2) perfective Aux, i.e. tense marker (with intransitive verbs): he is changed ‘he has become different’ (Rydén & Brorström 1987: 24)

This problematic ambiguity might also have affected actual usage and thus partly account for the greater use of be in contexts of recentness, though it has no effect on the use of have in contexts of pastness. Compared to the general be/have distribution in Rydén & Brorström’s corpus of letters between 1750 to 1800 (Rydén & Brorström 1987: 196) Priestley’s usage is neither conservative nor progressive. His usage is illustrative of the general usage of the educated, literate middle class from which he came, and from which Rydén & Brorström derive their data (1987: 215–216). As shown in the quotation in the introduction to this paper, Priestley gives no prescription for any standard usage of either be or have because according to him it depends on the context which auxiliary is more appropriate with the main verb in question. This is indeed reflected in his own usage in his letters: Priestley generally uses have when there is a more definite sense of pastness or completion and be when this is less so. This corresponds to what he says in his grammar, and can mean one of two things. Either Priestley “cleaned up” his usage after writing his grammar or he described his own usage in his grammar, thus in essence being his own informant



Robin Straaijer

considering matters of usage and grammaticality. The first option is possible since 97  percent of the letters (corresponding to 94  percent of the total number of words) in the JPLC date from after 1768, the year Priestley published the second edition of his grammar. However, I am greatly inclined to say that the latter is the case, since Priestley’s usage seems too consistent to have been influenced much by writing his grammar, rather the opposite in fact. We can therefore conclude that with regard to auxiliary option with mutative intransitive verbs, Priestley lives up to his reputation as a descriptivist, and that his description reflects his own usage. References Baugh, Albert C. & Thomas Cable. 2002. A History of the English Language. 5th edition. Upper Saddle River, New Jersey: Prentice Hall. Buchanan, James. 1762. The British Grammar: Or, an Essay, in Four Parts, Towards Speaking and Writing the English Language Grammatically, and Inditing Elegantly. London: printed for A. Millar. Bryan, William F. 1923. “Notes on the Founders of Prescriptive English Grammar”. Manly Anniversary Studies in Language and Literature. Chicago: Chicago University Press. 383–93. Dury, Richard. 2006. “A Corpus of Nineteenth-Century Business Correspondence: Methodology of Transcription”. Business and Official Correspondence: Historical Investigations ed. by Marina Dossena & Susan M. Fitzmaurice, 193–205. Bern: Peter Lang. ECCO = Eighteenth Century Collections Online. Gale Group. . Greenwood, James. 1711. An Essay Towards a Practical English Grammar. London: printed by R. Tookey, and sold by S. Keeble, J. Lawrence, J. Bowyer, R. and I. Bonwick, and R. Halsey. Harris, James H. 1751. Hermes: or, a Philosophical Inquiry Concerning Language and Universal Grammar. London: printed by H. Woodfall, for J. Nourse, and P. Vaillant. Hodson, Jane. 2006. “The Problem of Joseph Priestley’s (1733–1804) Descriptivism”. Historiographia Linguistica: International Journal for the History of the Language Sciences 33.57–84. Kytö, Merja. 1997. “Be/have + Past Participle: The Choice of the Auxiliary with Intransitives from Late Middle to Modern English”. English in Transition: Corpus-based Studies in Linguistic Variation and Genre Styles ed. by Matti Rissanen, Merja Kytö & Kirsi Heikkonen, 17–85. Berlin & New York: Mouton de Gruyter. Leonard, Sterling Andrus. 1929. The Doctrine of Correctness in English Usage, 1700–1800. Madison, Wis.: Wisconsin University Press. Lowth, Robert. 1762. A Short Introduction to English Grammar. London: printed by J. Hughs; for A. Millar; and R. and J. Dodsley. ODNB = Oxford Dictionary of National Biography Online. . Priestley, Joseph. 1761. The Rudiments of English Grammar; Adapted to the Use of Schools. With Observations on Style. London: printed for R. Griffiths.



Be/have variation with mutative intransitive verbs  Priestley, Joseph. 1768. The Rudiments of English Grammar, Adapted to the Use of Schools; With Notes and Observations, for the Use of Those Who Have Made Some Proficiency in the Language. London: printed for T. Becket and P.A. De Hondt, and J. Johnson. Rydén, Mats & Sverker Brorström. 1987. The BE/HAVE Variation with Intransitives in English: with Special Reference to the Late Modern Period. Stockholm: Almqvist & Wiksell. Scott, Mike. 2008. WordSmith Tools 5.0. . Straaijer, Robin. 2009. “Deontic and Epistemic Modals as Indicators of Prescriptive and Descriptive Language in the Grammars by Joseph Priestley and Robert Lowth”. Current Issues in Late Modern English ed. by Ingrid Tieken-Boon van Ostade & Wim van der Wurff, 57–88. Bern: Peter Lang. Tieken-Boon van Ostade, Ingrid. 1999. “Of Formulas and Friends: Expressions of Politeness in John Gay’s Letters”. Thinking English Grammar ed. by Guy A.J. Tops, Betty Devriendt & Steven Geukens, 99–112. Leuven: Peeters. Tieken-Boon van Ostade, Ingrid. 2002. “Robert Lowth and the Corpus of Early English Correspondence”. Variation Past and Present: VARIENG Studies on English for Terttu Nevalainen ed. by Helena Raumolin-Brunberg, Minna Nevala, Arja Nurmi & Matti Rissanen, 161–72. Helsinki: Société Néophilologique. Tieken-Boon van Ostade, Ingrid. 2006. “Eighteenth-Century Prescriptivism and the Norm of Correctness”. The Handbook of the History of English ed. by Ans van Kemenade & Bettelou Los, 539–557. Malden, Mass.: Blackwell.

On the idiomatization of “give + O + to” constructions Minoji Akimoto*

Aoyama Gakuin University This paper discusses the idiomatization process of the “give + O + to” construction (e.g. give rise to and give way to). Investigation of this construction from Middle English to Present-Day English based on various corpora shows that, in the process of change of “give + indirect object + direct object” to “give + O + to”, the latter construction has strengthened its idiomatic nature leading to the development of many idioms based on this construction in PresentDay English. The function of eventive objects without suffixes (e.g. credit and rise) as direct objects is essential in this process of idiomatization. Of various factors contributing to the idiom formation of the “give + O + to” construction, weakening of nouniness of the direct object, construction and frequency, and French influence play the most important roles.

1. Introduction “Give + O + to” idioms such as give birth to, give occasion to or give rise to1 are abundant in Present-Day English. In Present-Day English, the verb give takes two types of construction, that is either “give + indirect object (Oi) + direct object (Od)” or “give + object (O) + to” (Quirk et al. 1985: 1208).2 Most idioms belong to the latter type. In earlier English, however, the “give + Oi + Od” construction was more common * I am grateful to Laurel Brinton, Ursula Lenker, Judith Huber and the anonymous referees for helpful comments and suggestions on earlier versions of this paper. Thanks are also due to Merja Kytö for enabling me to make use of the ARCHER Corpus. 1. For illustration, I give some frequent examples of the “give + O + to” construction in the BNC: give rise to (1,856), give way to (817), give birth to (414), give credence to (58), give voice to (55), give the lie to (47), give vent to (45), and give place to (32). 2. After Quirk et al. (1985: 1208), I shall represent “give + indirect object + direct object” as “give + Oi + Od”, but in the case of “give + O + to”, although O is here a direct object, I shall represent it as neutral.



Minoji Akimoto

(see Denison 1993:  107; Visser 1970:  621; Allen 2006 and McFadden 2006). From around the Middle English period, the “give + O + to” construction began to be frequent, and hence productive. Most previous discussions have considered these changes to be grammatical change, that is, a result of the loss of the case endings, without any reference to the further development of the “give + O + to” construction with idiomatic nature. I shall investigate some factors which have led to the development of idiomatic “give + O + to” patterns. This paper is organized as follows. In Section 2, I will survey the state of the “give + O + to” construction before Middle English; this survey is based on the Middle English Dictionary (MED). In Section 3, I will first describe “give”‑constructions from Middle English to Present-Day English, and then discuss the idiomatization process of the “give + O + to” construction, mainly based on data from the OED, the Helsinki Corpus, ARCHER and FLOB. In Section 4, I will discuss some factors contributing to the idiomatization of this construction. 2. A general survey of the “give + O + to” construction before Middle English Quirk & Wrenn (1957: 64–65) say: “In late OE, tō came to be used with the indirect object just as in Mod.E; thus gyfan (tō) ænig ‘give (to) anyone’”.3 Visser (1970: 621–635) shows that in Old English the indirect object + direct object construction is common, the indirect object being expressed in the dative and the direct object in the accusative case. He further states that the indirect object refers to a person and the direct object to a thing. As regards Middle English, Kellner (1957:  271) states: “[...] throughout the thirteenth century, to replacing the dative is far from being generally accepted”. Denison (1993: 105–107) describes the process of change of benefactives marked with the dative to those marked with to which, he says, became common beginning in the fourteenth century. From these quotations, it appears that the Middle English period can be considered as the first stage of the “give + O + to” construction. Therefore, I shall look into the “give + O + to” patterns in the Middle English Dictionary (MED). In the MED (s.v. yēven), twenty-seven meanings of give are identified. In the following, I will give some examples of subdivided meanings. The numbers in parentheses correspond to those of the entry meanings in the MED. (1) ‘to give (sth.) gratuitously and unconditionally’ (1a.(b)) A frend Šeueþ a present to anoþir. (?a1425 Orch. Syon 159/35) ‘A friend gives a present to another.’ 3. Mitchell (1985: 512‑513), however, casts doubt on the existence of giefan to ænigum.

Idiomatization of “give + O + to” constructions 



(2) ‘furnish’ (1d.(g)) Þis lampe Šefth gret lyŠt To euery creature abouŠt hit dwellynge.  (c1450 St. Editha 958) ‘This lamp gives a great light to every creature dwelling about it’ (3) ‘give (credence to sb. or sth.)’ (8.(a)) They yauen ful credence Vn to hir word. (c1380 Chaucer CT. SN. G.415) ‘They gave full credence to her word.’ (4) ‘issue’ (13.(d)) You shall ... geve warninge to the master and wardens therof.  (1463–4 Doc. in Welch Hist. Pewterers Lond. 31) ‘You shall ... give warning of it to the master and wardens.’ (5) ‘bring life’ (9a.(e)) He þat Šeueþ liif to þe tre ond to þe braunchis is þe rote.  (?a1425 Orch. Syon 37/23) ‘He that gives life to the tree and to the branches is the root.’ (6) ‘to do (battle)’ (14.(b)) He hath gyven bataill to the same kyng, and putt hym to flight.  (a1475 Fortescue Gov. E. 129) ‘He has given battle to the same king and put him to flight.’ (7) ‘to give a hearing (to someone’s request)’ (20.(b)) Še list ... to Ševen audience To my request.(c1425 (a1420) Lydg. TB 1.1418) ‘You like [...] to pay attention to my request.’ (8) ‘give in’ (21b.(a)) The chiefteynes han ... yove way to hire enemyes & made the peopil pray to him. (a1500 (1413) Pilgr. Soul 4.32. 78b) ‘The chieftains have given way to their enemies and made the people pray to them.’ 3. Analysis of “give + O + to” constructions 3.1

Rivalry among the give constructions

I would now like to trace the state of “give + Oi + Od”, “give + O + to” and other constructions4 from Middle English to Present-Day English periods on the basis of the Helsinki Corpus (hereafter HC), the ARCHER Corpus (hereafter AC) and 4. By other constructions I mean all constructions with give except “give + Oi + Od” and “give + O + to” constructions, such as give a view and give away.



Minoji Akimoto

the FLOB Corpus.5 Note that the frequencies given below differ in terms of the size of the corpora, and that only the percentages can reflect some trends of change. The following are some examples of each pattern given above. The examples of “give + Oi + Od” will be cited in the later discussion. Table 1.  Percentages of three categories of “give” constructions from Middle English to Present-Day English

ME

EModE

LModE

PDE

I II III IV I II III I II III IV

give + Oi + Od

give + O + to

  33   41   39   58   28   53 136 168 397 346 286 422

   0   13   25   18   11   11   29   43 138 135 128 138

(62%) (46%) (26%) (33%) (26%) (35%) (42%) (49%) (39%) (31%) (33%) (33%)

(0%) (14%) (16%) (10%) (10%)   (7%)   (9%) (12%) (13%) (12%) (14%) (10%)

Others6   20   35   84   96   69   85 155 128 468 618 450 713

(37%) (39%) (57%) (56%) (63%) (57%) (48%) (37%) (46%) (56%) (52%) (56%)

total    53    89   148   172   108   149   320   339 1003 1099   864 1273

(9) Examples with “give + O + to”: a. ... or Place for Devotion, with Images, which gave occasion to Doubt if ever for that End ... (1698 A New Account of East India) [HC] 5. The Helsinki Corpus is divided into the following periods: ME I (1150–1250, 113,010 words), ME II (1250–1350, 97,480 words), ME III (1350–1420, 184,230 words), ME IV (1420–1500, 213,850 words), EModE I (1500–1570, 190,160 words), EModE II (1570–1640, 189,800 words), EModE III (1640–1710, 171,040 words). The ARCHER Corpus covers seven subperiods of 50 years, spanning the time between 1650 and 1990. Each 50-year subcorpus includes roughly 20,000 words per register, containing ten texts of approximately 2,000 words each. The entire corpus totals some 1,700,000 words. Approximately two-thirds of the corpus are British English, and the one-third is American English. The FLOB Corpus (short for Freiburg version of the Lancaster-Oslo-Bergen Corpus) contains 500 samples of approximately 2,000 words each, totalling approximately one million words. As regards the Helsinki Corpus and the ARCHER Corpus, see Kytö (1996) and Biber at al. (1994) respectively. For practical reasons, I chose between 10% and 20% of the many examples of the “give + O + to” construction in each of these corpora. 6. The high frequency of this category includes the non-finite forms of give, its intransitive uses, and its phrasal verb uses, such as give up and give away. Particularly, the development of phrasal verbs since Late Modern English deserves attention. Of the give phrasal verbs in Present-Day English (84 examples), give up (37) is the most frequent.

Idiomatization of “give + O + to” constructions 



b. I endeavour to regulate my conduct so as to avoid giving offence to any man ... (1778 reev. f3) [AC] c. Next morning the girl was taken with labour, and soon gave birth to a dead child ... (1864 mack. m6) [AC] (10) Other examples: a. ... and no man came yn to gyve evydence agaynste me. (1553 The Autobiography of Thomas) [HC] b. You have not heard a quantum sufficit to render you competent to give a decisive opinion; besides you hear with passion and prejudice. (1789 lows. d4) [AC] c. Adam gave up making her understand his scruples of grammar, at least until she should recover from the idiocy of her honeymoon. (1878 eggl. F7) [AC] The following points are worth making: 1. While the “give + O + to” construction is not attested in the ME I period (1150–1250), this construction has been steady from ME II up to the present. 2. In the “give + Oi + Od” construction, more than two thirds of Oi are pronouns. In Table 1, Ois include both nominal and pronominal objects. The second point is demonstrated in Table 2, which shows the frequency of non-pronominal Ois among Ois in the “give + Oi + Od” construction. Table 2.  Number of attestations of non-pronominal Ois give + Oi + Od ME

EModE

LModE

PDE

I II III IV I II III I II III IV

   7    4   13    4    5    7   20   35   68   86   77 167

(21%)   (9%) (33%)   (6%) (17%) (13%) (14%) (20%) (17%) (24%) (26%) (39%)

total   33   41   39   58   28   53 136 168 397 346 286 422

Table 2 illustrates the predominance of pronominal indirect objects. This tendency, which corresponds to McFadden’s (2006: 119) research on the word order in Middle English, has continued from Early Middle English to Present-Day English.



Minoji Akimoto

3.2

The “give + O + to” construction

The following tables 3 and 4 show the breakdown of objects in the “give + O + to” construction; divisions are made between eventive objects and other nouns. The reason for this division of nouns is that the eventive object in combination with give tends to be idiomaticized because of its abstract nature (see below, Section 4), while the non-eventive object noun usually does not constitute an idiom. Table 3.  Percentages of eventive and other object nouns Eventive object ME

EModE

LModE

I II III IV I II III I II III IV

PDE

 0  5  7  6  8  5 19 28 55 65 55 88

  (0%) (38%) (28%) (33%) (72%) (45%) (65%) (68%) (39%) (48%) (42%) (73%)

other noun

total

 0  8 18 12  3  6 10 13 83 70 73 32

   0   13   25   18   11   11   29   41 138 135 128 120

  (0%) (61%) (72%) (66%) (27%) (54%) (34%) (31%) (60%) (51%) (57%) (26%)

Table 4.  Numbers of eventive objects with/without suffixes

ME

EModE

LModE

PDE

I II III IV I II III I II III IV

suffix

no suffix

total

 0  1  2  3  3  2  4  7 18 22 19 33

 0  4  5  3  5  3 15 21 37 43 36 55

 0  5  7  6  8  5 19 28 55 65 55 88



Idiomatization of “give + O + to” constructions 

Quirk et al. (1985: 750–752) mention an eventive object, explaining that it is “semantically an extension of the verb and bears the major part of the meaning” (750). They distinguish three types: 1) a noun which has the same form as a verb (e.g. fight), 2) a suffixed noun (e.g. argue vs. argument), and 3) a noun not derived from a verb but semantically equivalent to the verb (e.g. effort). Thus, functionally, deverbal nouns which stem from their corresponding verbs with or without suffixes are included in the category of eventive objects. (11) Examples with eventive objects with suffix: a. This mayster gevyth commaundementes to the childe in his growynge age. (d. 1497 In Die Innocencium) [HC] ‘This master gives commandments to the child in his growing age.’ b. ... if my Lady would give Entertainment to one (ˆHicksˆ), and another Person ... (d. 1685 The Trial of Lady Alice Lisle) [HC] c. ... since he hath submitted to the King in everything, and given all satisfaction to the Court that he was able. (1654 mer 2. n1) [AC] d. In the face of the threat to Transylvania, no one gave much thought to the southern Dobrudja ... (1962 mann. f9) [AC] (12) Examples with eventive objects without suffix a. ... many oþer seyntys þat ben in Hevyn, whech Ševyn gret worshep to me for þe grace þat ... (a1438 Margery Kempe) [HC] ‘... many other saints who are in heaven, who give great worship to me for the grace that ...’ b. what arte thou that we maye geve an answer to them that sent vs ...(d. 1534 The New Testament) [HC] c. The federal constitution does not affect the right of the state: for though it gives effect to the treaty of peace ... (1795 para. 14) [AC] d. ‘Well, would you mind confiding to me the circumstance that gave rise to the reflection?’ (1895 mach. f6) [AC] e. In addition, the presence of a family history gave a clue to the possibility that ... (1985 wier. m0) [AC] From the tables given above, we see a steady increase of deverbal nouns from the Early Modern English period onward, although the increase is by no means straight, with ups and downs depending on the centuries, possibly because of different styles and registers of each century. Since the LModE period, those nouns without suffixes, such as vent, way and rise have become frequent, which sets the stage for the idiomatic nature of the “give + O + to” construction. The nouns forming the more idiomatic phrases are those without suffixes. On the other hand, those nouns with suffixes tend not to become idiomaticized possibly



Minoji Akimoto

because they suggest more nouniness or heavy nouns whose property is incompatible with their idiomatic nature7 (for details, see Section 4). A further related issue partly stemming from the assumption that the “give + O + to” pattern forms an idiomatic construction is raised by Hudson’s (1992: 262) finding that the Od is part of an idiom, but the Oi rarely is. Usually the indirect object is an animate object, which means that an idiom does not consist of “verb + animate object”. If this is true, it can be said that a shift from the “give + Oi + Od” construction to the “give + O + to” construction in English has paved the way for the formation and development of give idioms based on the “give + O(inanimate) + to” construction. That is, despite the order of “give + Oi + Od”, an idiom consists of “give + Od”8, and therefore the Od must be closely related to the verb. In this respect, it can be said that the “give + O + to” construction, once established, has become a productive basis for the formation of idioms.9 4. Some factors contributing to the idiomatization of the “give + O + to” construction Semantically, idiomatization is a process whereby meaning becomes more opaque, and syntactically, the phrase unity becomes more fixed than before. One characteristic of idiomatization is the gradual abstractness of nouns whose conceptualization is more or less equivalent to that of decrease in nouniness. In view of these semantic and syntactic properties, the idiomatic patterns, such as give birth to and give rise to, should be kept apart from idioms such as kick the bucket and shoot the breeze. Nunberg et al. (1994), for instance, distinguish “idiomatically combining expressions”, whose meanings are distributed among their parts, from “idiomatic phrases”, which do not contribute their meanings to their components. The idiomatic give construction under discussion, such as give birth to and give vent to, compared with such an idiom as kick the bucket, shows more transparency in meaning, but syntactically fixed structure, i.e. *Rise should be given to the rumour, although there are degrees of fixedness. In this respect, the “give + O + to” construction is closer to the idiomatically combining expression type. These two types of idiom share the 7. Birth, for instance, is a deverbal noun, but the form does not go against this statement. The essential point concerns the length (or heaviness) of the noun after the suffix is added. 8. See Tomlin (1986:  89–91) and Nunberg et al. (1994:  525–527). Larson (1988:  339–341) holds a different opinion of this structure. 9. The placement of the object immediately after the verb also agrees with what Haiman (1983: 782–783) calls ‘iconic motivation’; e.g. the linguistic distance between expressions corresponds to the conceptual distance between them.

Idiomatization of “give + O + to” constructions 



abstractness of nouns and syntactic restraints, differing in their transparency. The following are some representative examples which are given chronologically. (13) ...wherefore you gave ear to my Lord (ˆCobhamˆ) for receiving Pensions ... (1603 The Trial of Sir Walter Raleigh) [HC] (14) What ... dangerous opinions soever their phancies might give birth to. (1662 Glanvill Lux Orient. Pref. 5) [OED] (15) O every thing, Madam, is to give way to Business ... (1697 Relapse) [HC] (16) ... wou’d you give the lye to Authority? (1723 bull. d2) [AC] (17) It has given rise to an infinite deal of whimsical theory. (1756 Burke Subl. & Beaut. III.xi (1759) 208) [OED] The following factors are some of those which seem to contribute to the formation of the idiomatic “give + O + to” construction. 4.1

Weakening of Nouniness

I have borrowed the term “nouniness” from Ross (1973: 141), who advocates the nouniness squish as follows: that > for to > Q > Acc Ing > Poss Ing > Noun. He refers to degrees of nouniness in Present-Day English. I have put his idea into the concrete/abstract, suffixed/non-suffixed cast in the idiomatization process of the “give + O + to” construction. Together with the semantic expansion of give and to, and frequency of use of the construction as a whole, decrease in the nouniness of the object noun is important for idiom formation. For an expression to become idiomaticized, there is a process which it must undergo, namely a shift from concreteness to abstractness. This shift is observed in the formal and grammatical properties of nouns. At first, a noun behaves as a concrete noun in its combination with give, but in the process of idiomatization, the noun gradually loses its nominal properties, including loss of the articles and plurality, or rejection of modification. When the noun has completely lost its nominal nature, and has been embedded into the “give + O + to” construction it begins to have idiomatic properties. I shall give some examples in which this process takes place. Before the idiom give voice to had been established, voice appeared under various nouny forms, as in (18) He that should give his voice unto Christ ... (1660 Hammond Wks. IV.510 (R.)) [possessive pronoun] – [OED] (19) The Tantivy of wild pigeons, flying by twos and threes athwart my view. gives a voice to the air. (1854 Thoreau Walden iv. 125)  [indefinite article] – [OED]



Minoji Akimoto

(20) Hail to the courage which gave Voice to its creed. (1855 Arnold Haworth Churchyard v) [OED] Examples (18) and (19) show the status of full NPs before they become idiomaticized, and example (20) shows an idiomatic phrase in the sense of loss of nominal properties.10 These examples show all properties of nominals, such as determiner/ possessive determiner modification. In due course, however, voice in give voice to changed its concrete nominal nature to abstract, less nominal nature. It is only around the middle of the nineteenth century that give voice to established itself as an idiomatic construction (as in example (20)). The decreasing degree of nouniness is very common in the type of “V + O + to” idioms. Other examples, not those of give idioms, include find fault with, make use of and set foot on/in.11 10. I have checked the various forms of voice in give voice to in the OED. Since the examples are not so many, I have given all the citations including the give voice type. As the examples show, the early use of voice in the “give + O + to” construction was more concrete and literal. I gave all forms of voice regardless of the “give + voice + to” and “give + voice” constructions, because what is important here are the various forms of voice which indicate their higher degree of nouniness. (i) More Tribes ... gave their voices to approove it. (1600 Holland Livy v. xxx) (ii) I will ... give my Voice ... (1617 Moryson Itin. Iii. 190) (iii) ... to give my voice touching ... (1634 Mass. Bay Rec. (1853) I. 117) (iv) He that should give his voice unto Christ ... (1660 Hammond Wks. IV. 510 (R.) (v) ... others would give no Voice. (1678 Sidney in S. Papers (1746) I. 153) (vi) ... to give my determining Voice before the Senate. (1715 J. Barker Exilius I. 83) (vii) Nor give voice to slavery’s jackall cry. (1823 Byron Juan ix.xxvi) (viii) The Tantivy of wild pigeons ... gives a voice to the air. (1854 Thoreau Walden iv. 125) (ix) Hail to the courage which gave Voice to its creed. (1855 Arnold Haworth Churchyard v) (x) ... until he has given voice to the three words ... (1905 Jewish Encycl. XI. 146/1) (xi) To find ... daily Press ... giving voice to facts ... (1913 Christian 3 Jan. 10/1) (xii) Mr. Lyford gives voice to a temper ... (1961 Manas 5 Apr. 1/2) (xiii) ... if any young hound gives a challenging voice to a bullock ... (1976 Shooting Times & Country Mag. 18/24 Nov.) (xiv) ... which hath in their time borne great reuerence to the church of God ... found great fault against ye decrees ... (1583–1600 Deloney, The Works of Thomas Deloney: 283) (xv) Why, Madam, the Bishop has set his Foot in it. (1710 Swift, Swift’s Polite Conversation: 59) (xvi) We complained of advantages which the men take of our youth, inexperience, sensibility, and all that ... (1771 Smollett, The Expedition of Humphry Clinker: 259) These forms are in the non-idiom stage, but with an increase in abstractness in the nouns, they have developed into idiomatic phrases.



11. The nouns in the following examples (i.e. fault, foot and advantage) show the status of full NPs as in the give case given above, with modifiers and a plural form.

Idiomatization of “give + O + to” constructions 



4.2

Construction and frequency

According to Goldberg (1995:  66, 89–90), there are basic sentence-level (argument) constructions which are extended in various ways. For instance, the ditransitive construction is the semantic extension (via metaphor) of the causedmotion construction called “Transfer of Ownership as Physical Transfer”. Although Goldberg does not mention idioms based on the ditransitive construction, her idea of metaphor will be useful for the discussion of idiom formation. During the shift of “give + Oi + Od” to “give + O + to” discussed before, the meaning of give has become weakened in such examples as give rise to and give way to. With the establishment of the “give + O + to” construction, this construction began to form a syntactic-idiomatic frame where each element is absorbed into the construction. Although the meaning of give in the construction has lost much of its ‘CAUSE-RECEIVE’ meaning, it has some semantic content (cf. Green 1974: 226–229). In contrast to an idiom, which by its nature is a fixed structure and, therefore, not so productive, the construction has the potential for further expansion. These notions seem contradictory, but at the same time it can be said that the formation of idioms takes place despite the constant tug-of-war between productivity and non-productivity.12 Frequency of use also plays an important role in establishing an idiomatic construction. As “give + O + to” begins to be used frequently, phrases involving give are created,13 and give and the following noun occur in sequence; the more often the structure is used, the tighter it will become (cf. Bybee & Hopper 2001: 14), and consequently a strongly combined sequence develops into routinized patterns. Routinized or idiomaticized patterns, once established and frequently used, resist change and often result in lexicalization.14 In the case of give idioms, firstly the token frequency, and then their type frequency lead to the creation of a construction in which the “give + O + to” pattern becomes entrenched. This construction gradually becomes solidified and in due course extends itself to be productive. Change from concrete to abstract is regular in the formation of an idiomatic construction, and together with the bleaching of 12. Recently some linguists have recognized more productive properties of idioms than before. Fillmore et al. (1988: 534) state that “... the realm of idiomaticity in a language includes a great deal that is productive, highly structured, and worthy of serious grammatical investigation.” 13. In Late Old English, giefan began to show an indication of the “give + deverbal noun” construction, although limited to the “give + answer” collocation. See Brinton & Akimoto (1999: 34–36). 14. Trousdale (2008), in a construction approach, discusses the development of the giveg­ erund composite predicate in relation to grammaticalization and lexicalization. For further discussion, see Brinton & Traugott (1995).

 Minoji Akimoto

give and to, the concrete noun often develops into an abstract noun. The series of examples (taken from the OED) given below illustrates my argument. (21) They ... made so grete bruyt that the moost hardyest of the paynyms gaf them waye. (1485 Caxton Chas. Gt. 193) ‘They ... made so great noise that the hardest of the pagans retreated from them’ (22) The coaches comming downe from the upper parts, give the way to those that come up. (1617 Moryson Itin. iii. 55) (23) They who, through weakness, gave way to the ill-designs of bad men [etc.] (1793 Burke Corr. (1844) IV. 143) (24) Don’t give way to despair so quickly. (1891 Strand Mag. May 552/2) Example (22) can be said to be at an intermediate stage to an idiomatic phrase in the sense that way appears with the. The meanings have become less literal, as example (24) shows. Note that there are stages at which rival patterns exist in parallel (that is, before the establishment of the “give + O + to” construction, the “give + Oi + Od” and “give + O + to” constructions appeared at the same time). In sum, on the basis of all these observations, I shall propose the following stages of change for the formation of “give + O + to” idioms. (A) give + indirect object + direct object ↓ (B) give + object + to + NP [+ concrete] ↓ (C) give + object + to + NP [+ abstract]

....... OE period ....... ME period

....... Late ModE period

CONSTRUCTION

....... Modern English

(A) to (B) is a stage where the construction of “give + indirect object + direct object” changes into that of “give + object + to” from the Old English to the Middle English periods. From the (B) to the (C) stages, the construction of “give + object + to” begins to be common and its frequency in the data increases. The noun in the construction changes its nature from concrete to abstract. This means that the noun, which could be accompanied by the definite/indefinite articles or take the plural form, begins to lose these properties, becoming a more and more abstract noun. At the same time, at this stage, semantic opacity of the elements comes into play. The meanings of each element, give, object and to, begin to become less transparent, tending to have idiomatic

Idiomatization of “give + O + to” constructions 



meaning, i.e. give way to = ‘yield’. At stage (C), the “give + object + to” pattern becomes settled and forms a construction with the idiomatic function being strengthened. Finally, the construction has been extended to produce similar idiomatic expressions. 4.3

French influence on the ‘give + O + to’ construction

In addition to these factors, French influence on the “give + O + to” construction during the Middle English period should not be overlooked. As Allen (2006: 214) mentions briefly in relation to the high incidence of to-datives in Ayenbite of Inwit, French influence seems to have played a role, particularly in the development of the “give + O + to” construction, which must have been promoted by the French verb donner and some other French- and Latin-borrowed verbs, such as donate and contribute (cf. Mustanoja 1960: 101–102, Visser 1970: 624). The French donner construction permits only the “donner + O + à” pattern, not the “donner + Oi + Od” pattern, except when the Oi is a pronoun clitic and the Od is a full NP, and this may have encouraged the English “give + O + to” construction. Prins (1952: 138–141) gives the following examples after the model of donner: a. b. c. d. e. f. g. h.

give audience – donner audience give chase – donner la chasse give congee to – donner congé à give credence to – donner créance à give example – donner exemple give leave – donner congé give order – donner ordre give points to – donner, rendre des points

Levin (1993: 46) gives a list of verbs of non-alternating to, such as we contributed our paycheck to her vs. *we contributed her our paycheck. They are primarily verbs borrowed from Latin. I checked those more or less related to the give class regarding the earliest citations of their “V + O + to” patterns based on the OED: address (1475), announce (1483), contribute (1653), convey (1574), deliver (1440), demonstrate (1650), dictate (1667), distribute (1746), explain (1726), introduce (1739), recommend (1400), report (1489), submit (1386), transfer (1516), and transport (1483). Comparing the dates of the to construction of the French loans with the development of the “give + O + to” construction, we understand that the verbs borrowed from French with to constructions must have contributed to the development of the “give + O + to” construction. It should be emphasized, however, that verbs borrowed from French must have enhanced the “give + O + to” construction, but that the development of this construction as idiomatic unit has taken place after this stage and has undergone



Minoji Akimoto

an independent process of idiomatization partly in the tendency from synthetic towards analytic structures in English, although this tendency would need separate discussion (see Smith 1996: 153–158). 5. Conclusion I have examined the formation of “give + O + to” idioms from the Middle English period to the present time on the basis of the OED CD-ROM, the Helsinki Corpus, the ARCHER Corpus, and the FLOB Corpus. In Section 2, I made a general survey of the state of the “give + O + to” construction before 1500 by way of examples in the MED. In Section  3, I demonstrated that the “give + O + to” patterns in contrast to the “give + Oi + Od” patterns were on the increase, paving the way for idiom formation where eventive objects including deverbal nouns without suffixes are essential. Short eventive nouns without suffixes are susceptible to being idiomaticized possibly because of their short forms with increasing frequency. In Section 4, some factors contributing to the formation of the idiomatic “give + O + to” construction were discussed. They are weakening of nouniness, construction and frequency, and French influence. On the whole, occurrences of the “give + O + to” construction have been increasing throughout the history of English. This construction, once established, becomes frequent and productive, generating many idiomatic expressions. This idiomatization process involves not only the bleaching of give and to, the semantic expansion of give, and the consequent decrease in the nouniness of the object, but it also involves the construction as a whole into which give enters. All these processes have been encouraged further under the influence of verbs borrowed from French. Sources ARCHER = A Representative Corpus of Historical English Registers. BNC = British National Corpus (100 million words) distributed by Shogakkan Company. HC = Helsinki Corpus of English Texts. Diachronic Part. Ed. by Matti Rissanen, Merja Kytö, and Collin M. Palander (1993). (Included in ICAME) MED = Middle English Dictionary. (1952–2001). Hans Kurath, Sherman M. Kuhn, John Reidy, and Robert E. Lewis (eds.) Ann Arbor: University of Michigan Press. http://ets.umdl.umich.edu/m/mec OED = Oxford English Dictionary. 1989. Oxford: Clarendon, 2nd ed. (also OED CD-ROM (Version 2.0, 1999) Deloney, Thomas. 1583–1600 (1912). The Works of Thomas Deloney ed. by Francis Oscar Mann. Oxford: Clarendon.

Idiomatization of “give + O + to” constructions 



Smollett, Tobias. 1771 (1992). The Expedition of Humphry Clinker ed. by Lewis M. Knapp, rev. by Paul-Gabriel Bouché. Oxford: Oxford University Press. Swift, Jonathan. 1738 (1963). Swift’s Polite Conversation, annotated by E. Partridge. London: Andre Deutsch.

References Allen, Cynthia L. 2006. “Case Syncretism and Word Order Change”. The Handbook of The History of English ed. by Ans van Kemenade & Bettelou Los, 201–223. Malden, Mass.: Blackwell. Biber, Douglas, Edward Finegan, Dwight Atkinson, Ann Beck, Dennis Burges & Jean Burges. 1994. “The Design and Analysis of the ARCHER Corpus: A Progress Report”. Corpora across the Centuries: Proceedings of the First International Colloquium on English Diachronic Corpora Cambridge, 25–27  March 1993 ed. by Merja Kytö, Matti Rissanen & Susan Wright, 3–6. Amsterdam & Atlanta: Rodopi. Brinton, Laurel J. & Minoji Akimoto. 1999. Collocational and Idiomatic Aspects of Composite Predicates in the History of English. Amsterdam & Philadelphia: John Benjamins. Brinton, Laurel J. & Elizabeth C. Traugott. 2005. Lexicalization and Language Change. Cambridge: Cambridge University Press. Bybee, Joan & Paul Hopper. 2001. Frequency and the Emergence of Linguistic Structure. Amsterdam & Philadelphia: John Benjamins. Denison, David. 1993. English Historical Syntax: Verbal Constructions. London: Longman. Fillmore, Charles J., Paul Kay & Mary Catherine O’Connor. 1988. “Regularity and Idiomaticity in Grammatical Constructions: The Case of let alone”. Language 64.501–538. Goldberg, Adele E. 1995. Constructions: A Construction Grammar Approach to Argument Structure. Chicago & London: The University of Chicago Press. Goldberg, Adele E. 2006. Constructions at Work. Oxford: Oxford University Press. Goldberg, Adele E. & Devin Casenhiser. 2006. “English Constructions”. The Handbook of English Linguistics ed. by Bas Aarts & April McMahon, 343–355. Oxford: Blackwell. Green, Georgia M. 1974. Semantics and Syntactic Regularity. Bloomington & London: Indiana University Press. Haiman, John. 1983. “Iconic and Economic Motivation”. Language 59.781–819. Hudson, Richard 1992. “So-called ‘Double Objects’ and Grammatical Relations”. Language 68.251–276. Kellner, Leon. 1957 [1892]. Historical Outlines of English Syntax. Annotated by Kikuo Miyabe. Tokyo: Kenkyusha Ltd. Kytö, Merja. 1996. Manual to the Diachronic Part of the Helsinki Corpus of English Texts. Third Edition. Helsinki: Department of English University of Helsinki. Larson, Richard K. 1988. “On the Double Object Construction”. Linguistic Inquiry 19.335–391. Levin, Beth. 1993. English Verb Classes and Alternations: A Preliminary Investigation. Chicago & London: The University of Chicago Press. McFadden, Thomas. 2006. “The Rise of the to-Dative in Middle English”. Syntactic Effects of Morphological Change ed. by David W. Lightfoot, 107–123. Oxford: Oxford University Press Mitchell, Bruce. 1985. Old English Syntax. Vol. I. Oxford: Clarendon Press.



Minoji Akimoto Mustanoja, Tauno F. 1960. A Middle English Syntax. Part I: Parts of Speech. Helsinki: Société Néophilologique. Nunberg, Geoffrey, Ivan A. Sag & Thomas Wasow. 1994. “Idioms”. Language 70:3.491–538. Prins, Anton A. 1952. French Influence in English Phrasing. Leiden: Universitaire Pers Leiden. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech & Jan Svartvik. 1985. A Comprehensive Grammar of the English Language. London: Longman. Quirk, Randolph & Charles L. Wrenn. 1957. An Old English Grammar. Second Edition. London: Methuen. Ross, John R. 1973. “Nouniness”. Three Dimensions of Linguistic Theory ed. by Osamu Fujimura, 137–258. Tokyo: TEC. Smith, Jeremy. 1996. An Historical Study of English: Function, Form and Change. London & New York: Routledge. Tomlin, Russell S. 1986. Basic Word Order: Functional Principles. London: Croom Helm. Trousdale, Graeme. 2008. “Constructions in Grammaticalization and Lexicalization: Evidence from the History of a Composite Predicate Construction in English”. Constructional Approaches to English Grammar ed. by Graeme Trousdale & Nikolas Gisborne, 33–67. Berlin & New York: Mouton de Gruyter. Visser, Fredericus Th. 1970. An Historical Syntax of the English Language. Part I. Leiden: E.J.Brill.

The clausal complementation of good in extraposition constructions The emergence of partially filled constructions An Van linden*

K.U. Leuven (University of Leuven) This article develops a functional synchronic‑diachronic description of the clausal complement patterns found with good in extraposition constructions (ECs), and compares these to the patterns found with other deontic-evaluative adjectives, such as appropriate, important and fitting. The adjectives studied can currently take either mandative complements expressing desired action, or propositional complements describing arguable claims. Good differs from adjectives such as appropriate and important in that it currently favours propositional to‑clauses. More specifically, I will argue that it occurs in two types of partially filled constructions in the sense of Goldberg (1995) featuring such complements, viz. the locative pattern and the knowledge/acquisition of knowledge (KAK) pattern. The diachronic data will reveal that good started to prefer propositional to‑clauses only recently, amongst others through the emergence of the KAK pattern in Late Modern English and its subsequent rise in frequency. In addition, the present‑day occurrence of purely evaluative adjectives like nice and great in the locative and KAK constructions suggests that analogy with this class of adjectives may have played a role as well.

* The research reported on in this article has been made possible by research grants OT/03/20/TBA and OT/04/12 of the Research Council of the University of Leuven, as well as the Interuniversity Attraction Poles (IAP) Programme – Belgian State – Belgian Science Policy, project P6/44 Grammaticalization and (inter)subjectification. In addition, it has been supported by the Spanish Ministry of Education and Science (grant no. HUM2007–60706/ FILO) and the European Regional Development Fund. I would like to thank the audience of the Fifteenth International Conference on English Historical Linguistics held in Munich for their generous feedback on the material I presented there, which constitutes the core of this article. I am also indebted to Anatol Stefanowitsch and Tim Van de Cruys for their help in setting up the collostructional analysis reported on here. Finally, I am grateful to the two anonymous referees and the editors of this volume for their detailed and helpful comments. Of course, any errors of fact or interpretation remain my own responsibility.



An Van linden

1. Introduction In Present-Day English (PDE), the adjective good occurs in a number of extra­ posed complement patterns, as shown in (1) to (3). This article focuses on the most frequent complement types in Present-Day English, viz. that‑ and to‑clauses, as in (1) and (2). If‑clauses, as in (3), are excluded from analysis.





(1) If we see people waving the Union Jack we say, “Put your flag away, this isn’t a Morrissey gig” says Jim Bob. I think it was good that Morrissey was questioned so much about racism in the press, but there’s so many other people who get away with stuff, like countless rap bands. (CB, ukmags)1 (2) The main body of your letter should state the problem, stick to the point and avoid repetition, again include any areas of information such as model numbers, dates etc. Undoubtedly it is good to be firm, but avoid rudeness or abuse as it will not help your complaint. (CB, ukmags) (3) He commented: “The league is certainly more interesting than when I came here in January. Then, the gap between Rangers and the rest was from here to Dover. The young lads though are now playing with so much freedom. It would now be good if John Collins stayed too. He is a good player and you know what you’ve got with him as opposed to someone that might be brought in.” (CB, today)

More particularly, this article aims at a functional description of the synchronic distribution of complements, and it also studies the diachronic developments by which the present system was fashioned. Therefore, in addition to the formal distinction between that‑ and to‑clauses, it also proposes a semantic distinction between mandative and propositional complements, which cross-cuts the formal distinction. Mandative complements as in (2) refer to potential actions that are desired to be actualized, whereas propositional complements as in (1) refer to propositional contents that are taken to be true. I will show that the Present-Day English distribution of clausal complement patterns with good differs from that with other deontic‑evaluative adjectives, such as proper, fitting or appropriate in that it currently prefers propositional to‑clauses. In fact, it will be argued that good occurs in two types of partially filled constructions in the sense of Goldberg (1995) featuring such complements, viz. the locative pattern and the knowledge/ acquisition of knowledge (KAK) pattern, illustrated in (4) and (5) respectively. 1. The synchronic data were extracted from the COBUILD corpus via remote log-in and are reproduced (in each case marked with CB) with the kind permission of HarperCollins Publishers. They are taken from the following subcorpora: ukephem, ukbooks, ukmags, ukspok, bbc, times, today, and sunnow (for more information, see http://www.collins.co.uk/cobuild/).

Good in extraposition constructions 





(4) “How are you?” I ask her. “It’s good to be here on this Earth,” she replies in low, rich, fruity tones. (CB, today) (5) We’ve never been shy to bring in young players in the past and it’s good to see England recognising their young talent now. (CB, ukmags)

The diachronic data will show that good started to favour propositional to‑clauses only recently, as up to Early Modern English its distribution of complement patterns did not differ so much from the ones observed with the other deontic‑evaluative adjectives studied. It will become clear that it is especially the emergence of the KAK pattern in Late Modern English and its subsequent increase in frequency that accounts for the synchronic high frequency of propositional to‑clauses with good. In addition, Present-Day English complement constructions with purely evaluative adjectives, such as great and nice, suggest that good has come to analogize with this semantically distinct class of adjectives to a considerable extent. The structure of this article is as follows. In Section 2, I will briefly introduce the data and methods used in this study. In Section  3, I will present the synchronic distribution of formal and semantic types of complements with good in extraposition constructions, and I will compare this to the distribution with a reference set of ten other deontic‑evaluative adjectives. Special attention will go to the locative and the KAK pattern. In Section 4, I will investigate the diachronic distribution of formal and semantic types of complements, and I will trace the origin and development of the two propositional patterns. In Section  5, I will expand on the two patterns as they are also found with adjectives characterized by purely evaluative meaning, which may hint at analogy as an additional explanation for the current distribution of complement patterns with good. In Section 6, finally, I will summarize the main findings and formulate some questions for further research. 2. Data and methods Presenting a synchronic-diachronic account of the complement patterns found with the adjective good, the following sections are based on detailed corpus study.2 Table 1 shows the synchronic and diachronic corpora used and the number of tokens retrieved for each period.3 2. This section is based on Van linden (2009: 62–72, and forthc.). 3. The Late Modern English data of good are limited to the adjective immediately followed by that, to and for, as the total number of tokens would otherwise have become unmanageable. For the Present-Day English data, I also used a special design of query to avoid as much noise as possible, including anticipatory it.



An Van linden

Table 1.  The corpora used for each subperiod with their number of words and attestations of good Subperiod of English Old English (OE) Middle English (ME) Early Modern English (EModE) Late Modern English (LModE) Present-Day English (PDE)

Time span

750–1150

1150–1500

1500–1710

1710–1920 roughly 1990–1995

Corpus

York-Toronto-Helsinki Parsed Corpus of Old English Prose (YCOE) Penn-Helsinki Parsed Corpus of Middle English, Second Edition (PPCME) Penn-Helsinki Parsed Corpus of Early Modern English (PPCEME) Corpus of Late Modern English texts (Extended version) (CLMETEV) Collins COBUILD Corpus (CB) (only British subcorpora)

Number of words (million)

Attestations of good

  1.45

1,733

  1.16

2,525

  1.79

2,438

15.01

685

42.10

1,241

In addition, this article also compares the clausal complement patterns with good to those of other adjectives that can take mandative as well as propositional complements. To find such adjectives expressing degrees of goodness or appropriateness, I used Roget’s Thesaurus (Dutch & Roget 1970) along with the online Oxford English Dictionary (OED). The adjectives in the Present‑Day English dataset are given in the bottom row of Table 2. This set served as a starting point for the diachronic onomasiological search for lexical items. I looked for Old and Middle English counterparts of the Present‑Day English adjectives in the online Thesaurus of Old English and the Middle English Dictionary. The adjectives thus identified were also searched for in the five corpora listed in Table 1, taking into account spelling variants. The numbers of tokens between brackets in Table 2 give the overall occurrence of the adjectives, but not necessarily in the complement constructions studied here. This study not only relies on qualitative and quantitative analyses of the diachronic and synchronic corpus data, but also uses a collostructional type of ana­ lysis, viz. a multiple distinctive collexeme analysis (Gries & Stefanowitsch 2004). Such an analysis starts with a particular construction, like the extraposition construction with adjectival matrices studied here, and “investigates which lexemes are strongly attracted or repelled by a particular slot in the construction (i.e. occur more frequently or less frequently than expected)”, like the to‑infinitive slot of the extraposition construction with the several adjectives (Stefanowitsch & Gries

Good in extraposition constructions 



Table 2.  The deontic-evaluative adjectives under investigation and their number of tokens Period

Adjectives

OE (487)

andfenge (23), arlic (5), (ge)beorh (1), (ge)beorhlic (6), bryce (3), (ge)cop (1), (ge)coplic (2), (ge)cweme (61), (ge)cynde (28), (ge)cyndelic (37), cynn (7), (ge)dafen (2), (ge)dafenlic (33), (ge)defe (4), (ge)defenlic (1), fremgendlic (3), fremful (10), fremfullic (2), geornlic (5), (ge)limplic (17), (ge)mæte (4), medeme (15), (ge)met (4), (ge)metlic (9), nytlic (7), nytt (28), nyttol (1), nytweorðe (33), nytweorðlic (2), (ge)radlic (3), rædlic (1), rihtlic (53), (ge)risenlic (14), (ge)risne (14), (ge)screpe (4), (ge)tæse (1), til (4), þæslic (14), (ge)þungen (25) able (33), aise (3), bicumelich (28), comely (3), commendable (2), competent (3), convenient (8), covenable (30), desiderable (5), desirable (1), expedient (5), fremful (6), goodly (29), helply (2), just (30), kendeli (37), lele (2), limplich (1), medeme (3), (i)mete (5), profitable (42), proper (4), (i)queme (62), rightful (133), semeli (18), servisable (2), skilful (11), vertuous (34) advantageable (1), appropriate (8), commendable (13), commodious (15), competent (14), convenient (192), covenable (2), desirable (13), expedient (27), fit (288), fitting (11), important (9), just (186), meet (120), pertinent (3), profitable (61), proper (137), rightful (4), servisable (9), shapely (1), skilful (32), suitable (27), useful (38), virtuous (107) appropriate (189), convenient (420), desirable (415), expedient (93), fit (951), fitting (81), important (1,784), meet (51), profitable (172), proper (2,361), suitable (391) appropriate (323), convenient (162), desirable (84), expedient (13), fit (306), fitting (78), important (2,598), profitable (40), proper (150), suitable (155)

ME (542)

EModE (1318)

LModE (6908) PDE (3909)

2003: 214). To calculate the association strength between a particular to-infinitive (I) and an adjective (A), relative to the other to‑infinitives and adjectives included in the analysis, we need four frequencies: i. ii. iii. iv.

the frequency of I in extraposition constructions with A, the frequency of I in extraposition constructions with adjectives other than A, the frequency of A with to‑infinitives other than I, and the frequency of to‑infinitives other than I with all adjectives other than A

(cf. Stefanowitsch & Gries 2003: 218). On the basis of these frequencies, a collexeme analysis computes a vast amount of probability tests (viz. Fisher’s exact tests), which for each adjective results in specific p‑values indicating the collostruction strength with each to‑infinitive. The present analysis is based on exhaustive extractions of the extraposed to‑clauses with good and ten similar deontic‑evaluative adjectives (as listed in Table 2) in the COBUILD corpus (see Table 3). These data will be used to adduce evidence for the frequency and entrenchment of the two constructions found with good in Present-Day English (cf. Goldberg 1999).

 An Van linden

Table 3.  The deontic-evaluative adjectives included in the collexeme analysis with their number of to-infinitives in COBUILD appropriate convenient

88 32

expedient fit

 8 49

good important

278 969

desirable

23

fitting

 6

profitable

   7

proper suitable

18  3

3. Towards a functional description of the Present-Day English data In this section, I will propose to distinguish between two semantic types of complement occurring with good (and other deontic‑evaluative adjectives), viz. mandative and propositional ones, and I will show that this distinction cross-cuts the formal distinction between that‑ and to‑clauses (Section 3.1). In addition, I will present the quantitative instantiation of these types of complements in Present-Day English (Section 3.2). Finally, I will concentrate on two recurrent patterns with propositional to‑complements that are typical of good in Present-Day English, viz. the locative pattern and the knowledge/acquisition of knowledge (KAK) pattern (Section 3.3). 3.1

Mandative versus propositional complements

From a functional perspective, that‑ and to‑clauses complementing matrices with good can be of two types.4 In a first type, they express desired action, as in (6) and (7) (cf. Wierzbicka 1988: 139 on verbal complementation; Biber et al. 1999: 673–674).



(6) I know a number of reputable journalists of various nationalities who have done as I did. In part, the motivation is professional: new sources, exchanges of information. In some cases, too, patriotism plays a part: it is obviously good for one’s country that its foreign policy should be based on information known to be sound. (CB, ukbooks) (7) For example, you may fear that there are going to be some compulsory redundancies at your place of employment. You obviously hope that you are not going to be one of them, but it is good to take precautions just in case you are. (CB, ukbooks)

In a second type, they contain a proposition that is presented as true, as in (8) and (9). Following Huddleston & Pullum (2002: 996) and Halliday (1994: 70), I refer

4. This section is based on Van linden & Davidse (2009: 177–187) and Van linden (2009: 208– 219, and forthc. ).

Good in extraposition constructions 



to these semantic types of complement as “mandative” and “propositional” ones respectively (cf. Van linden & Davidse 2009).



(8) WHILE it is good that Will Carling has been reinstated as captain of the England rugby team, let us not blame RFU president Dennis Easby for the action he took. (CB, today) (9) Well I think we’ll have a word with company name’s about that or you know I I don’t honestly think I’ve heard them s mention anybody else’s name really in in great seriousness. But er it is good to be here and it is good to see As I say I’m sort of sitting very much and looking on the outside and I’m seeing some a lot of different sort of points of view coming out and a lot of different ideas (CB, ukspok)

Crucially, mandative and propositional complements differ in terms of the factuality status of the State of Affairs (SoA) referred to in the complement. Mandative complements invariably involve potential or virtual SoAs, which have not yet been actualized at the moment of assessment. In (6), for instance, the context does not give us any clues as to whether it is the case that the foreign policy of one’s country is based on information known to be sound. Rather, it is assessed as desirable that in general this should be the case. Therefore, just like that‑clauses with a subjunctive or deontic modal finite form (cf. should in example (6)) (Huddleston & Pullum 2002: 996), to‑infinitives as in (7) make excellent coding forms of mandative complements, since they are non‑finite and non‑tensed forms (cf. Bolinger 1967: 351–352). In the case of indicative that‑clauses, the potentiality and desirability of the situation have to be inferred from the context. Propositional complements, by contrast, involve situations that are presented as presupposed true, as “fact”. What is essential to a propositional complement is that it is a circumscribed claim that can be agreed with or not. For this, it needs to have a reference point in the deictic centre shared by speaker and hearer (Halliday 1994: 75; Langacker 1991: 195). In that‑clauses, it is finite tensed or epistemic modal VPs that give the proposition such a reference point, such as the present perfect form has been reinstated in (8). In the case of infinitival complements, the temporal anchoring of the proposition is brought about indirectly. In (9), the locative adverb here indirectly indicates that the actualization of the situation in the to‑complement is simultaneous with the moment of speaking, as it deictically locates the proposition in the here‑and‑now of the speech event. In addition, propositional complements need to have a subject in terms of which the truth of the proposition is asserted or, in this case, presupposed (Halliday 1994: 76–77), such as Will Carling in (8). In (9), the context shows that the subject of the propositional to‑clause is coreferential with the speaker. Although to‑infinitives clearly are less straightforward coding means of propositional complements,

 An Van linden

they have nevertheless become most frequent with good in Present‑Day English (see Section 3.2). The difference in factuality status of the SoA in the complement implies a different interpretation of the adjectival matrix. In fact, only potential SoAs, as in (6) and (7), can be regarded as morally desirable in the true sense of the word, as these SoAs have not yet been actualized. Deontic constructions with a potential SoA in their complement thus have a volitional flavour: the assessor wants the SoA to be actualized (on the basis of moral arguments) (cf.  Kiefer 1997:  242; McGregor 1997: 222; Verstraete 2005: 1405–1406). In (6), for example, it is the speaker’s moral conviction that the information journalists deliver to their country’s politicians should be correct; (s)he wants journalists to come up with sound information only. The inherent potentiality of mandative complements thus forces a deontic/ volitional interpretation onto the adjectival matrix. By contrast, complement constructions with presupposed SoAs lack this volitional flavour. In fact, it makes little sense to desire the actualization of a SoA that has already been actualized (temporal relation of anteriority, as in (8)), is being actualized (temporal relation of simultaneity, as in (9)), or is bound to be actualized (temporal relation of posteriority, as in (10) below) (cf. McGregor 1997: 221). Therefore, speakers cannot assess such SoAs as desirable, but they can only evaluate them as being or having been good. Again, the factuality status of the SoA under assessment forces a particular interpretation onto the adjectival matrix, viz. an evaluative meaning. (10) It is going to be fascinating next season with the two big guns, Arsenal and United, head‑to‑head at the top of the Premiership and in the European Cup. It can only be good for English football that so much quality will be on view in the Champions League, which rival managers Arsene Wenger and Alex Ferguson will be desperate to win. (CB, sunnow) Finally, the examples above also show that the semantic distinction between mandative and propositional complements does not systematically correlate with the formal distinction between that‑ and to‑clauses on a one‑to‑one basis. The two semantic types can be coded by the two formal types of complement. In the following section, I will point out which combinations of meaning and form are most frequent in the Present-Day English data. 3.2

The Present-Day English distribution of types of complements

The Present‑Day English data show that the synchronic distribution of clausal complements with good differs from that with the ten other deontic‑evaluative adjectives in a number of ways. Table 4 details the overall frequencies of the semantic subtypes with good, cross‑classified with their formal codings.

Good in extraposition constructions 



Table 4.  The overall distribution of propositional and mandative complements with good in Present-Day English COBUILD good

prop prop/mand mand

n

N

%

that to

  55 154

0.13 0.37

that

   2

0.0048

0.6

to

   3

0.0071

0.9

that

   4

0.010

1.2

to

total

% of semantic type

% form per semantic type

61.7

26.3 73.7

16.2 45.4

121

0.29

35.7

339

0.81

100.0

1.5 36.9

40.0 60.0 3.2 96.8

100.0

–

Table 5.  The overall distribution of propositional and mandative complements with the ten deontic-evaluative adjectives (cf. Table 2) in Present-Day English COBUILD n prop prop/mand mand total

that to that to that to

  103     5    17     5   452 1193 1775

Ten deontic-evaluative adjectives N % % of semantic type 0.24   0.012   0.040   0.012 1.07 2.83 4.22

  5.8   0.3   1.0   0.3   25.5   67.2 100.0

  6.1   1.2   92.7 100.0

% form per semantic type 95.4   4.6 77.3 22.7 27.5 72.5 –

Table 5 does so for the other deontic‑evaluative adjectives. In Tables 4, 5, 6, 11 and 12, “n” refers to absolute frequencies, “N” to normalized frequencies per 100,000 words,5 and “%” to relative shares. The most striking difference between good and the other deontic‑evaluative adjectives concerns the semantic type of complement they prefer. Good combines most frequently with propositional complements (61.7  percent) (cf.  Table  4). A Fisher’s exact test (cf.  Pedersen 1996) shows that this preference is statistically significant, with Fisher’s exact p = 0.0005131. The other adjectives, by contrast, 5. The normalized frequencies (per 100,000 words) in Tables 4, 5, 11 and 12 have been rounded up to two decimal places, or, in the case of figures with larger decimals, to at least two significant digits.

 An Van linden

occur in an overwhelming majority of cases with mandative complements (92.7 percent) (cf. Table 5). This preference is even more statistically significant, with Fisher’s exact p < 2.2E‑16. In addition, Fisher’s exact tests run on the distribution of semantic types of complements across Tables 4 and 5 confirm this difference in preferences as highly statistically significant, with Fisher’s exact p < 2.2E‑16 for both propositional and mandative types.6 Within the type of propositional complements, good prefers to‑infinitives (73.7 percent) (again a significant preference, with Fisher’s exact p = 1.220E‑08), while the other adjectives clearly prefer that‑clauses (95.4  percent) (a highly significant preference, with Fisher’s exact p < 2.2E‑16). The following section, therefore, focuses on the propositional to‑clauses found with good. Two patterns with propositional to‑clauses

3.3

Closer study of the extraposition constructions with good taking propositional to‑clauses reveals that a distinction can be made between two major patterns, viz. the locative pattern and the knowledge/acquisition of knowledge (KAK) pattern.7 The frequencies of these patterns are shown in Table 6. Table 6.  The types of propositional to-clauses with good in COBUILD Type of propositional pattern

n

%

locative pattern KAK pattern

  48   77

31.2 50.0

  (i)  KAK with secondary complement

  60

39.0

  (ii)  KAK without secondary complement

  17

11.0

other

  29

18.8

total

154

100.0

3.3.1 The locative pattern The locative pattern, which accounts for 31.2 percent of the propositional to‑clauses found with good (cf. Table 6), is a specific subtype of evaluative construction in which speakers assess their or someone else’s spatio‑temporal location as good. Expressions of spatio‑temporal location are taken to include not only true locative phrases (‘to be at a particular place’), as in (11), 6. The result for the ambiguous cases, however, is not statistically significant (Fisher’s exact p = 0.7906). 7.

This section is based on Van linden (2009: 254–263, and forthc.).



Good in extraposition constructions 

(11) Deirdre’s luck changed thanks to The Sun’s spectacular crusade, which captured the imagination of the nation. [...] Deirdre wrote a touching letter of gratitude to campaigning Sun readers while she was banged up in prison. And last night she gleefully clutched a special Weatherfield edition of The Sun reporting her release. She said: “It’s just so good to be home. I owe a big thankyou to The Sun’s brilliant campaign.” (CB, sunnow) but also associative expressions (‘to be with someone’), as in (12), (12) But O’Brien survived it. Some ‘friends’ couldn’t believe I had a by‑pass. They said the way I wrote about their team suggested I couldn’t have a heart. But I’ll let you in on a secret. It’s good to be back amongst my hurling friends again. I’m looking forward to the Championship. (CB, sunnow) and perception expressions (‘to see someone’), as in (13) (rather than to see a particular SoA, see Section 3.3.2). (13) “Excuse me, could you tell when the next 406 bus leaves for Santa Ana?” a voice said behind him. The phrase. His contact. “I’m afraid the 406 doesn’t go to” Marlette stopped abruptly when Hector Amaya stepped into view. “It’s good to see you again, Marlette,” Amaya said with an icy smile. Marlette’s mind was racing. Was Amaya his contact? Or had he been set up? (CB, ukbooks) What is characteristic of this pattern and common to the three examples above is that the evaluative assessment is simultaneous with the realization of the SoA referred to in the propositional to‑clause. This temporal relation of simultaneity is implied by the locative meaning, which is fairly straightforward in expressions such as (11) and (12). For perception expressions as in (13), it has been argued that they have a locative component to their meaning as well: there must be some association in terms of spatio-temporal location between the perceiver and the perceived entity (cf. Wierzbicka 1980: 99–114), as in an act of perception, “a stimulus of some kind, e.g. visual, auditory, or tactile, comes in contact with a sense organ of the perceiver” (Foley & Van Valin 1984: 48). In addition to this relation of simultaneity, the locative meaning also implies that the understood subject of the to‑infinitive has specific reference; typically it is coreferential with the speaker (cf. (11) to (13)). The finding that this locative pattern is typical of good (i.e. not attested with other deontic-evaluative adjectives) is reflected by the results from the multiple distinctive collexeme analysis. Table  7 shows the ten collexemes that are most strongly attracted to the to‑infinitive slot of the extraposed to‑infinitive construction with good. Table  8 gives the results of the same analysis, with the lexical

 An Van linden

Table 7.  The collexemes most strongly attracted to the to-infinitive slot of the extraposed to-infinitive construction with good Collexeme see be_locative talk know hear go have meet find be_noun

Obs. Freq. in Obs. Freq. in A B 65 21 14 29  6  7  6  4  4  6

11  1  1 33  1  3  2  0  1  6

Distinctive for:

Fisher Yates p-value

Bonferroni correction

A A A A A A A A A A

1.57E-33 2.96E-13 9.07E-09 1.55E-05 7.49E-04 1.85E-03 2.42E-03 2.63E-03 1.08E-02 3.47E-02

1.73E-32 3.25E-12 9.98E-08 1.70E-04 8.23E-03 2.04E-02 2.67E-02 2.89E-02 1.18E-01 3.81E-01

Table 8.  The process types attracted to the to-infinitive slot of the extraposed to-infinitive construction with good Collexeme: process types perception location intensive affection

Obs. Freq. in Obs. Freq. in A B 73 21 14  4

17 12 56 15

Distinc-tive for:

Fisher Yates p-value

Bonferroni correction

A A A A

1.12E-40 1.15E-08 4.43E-01 4.91E-01

1.23E-39 1.26E-07 4.88E+00 5.40E+00

items collapsed into process types (cf. Halliday 1994: Chapter 5). Crucially, the smaller the p‑value, the stronger the collostruction strength. It can be seen in Table 7 that the verbs attracted most strongly to the construction studied here (viz. A in the Tables) are see (p = 1.57E‑33) and be followed by a locative or associative expression (p = 2.96E‑13). The list also includes the verb meet, which is often found in the locative pattern as well (p = 2.63E‑03). The table thus shows that the frequencies of see, be‑locative and meet are significantly higher than what would be expected on a chance level (with α = 0.05 as the stan­ dard level of significance, cf. Stefanowitsch and Gries 2003: 239, note 6). Table 7 also demonstrates that these results remain statistically significant after the Bonferroni correction.8 Table 8 shows similar results for the process types. In fact, the 8. The Bonferroni correction is a “post hoc comparison” or adjustment that is often performed in multiple testing applied to the same dataset (cf. Rietveld & Van Hout 2005: 65), such



Good in extraposition constructions 

frequencies of perception and location are the only ones that are significant, even at corrected level. Thus, compared to the other ten adjectives included in the multiple distinctive collexeme analysis, good stands out as preferring perception and locative processes in the extraposed to‑infinitive construction. However, it should be noted that not all instances of the verb see (and the perception type) are examples of the locative use discussed here. It will become clear in the next section that see is also used in the knowledge/acquisition of knowledge pattern. 3.3.2 The knowledge/acquisition of knowledge pattern The second pattern of propositional to‑clauses found with good is concerned with the positive evaluation of knowing or getting to know a particular propositional content (knowledge or acquisition of knowledge, henceforth KAK, cf.  Noonan 2007: 129–130). This pattern accounts for 50 percent of the propositional to‑clauses observed with good (cf. Table 6) and can be thought of as a “construction” in the Construction Grammar sense, specifically in the sense of Goldberg (1995, 1996): A construction is [...] a pairing of form with meaning/use such that some aspect of the form or some aspect of the meaning/use is not strictly predictable from the component parts or from other constructions already established to exist in the language. (Goldberg 1996: 68)

The KAK pattern is very frequent with good, and it is also found with important, albeit only once. Consider the following examples: (14) He said: “[...] It is nice to get something like this after people have judged your performances. I was runner‑up in Barnsley’s Player of the Season to Ashley Ward. This is a nice consolation. It is good to know people have thought you played well.” (CB, sunnow) (15) The young pastors, now elderly men, had discharged the responsibility vested in them beyond my father’s hopes and prayers. Christian faith and worship had been resurrected and emerged stronger than ever. It is good to read that this is being sustained, and we could imagine ourselves sharing this wonderful Eastertide in Wenzhou. (CB, times) (16) He revealed: “It started slowly at first, but then worked up until I was flat out. I felt much leaner and sharper. It was good to hear people telling me I was looking much better – particularly the manager.” (CB, sunnow) as, for example, the eleven collostructional analyses here. It is used because uncorrected results of multiple testing may falsely give the appearance of significance, as 1 out of 20 probability tests will appear to be significant at the α = 0.05 level purely due to chance (Stefanowitsch, personal communication). I thus multiplied the Fisher exact p-values by the number of tests run, viz. 11, to arrive at the corrected p-values.

 An Van linden

(17) It is important to see UK base financial markets on a world basis following the recent spread of “global” or “round the clock” trading from foreign exchange to securities, financial futures and commodities. This has been made possible by improved satellite based communications and the deregulation of financial markets. (CB, ukephem) The examples above all involve an extraposed to‑infinitival subject consisting of a KAK predicate and a secondary propositional complement. Note that the non‑deliberate perception verbs see and hear with participial complements are included in this KAK use as well (even if they are typically regarded as expressing immediate perception; cf. Noonan 2007: 142–144), as sensory perception essentially implies acquisition of knowledge. What is crucial to the KAK pattern is that it is not so much the SoA encoded by the to‑infinitive that is evaluated as good or important. Rather, the construction as a whole expresses the speaker’s positive evaluation of the propositional content of the secondary complement.9 In this sense, the meaning of the KAK pattern cannot be compositionally derived from its constituent parts, and thus forms a construction in the Construction Grammar sense. However, it should be noted that not all examples with to‑clauses containing KAK predicates are propositional in nature. In fact, the following examples feature mandative to‑clauses. What is characteristic of these examples is that the understood infinitival subjects have arbitrary reference, whereas those in the KAK pattern have specific reference, viz. they are coreferential with the speaker (cf. (14) to (17)), just like those in the locative pattern (see Section 3.3.1 above). (18) “I’ve this urge to see something of the world.” “Some of the other cities, you mean.” “Right. Some other cities.” “Well, why not?” asked Soniff, expansively. “The Purples have affiliates in a lot of the cities, and it’s always good to see the way things get done other places.” (CB, ukbooks)

9. A similar pattern has been noted by Mair (1990:  25) with matrix predicates assessing truth and probability, such as true, obvious and probable. These matrices are typically complemented by propositional that-clauses, but they also occur in constructions with extraposed to-clauses containing utterance or propositional attitude predicates (e.g. say or believe) and secondary that-complements, as in “I often think”, Treece said rather smugly, “that it’s equally true to say that genius is an infinite capacity for faking pains.” (W.16.2.107–1) (Mair 1990: 25 (23)). In this example, it is not the act of saying that is equally true, but rather the propositional content of the secondary that-clause (see also Herriman 2000: 591). This is why Mair calls this construction “slightly incongruous” (1990: 25). In both the KAK and in Mair’s pattern, it is this incongruity that makes the meaning of the whole construction unpredictable from its constituent parts.



Good in extraposition constructions 

(19) What kind of input have the police or other departments had? <M02> Erm again it’s it’s it’s looking at the effectiveness of the the solutions that we’re <M01> Mhm. <M02> we’re putting forward. Erm other people’s experience may be that those things have been tried elsewhere and do or don’t work and it’s important to know that. (CB, ukspok) Apart from specific reference of the infinitival subjects, the KAK pattern shares some further properties with the locative pattern. Importantly, the KAK construction also involves simultaneity of evaluative assessment and realization of the SoA referred to in the to‑clause. This property might be related to the presence of a locative component in the KAK pattern as well. In functional accounts, for instance, it has been argued that the semantics of KAK predicates includes a locative element, albeit indirectly: in the first place they have a possessive component to their meaning, which in turn implies a locative aspect (Wierzbicka 1980: 105– 114; Foley & Van Valin 1984: 49). However, in cases like (20), the realization of the KAK to‑clause must be interpreted as being both anterior to and simultaneous with the evaluative assessment in the here‑and‑now of the speech event. (20) Tracker funds are the cheapest and most straightforward of all equity investments. They simply invest in a basket of shares which replicate the performance of the chosen index. They gained prominence last year when Virgin rocked the market with the launch of its low-cost index-tracking Pep, which is sold over the phone in the same way as Direct Line already sells insurance. Virgin said: “It is good to see another company with a good reputation coming into the market and showing that simple, low-cost products are the way the industry must go. As well as low charges, tracker funds also have investment performance on their side.” (CB, times) In (20), the “seeing-event” evaluated as good arguably consists of a range of successive “seeing-events” in a time span that started before and continues into the moment of evaluation, or, in other words, the propositional content evaluated as good is not an event that can be seen at a single glance. The finding that the events referred to (viz. coming into the market with low‑cost products, and gaining profits from this business) have taken place in a specific time span rather than at a specific moment suggests that in the KAK construction the meaning of the predicate see may differ from that in the locative pattern (see also De Smet & Cuyckens 2007 on other patterns of combined complementation with see). In any case, the participial form of the secondary complements (another company with a good reputation coming and showing) suggests that see is still used in its immediate perception sense rather than in its (strictly KAK) ‘realize, understand’ sense (cf. Noonan 2007: 72–73, 129–130, 142–143).

 An Van linden

Finally, the observation that the KAK pattern is more typical of good than of important or the other deontic-evaluative adjectives of the reference set is confirmed by the results of a multiple distinctive collexeme analysis. Table 9 not only shows the ten collexemes most strongly attracted to the to‑infinitive slot of the extraposed to‑infinitive construction with good (cf.  Table  7), but also the ten collexemes most strongly repelled by it. Table 10 does the same with the process types (cf. Table 8). However, a distinction is made here between perception predicates complemented by a secondary proposition (perception_comp) and those without clausal complement (perception (proper)), so as to give some idea about the ratio of locative versus KAK uses (see Section 3.3.1 above). Table 9.  The collexemes most strongly attracted to and repelled by the to-infinitive slot of the extraposed to-infinitive construction with good Distinctive for A Distinctive for B Collexeme Obs. Obs. Fisher Bonferroni Collexeme Obs. Obs. Fisher Bonferroni Freq Freq Yates correction Freq Freq Yates correction in A in B p-value in A in B p-value see be_locative talk know hear go have meet find be_noun

65 21 14 29  6  7  6  4  4  6

11  1  1 33  1  3  2  0  1  6

1.57E-33 2.96E-13 9.07E-09 1.55E-05 7.49E-04 1.85E-03 2.42E-03 2.63E-03 1.08E-02 3.47E-02

1.73E-32 3.25E-12 9.98E-08 1.70E-04 8.23E-03 2.04E-02 2.67E-02 2.89E-02 1.18E-01 3.81E-01

remember understand ensure realize recognize make_sure note say look keep_cont

1 0 0 0 0 0 1 0 0 1

46 22 19 18 15 15 21 14 13 18

6.47E-05 3.25E-03 7.15E-03 9.29E-03 2.04E-02 2.04E-02 2.47E-02 2.65E-02 3.43E-02 4.79E-02

7.12E-04 3.57E-02 7.86E-02 1.02E-01 2.24E-01 2.24E-01 2.72E-01 2.91E-01 3.78E-01 5.26E-01

It is clear from Table 9 that the KAK predicates rank high in the list of attracted collexemes (see, know, hear). It is also telling that the cognition/knowledge verb know is a strongly attracted item, whereas the cognition verbs remember, understand, realize and recognize are strongly repelled items (however, only the first two at corrected level as well). The findings for these four verbs explain why the category of cognition verbs – despite the result of know – ends up as a strongly repelled process type in Table 10. This table also shows that both the category of perception proper and that of perception with clausal complement are strongly attracted; the first one two orders of magnitude more so than the second one. However, the first category still includes examples of the KAK pattern, for instance those which express the perceived SoA as an action nominal rather than as a clausal complement, as in (21) below.

Good in extraposition constructions 



Table 10.  The process types attracted to and repelled by the to-infinitive slot of the extraposed to-infinitive construction with good Collexeme

perception perception_ comp location intensive affection

Distinctive for A Distinctive for B Obs. Obs. Fisher Bonfer- Collexeme Obs. Obs. Fisher Bonferroni Freq Freq. Yates roni Freq Freq Yates correction in A in B p-value correction in A in B p-value 43 30

14 2.31E−21 2.54E−20 cognition   3 1.35E−19 1.48E−18 utterance

45  4

437 1.07E-11 1.18E−10   62 2.57E-03 2.82E−02

21 14  4

12 1.15E−08 1.26E−07 material 102 56 4.43E−01 4.88E+00 possession   15 15 4.91E−01 5.40E+00 behavioural    0

509 4.92E-02 5.41E−01 94 1.00E-01 1.10E+00   1 8.12E-01 8.94E+00

It + COP + is was

ADJ good important

+ to +

KAK predicate know see hear read

+

object of KAK predicate that-clause participial clause action nominal NP referring to proposition

Figure 1.  The KAK pattern as a partially filled construction

(21) Mm. And it was quite good when we went to get my sister and my mum ‘cos my nephew’s like growing up really quick and it They hadn’t seen him for a while and it was really good to see you know their reaction to when they see the baby all grown up and doing all these different things.  (CB, ukspok) In general, the collostructional analysis thus provides further evidence for considering the KAK pattern as a partially filled construction with a restricted number of lexical elements occurring in two of the six slots (in boxes), as presented in Figure 1 above. 4. Diachronic developments This section concentrates on the diachronic development of the distribution of complements with good, and compares it to the complement patterns observed

 An Van linden

for the deontic-evaluative adjectives of the reference set. More specifically, it traces the emergence of the two propositional patterns discussed above, viz. the locative and the KAK pattern. The distribution of formal and semantic types of complements with good and the other deontic-evaluative adjectives has witnessed a number of changes across the various historical periods. We can note that some changes affected both good (see Table 11) and the deontic-evaluative adjectives (see Table 12), whereas other developments seem to be restricted to good. The main change common to good and the other deontic-evaluative adjectives relates to mandative complementation. Tables 11 and 12 show that with all adjectives studied this type of complementation shifted from a predominance of that‑clauses in Old English to one of to‑infinitives in Middle English, a development parallel to that of complements of verbs with a volitional element described by Table 11.  The diachronic distribution of propositional and mandative complements of good good

Fr

that

prop

to

total

that

mand

to

total

750– 950

950– 1150

1150– 1350

1350– 1500

1500– 1710

1710– 1920

1990– 1995

n

1

4

0

1

0

2

55

N

0.29

0.36

0.00

0.12

0.00

0.013

0.13

%

100.0

66.7

–

50.0

–

9.1

26.3

n

0

2

1

1

0

20

154

N

0.00

0.18

0.28

0.12

0.00

0.13

0.37

%

–

33.3

100.0

50.0

–

90.9

73.7

n

1

6

1

0

0

22

209

N

0.29

0.54

0.28

0

0.00

0.15

0.50

%

100.0

100.0

100.0

–

–

100.0

100.0

n

6

15

7

14

16

15

6

N

1.74

1.36

1.99

1.74

0.89

0.010

0.014

%

66.7

93.8

41.2

37.8

21.3

30.6

4.6

n

3

1

10

23

59

34

124

N

0.87

0.090

2.84

2.86

3.29

0.23

0.29

%

33.3

6.3

58.8

62.2

78.7

69.4

95.4

n

9

16

17

37

75

49

130

N

2.61

1.45

4.83

4.60

4.18

0.33

0.31

%

100.0

100.0

100.0

100.0

100.0

100.0

100.0

Good in extraposition constructions 



Table 12.  The diachronic distribution of propositional and mandative complements of the adjectives of the reference set reference set

that

prop

to

total

that

mand

to

total

Fr

750– 950

950– 1150

1150– 1350

1350– 1500

1500– 1710

1710– 1920

1990– 1995

n

1

5

0

1

6

8

103

N

0.29

0.45

0.00

0.12

0.33

0.053

0.24

%

100.0

83.3

-

100.0

100.0

80.0

95.4

n

0

1

0

0

0

2

5

N

0.00

0.090

0.00

0.00

0.00

0.013

0.012

%

–

16.7

–

–

–

20.0

4.6

n

1

6

0

1

6

10

108

N

0.29

0.54

0.00

0.12

0.33

0.067

0.26

%

100.0

100.0

–

100.0

100.0

100.0

100.0

n

11

32

2

10

57

193

469

N

3.19

2.90

0.57

1.24

3.18

1.29

1.11

%

78.6

100.0

100.00

47.6

39.0

21.8

28.1

n

3

0

0

11

89

691

1198

N

0.87

0.00

0.00

1.37

4.96

4.60

2.85

%

21.4

–

–

52.4

61.0

78.2

71.9

n

14

32

2

21

146

884

1667

N

4.05

2.90

0.57

2.61

8.14

5.89

3.96

%

100.0

100.0

100.00

100.0

100.0

100.0

100.0

Los (2005) (cf. Van linden 2009: 163–178, 2010 and forthc.).10 The tables also indicate that – unlike with some volitional verbs (cf. Rohdenburg 1995) – this replacement of the that‑clause by the to‑infinitive has not run its full course; the mandative that‑clause continues to be a (minor) coding option in Present‑Day English. The most important differences between the two datasets pertain to the relative frequencies of the two semantic types of complements and, within propositional complementation, to the relative frequencies of the two formal types of clauses. Firstly, if we look at the relative shares of mandative and propositional complements, Tables 11 and 12 show that good and the other deontic-evaluative 10. The to-infinitives referred to as appearing in the clausal complements of the adjectives studied here include inflected infinitives preceded by to in Old English, inflected infinitives preceded by to, te or forto in Middle English, and those in for-NP-to-infinitive constructions from Late Middle English onwards. The data also include two postverbal bare infinitives in Middle English.

 An Van linden

adjectives studied differ most markedly in Present‑Day English (see also Tables 4 and 5 in Section 3.2 above). Up to Early Modern English, by contrast, the relative shares of the two semantic types of complements are very similar; Fisher’s exact tests demonstrate that the distribution of mandative and propositional complements with good and the other deontic-evaluative adjectives does not show significant differences (with Fisher’s exact p‑values ranging from p = 0.1815 to p = 1). In Late Modern English, good combines significantly more often with propositional complements than the other deontic-evaluative adjectives do (Fisher’s exact p < 2.2E‑16), but overall it still patterns more frequently with mandative (69 percent) than with propositional complements (39 percent). By Present‑Day English, however, this distribution has almost been reversed, as good has come to favour propositional complements (62 percent) over mandative ones (38 percent), whereas the deontic-evaluative adjectives still prefer mandative clauses (94 percent) (again Fisher’s exact p < 2.2E‑16, cf. Tables 4 and 5). Secondly, if we consider the diachronic distribution of the formal types of propositional complements, we can observe another difference. More specifically, the development of the distribution of propositional that‑ and to‑clauses with good seems to mirror the development found for the mandative complements: even though the data are few, we can note a shift from a prevalence of that‑clauses to one of to‑clauses, which has been accomplished by Late Modern English (see Table 11). With the other deontic‑evaluative adjectives, by contrast, that‑clauses remain the preferred type of propositional complement throughout the various periods (ca. 93 percent on average, see Table 12). The finding that the predominance of propositional to‑clauses with good is a recent development ties in with the origin and development of the locative and KAK pattern. In fact, the KAK pattern only emerged in the Late Modern English period and rapidly increased in frequency to become the predominant one in the Present-Day English data (cf. Table 6). In the examples (22) and (23), dating from the early twentieth century, the KAK predicates are perception verbs, which are complemented by participial clauses. No such examples have been found in the Old, Middle, or Early Modern English period. (22) What a wholesome thing to have Mr. Henley, for example, at that in the place of some of the several specialists who will lecture you so admirably on the Troubadours! How good to hear Mr. Frederic Harrison (with some one to follow) adjusting all our living efforts to the scale of the divine Comte, and Mr. Walkley and Mr. Herbert Paul making it perfectly clear that a dead dog is better than a living lion, by demonstrations on the lion.  (CLMETEV 1902–03 Wells, Mankind in the making)



Good in extraposition constructions 

(23) “It’s nothing.” She went in and tore it up, and then began to write – a very short letter, whose gist was “Come and save me.” It is not good to see your wife crying when she writes – especially if you are conscious that, on the whole, your treatment of her has been reasonable and kind.  (CLMETEV 1905 Forster, Where angels fear to tread) The locative pattern, by contrast, is already attested in Old English, as illustrated in (24). A Middle English translation of the same Bible verse is given in (25). Both examples involve a clear locative phrase, viz. her/here (cf. (9)). (24) Þa cwæð Petrus to him, Drihten, god ys us her to Then said Petrus to him, Lord, good is to.us here to beonne; Gyf þu wylt uton wyrcean her be If you want.prs.ind go.prs.sbjv.pl make.inf here þreo eardungstowa three dwelling.places (YCOE 1000–1050 Mt (WSCp) 17.4) ‘Then Peter said to him: “Lord, it is good for us to be here; if you want it, let us make here three dwelling‑places.”’ (25) And þerfore seyde seynt Peter: domine bonum est and therefore said saint Peter: Lord.voc good.nom.n is nos hic esse, faciamus hic tria we.acc here be.inf, make.prs.sbjv.1pl here three.acc tabernacula, þat is to seye: Lord it is gode for vs dwelling.place.acc.pl, that is to say: Lord it is good for us to ben here, make wee here .iij. dwellyng places. to be.inf here, make.prs.sbjv we here three dwelling places  (PPCME ?a1425 (c1400) Mandev.(1) (Tit C.16) 76) ‘And therefore Saint Peter said: “Lord, it is good for us to be here, let us make three tabernacles here”, that is to say: “Lord, it is good for us to be here, let us make three dwelling‑places here.”’ In spite of its early emergence, the locative pattern never was very frequent up to Late Modern English, as can be inferred from the low absolute numbers of propositional to‑clauses with good (cf. Table 11). Presumably, its rise to about one third of the propositional to‑clauses found with good in Present‑Day English (cf. Table 6) can be explained as being promoted by the emergence and rise of the KAK pattern. In addition, as suggested in the next section, analogy with purely evaluative adjectives may have played a role.

 An Van linden

5. Locative and KAK patterns with purely evaluative adjectives Even if among the deontic‑evaluative adjectives studied here, the locative and KAK patterns are typical of good, it is certainly not the only adjective that occurs in these patterns in Present‑Day English.11 Interestingly, the adjectives in the examples below are generally held to express only evaluative meaning (cf. Noonan 2007: 127–129), that is, unlike the adjectives studied here, they do not occur with both mandative and propositional complements.12 (26) Those two putts must have flushed away some of the negative thoughts that have been bedevilling him; at last, the flashing Ballesteros smile was back. “It’s nice to be here for the weekend,” he said. One would think the sponsors were fairly happy about it as well. (CB, times) (27) Last June eight years down the line he completed an MA in International Politics and Relations at Aberdeen University. “it’s it’s great to see you here today at this ceremony. You’ve actually been nominated for a regional award so you’re a regional finalist.” (CB, ukspok) (28) I can’t get used to the idea of Marcia, spending the war – well “On the other side? And courted by German officers? Your little sister? Is that what you mean?” “I suppose that it is.” But it was both wonderful and poignant to hear that Marcia was well. (CB, ukbooks) (29) And the big story is still the situation in Iraq and particularly the situation facing the Kurds as they flee into the mountains away from Saddam Hussein’s armies. [...] And things are pretty bad really for the Kurds at the moment. And er it’s interesting to see Mrs Thatcher’s been weighing in on their behalf. (CB, ukspok) (30) “Everybody loves him round here and they don’t want me to upstage him. Let’s just say I’ll creep up as close as I can.” Tony was present to see the action and said: “It’s great to see Michael carrying on where I left off. The

11. This section is based on Van linden (2009: 265–267, and forthc.). 12. In terms of Noonan’s (2007: 120–145) semantic classification of complement-taking predicates, the adjectival matrices central to this article, like be good or be proper, belong, when combined with a mandative complement, to the type of desiderative predicates like want (2007: 132–137), or modal predicates like ought, should or must (2007: 137–139). When combined with a propositional complement, however, these adjectival matrices belong to the type of commentative predicates (Noonan 2007: 127–128), which “provide a comment on the complement proposition which takes the form of an emotional reaction or evaluation (regret, be sorry, be sad) or a judgement (be odd, be significant, be important)” (cf. Van linden & Davidse 2009). The adjectival matrices in examples (26) to (31), by contrast, act inherently as commentative predicates, and cannot be used as desiderative or modal ones (cf. Noonan 2007: 127–129).

Good in extraposition constructions 



race had a few problems but, at the same time, the racing was excellent and it’s a unique track and occasion.” (CB, sunnow) (31) Benn has always said that 1996 would be his last year, and he would like to go out with the final tear‑up with Jones. It was not surprising to see Benn wanting to retire after a hard campaign over nine years and after being comprehensively outpointed by an ageing opponent Malinga is 36 who is himself contemplating retirement. (CB, times) Examples (26) and (27) with nice and great instantiate the locative pattern. These examples both have anticipatory it and a copular finite, but the locative pattern is also often found without these elements, e.g., nice to meet you, or good to see you (not included in the analyses here). Arguably, these locative expressions have become semi-formulaic phrases typically used in face‑to‑face communication. The examples given here also suggest that the locative pattern is restricted to adjectives expressing degrees of likeability (cf. Nuyts 2006: 12: “the degree of the speaker’s (or someone else’s) liking or disliking of the state of affairs”). We might thus be led to conclude that in the locative pattern good has come to express likeability rather than moral evaluation, and hence, that the locative pattern may qualify as a partially filled construction, just like the KAK pattern, in which a specific constructional make‑up is paired with a particular meaning that is not fully predictable from the component parts. The locative construction can be visualized as in Figure 2. It + COP + is was

ADJ good great nice

+ to +

Predicate with locative component be + locative phrase be + associative phrase see + someone

Figure 2.  The locative pattern as a partially filled construction

Examples (28) to (31) above illustrate the KAK pattern with the adjectives wonderful, poignant, interesting, great and surprising. The KAK predicates include hear, know and see, and the secondary propositional complements take the form of a that‑clause in (28) and (29), and that of a participial clause in (30) and (31). It is clear from the examples that the types of evaluative meaning expressed by KAK constructions include not only likeability but also expectability and significance. This finding explains, for instance, why important is found in this pattern, but not in the locative pattern. It also suggests that in this construction good has kept its general moral undertone, more so than in the locative pattern.13 More generally, 13. However, as rightly noted by a referee, in some KAK constructions good can be assigned a ‘likeability-rather-than-moral-evaluation-reading’, like, for instance, in (16).

 An Van linden

all examples above, featuring adjectives of a semantic class distinct from the one studied here, offer further evidence for establishing the locative and KAK use found with good (and important) as true patterns or constructions, which are (getting) entrenched in the language (cf. Hopper’s “Emergent Grammar”; Hopper 1987, 1998). 6. Conclusions In this article, I have developed a functional account of the clausal complement patterns with good, and I have compared these to a reference set of ten other deontic‑evaluative adjectives such as appropriate, fitting and important. In doing so, I assumed distinctions between semantic types of complements, viz. mandative versus propositional, and formal types, viz. that‑clauses versus to‑clauses, which I found do not correlate with each other on a one‑to‑one basis. The Present‑Day English data showed that good differs from the other deontic‑evaluative adjectives in that it currently favours propositional to‑clauses, whereas those adjectives prefer mandative to‑clauses. In fact, the synchronic data revealed that good is frequently used in two patterns with propositional to‑clauses, viz. the locative and the knowledge/acquisition of knowledge (KAK) pattern. In the first one, the speaker assesses someone’s spatio‑temporal location as good, and in the second one, the speaker expresses his/her positive evaluation of a particular propositional content. On the basis of qualitative and quantitative (amongst others, collostructional) analyses of corpus data, these patterns were characterized as partially filled constructions in the sense of Goldberg (1995). The diachronic data revealed that good’s current preference for propositional to‑clauses is a rather recent phenomenon, which could be detected as incipient in the Late Modern English period, and which has established itself firmly only in Present‑Day English. In fact, up to Early Modern English good clearly patterned like the other deontic‑evaluative adjectives, favouring mandative complements which themselves underwent a shift in predominant coding form, viz. from that‑clauses to to‑clauses (cf. Van linden 2010). In Late Modern English, good still preferred mandative complements, but to a significantly lesser degree than the other deontic‑evaluative adjectives. In the same period, the KAK pattern emerged, which probably also promoted the use of the locative pattern, already attested as of Old English. It is only in Present-Day English that good combines in an overwhelming majority of cases with propositional to‑clauses, especially in the locative and KAK patterns. The occurrence of these two patterns with purely evaluative adjectives like nice, great, wonderful, poignant and interesting in Present-Day English suggests that this class of evaluative adjectives may have set the example

Good in extraposition constructions 



for good. That is, whereas good patterned much like the other deontic-evaluative adjectives studied here up to Early Modern English, it may have come to analogize with a different semantic class of adjectives. The question remains of course why good changed its course of development in the first place, and, if analogy did play a part in this shift, it still needs to be investigated when and how exactly the analogies came about. References Biber, Douglas, Stig Johansson, Geoffrey Leech, Susan Conrad & Edward Finegan, eds. 1999. Longman Grammar of Spoken and Written English. London: Longman. Bolinger, Dwight. 1967. “The Imperative in English”. To Honor Roman Jakobson. Essays on the Occasion of his Seventieth Birthday. Volume 1. (= Janua Linguarum: Series Maior XXXI), ed. by Roman Jacobson, 335–362. The Hague & Paris: Mouton. De Smet, Hendrik & Hubert Cuyckens. 2007. “Diachronic Aspects of Complementation: Constructions, Entrenchment, and the Matching Problem”. Managing Chaos: Strategies for Identifying Change in English ed. by Christopher Cain & Geoffrey Russom, 187–214. Berlin & New York: Mouton de Gruyter. Dutch, Robert A. & Peter M. Roget. 1970. Roget’s Thesaurus of English Words and Phrases. London: Longman. Foley, William A. & Robert D. Van Valin. 1984. Functional Syntax and Universal Grammar. (= Cambridge Studies in Linguistics, 38). Cambridge: Cambridge University Press. Goldberg, Adele. 1995. Constructions: A Construction Grammar Approach to Argument Structure (Cognitive Theory of Language and Culture). Chicago: University of Chicago Press. Goldberg, Adele. 1996. “Construction Grammar”. Concise Encyclopedia of Syntactic Theories ed. by Keith Brown & Jim Miller, 68–71. New York, New York: Pergamon. Goldberg, Adele. 1999. “The Emergence of the Semantics of Argument Structure Constructions”. The Emergence of Language ed. by Brian MacWhinney, 197–212. Mahwah, New Jersey: Lawrence Erlbaum. Gries, Stefan Th. & Anatol Stefanowitsch. 2004. “Extending Collostructional Analysis: A Corpus‑based Perspective on ‘Alternations’”. International Journal of Corpus Linguistics 9:1.97–129. Halliday, Michael A. K. 1994. An Introduction to Functional Grammar. 2nd edition. London: Arnold. Herriman, Jennifer. 2000. “Extraposition in English: A Study of the Interaction between the Matrix Predicate and the Type of Extraposed Clause”. English Studies 81:6.582–599. Hopper, Paul J. 1987. “Emergent Grammar”. Berkeley Linguistics Society 13.139–157. Hopper, Paul J. 1998. “Emergent Grammar”. The New Psychology of Language: Cognitive and Functional Approaches to Language Structure ed. by Michael Tomasello, 155–175. Mahwah, New Jersey: Lawrence Erlbaum. Huddleston, Rodney & Geoffrey Pullum. 2002. The Cambridge Grammar of the English Language. Cambridge: Cambridge University Press. Kiefer, Ferenc. 1997. “Presidential Adress–Modality and Pragmatics”. Folia Linguistica 31:3–4.241–253.

 An Van linden Kurath, Hans & Robert E. Lewis, eds. 1963–2001. Middle English Dictionary. Ann Arbor: University of Michigan Press. . Langacker, Ronald. 1991. Foundations of Cognitive Grammar. Vol. 2: Descriptive Application. Stanford: Stanford University Press. Los, Bettelou. 2005. The Rise of the To‑infinitive. Oxford: Oxford University Press. Mair, Christian. 1990. Infinitival Complement Clauses in English. Cambridge: Cambridge University Press. McGregor, William B. 1997. Semiotic Grammar. Oxford: Clarendon Press. Noonan, Michael. 2007. “Complementation”. Language Typology and Syntactic Description: Vol. 2: Complex Constructions. 2nd edition ed. by Timothy Shopen, 52–150. Cambridge: Cambridge University Press. Nuyts, Jan. 2006. “Modality: Overview and Linguistic Issues”. The Expression of Modality (= The Expression of Cognitive Categories, 1) ed. by William Frawley, 1–26. Berlin & New York: Mouton. Oxford English Dictionary: . Pedersen, Ted. 1996. “Fishing for exactness”. Proceedings of the South Central SAS(c) User Group 96.188–200. Rietveld, Toni & Roeland Van Hout. 2005. Statistics in Language Research: Analysis of Variance. Berlin & New York: Mouton de Gruyter. Roberts, Jane & Christian Kay, eds. 2000. Thesaurus of Old English. London: King’s College London. . Rohdenburg, Günter. 1995. “On the Replacement of Finite Complement Clauses by Infinitives in English”. English Studies 76.367–388. Stefanowitsch, Anatol & Stefan T. Gries. 2003. “Collostructions: Investigating the Interaction of Words and Constructions”. International Journal of Corpus Linguistics 8:2.209–243. Van linden, An. 2009. Dynamic, Deontic and Evaluative Adjectives and their Clausal Complement Patterns: A Synchronic‑Diachronic Account. Ph.D. dissertation, University of Leuven. Van linden, An. 2010. “The Rise of the To‑infinitive: Evidence from Adjectival Complementation”. English Language and Linguistics 14:1.19–51. Van linden, An. Forthcoming. The Development of Deontic and Evaluative Meanings in English Adjectival Constructions (Topics in English Linguistics). Berlin & New York: Mouton de Gruyter. Van linden, An & Kristin Davidse. 2009. “The Clausal Complementation of Deontic-evaluative Adjectives in Extraposition Constructions: A Synchronic‑Diachronic Approach”. Folia Linguistica 43:1.171–211. Verstraete, Jean‑Christophe. 2005. “Scalar Quantity Implicatures and the Interpretation of Modality. Problems in the Deontic Domain”. Journal of Pragmatics 37:9.1401–1418. Wierzbicka, Anna. 1980. Lingua Mentalis: The Semantics of Natural Language. Sydney: Academic Press. Wierzbicka, Anna. 1988. The Semantics of Grammar. Amsterdam & Philadelphia: John Benjamins.

part ii

Modality and (marginal) modals

The ‘fail to’ construction in Late Modern and Present-Day English Thomas Egan*

Hedmark University College This paper traces the development of the ‘ fail to’ construction over the last three hundred years. In the eighteenth century, almost 95 percent of tokens of ‘ fail to’ were negated. In corpora from the late twentieth century, on the other hand, fewer than 4 percent of all tokens of ‘ fail to’ are negated. The non‑negated ‘ fail to’ construction may encode unsuccessful effort or neglect of duty on the part of the subject, or it may encode disappointment of the speaker’s expectations. It may even encode negation pure and simple. Special attention is paid to the growth in these uses of the construction in the nineteenth century. The question of whether or not ‘ fail to’ is in the process of grammaticalizing is also addressed.

1. Introduction The ‘ fail to’ construction dates back to the fourteenth century. (1) and (2) are from Chaucer. (1) For she hath taught hym how he shal nat fayle The Fles to wynne. (The Legend of Good Women 1646–7) ‘For she has taught him how he cannot fail to win the fleece.’ (2) And preyed hem to laboure in this nede, And shapen that he faille nat to spede.(The Merchant’s Tale IV (E) 1631–2) ‘And asked them to exert themselves in this matter and fix things so he could not fail to succeed.’ * I would like to thank Anne-Line Graedler, Ursula Lenker and two anonymous referees for their helpful and constructive comments. In addition I would like to thank Geneviève Girard-Gillet for helping me figure out the tortuous history of faillir and Christine Meklenborg Salvesen for providing me with thousands of concordance lines from the Corpus de la littérature médiévale française and for checking my interpretation of some of these.

 Thomas Egan

Both of these examples presuppose that the subject of the matrix verb makes an effort to realize the situation in the complement clause. And in both the matrix verb is negated.1 Indeed, the earliest non‑negated example of the construction in the OED dates from 1810. In this respect the OED accurately reflects the situation before the nineteenth century. In the Helsinki Corpus, for example, twelve of thirteen tokens of the construction are negated. In the Corpus of Early English Correspondence Sampler all of the twenty‑four tokens are negated. Today, however, the picture with respect to polarity is very different. (3) and (4) are recent examples of the construction.2 (3) NOW WE KNOW why the government failed to stop 9/11.  (http://www.anthraxinvestigation.com/OddsEnds.html) (4) The thing the academics who push the semantic web fail to consider (most of the time) is that the Real World does not function like their Ideal World. (http://developers.slashdot.org/article.pl?sid=07/03/21/0235208&from=rss) Neither (3) nor (4) is negated. Moreover, neither token presumes that the subject of the matrix verb makes an effort to realize the situation in the complement clause. In (3), the government was totally unaware of the existence of a plot to attack New York. In (4), the academics in question have not expended any effort in considering whether the Real World functions like their Ideal World. The majority of instances of the construction in my data before the nineteenth century resemble (1) and (2) in being negated. Some two hundred years ago the negated form began to lose ground rapidly. This development will be traced in Section  3, after the various senses of the construction have been 1. On reading an earlier version of this paper, Ursula Lenker raised the question of whether or not the construction is borrowed from French in this form, with the negation already in place. Faillir is a very polysemous verb in Old French. It was already in the process of splitting into falloir, meaning ‘have to’, and faillir, which means ‘almost do’ when it occurs with an infinitive complement. Negated faillir + infinitive in Old French could be used in non-impersonal constructions in the same sense as ‘not fail to’. The Corpus de la littérature médiévale française contains several examples of this negated construction from the thirteenth and early fourteenth centuries. Among these are (i) Qui l’apele de cuer sanz fausseté, / ja ne faudra a avoir repentance (Thibaut de Champagne (written before 1253)) ‘(he) who appeals to her from the heart without duplicity will never fail to gain forgiveness’ and (ii) et vous touz, juges, ne failliez pas a ce faire. (Miracle de Saint Lorens (written before 1339)) ‘and all of you, judges, don’t fail to do this’. It would thus appear likely that the negated ‘ fail to’ construction was borrowed in that form. The question remains, of course, as to why its non-negated counterpart was not borrowed at the same time. 2. All tokens from the World Wide Web were downloaded using WebCorp on 27.05.2008. apart from (44) to (46), downloaded on 22.12.2009.

The ‘ fail to’ construction 



presented in Section 2. Section 4 takes a close look at the rise of the non-negated form in the nineteenth century and Section 5 considers the question of whether or not the construction is in the process of grammaticalizing. Section 6 contains a brief summary of the discussion. 2. The main senses of ‘fail to’ The ‘ fail to’ construction is used with the following four basic senses. The first three definitions are taken from The New Oxford Dictionary of English (Pearsall & Hanks 1988): – – – –

be unsuccessful in achieving one’s goal neglect to do something behave in a way contrary to hopes or expectations by not doing something not do something.

The senses are progressively more subjective in the sense of that term used by Traugott & Dasher, according to whom subjectivity involves the grounding of a predication in the speaker’s perspective (Traugott & Dasher 2002:  6). The first sense encodes the non-realization of the subject’s own goals, expressed in the complement clause. The second encodes the non-realization of what the speaker encodes as the subject’s obligations, if not his or her goals. In the third, the subject’s own goals are irrelevant: the construction merely encodes the disappointment of the speaker’s expectations. In the fourth, ‘ fail to’ functions as a negation marker. The last two senses are more subjective than the first two insofar as they make no explicit or implicit reference to effort or duty on the part of the subject. They will both be classified as “subjective”, the first two senses as “objective”. Subjectivity is, of course, not an absolute: thus subjective should be understood in the sense of “towards the subjective end of the subjective-objective cline”, objective understood as “towards the objective end of the cline”. Table 1 below shows the four basic senses of both ‘ fail to’ and ‘not fail to’. Examples (5) to (11), taken from various corpora in the ICAME collection, illustrate the eight senses defined in Table 1. We begin with the four objective senses.

(5) He failed to set the hook at least half the time, not from being slow, but because he was overeager – too fast. (Frown P13 193) (6) The Super Sleuth, Jane Marple, never seems to fail to find her villain, and an audience as well. (ACE) (7) The shortage was discovered after Huff failed to report for work on Sept. 18. (Brown A25 1880)

 Thomas Egan

Table 1.  The basic senses of ‘ fail to’ and ‘not fail to’ Basic senses

Non-negated ‘fail to’

Negated ‘fail to’

Subjective/ Objective

Effort

Tried + did not succeed (see example 5) Did not fulfil duty (see example 7)

Tried + succeeded (see example 6) Fulfilled duty (see example 8)

Objective

Did not live up to speaker’s expectations (see example 9) Did not do (see example 11)

Lived up to speaker’s expectations (see example 10) Did (see example 11)

Neglect Expectation

Negation



Objective Subjective

Subjective

(8) There was the day Uncle Izaak had, in an unexpected grandiose gesture, handed over the pretty sloop to Abel for keeps, on condition that he never fail to let his brother accompany him whenever the younger boy wished. (Brown K23 1560)

(5) and (6) illustrate the two “effort” senses. In (5), the subject makes a conscious effort to bait his hook, an effort that often proves unsuccessful. In (6), on the other hand, the subject not only sets out to identify the culprit, but always succeeds in doing so. (7) and (8) illustrate the two “neglect” senses. In (7), the subject, although obliged to turn up for work, omits to do so. In (8) the subject assumes the obligation to allow his brother to accompany him. Tokens like (7) and (8) are labelled “objective” since the obligation in question is not just in the eye of the speaker, but is both known to, and assumed by, the subject. (9) Traffic frequently has failed to measure up to engineers’ rosy estimates.  (Brown A43 0650) (10) You may know nothing at all about a society; but you cannot fail to recognize this specific type of hilarity. (LOB G77 55) (11) “In typical cases women fail to exhibit any sexual overvaluation towards men; but they scarcely ever fail to do so towards their own children.”  (Frown G65 162) (9) to (11) are very different from (5) to (8) with respect to the question of whether the subject makes an effort to achieve the goal in the complement clause. In (9), for example, the subject, traffic, neither makes an effort to, nor has a duty to, comply with the subjective expectations of the engineers. In (10), it is the expectation of the speaker with respect to the impossibility of the subject’s not recognizing the hilarity in question that is encoded. The two tokens in (11) differ from those

The ‘ fail to’ construction 



in (9) and (10) in that the element of expectation is backgrounded, if indeed it may be detected at all. (11) may felicitously be paraphrased: “In typical cases women do not exhibit any sexual overvaluation towards men; but they always do so towards their own children.” In (11), fail to functions as a marker of negation in the first clause, scarcely fail to as an emphatic marker in the second. These are the eight basic senses to which all the tokens in the various corpora investigated in this paper have been assigned. In the next section we will see how the four negated senses have gradually given way to their non-negated counterparts. 3. The loss of negation Figure 1 contains percentage totals for the number of tokens of ‘ fail to’ with positive and negative polarity in the 1850–1920 component of the Corpus of Late Modern English Texts (CLMET: see De Smet & Cuckyens 2005) and a selection of twentieth and twenty-first century corpora. 100.0 90.0 80.0 70.0 60.0 50.0 40.0 30.0 20.0 10.0

Webcorp

BNC

Wellington

ACE

Frown

FLOB

Brown

LOB

non-negated negated CLMET 1850–1920

0.0

Figure 1.  Percentages of tokens of non-negated and negated ‘ fail to’ in a selection of corpora from CLMET 1850–1920 to WebCorp

 Thomas Egan

The six corpora in the ICAME collection in Figure 1 all contain one million words. The number of tokens of the ‘ fail to’ construction varies from seventy-four in the case of LOB, which contains British English from the early 1960s, to one hundred in FLOB, which contains British English from the early 1990s. The American corpora from the corresponding periods, Brown and Frown, contain ninety-one and ninety-nine tokens respectively. The Australian Corpus of English (ACE), from the mid 1980s, contains seventy-seven instances of the construction and the Wellington Corpus from New Zealand, from the latter half of the 1980s contains seventy-eight. The figures for the BNC (the British National Corpus) are based on a one thousand token random sample of the verb fail which included 729 tokens of the construction, while a search of the World Wide Web, using the WebCorp search engine, returned 380 relevant tokens. It is obvious from Figure 1 that there has been a decline in the incidence of negated ‘ fail to’ in the twentieth century. Moreover, the negated construction would appear to be even rarer in Present-Day English. Only 5 of 380 tokens of ‘ fail to’ downloaded using WebCorp on 27.05.08 were negated. Moreover, all five of them were biblical quotations. One of these is cited as (12). (12) 1 Thessalonians 1:3 For we never fail to remember your works of faith and labours of love and your persistent and unwavering hope in our Lord Jesus Christ in the presence of our God and Father;  (http://bibletab.com/f/fail.htm) The reduction in negated tokens shown in Figure 1 is the continuation of a process that seems to have begun in the nineteenth century, as may be seen from the normalized frequencies in the three sub-corpora of the Corpus of Late Modern English Texts in Figure 2. In the first period covered by the Corpus of Late Modern English Texts, 1710– 1780, eighty-five of all ninety-one tokens are negated. (13) to (15) may be taken as representative examples. (13) Jerome returned to the Prince, and did not fail to repeat the message in the very words it had been uttered.(Walpole, The Castle of Otranto, Chapter 3) (14) Such an unexpected visit could not fail to affect me with surprise and consternation.(Smollett, The Adventures of Peregrine Pickle, Chapter LXXXI) (15) Such evils as these will not fail to render us miserable, though they have little tendency to diminish pride.  (Hume, A Treatise of Human Nature, Part IV, Sect. VI) (13) instantiates the “effort” sense of ‘ fail to’, shading into the “neglect” sense. It is clearly objective in that Jerome is encoded as having made a conscious effort to convey the message. (14) and (15), on the other hand, are subjective in sense. No

The ‘ fail to’ construction 

4.5 4 3.5 3 2.5 2 1.5 1 0.5

non-negated negated

0 1710–1780

1780–1850

1850–1920

Figure 2.  Negated and non-negated tokens per 100,000 words of ‘ fail to’ in CLMET 1710–1780, 1780–1850 and 1850–1920

effort has been expended by the visit in the former or the evils in the latter. They thus differ from the Chaucerian examples in (1) and (2), which are both objective. The subjective reading seems to have developed in the course of the sixteenth century and accounts for some two thirds of the tokens in CLMET 1710–1780. In the second period of CLMET, 1780–1850, 37 of 147 tokens are negated. This represents a reduction in the proportion of negated tokens from over 90 percent to some 75  percent. (16) and (17) represent the still predominant negated form, (18) its increasingly popular non-negated counterpart. (16) The duke himself shot admirably, and never failed to hit the bulls-eye.  (Ainsworth, Windsor Castle, Chapter III) (17) The attacks of illness which arise from miasma never fail to appear most mysterious. (Darwin, The Voyage of the Beagle, Chapter XVI) (18) Thrice she attempted to speak, and thrice her voice failed to penetrate the folds of the heavy door.  (Bulwer-Lytton, The Last Days of Pompeii, Chapter XIV) (16) resembles (13), both in being negated and in being subject to an objective construal. (17) resembles (14) and (15) in being negated and subjective. (18) is an example of the type of usage that is on the increase in the early nineteenth century. It is not negated and it is to be construed objectively, the voice being linked metonymically to its owner, who made an effort to be heard through the door. It is from the second to the third period of CLMET, from before to after 1850, that we find the most significant change in the ratio of negated to non-negated

 Thomas Egan

tokens (P = 0.0000: Pearson’s chi-sq. = 87.35). In the first half of the nineteenth century, roughly three-quarters of all tokens are negated. Towards the end of the century the reverse is the case, with just 47 of a total of 196 being now negated, as may be seen in Figure 2. Examples (19) to (21) may be taken as typical of usage around the turn of the century. (19) Only a few of the Polygonal Class fail to pass the Final Test or Leaving Examination at the University. (Abbott, Flatland, Part 1, Section 6) (20) I fail to see why Mr. Short should be privileged.  (Grossmith, The Diary of a Nobody, Chapter XIX) (21) Let us not dwell on it, on all the average civilized man still fails to achieve;  (Wells, Mankind in the Making, Chapter V) (19) instantiates the “effort” sense. Students try to pass examinations. In this case most of them are successful. (20) and (21), on the other hand, are subjective. In (20) there is no reason to assume that the speaker has made a serious effort to understand the reasons behind Mr. Short’s position of privilege. It could be paraphrased aptly by ‘I do not see why Mr. Short should be privileged’. Similarly in (21), Wells regrets that the civilized man does not live up to his (Wells’) expectations, not that the civilized man has not reached his own goals. The decline in the use of the negated form was most marked in the nineteenth century, and it has continued to the present day, as may be seen in Figure 1. The difference in Figure 1 between the percentages for LOB and Brown may indicate that the change was more advanced in American English in the 1960s, with 15 of 74 tokens being negated in the British corpus as opposed to only 6 of 91 in its American counterpart. However, the degree of conformity between the results for the British, American, Australian and New Zealand corpora shows that we are here witness to a change in all major varieties of English. By the last decades of the twentieth century, the average proportion of instances that was negated ranged from two percent in the case of the Frown corpus of American English to five percent in the case of the FLOB corpus of British English. Moreover this development would seem to have reached a climax in the early twenty-first century, when, as we have seen, the only negated tokens returned by WebCorp were biblical quotations. In the next section we will look more closely at the nineteenth century since it witnessed the most significant fall-off in the use of negated ‘ fail to’ and rise in the use of its non-negated counterpart. 4. Some new sorts of failure The decline in the use of the negated form in the nineteenth century affected all uses of the form to a similar degree. They all declined but none of them disappeared

The ‘ fail to’ construction 



altogether. All four semantic types in Table  1 continue to occur. In CLMET 1780–1850 the three most common forms could not fail, cannot fail and never fail account for some 70 percent of all tokens. In CLMET 1850–1920 the same three forms account for 60 percent. Other forms, such as will not fail to, would not fail to, did not fail to and seldom fail to also continue to be used, though with ever declining frequency, right up to the end of the twentieth century. The situation for the non-negated form is rather different in that, while established uses become more widespread, the nineteenth century also sees ‘ fail to’ being used to encode new sorts of failure. Both subjective and objective forms increased, as may be seen in Figure 3. To begin with the subjective readings, one area in which there is a marked increase is in the use of ‘ fail to’ with predicates of perception and understanding. We have seen one such token in (20). Performative uses such as (20) and (22) to (23) become common in the late nineteenth century. (22) “I fail to see the connection,” said Leonard, hot with stupid anger.  (Forster, Howards End, Chapter XVI) (23) “I am afraid I fail to follow you, Mr. Bruff,” I said, modestly.  (Collins, The Moonstone, Chapter III) Although I have placed (22) and (23) with (20) at the subjective end of the objective-subjective cline described in Section 2, an analysis along the lines of Traugott & Dasher (2002) would go further and classify all three examples as intersubjective. 2.5 2 1.5 1 0.5 objective subjective

0 1780–1850

1850–1920

Figure 3.  Subjective and objective tokens of non-negated ‘ fail to’ per 100,000 words

 Thomas Egan

Traugott & Dasher state that, in their view “intersubjective meanings crucially involve social deixis (attitude toward status that speakers impose on first person – second person deixis)” and that “what is said implies more is meant” (Traugott & Dasher 2002: 23). In (22) the speaker not only says that he finds it impossible to see the connection, but he implies that no such connection exists. In (23) the speaker not only says that he cannot understand his addressee, but implies that such understanding would be difficult to achieve, at the same time hedging this objection by superficially attributing its cause to himself. Similarly, in (24), when the speaker instructs the addressee to fail to see the point, he is not advocating a form of cognitive inability but rather one of intersubjective behaviour. (24) When he says anything laughable, fail to see the point and don’t smile, and speak of Him before those who will report your talk as “that fantastical man,” or “that Sergeant What’s-his-name.”  (Hardy, Far from the Madding Crowd, Chapter XXIX) (25) “Of course,” said the troop horse, “everyone is not made in the same way, and I can quite see that your family, on your father’s side, would fail to understand a great many things.”  (Kipling, The Jungle Book, “Her Majesty’s Servants”) Not all instances of fail to see/follow/understand are intersubjective, however. (25) is an example in which the construction functions subjectively rather than intersubjectively. Figure 4 contains normalized frequencies for non‑negated ‘ fail to’ with “predicates of seeing and understanding” in the three subcorpora of CLMET. Together, these account for over a third of the subjective tokens in the period 1850–1920. As may be seen in Figure 3, the increase in the number of objective tokens is less than that of subjective ones. This increase takes two forms. In the first place, there is an increase in the number of tokens encoding what one might call more “local” failure, by which I mean failure in the carrying out of a specific task or achieving a specific result on a particular occasion. This sort of failure is instantiated in (18) and (19) above and (26) to (29) below. (26) I failed to find Lord Hilton at his house, but I was told he was expected from London by the six o’clock train from Waterloo.  (Wells, The War of the Worlds, Chapter Three) (27) It appeared that when Prince Otto met me after my interview with Prince Ernest, he did his best to provoke a rencontre, and failing to get anything but a nod from my stunned head, betook himself to my University.  (Meredith, The Adventures of Harry Richmond, Chapter XXXII)

The ‘ fail to’ construction 

0.7 0.6 0.5 0.4 0.3 0.2 0.1

fail to + perception fail to + understanding

0 1710–1780

1780–1850

1850–1920

Figure 4.  Tokens of ‘ fail to’ per 100,000 words with predicates of perception/understanding in CLMET 1710–1780, 1780–1850 and 1850–1920

In (26) the subject attempts to see Lord Hilton on a specific occasion. This attempt was unsuccessful. Similarly in (27) the subject did not succeed in provoking a rencontre on a specific occasion. (26) and (27) differ from (19) in that the failed endeavour is not competitive in nature. Failure in competition would seem to be a new use, instantiated in (19) and (28) to (29). (28) It is from these specimens of the refuse of our Nobility that the great Tumults and Seditions of past ages have generally derived their leaders; and so great is the mischief thence arising that an increasing minority of our more progressive Statesmen are of opinion that true mercy would dictate their entire suppression, by enacting that all who fail to pass the Final Examination of the University should be either imprisoned for life, or extinguished by a painless death. (Abbott, Flatland, Part 1, Section 6) (29) Erotic had failed to win the Lancashire Cup.  (Galsworthy, The Man of Property, Chapter VII) While horse-racing dates back to the Restoration, public examinations only became common in the nineteenth century. We would therefore not expect to encounter a token like (28) in an earlier era. It is, however, not only in the academic world that we now witness more people failing. There is a marked general increase in the incidence of animate subjects during the period, as may be seen in Figure 5, which contains normalized frequencies for (both negated and non‑negated) ‘ fail to’.

 Thomas Egan 4.00 3.50 3.00 2.50 2.00 1.50 1.00 0.50

animate inanimate

0.00 1710–1780

1780–1850

1850–1920

Figure 5.  Tokens of ‘ fail to’ per 100,000 words with animate and inanimate subjects in CLMET 1710–1780, 1780–1850 and 1850–1920

The sort of failure we have looked at thus far is local, restricted to a specific occasion. What is new in the latter half of the nineteenth century is the predication of failure by people on a more global scale, as in (21) and (30) to (32). (30) The professional has not, he might well say, the leisure and freedom from money anxieties which will let him devote himself to his art in singleness of heart, [...]; he has got to square everyone all round and will assuredly fail to make his way unless he does this; if, then, he betrays his trust he does so under temptation. (Butler, Notebooks, “Amateurs and Professionals”) (31) But if by some miracle we could use the Casual Wards as a means of providing for all those who are seeking work from day to day, [...] they would utterly fail to have any appreciable effect upon the mass of human misery with which we have to deal.    (Booth, In Darkest England and the Way Out, Chapter 9) (32) It is very doubtful if these most favourable conditions fall to the lot of more than a quarter of the children born to‑day even in England, where infant mortality is at its lowest. The rest start handicapped. They start handicapped, and fail to reach their highest possible development.  (Wells, Mankind in the Making, Chapter III) Neither (21) nor (30) to (32) encodes failure on one particular occasion. Rather the subjects are seen to fail comprehensively. All four tokens are concerned with the question of possible improvement to the human condition, either on an individual or collective level. In (32), for instance, we see how whole lives may be



The ‘ fail to’ construction 

portrayed as failures, an idea we also see reflected in the uses of the noun failure in (33) to (38). (33) “I consider my life to some extent a failure,” said Knight again after a pause. (Hardy, A Pair of Blue Eyes, Chapter XIX) (34) And as to her life. Failure, failure through headstrong blindness and self‑will, resulting in the agony of the innocent.  (Yonge, The Clever Woman of the Family, Chapter XX) (35) He earned enough for his actual needs, and was under no pressing fear for the morrow, so long as his faculties remained unimpaired; but there was no disguising from himself that his life had been a failure.  (Gissing, New Grub Street, Chapter VII) There are no instances prior to 1850 in either the OED or in CLMET in which failure is predicated of people’s lives. After 1850 we find that not only people’s lives but also individuals themselves may be considered “failures”, as in (36) to (38). (36) It is nearly certain that as a rule they make less agreeable boys and girls, but to me at any rate it is not nearly so certain that they make adult failures. (Wells, Mankind in the Making, Chapter IV, sec. 3) (37) ‘Of course, Edwin, if you make up your mind that you are a failure, you will end by being so. (Gissing, New Grub Street, Chapter IV) (38) “I have been a failure,” said Miss Bartlett, as she struggled with the straps of Lucy’s trunk instead of strapping her own.  (Forster, A Room with a View, Chapter VII) This use of failure is new in the late nineteenth century and reminiscent of the late twentieth century use of loser in a global sense.3 Not only do these new uses of the noun failure parallel the new uses of the ‘ fail to’ construction, the noun also expands in incidence at a similar rate to non‑negated ‘ fail to’, as may be seen in Figure 6. To sum up, while it is impossible at this stage to advance an explanation for the growth in the use of non‑negated ‘ fail to’ in the nineteenth century, we have noted three developments. The period witnessed – an increase in formulaic intersubjective utterances (I fail to see/understand) – an increase in animate subjects – an increased focus on personal failure (failed lives and individuals). Whatever the reasons for these changes, the increase in the use of the non‑negated form has continued to the present day, to the extent that one may ask whether

3. I am grateful to Juliet Munden for pointing out this parallel to me.

 Thomas Egan 6.00 5.00 4.00 3.00 2.00 1.00 fail to (non-negated) failure

0.00 1710–1780

1780–1850

1850–1920

Figure 6.  Tokens of non‑negated ‘ fail to’ and failure per 100,000 words in CLMET 1710–1780, 1780–1850 and 1850–1920

‘ fail to’ is in the process of grammaticalizing as a negation marker. This question is addressed in the next section. 5. Is ‘ fail to’ grammaticalizing? We have now seen that there has been a general increase in the use of non‑negated ‘ fail to’ in the last two centuries. We have also noted that one of the four main uses of the form is to encode negation pure and simple, as in examples (4), (11), (22) and (23). In these and similar instances there is no suggestion of the subject’s having made an effort to achieve some goal or having neglected to perform some duty. Given the amount of semantic bleaching involved in the “negation sense” compared to the original “effort sense”, instantiated in (1) and (2), the question naturally arises as to whether the ‘ fail to’ construction is in the process of grammaticalizing. Much has been written about grammaticalization since Hopper & Traugott (1993) produced the standard introduction to the topic. Traugott & Dasher (2002) describe grammaticalization as being “properly conceived as the change whereby lexical material in highly constrained pragmatic and morphosyntactic contexts is assigned functional category status, and where the lexical meaning of an item is assigned constructional meaning” (Traugott & Dasher 2002: 81). According to Boye & Harder (2009) the assumption of functional category status by a full lexical item is a change in which “lexical elements go from being used to convey primary information



The ‘ fail to’ construction 

to being used predominantly to encode secondary information. [...] Grammaticalization resides basically in the coding of secondary information status” (Boye & Harder 2009: 32). Boye & Harder’s formulation in terms of primary and secondary information has the advantage that it is easy to operationalize. In order to investigate the extent to which ‘ fail to’ encodes primary or secondary information I searched WebCorp for tokens of ‘ fail to’ followed by did. Tokens in which did functioned as an anaphoric pro‑form were extracted. In some of these, such as (39) to (41), the matrix verb failed is “discursively secondary” to the predicate in the complement clause. (39) When they launched it, everyone from engineers to Communist Party big shots failed to realize its importance. Only Korolyov did. (http://www.columbiamissourian.com/stories/2007/10/07/Space-remains-a/) (40) The simpler N,N’-bis(salicylidene)-ethylenediaminocobalt(II) [Co(Salen)2] (16) failed to catalyze deoxygenations in THF but did in DMF. (http://www.ingentaconnect.com/content/els/00404039/1999/00000040/ 00000050/art01880) (41) Zidane spearheaded the semi‑final assault on Croatia, and although he failed to score, Thuram did twice to Suker’s one – and booked the hosts a place in the final against the World Champions.  (http://www.travelnotes.org/Football/WorldCup/France.htm) In the three examples the pro‑form did refers to the complement predicates realize, catalyze and score rather than the matrix verb failed. This reference is signalled by only in (39), but in (40) and although in (41). Note that while (39) and (40) involve bleaching, this is not the case to the same extent for (41), which encodes an objective effort on the part of the footballers. However, although the meaning of fail in (41) is not as bleached as it is in the two other examples, it is still backgrounded compared to the meaning of score. The material downloaded from WebCorp also included tokens such as (42) and (43), in which failed is not backgrounded. (42) The bomb, however, failed to explode, and so did three others planted on vehicles he chose at random.  (http://www.newsregister.com/news/story_print.cfm?story_no=213941) (43) Sites that failed to generate STOC did so consistently, while those at the remaining sites generated STOC without failure.  (http://ajpcell.physiology.org/cgi/content/short/287/6/C1577) In (42) and (43), the pro‑form did refers to failed rather than to explode and generate. This interpretation is signalled in both examples by (and) so. If we were to replace and so in (42) by but, the natural reading would be that did would then

 Thomas Egan

refer to explode rather than failed. Moreover, if ‘ fail to’ had grammaticalized in (42), so would have to be replaced by neither, as in (44) to (46). (44) Even Albeau failed to make final and neither did Pritchard as the event lead was left on a knife‑edge.  (http://www.gazettelive.com/features/feature_1.htm) (45) The ref who was well behind the ball, failed to see anything, and neither did the linesman. (http://www.merseyworld.com/rearguard/11com98.html) (46) The young writer hired to update her work never visited the places in question, but had used websites or phone calls, failed to double check – and neither did the publishers.  (http://www.nomadtours.co.za/guidebooksdontbelieveeverythi.html) In all three of these examples the pro‑form did does not refer to fail but to the complement predicates. In this respect they resemble examples (39) to (41) rather than (42) and (43). We would therefore seem to be faced with a situation in which ‘ fail to’ has grammaticalized in some contexts, but not in others, and we are dependent on the immediate co‑text for information on which to base our choice between the two interpretations. If  ‘ fail to’ is partly grammaticalized, the next question is when this process started. The criterion of semantic and syntactic backgrounding, which serves to distinguish (39) to (41) from (42) and (43) is less easy to operationalize in the case of historical corpora. There are two main reasons for this. The first is that tokens in which the construction is followed by a pro‑form or by ellipsis are relatively rare. Even WebCorp only returns a handful of relevant tokens. The second reason is the presence in the historically dominant ‘not fail to’ construction of two morphemes encoding non‑realization of the complement situation. In such cases, a contrastive conjunction or conjunct would serve to indicate the cancellation of the most external negation marker, of not/never/seldom rather than either fail or the complement predicate. In the absence of historical tokens which could furnish us with syntactic evidence of discursive foregrounding/backgrounding, we are thrown back on our intuition and other text‑linguistic evidence. (47) to (49) are from CLMET 1710–1780. (47) In our arrangement of bodies we never fail to place such as are resembling, in contiguity to each other, or at least in correspondent points of view. (Hume, A Treatise of Human Nature, part IV, Sect. V) (48) Nations, accordingly, seldom fail to retaliate in this manner.  (Smith, The Wealth of Nations, Chapter II) (49) [...] the young gentleman, taking the advantage of being alone with the physician, recapitulated all the affronts he had sustained from the painter’s petulance, aggravating every circumstance of the disgrace, and advising

The ‘ fail to’ construction 



him, in the capacity of a friend, to take care of his honour, which could not fail to suffer in the opinion of the world, if he allowed himself to be insulted with impunity, by one so much his inferior in every degree of consideration. (Smollett, The Adventures of Peregrine Pickle, Chapter LXIII) How much in the way of meaning would be lost if we replaced never fail to place in (47) with always place, seldom fail to retaliate in (48) with almost always retaliate and could not fail to suffer in (49) with must needs suffer? The less one feels is lost by the substitution the more one interprets fail as semantically bleached and the more one would be inclined to view ‘ fail to’ as grammaticalized. Note too that, if one concludes that ‘ fail to’ is grammaticalized in these examples, this leads to the conclusion that the process of grammaticalization is not directly related to the loss of negation. Given that bleaching is by definition most evident in subjective tokens one could also ask whether the ratio of these to objective tokens has altered over the last several hundred years. Figure 7 contains details of this ratio in CLMET and a selection of contemporary corpora. From the data in the figure we can conclude that the ratio of subjective to objective tokens of ‘ fail to’ has remained much the same over the last three centuries. There has always been a sufficient critical mass of objectively‑construed usages in both Late Modern and Present‑Day English to resist the gravitational pull towards subjectification and intersubjectification that often accompany grammaticalization. One should of course add that there is no vacuum in the system of English negation at present to motivate the recruitment of a new negation marker. 100 90 80 70 60 50 40 30 20 10 0 Wellington

ACE

Frown

FLOB

1850–1920

1780–1850

1710–1780

objective subjective

Figure 7.  Ratio of objective to subjective tokens of ‘ fail to’ in CLMET, FLOB, Frown, ACE and Wellington corpora (percentages)

 Thomas Egan

6. Summary and conclusion This paper has shown that there has been a steady reduction in the use of the negated ‘ fail to’ construction over the past three hundred years. This reduction has been accompanied by an increase in the use of non‑negated ‘ fail to’. We have seen that there was a corresponding increase in the use of the noun failure in the nineteenth century, when both the verb and the noun were increasingly used to encode failure on a global scale. We have also noted that the negated form has practically disappeared from present‑day usage and that the now dominant non‑negated form has grammaticalized in at least some uses to the extent that it can function as a marker of negation. As pointed out in the previous section we have at present no need of such a marker in English. Should one ever be required, however, ‘ fail to’ would fit the job description to a tee. Sources ACE = Australian Corpus of English. The ICAME Corpus Collection on CD-ROM, version 2 (1999). Bergen: Aksis. BNC = British National Corpus: on CD-Rom (2001). Oxford: Oxford University Computing Services. Brown = Corpus of Present-Day Edited American English (1979 [1964]). The ICAME Corpus Collection on CD-ROM, version 2 (1999). Bergen: Aksis. CLMET = Corpus of Late Modern English Texts. See De Smet & Cuyckens (2005). Corpus de la Littérature Médiévale des Origines au 15e Siècle. See http://www.classiques-garnier.com/ CEECS = Corpus of Early English Correspondence Sampler (1998). The ICAME Corpus Collection on CD-ROM, version 2 (1999). Bergen: Aksis. FLOB = The Freiburg – LOB Corpus of British English (1998). The ICAME Corpus Collection on CD-ROM, version 2 (1999). Bergen: Aksis. Frown = The Freiburg – Brown Corpus of American English (1999). The ICAME Corpus Collection on CD-ROM, version 2 (1999). Bergen: Aksis. Helsinki = Helsinki Corpus of English Texts (1996). The ICAME Corpus Collection on CD- ROM, version 2 (1999). Bergen: Aksis. LOB = Lancaster-Oslo-Bergen Corpus of British English (1978). The ICAME Corpus Collection on CD-ROM, version 2 (1999). Bergen: Aksis. WebCorp (1999–2009): Birmingham City University. See http://www.WebCorp.org.uk/. Wellington = Wellington Corpus of Written New Zealand English (1993). The ICAME Corpus Collection on CD-ROM, version 2 (1999). Bergen: Aksis.

The ‘ fail to’ construction 



References Benson, Larry D., ed. 1988. The Riverside Chaucer. 3rd edn. Oxford: Oxford University Press. Boye, Kasper & Peter Harder. 2009. “Evidentiality: Linguistic Categories and Grammaticalization”. Functions of Language 16:1.9–43. De Smet, Hendrik & Hubert Cuyckens. 2005. “Pragmatic Strengthening and the Meaning of Complement Constructions”. Journal of English Linguistics 33:1.3–34. Hopper, Paul J. & Elizabeth Closs Traugott. 1993. Grammaticalization. Cambridge: Cambridge University Press. OED. 1994. The Oxford English Dictionary: On Compact Disc. Oxford, Oxford University Press. Pearsall, Judy & Patrick Hanks. 1998. The New Oxford Dictionary of English. Oxford: Clarendon Press. Traugott, Elizabeth Closs & Richard B. Dasher. 2002. Regularity in Semantic Change. Cambridge: Cambridge University Press.

The interplay of modal verbs and adverbs A history of mæg eaþe Jerzy Nykiel*

University of Silesia As Hoye (1997) points out, harmonic combinations of modal verbs and modal adverbs are not uncommon in Present-Day English. The present paper explores semantic and syntactic aspects of one of such combinations, namely mæg eaþe ‘may easily’, from a diachronic perspective. The collocation is attested at the earlier stages of the development of the language, that is, in Old English and Middle English. I aim to show that as early as in Old English, the adverb eaþe ‘easily’ helps to reinforce possibility-based meanings of mæg ‘may’, including epistemic possibility. The Middle English range of contexts in which mæg eaþe ‘may easily’ occurs is more limited, which ultimately leads to the demise of the combination toward the end of the Middle English period. I also examine the issue of the extent to which the collocation is lexicalized in the respective periods, especially with evidence from such factors as decomposition of meaning, productivity and substitutability.

1. Introduction In this article I seek to look into the history of the short-lived collocation mæg eaþe ‘may easily’ found in Old English and Middle English texts. The nature of the collocation, that is, the fact that it is composed of a pre-modal verb1 and a modal adverb, invites comparisons with Hoye’s (1997) account of modal verb – adverb constructions in Present-Day English. He shows that the meanings of * I wish to thank the editors of this volume and two anonymous referees for their valuable comments and suggestions. All errors and inadequacies remain, of course, mine. 1. I follow Lightfoot (1979, 1991) and Traugott (1992) and use the term “a pre-modal verb” which stands for “a predecessor of a PDE modal auxiliary”. The term has syntactic, rather than semantic, significance in that it implies that the Old English and Middle English ancestors of the PDE modals, e.g. cann ‘can’, mæg ‘may’, sceal ‘shall’, are not full-fledged auxiliaries. The same verbs are attested with both modal and non-modal semantics in Old and Middle English.

 Jerzy Nykiel

modals are often subject to modification, achieved through the use of an accompanying adverb. It is, however, Lyons (1977: 807) who first casts a wider theoretical net over similar expressions. He classifies them as “harmonic” or “non-harmonic”, depending on whether the two members of the modal verb – modal adverb collocation are semantically compatible or not. In modally harmonic collocations two distinct modal forms, in harness with each other, converge to produce a single modal meaning. In

(1) Now it may easily happen that the activation energy is large compared with the transition temperature. (W.9.9.153, quoted in Hoye (1997: 167))

may and easily both contribute to one meaning of possibility. The fusion of compatible modalities supplied by an adverb and a modal verb in one clause is called “semantic harmonization” by Shibasaki (2004: 392). These tenets resonate in both synchronic (e.g.  Hoye 1997 and Wärnsby 2006) and diachronic (e.g.  Molencki 2003; Shibasaki 2004, 2009) studies of cases of the modal adverb – modal verb co-occurrence. Non-harmonic combinations, on the other hand, by virtue of their internal modal incompatibility, receive a two-clausal interpretation. Thus, sentence (2)

(2) Certainly it may rain in the evening.

can be paraphrased as

(2) a. It is certainly the case that it may rain in the evening.

This interpretation follows from Lyons’ (1977: 808) observation that “no simple utterance may contain more than a single subjective epistemic modality [...]”. This study, with its focus being on mæg eaþe ‘may easily’, is also a study of possibility, both epistemic and non-epistemic. This paper follows the conceptualization of possibility offered by van der Auwera & Plungian (1998). The resultant terminology and divisions implemented in my considerations of possibility follow in Table 1. Table 1.  The modal meaning of possibility according to van der Auwera & Plungian (1998: 82) Non-epistemic possibility Participant-internal possibility (ability, dynamic possibility, capacity)

Participant-external possibility Non-deontic possibility

Deontic possibility (permission)

Epistemic possibility

Mæg eaþe 



According to this approach, possibility branches into two major areas: non-epistemic and epistemic possibility. Within non–epistemic possibility, participantinternal possibility (henceforth PI possibility) corresponds to the traditionally acknowledged notions of ability, dynamic possibility and capacity, as in (3):

(3) Your brother can run faster than you.

Participant-external possibility is divided into non-deontic possibility and deontic possibility or permission. The former comes into play when enablement is contingent upon some more or less nebulous Agent-external circumstances, e.g.

(4) You can see this painting in the local museum.

The fact that the painting is displayed in the museum opens up the possibility for the participant to see the painting. A context of deontic possibility (permission) calls for the speaker, or optionally some other person of authority or a set of regulations, as the source of enablement. The participant is free to proceed in a given way, an external source granting the possibility, as in (5):

(5) You can sit down now.

Permission will play no role in this study, however. Finally, epistemic possibility has to do with a speaker’s assessment of a state of affairs as (un)likely to happen or be the case, e.g.

(6) Johnny may win the race one day.

When analyzing data, however, one encounters instances which are suggestive of more than just one type of possibility. Coates (1983: 14ff.) speaks of “indeterminacy” in such cases. Indeterminacy can take the form of (a) “gradience”, when one meaning of a modal form shades into another and a given instance does not show the most characteristic features of either, (b) “ambiguity”, when the context provides too few clues to deduce the type of possibility meant, or (c) “merger”, when an instance easily lends itself to two interpretations, yet neither has to be chosen as both are equally viable. Although I have tried to give unambiguous illustrations, some of the examples of mæg eaþe ‘may easily’, which follow in the remaining part of this study, show signs of indeterminacy. 2. Etymology The adverb of manner eaþe (ME eth(e), EME eath) ‘easily’, rich in cognates in other Germanic languages, e.g. Old Saxon ôðo ‘easily, perhaps’, Old High German

 Jerzy Nykiel

ôdo ‘perhaps’, Old Norse auð‑, as in auð‑gorr ‘easy to do’, was active in English since the written beginnings of the language but became obsolete in the eighteenth century, according to the data given in the OED, s.v. eath. The dictionary locates the latest example of eath ‘easily’ in 1748. The latest occurrences in Scottish English are about a hundred years older. The only period in which the adverb in question enjoys respectable frequency is Old English. In the extant Old English texts collected in the DOE corpus I have identified 247 instances of eaþe ‘easily’, including all the spelling variants of the adverb as listed by the DOE. Since the meaning of the adverb remains relatively stable until its demise, I cite the senses of eaþe ‘easily’ as given by the DOE s.v. eaþe: Meanings of eaþe (adv) ‘easily’ according to the DOE: 1. 2. 3. 4. 5.

easily, with little or no difficulty in comparative: more comfortably, more at ease with equanimity: easily, calmly without hesitation or reluctance, readily, willingly used to denote likelihood or possibility: easily, possibly, likely, well

In the group of the 247 instances of OE eaþe ‘easily’, the adverb can be found in the company of mæg ‘may’ 178 times. Eaþe ‘easily’ co-occurs with the pre-modal in all persons and both tenses, i.e. present and past, so I use the Old English 1st/3rd person singular present form mæg ‘may’, rather than the infinitive form magan, while referring to the collocation in general. Mæg eaþe ‘may easily’ continues well into Middle English yet fails to survive beyond the fifteenth century. The latest example I have found is (7): (7) Oone worde might thou speke ethe, yit might it do the one word may thou speak easily yet might it do thee som letht, some mitigation (a1500(a1460) Towneley Pl. (Hnt HM 1)232/141) ‘You can easily say one word, it might still mitigate your case.’ It seems that the key motivation behind the disappearance of mæg eaþe ‘may easily’ is the appearance of the French loanword esi (OF aaisié, aisé, a past participle of aiser, aisier ‘to put at ease’ according to the OED s.v. easy; PDE easy) in early Middle English. The adjective esi ‘easy’ gives rise to the adverb esili ‘easily’, the first instance of which dates back to the year 1300 according to the OED and MED. Eaþe ‘easily’ is ousted from collocating with mæg ‘may’ by the end of the Middle English period, its function being taken over by esili ‘easily’. This argument finds further support in Shibasaki’s (2009: Table 5) analysis of the instances of may easily in the

Mæg eaþe 



University of Virginia Corpus. Beginning with the year 1500 the frequency of the collocation may easily is shown to grow apace with each subsequent century. The other collocate, that is the pre-modal verb mæg ‘may’, is a member of the preterite-present class, a common Germanic inheritance which is a result of reinterpretation of Proto-Indo-European perfective and stative forms of verbs as present forms (see Prokosch 1939; Lass 1994; Kotin 2008). As shown by, for instance, Kytö 1991 and Nykiel 2010, Old English meanings of mæg ‘may’ center around the notion of PI and non-deontic possibility. The former type, that is PI possibility (general capacity residing in the Agent, ability), is illustrated in example (8): (8) oððe hwa is swa heardheort þæt ne mæg wepan swylces or who is so hardhearted that not may weep of-such ungelimpes? misfortune (ChronE 1086.11) ‘or who is so hardhearted as not to weep at such misfortune?’ where hardheartedness as a feature of the participant’s character can prevent them from weeping. In (9) the possibility to earn eternal bliss is contingent upon participants-external factors, which could be described as the nature of the Christian doctrine, thus the example represents non-deontic possibility: (9) Hwæt we nu gehyraþ þæt we magon mid þære soðan what we now hear that we may with that true hreowe ece blisse geearnian. penitence eternal bliss earn (HomS 8 197) ‘We have now heard that we can earn eternal bliss with true penitence.’ At the same time, already in Old English mæg ‘may’ shows signs of encroachment upon the territory of epistemic possibility and permission, a trend which continues in the later periods (see, for example, Tellier 1962, Goossens 1987, Bybee et al. 1994, Nykiel 2010). Since the clearest cases of epistemic mæg ‘may’ are among examples of mæg eaþe ‘may easily’, I consider it feasible to exemplify epistemic possibility of mæg ‘may’ while discussing the epistemic use of mæg eaþe ‘may easily’ in Old English in Section 4. 3. Framework and the origin of the collocation Instances of eaþe may ‘easily may’ are readily available throughout the Old English period, which makes it difficult to establish the genesis of the collocation. As a starting point, it can be safely assumed that the sense ‘possibly, perhaps’ of eaþe

 Jerzy Nykiel

‘easily’ is diachronically later than the other meanings of the adverb. This assumption is in accordance with Hanson’s (1987) insight into the semantic development of epistemic and evidential adverbs such as possibly or apparently, which enter the language in the Middle English period. As a rule, an epistemic extension of an adverb follows an earlier non-modal meaning.2 Eaþe ‘easily’ is, thus, initially an adverb of manner meaning ‘easily, with little difficulty’ (cf. sense 1 of eaþe ‘easily’ in Section 2), which comes to frequently modify the verbal complement of mæg ‘may’ of PI and non-deontic possibility. As a result, two meanings emerge: ability to perform an action effortlessly and Agent-external possibility to perform an action effortlessly. The ideas of PI/non-deontic possibility and ease are harmonic, they complement each other, which is reflected in the fact that eaþe ‘easily’ does not consistently co-occur with any other pre-modal in Old and Middle English. Due to common co-occurrence eaþe ‘easily’ comes to be associated more with mæg ‘may’ than with its complement and starts to reinforce the possibility of mæg ‘may’. The co-occurrence continues even though the meaning of mæg ‘may’ treads along the path toward epistemicity. When mæg ‘may’ starts to lend itself to an interpretation in terms of epistemic possibility, infrequently as it happens in Old English, eaþe ‘easily’ is still collocated with the pre-modal. In this context eaþe ‘easily’ becomes a truly modal adverb in that it is an emphasizer of the epistemic possibility of mæg ‘may’ in the sense of Lyons (1977) and Hoye (1997). Two formal issues are noteworthy in connection with the development of mæg eaþe ‘may easily’ sketched above. Firstly, the shift in the function of eaþe ‘easily’ from a modifier of the complement of mæg ‘may’ to a modifier of mæg ‘may’ itself can be seen as a transition of eaþe ‘easily’ from, using Huddleston & Pullum’s (2002: 576) terminology, a “VP-oriented adjunct” to a “clause-oriented adjunct”. This trajectory of change partly coincides with that attested by Traugott (1995) in the case of adverbs which evolve into pragmatic markers. The first shift on the trajectory, that is, “Internal Adverb > Sentential Adverb”, is the one experienced by eaþe ‘easily’ in that the adverb can no longer be associated with the VP. The second issue concerns the mechanism behind the shift of eaþe ‘easily’. It seems that we have to do with generalization of invited inferences as formulated by Traugott & Dasher (2005). The contexts of PI and non-deontic possibility, where eaþe emphasizes the ease with which an action can be performed, produce the inference that the action is likely to happen. The inference comes closest to being generalized when eaþe accompanies epistemic mæg.

2. Fischer (2007: 276) shows, however, that there are exceptions to this rule. Presumably, for example, came to be used as a sentential/modal adverb without being first used as a VP adverb. It is argued that it is analogy that is responsible for this development.

Mæg eaþe 



The first instances of mæg eaþe ‘may easily’ can be traced back to early Old English and then they are scattered all over the Old English and Middle English periods. Some illustrations follow in examples (10) through (13): (10) Oððe eft ðæt he ryhtlice 7 stiðlice wrecan sceolde, or else when he justly and sternly avenge should ðæt he ðæt ne forielde, ðylæs se ryhtwislica anda that he that not delay lest that good zeal acolige, ðæt he hit eft sua eaðe wrecan ne mæge weaken that he it again so easily avenge not may  ((ca880) CP 20.149.24) ‘Or else when he should justly and sternly take revenge, let him not delay, lest the good zeal should weaken so that he is not able to avenge the damage so easily any more.’ (11) ac we sceolon secgan and forswigian ne durron þa but we should say and keep-quiet-about not dare that halgan lare, þe se hælend tæhte: Seo lare mæg holy teaching, which the lord taught that teaching may eaðe unc emlice seman. easily us equally bring-about-agreement ((950–1050) ÆLet 5 8) ‘But we should talk and we dare not keep quiet about the holy teaching that the Lord gave us. The teaching can easily settle our dispute in a fair way.’ (12) “Seli wif, what eilleþ þe?” “Bote eþe mai I sori be: poor woman what ails thee but easily may I sorry be Ich heuede a douter feir and fre, Feiror ne I had a daughter lovely and gracious lovelier not miggtte no mon se. might no man see ((c1272–83) Dame Sirith 340) ‘“Poor woman, what ails you?” “But I can easily be sorry. I had a daughter, lovely and gracious, no one could see a lovelier one.’ (13) þe kyng hade too fewe Ageyn so monie schrewe the king had too few against so many villains So fele myhten eþe [Cmb: yþe] bringe þre to deþe. so many might easily bring three to death  ((c1325) Horn (Hrl2253) 61) ‘The king had too few against so many villains, so many could easily bring three to death.’

 Jerzy Nykiel

Tellier (1962: 90) calls the collocation a stereotyped expression in Old English, although, on top of giving a few examples, he does not elaborate on the matter. The stereotyped character hints at lexicalization of mæg eaþe ‘may easily’, which turns out to be generally weak, with the notable exception of the contexts where the expression is used epistemically or to gloss Latin forte ‘by chance, accidentally’ and forsitam ‘perhaps, probably’ in the Lindisfarne and Rushworth Gospels. The extent to which mæg eaþe ‘may easily’ is lexicalized is the focus of Section 6. In Sections 4 and 5 I discuss the semantics of the collocation in Old and Middle English respectively. 4. Mæg eaþe ‘may easily’ in Old English It is most plausible, and chronologically justified, to start an analysis of the semantics of mæg eaþe ‘may easily’ with the Old English examples as it is the period when the collocation has the richest range of uses and Middle English stops short of bringing any innovation. In accordance with Hoye’s (1997:  73ff.) and Nuyts’s (2001: 200) assessments, in a modal verb – modal adverb collocation, the verb is central in that it determines the meaning of the construction while the adverb plays the role of the verb’s “satellite”. It follows that the examples of mæg eaþe ‘may easily’ in my analysis will be arranged according to the relevant meaning areas of mæg ‘may’, that is, PI possibility, non-deontic possibility, and epistemic possibility. The area of PI possibility, that is possibility contingent on a conscious Agent, is abundantly represented in the examples below. In (14), as well as in (15), the Agent is God and his divine power to work miracles stands behind the PI possibility. If eaþe ‘easily’ is not adjacent to the pre-modal, which is the case in (14), it is not possible to say whether the adverb is still a VP-oriented adjunct or a clause-oriented adverb. (14) Swa Dryhten mæg, ana ælmihtig, eadigra gehwone wið as Lord may one almighty blessed each against earfeþum eaðe gescildan. tribulations easily shield ((960–990) GuthA 554) ‘As the Lord, the one Almighty, can easily shield each blessed one against tribulations.’ With much probability the meaning intended is ‘The Lord can shield each blessed soul easily ...’, with eaþe ‘easily’ functioning as a VP modifier, but there is an invited inference which suggests another reading, namely ‘The Lord can easily shield ...’, with eaþe ‘easily’ emphasizing mæg ‘may’. Example (15) has mihte ‘might’ and eaðe ‘easily’ used next to each other, which clearly casts the adverb in the role of the emphasizer of the verb:

Mæg eaþe 



(15) He gesceop ealle menn, and sume arærde, se ðe eaðe he created all men and some raised that who easily mihte ealle gif he wolde deade aræran þurh his might all if he would dead raise through his drihtenlican mihte; divine power ((c1000) ÆHom 6 126) ‘He created all men and brought some back to life, he who could easily bring everyone back to life if he wished to through his divine power;’ In (16), however, the VP has its own adjunct, i.e. genoh ryhte ‘well enough’. Given this, eaðe ‘easily’ can only modify mæg ‘may’. (16) Swiðe eaðe mæg on smyltre sæ ungelæred scipstiera very easily may on calm sea unlearned steersman genoh ryhte stieran, ac se gelæreda him ne getruwað on enough well steer but that learned him not trusts on ðære hreon sæ 7 on ðæm miclan stormum. that rough sea and in those great storms ((c880) CP 9.57.25) ‘An unskilled steersman can very easily navigate well enough when the sea is calm but a learned steersman does not trust him on the rough sea or in violent storms.’ Also common is the type where the enabling conditions stay outside the Agent and can be related to some existing states of affairs, which makes the type a straightforward case of non-deontic possibility. Illustrations follow in (17) through (19): (17) Hi þa ealle glædmode begunnon to ceorfenne þone heagan they then all cheerfully began to cut-down that high pinbeam, and he wæs ahyld on ane healfe þæt man eaðe pine-tree and he was bent on one side that man easily mihte witan hwider he sigan wolde, and hi setton Martinum might know whither he fall would and they set Martin þær foran ongean, þæt he hine offeallan sceolde. there in-front against that he him fall-upon should  ((993–998) ÆLS (Martin) 406) ‘Then they cheerfully began to cut down the high pine-tree and it was bent on one side so that one could easily know where it would fall. They placed Martin opposite the tree so that it would fall upon him.’ In (17) eaþe mihte ‘might easily’ comes in a clause of purpose. The action of cutting down a tree was carried out by a group of idolaters in such a way as to make

 Jerzy Nykiel

it quite possible for them to know where the tree would fall. The enablement of knowing arises outside the subject man ‘one’ of eaðe mihte ‘might easily’, it stems from the way in which the cutting is performed. Along similar lines, the possibility for the Agents to take refuge in the cave in example (18) stems from the location of the cave on the hill, hence this is also a case of non-deontic possibility: (18) Betere we ahreddon us sylfe of ðissere burhware; better we save us self from these townspeople’s gehlyde; faran us into þam mycclan scræfe her geond on clamor go us into that big cave here through on Celian dune, and ðær we magon full eaðe on genere wunian Celius hill and there we may full easily on refuge stay oððæt eft se casere into þissere birig fare; until again that emperor into these people’s town come  ((c 1000) LS 34 (SevenSleepers) 180) ‘We had better save ourselves from the clamor of the townspeople and go through here to the big cave on the hill called Celius ... we can very easily take refuge there until the emperor comes back to their town.’ In (19) we find eaðe mæg ‘may easily’ interpretable as ‘it is objectively possible in this reality (for a human to find out how to kill themselves)’, it is the nature of the reality that brings it within the confines of possibility. (19) Eaðe mæg se mann findan hu he hine sylfne amyrre ac easily may that man find how he him self kill but we sceolan witan þæt nan sylfcwala, þæt is agenslaga, ne we should know that no self-killer that is self-slayer not becymð to Godes rice. comes to God’s kingdom (ÆAbusMor 10) ‘One can easily find out how to kill oneself but we should be aware that those who commit suicide do not go to God’s kingdom.’ If the location of the source of the possibility outside the Agent is responsible for classifying the above examples as expressing non-deontic rather than PI possibility, the fact that the speakers make the judgments based on the generally shared knowledge rather than on their individual beliefs rules out any epistemic interpretation. Eaþe ‘easily’ each time seems to have the effect of enhancing the possibility, yet, crucially, not in the sense that the possibility that an action will take place is increased. Rather, it is the ease and smoothness with which the action can proceed that are highlighted. If the adverb eaþe ‘easily’ is ever mentioned in linguistic literature, it is in connection with the emerging epistemic sense of mæg ‘may’ in Old English. Goossens

Mæg eaþe 



(1982: 78) and Kytö (1991: 150) argue that the first instances of epistemic mæg ‘may’ can be found among those of mæg eaþe. Warner (1993) cautions, however, that the collocation does not guarantee the presence of an epistemic meaning. This research supports this conclusion. Both in Old English and Middle English the epistemic senses of mæg eaþe ‘may easily’ are in the minority, always outnumbered by instances of PI and non-deontic possibility. Obviously, any far reaching assumptions indicating that epistemicity of mæg ‘may’ has a source in mæg eaþe ‘may easily’ have to be abandoned given that epistemic mæg eaþe ‘may easily’ virtually disappears with the end of Old English while may still functions only marginally as an epistemic throughout the Middle English period, as Goossens (1982) claims. Rather, mæg eaþe ‘may easily’ is one of a number of means through which mæg ‘may’ flirts with an epistemic flavor. If eaþe ‘easily’ is actually paired with epistemic mæg ‘may’, it serves to give the pre-modal a stronger epistemic import, that is, to make the possibility higher. The two clearest epistemic examples that I have identified follow in (20) and (21). These two show strongly subjective assessments of possibility by the speaker. The second speaker in (20) gives his opinion about the statement about fate made by the first speaker. Note that the clause with eaþe ... mæge ‘may easily’ is preceded by Ic wene þæt ‘I think that’, one of the earliest epistemic markers in English (see Goossens 1982) and an ancestor of first-person epistemic parentheticals (see Brinton 1996), while in the following clause the verb takes the epistemicity-marking subjunctive form þynce ‘seem’. The collection of the three epistemic signals on the one hand helps dispel any doubts about the epistemic possibility meaning of (20), but on the other hand shows that mæg eaþe ‘may easily’ itself is not a firmly established marker of epistemicity. (20) ða cwæð he: Ic wille secgan þæt ælc wyrd bio good, then said he I will say that each fate is good sam hio monnum good þince, sam hio him yfel þince whether she to-men good seem or she them evil seem ða cwæð ic: Ic wene þæt hit eaðe swa bion mæge, þeah then said I I think that it easily so be may though us hwilum oðer þynce. us at-times otherwise seem ((c880) Bo 40.137.4) ‘Then he said, “I want to say that every fate is good, whether men consider it good or bad.” I said to this, “Methinks it may easily be so, though we may at times deem otherwise.”’ Example (21) shows a piece of a conversation between Antiochus and Achaius, who are not certain whether the man whom they have just met is their friend Eustace or not. An assessment of epistemic possibility based on the evidence available, that is after examining the man, is made by one of them through the use of eaþe mihte ‘may easily’:

 Jerzy Nykiel

(21) Hi þa geornlice hine beheoldon and hine be dæle they then eagerly him examined and him by part oncneowan and cwædon, gelic is þæs man þam men recognized and said similar is this man that man þe wit secað, eaðe he hit mihte beon. Ða cwæð which we-two seek easily he it may be then said se oðer, Ic wat þæt he hæfde ane dolhswaðe on his that other I remember that he had one scar on his hneccan þæt him gelamp iu on gefeohte. neck that him occurred formerly on fight((c1000) LS 8 (Eust) 265) ‘They examined him eagerly, to some extent recognized him and said, “This man is similar to the one we are seeking, it may easily be him.” The other man said, “I remember that he had a scar on his neck which he had once gained in battle.”’ A discussion of epistemic possibility of mæg eaþe ‘may easily’ would not be complete without mentioning the use of the collocation in glosses capturing the meaning of Latin forte ‘by chance, accidentally’ and forsitam ‘perhaps, probably’ in the Lindisfarne and Rushworth Gospels. This is the only context in which mæg eaþe ‘may easily’ is consistently used with an epistemic function and the two collocates are inseparable, invariably followed by a clause. Overall, there are seventeen such instances, two of which are given below as examples (22) and (23): (22) L: iube ergo custodiri sepulchrum usque in diem tertium ne forte ueniant discipuli eius et furentur eum et dicant plebi surrexit a mortuis OE: gehat forðon gehalda vel þætte sie gehalden command therefore hold-secure or that be held-secure

byrgenn oð ðone ðirde doege eoðe mæg þæt hia cyme tomb until that third day easily may that here come



ðegnas his 7 hia forstela hine 7 cuoða ðæm folce disciples his and here steal him and tell those people



arisa vel aras from deadum (...) risen or rose from dead ((960) MtGl (Li) 27.64) ‘Therefore command that the tomb be kept secure until the third day; his disciples may easily go and steal him away, and tell the people, “He has risen from the dead.”’

(23) L: Cum autem uadis cum aduersario tuo ad principem in uia da operam liberari ab illo ne forte trahat te apud iudicem et iudex tradat te exactori et exactor mittat te in carcerem

Mæg eaþe 



OE: miððy uutedlice ðu gast vel gegað mið wiðerworde when indeed thou go or go with adversary

ðinum to aldormen on woeg sel geornlice þætte ðu se thy to magistrate on way give eagerly that thou be



gefreod from him eaðe mæg vel ðylæs genime ðec mið liberated from him easily may or lest take thee against



doema vel gelædæ ðec to dome 7 se doemere seleð. judge or lead thee to judgment and that judge gives

ðec ðæm æfgroefe 7 se æfgroefa sendað ðec in carcern thee that officer and that officer sends thee in prison  ((960) LkGl (Li) 12.58) ‘Thus, when you go with your adversary before a magistrate, on the way make an effort to be liberated from him, or you may easily be taken before the judge, and the judge sends you to the officer, and the officer sends you to prison.’ Interestingly, the Gospels contain ðylæs ‘lest’, side by side with mæg eaþe ‘may easily’, as an optional gloss for ne forte ‘lest by chance, accidentally’ in fourteen cases. This can be seen in example (23). A possible reason for the optional use of ðylæs ‘lest’ is that the epistemic possibility here concerns a proposition whose realization is not desired by the speaker. It is a possibility that something unpleasant will happen, which does not fit in with the epistemic possibility of mæg eaþe ‘may easily’ in the majority of examples where the speaker hopes that the proposition will be realized or is, at least, indifferent. López-Couso (2007: 14) calls lest as used here an “adverbial subordinator [...], introducing clauses of NEGATIVE PURPOSE with the meaning ‘so that ... not, for fear that’”. It should also be noted that throughout the Old English period mæg eaþe ‘may easily’ has a competitor in the form of the collocation mæg eaþelice ‘may easily’. The adverb eaþelice ‘easily’ (ME ethelich) is derived from the adjective eaþelic ‘easy’, itself a derivative of the adverb eaþe ‘easily’ (cf.  BT, s.v. æðelic). According to the DOE, eaþe and eaþelice are synonyms and the latter also enters the collocation with mæg ‘may’. A check of the DOE corpus data indicates that eaþelice ‘easily’ is attested 46 times in the vicinity of mæg, which accounts for 29 per cent of all the appearances of eaþelice in Old English – overall eaþelice, with different spelling variants included, has 160 attestations.3 3. It can be added that four Middle English instances of mai ethelich ‘may easily’ can be found in the MED, the last one dating back to the year 1398. Nevertheless, given the scarcity of the examples and the fact that no new semantic or syntactic developments are attested, no mention is made of the collocation in Section 5, which is devoted to Middle English.

 Jerzy Nykiel

Mæg eaþe and mæg eaþelice ‘may easily’ occur in very similar contexts to the extent that in the area of PI possibility they seem virtually interchangeable. In example (24) Hwi hingrað ðe: gif ðu godes sunu sy. wend þas why hungers-it you if thou God’s son be change these stanas to hlafum. and et; Eaðe mihte God se ðe awende stones to loaves and eat easily might God that that changed wæter to wine. and se ðe ealle gesceafta of nahte water to wine and that that all creatures from nothing geworhte: eaðelice he mihte awendan þa stanas to hlafum: made easily he might change those stones to loaves ac he nolde nan ðing doon be ðæs deofles tæcunge; but he would-not no thing do by that devil’s order  ((990–995) ÆCHom I, 11 267.47) ‘‘‘Why are you starving? If you are the Son of God, turn these stones to bread and eat.” God could easily, he who turned water into wine and the one who created all creatures out of nothing, he could easily have turned the stones to bread but he would do nothing by the Devil’s direction;’ Ælfric uses eaðelice he mihte as a stylistic variant of eaðe mihte God, probably in order to avoid repetition of the latter. Mæg eaþelice ‘may easily’, not unlike mæg eaþe has a leaning toward contexts in which a PI possibility reading follows from the designation of a powerful deity as the subject and the focus on his capacity to act. The same context was addressed with reference to examples (14) and (15) above. Instances of non-deontic possibility with mæg eaþelice ‘may easily’ also occur. Example (25) is a case in point (25) Heo (...) sæde, þæt heo wolde gesecan þone ealdormann and she said that she would seek that ruler and hine gewissian to his agenum willan, hu he eaðelice mihte him instruct to his own will how he easily might þæt manncynn berædan butan frecednysse his that nation take-by-treachery without danger his agenes folces. own people’s ((ca 1000) ÆHomM 15 239) ‘She (...) said that she wanted to seek the ruler and show him, with his permission, how he could easily take that nation by treachery without any danger for his own people.’ in that the possibility to take the nation by treachery is shown to originate outside the subject/Agent.

Mæg eaþe 



As regards an epistemic function, mæg eaþelice ‘may easily’ is not used in that way. Mæg eaþelice ‘may easily’ may never really extend to epistemic possibility as such but it develops a use which produces an inference of epistemicity. In Bybee’s (1988) view, root (non-deontic) possibility of may invites an inference of epistemic possibility when the modal has a wide scope, that is, when the Agent is not expressed or is generic. This is exactly the case with example (26): (26) Ne sy þis on nanum gamete gedyrstleht fram not be this on no measure undertaken from mynecenehades manna, forði of þam mæg swiðe monastic-state’s men because from that may very mycel saca yðelice arise. great scandal easily arise ((950–1050) BenRW 69.141.23) ‘Do not let it (a monk defending another monk) be undertaken in any measure by men of the monastic state, because grave scandal can easily arise from that.’ In summary, mæg eaþe ‘may easily’ shows a variety of uses in Old English. It goes a long way toward expressing three types of possibility, namely PI possibility, non-deontic possibility and epistemic possibility. In the context of PI possibility eaþe ‘easily’ tends to be employed as a VP-oriented adjunct, which invites an inference of clausal orientation. When non-deontic possibility is involved, the inference is stronger while the epistemic examples only allow an interpretation of eaþe ‘easily’ as a clause-oriented adjunct, that is an emphasizer of epistemic mæg ‘may’. The adverb eaþe ‘easily’ itself develops the meaning of epistemic possibility in the process. The epistemic function of mæg eaþe ‘may easily’ may be attested rarely but it is fairly uncontroversial to the extent that this collocation is selected to render the epistemic meaning of Latin forte ‘by chance, accidentally’ and forsitam ‘perhaps, probably’ in the Old English gospels. 5. Semantics of mæg eaþe ‘may easily’ in Middle English Over the course of Middle English instances of mæg eaþe ‘may easily’ are rare, with the Helsinki Corpus offering none and the MED adducing twenty seven. It is the areas of PI and non-deontic possibility that still exercise a strong hold over the collocation as illustrated in examples (27) and (28) respectively. (27) þou art mi man, & al mi trest is þe vpan. thou are my man and all my trust is thee upon

 Jerzy Nykiel

Nou þou migt wel eþe Arede me fram þe deþe. now thou may well easily save me from the death  ((c1330) Floris (Auch)) ‘You are my man, and all my trust is in you. Now you can easily save me from death.’ (28) Here byfore he myghte ethe [Cmb: eyth] Sone hafe mad me here before he might easily soon have made me asethe. compensate ((c1440) Degrev. (Thrn) 489) ‘Heretofore he could easily have made me compensate.’ It is, however, significantly more difficult to stumble upon a case of epistemic possibility. The best instance I have found is that shown in (29): (29) Hie iseð bineðen hem deflen þe hem gredeliche kepeð. they see beneath them devils who them greedily keep and beð swiðe of grisen; and ful eaðe mugen. and are very afraid and full easily may  ((a1225(?a1200)) Trin.Hom. (Trin-C B.14.25) 173) ‘They will see beneath them devils, who will greedily seize them, and they will be greatly terrified and very easily may be.’ and still it is controversial as it blends an epistemic reading with a non-deontic possibility reading. It is not entirely clear whether and ful eaðe mugen ‘and very easily may’ indicates possibility resulting from the disposition of the circumstances or an epistemic judgment. Since the two interpretations do not exclude each other, the example can be taken to be a case of merger as formulated by Coates (1983: 16). With neither option being ruled out, the sentence is still comprehensible. It is noteworthy, however, that eaþe ‘easily’ itself is not instrumental whatsoever in differentiating between a non-epistemic and epistemic meaning of mæg ‘may’. What is more, given that among the Middle English instances of mæg eaþe ‘may easily’ one does not find any clearer epistemic examples than (29) and even this sentence comes from a Middle English version of Old English homilies, it must be concluded that the most innovative usage of eaþe mæg ‘may easily’ in Old English, that is epistemic, is lost first in Middle English, just in time for the collocation to fall out of use by the end of Middle English. As was mentioned in Section 2, in Middle English eaþe ‘easily’ is gradually replaced with esili ‘easily’ in the collocation with mæg ‘may’. The new collocation makes its first written appearance in 1384 and has only nine attestations in the Middle English texts found in the Helsinki Corpus and MED. None of these attestations

Mæg eaþe 



conveys an epistemic judgment, which suggests that the lexical replacement blocks the development of the epistemic use of the collocation. Obviously, this assessment can only be tentative given the overall low number of occurrences of eaþe ‘easily’ and esili ‘easily’ collocated with mæg ‘may’ in Middle English. 6. To what extent is it lexicalization? As Hoye (1997: 164ff.) points out, some harmonic combinations of modal verbs and modal adverbs in Present-Day English, e.g. may easily or may well, develop an idiomatic status. In this section I look into traces of lexicalization present already in Old and Middle English in the case of mæg eaþe ‘may easily’. While offering a typology of lexicalization, Himmelmann (2004: 27) points to idiomatization and univerbation as kinds of lexicalization applicable to collocations. As the two elements in mæg eaþe ‘may easily’ remain morphologically free and they are never univerbated, the type of lexicalization involved here would be idiomatization. Features associated with idiomatization are usually loss of compositional meaning, loss of productivity and loss of substitutability, as noted by Brinton & Traugott (2005: 55). There is no decrease in productivity and substitutability in the case of mæg eaþe ‘may easily’. The collocation occurs in all persons and tenses and the adverb eaþe ‘easily’ can assume comparative (example (30)) and superlative forms (example (31)) when paired with mæg ‘may’: (30) þe þe Godes wille [?deð], ðe eiðer he mai that which God’s will does the more-easily he may him finde him find ((?c1250) PMor. (Eg 613(1)) 88) ‘the one who acts according to God’s will can more easily find him.’ (31) þa wæs nergendes þeowen þrymful, þearle gemyndig hu then was savior’s servant strong very mindful how heo þone atolan eaðost mihte ealdre benæman. she that terrible-one most-easily might of-life deprive ær se unsyfra, womfull, onwoce before that impure evil awake (Jud 73) ‘[The lady], the strong servant of the Savior, was very mindful of how she could most easily deprive the most terrible one of life, before that wicked and evil man awoke.’ This is a clear indication that the adverb does not lose productivity, it sits well in the class of gradable adverbs and behaves like one. What is more, it is not uncommon that eaþe ‘easily’, when operating in mæg eaþe ‘may easily’, is modified both

 Jerzy Nykiel

in Old and Middle English. In Old English the modifiers are adverbs such as swa ‘so’, swiðe ‘very’, ful(l) ‘very’, and naht ‘not’ while the set of Middle English modifiers comprises ful ‘very’ and wel ‘well’. All these forms with the exception of naht ‘not’ can be classified, following Hoye (1997: 169ff), as amplifiers, or, following Paradis (2008:  321) as boosters. They have the effect of enhancing the quality which they modify. When paired with eaþe ‘easily’, naht ‘not’, through occurrence in negative clauses, is a downtoner, or a diminisher in Paradis’s (2008: 321) terminology, that is, it weakens the meaning of eaþe ‘easily’, as in example (32): (32) uah and racha sind ebreisce interiectiones, and ælc uah and racha are Hebrew interjections and each þeod hæfð synderlice interiectiones, ac hi ne magon language has particular interjections but they not may naht eaðe to oðrum gereorde beon awende. not easily to other languages be translated((c1000) ÆGram 279.18) ‘uah and rach are Hebrew interjections, and every language has special interjections but they cannot be easily translated into other languages.’ Mæg eaþe ‘may easily’ does not show any deficiency with regard to substitutability, either. The synonymous adverb eaþelice ‘easily’ can be used in lieu of eaþe ‘easily’, the order of the two collocates can alternate between eaþe mæg and mæg eaþe, and the two collocates can be separated. As for decomposition of meaning, the process does apply when OE mæg eaþe ‘may easily’ is used epistemically. In the company of epistemic mæg ‘may’, eaþe ‘easily’ loses its original semantics in favor of the inference of possibility or likelihood derived from the use of mæg ‘may’. In such examples eaþe ‘easily’ comes to indicate possibility, or it simply emphasizes the possibility of mæg ‘may’. This can be seen as decomposition of meaning. Concurrently, this assessment of lexicalization fails to reflect the extent of lexicalization pertaining to the use of mæg eaþe ‘may easily’ in the glosses in the Lindisfarne and Rushworth Gospels, as illustrated in examples (22) and (23). In this case loss of both productivity and substitutability takes place in as far as this use of mæg eaþe ‘may easily’ requires the fixed order of the collocates and does not tolerate different forms of the verb or modification of the adverb. Obviously, the fact that such consistency can be found only in the glosses makes one approach these occurrences of mæg eaþe ‘may easily’ with caution. On the other hand, the authors of the glosses selected this collocation, and not any other, to render epistemic possibility, which may indicate that mæg eaþe ‘may easily’ was much more lexicalized in everyday communication than what we can see in the extant Old English texts.

Mæg eaþe 



7. Conclusion With this study, I hope to have shown the evolution path of the collocation mæg eaþe ‘may easily’ in Old and Middle English. Occurrences of the collocation are already widespread in early Old English and the data available make it impossible to trace the exact stages through which mæg eaþe ‘may easily’ comes into being. I suggested that the juxtaposition of the pre-modal verb mæg ‘may’ and the adverb eaþe ‘easily’ results from reinterpretation of the VP-oriented adverb eaþe ‘easily’ as a clause-oriented adverb. The reinterpretation of the adverb has an inferential basis. So does the development of the epistemic possibility meaning of eaþe ‘easily’, which enables the adverb to emphasize the epistemic possibility of mæg ‘may’. In this way mæg ‘may’ and eaþe ‘easily’, already harmonic in the area of PI and non-deontic possibility, become harmonic when it comes to epistemic possibility as well. The only meaning of mæg ‘may’ not conveyed by mæg eaþe ‘may easily’ is permission. The evolution of mæg eaþe ‘may easily’ in Middle English slows down and comes to a halt by the end of the period as eaþe ‘easily’ is gradually replaced with esili ‘easily’. The overall popularity of mæg eaþe ‘may easily’ decreases and clearly epistemic examples are nowhere to be found. Especially in the Old English data there are clues pointing to some idiomatization of mæg eaþe ‘may easily’: the adverb undergoes decomposition of meaning when paired with epistemic mæg ‘may’. At the same time, however, there can be no denying that we have to do with a lexicalization process at a very early stage as mæg ‘may’ and eaþe ‘easily’ behave like a productive pre-modal and a productive adverb respectively and there is no loss of substitutability. Mæg eaþe ‘may easily’ as used in the glosses to capture the meaning of Latin forte ‘by chance, accidentally’ and forsitam ‘perhaps, probably’ in the Old English Gospels has been singled out, as idiomatization in this case also involves loss of productivity and loss of substitutability. Dictionaries Bosworth, Joseph & T. Northcote Toller, eds. = BT. 1898. An Anglo-Saxon Dictionary. 1921. Supplement by T. Northcote Toller. 1972. Enlarged Addenda and Corrigenda by Alistair Campbell. Oxford: Clarendon Press. Cameron, A., A. Crandell Amos, Antonette diPaolo Healey, J. Holland, D. McDougall, I. McDougall, N. Speirs & P.Thompson, eds. = DOE. 2003. The Dictionary of Old English on CD-ROM. The University of Toronto. Available on-line at: Kurath, Hans & S. Kuhn, eds. = MED. 1956–2001. Middle English Dictionary. Ann Arbor: University of Michigan Press. Simpson, J. A., E. Simon & C. Weiner, eds. = OED. 1989. Oxford English Dictionary. Second Edition. Oxford: Clarendon Press.

 Jerzy Nykiel

References Brinton, Laurel J. 1996. Pragmatic Markers in English. Grammaticalization and Discourse Functions. Berlin & New York: Mouton de Gruyter. Brinton, Laurel J. & Elizabeth C. Traugott. 2005. Lexicalization and Language Change. Cambridge: Cambridge University Press. Bybee, Joan L. 1988. “Semantic Substance vs. Contrast in the Development of Grammatical Meaning.” Proceedings of the Annual Meeting of the Berkeley Linguistics Society 14.247–264. Bybee, Joan L., Revere Perkins & William Pagliuca. 1994. The Evolution of Grammar: Tense, Aspect and Modality in the Languages of the World. Chicago & London: The University of Chicago Press. Coates, Jennifer. 1983. The Semantics of the Modal Auxiliaries. London & Sydney & Dover & New Hampshire: Croom Helm. Fischer, Olga. 2007. Morphosyntactic Change. Oxford: Oxford University Press. Goossens, Louis. 1982. “On the Development of the Modals and of the Epistemic Function in English.” Papers from the 5th International Conference on Historical Linguistics ed. by Anders Ahlqvist, 74–84. Amsterdam & Philadelphia: John Benjamins. Goossens, Louis. 1987. “Modal Tracks: The Case of magan and motan.” Studies in Honour of Rene Derolez ed. by Anne-Marie Simon-Vandenbergen, 216–236. Gent: Vitgeuer. Hanson, Kristin. 1987. “On Subjectivity and the History of Epistemic Expression in English.” Chicago Linguistic Society 23.133–147. Himmelmann, Nikolaus P. 2004. “Lexicalization and Grammaticalization: Opposite or Orthogonal?” What Makes Grammaticalization? ed. by Walter Bisang, Nikolaus P. Himmelmann & Björn Wiemer, 21–42. Berlin & New York: Mouton de Gruyter. Hoye, Leo. 1997. Adverbs and Modality in English. London & New York: Longman. Huddleston, Rodney & Geoffrey Pullum. 2002. The Cambridge Grammar of the English Language. Cambridge: Cambridge University Press. Kotin, Michail L. 2008. “Aspects of a Reconstruction of Form and Function of Modal Verbs in Germanic and Other Languages.” Modality-Aspect Interfaces ed. by Werner Abraham & Elisabeth Leiss, 371–384. Amsterdam & Philadelphia: John Benjamins. Kytö, Merja. 1991. Variation and Diachrony, with Early American English in Focus. Frankfurt am Main, Bern, New York & Paris: Peter Lang. Lass, Roger. 1994. Old English: A Historical Linguistic Companion. Cambridge: Cambridge University Press. Lightfoot, David. 1979. Principles of Diachronic Syntax. Cambridge: Cambridge University Press. Lightfoot, David. 1991. How to Set Parameters: Arguments from Language Change. Cambridge, Mass.: MIT Press. López-Couso, María José. 2007. “Adverbial Connectives within and beyond Adverbial Subordination.” Connectives in the History of English ed. by Ursula Lenker & Anneli MeurmanSolin, 11–29. Amsterdam & Philadelphia: John Benjamins. Lyons, John. 1977. Semantics. Cambridge: Cambridge University Press. Molencki, Rafał. 2003. “What Must Needs Be Said About must needs.” English Modality in Context. Diachronic Perspectives ed. by David Hart, 71–88. Bern: Peter Lang. Nuyts, Jan. 2001. Epistemic Modality, Language, and Conceptualization. Amsterdam & Philadelpha: John Benjamins.



Mæg eaþe  Nykiel, Jerzy. 2010. Expressing Obligation in Old English. Katowice: Wydawnictwo Uniwersytetu Śląskiego. Paradis, Carita. 2008. “Configurations, Construals and Change: Expressions of DEGREE.” English Language and Linguistics 12:2.317–343. Prokosch, Eduard. 1939. A Comparative Germanic Grammar. Philadelphia: Linguistic Society of America. Shibasaki, Reijirou. 2004. “Patterns of Semantic Harmonization in English: The Case of may well.” Proceedings of the 29th Annual Meeting of the Berkeley Linguistics Society ed. by Pawel M. Nowak, Corey Yoquelet & David Mortensen, 391–402. Berkeley: Berkeley Linguistics Society. Shibasaki, Reijirou. 2009. “Another Look at the Development of Epistemic Meanings: A Historical Collocational Approach.” Studies in Modern English 25.63–84. Tellier, André. 1962. Les Verbes Perfecto-présents et les Auxiliaires de Mode en Anglais Ancien. Paris: Klincksieck. Traugott, Elizabeth Closs. 1992. “Syntax.” The Cambridge History of the English Language. Vol. 1 ed. by Richard M. Hogg, 168–286. Cambridge: Cambridge University Press. Traugott, Elizabeth Closs. 1995. “The Role of the Development of Pragmatic Markers in a Theory of Grammaticalization”. Paper Presented at the Twelfth International Conference on Historical Linguistics, Manchester. Traugott, Elizabeth Closs & Richard B. Dasher. 2005. Regularity in Semantic Change. Cambridge: Cambridge University Press. van der Auwera, Johan & Vladimir Plungian. 1998. “Modality’s Semantic Map.” Linguistic Typology 2.79–124. Warner, Anthony R. 1993. English Auxiliaries: Structure and History (= Cambridge Studies in Linguistics, 66). Cambridge: Cambridge University Press. Wärnsby, Anna. 2006. (De)coding Modality. The Case of Must, May, Måste and Kan. Lund: Lund University Press.

Current change in the modal system of English A case study of must, have to and have got to Joanne Close & Bas Aarts*

University of Leeds & University College London This paper takes the variation between must, have to and have got to as a window through which to view changes in the modal system in Present-Day British English (1960s–1990s). The results from this study show a dramatic decrease in frequency of the core modal must and a significant increase in frequency of the semi-modal have to in the Diachronic Corpus of Present-Day Spoken English (DCPSE). Changes in the modal system affect both epistemic and root uses of must, although have to is only an active rival to root must; epistemic instances of have to (and have got to) are rare in the corpus. We suggest that a tendency to avoid expressions of strong commitment is the cause of the decline in must. This is supported by similar falls in the frequency of should and ought found by Leech et al. (2009).

1. Introduction The class of English modal verbs have undergone a number of important changes in the history of English, and continue to change in Present-Day English (PDE). Studies on current change suggest that the modals are decreasing in frequency, becoming monosemous, and are facing possible competition from rival semimodals which are increasing in frequency (see Krug 2000; Leech 2003; Smith 2003; Mair & Leech 2006; Mair 2006; Leech et al. 2009). This paper investigates the variation between the core modal must and the semi-modals have to and have got to in the Diachronic Corpus of Present-Day

* This research was carried out as part of the project The changing verb phrase in PresentDay British English at the Survey of English Usage, UCL, funded by the UK Arts and Humanities Research Council (grant number AH/E006299/1). We gratefully acknowledge the AHRC’s support.

 Joanne Close & Bas Aarts

Spoken English (see Section 3.1).1 We use the label “semi-modal” for have to and have got to in order to highlight that these forms share some of the properties common to must, a member of the class of core modals.2 The questions we address in this paper are: (i) is the core modal must decreasing in frequency? (ii) is there evidence for “competition” between must and the semi-modals have to and have got to? (iii) is must becoming monosemous?, and (iv) what is the motivation for the change(s)? The variation between the three forms is a window onto the changing modal system and, while the results cannot necessarily be generalized to other modals, the study will offer an insight into areas of the modal system susceptible to change and identify possible motivating factors behind these changes. This paper is organized as follows: Section 2 reviews some of the existing work on current change in the modal system; Section 3 presents the corpus and describes the data collection method, including decisions taken on which data to exclude from the current study; Section 4 discusses the main findings and investigates possible motivations for the changes that have occurred; Section 5 is the conclusion. 2. Current change in the modal system of English 2.1

Observing current change

Throughout the history of English, the modal verbs have undergone a number of significant changes, resulting in a class of verbs characterized by a number of syntactic properties which are not shared by main verbs (see Traugott 1989; Lightfoot 1991; Warner 1993). More recently, there has been a growing interest in current change in the modal system of English (Krug 2000; Leech 2003; Smith 2003; Mair 2006; Leech et al. 2009), as well as more generally. It is hoped that research into this phenomenon will lead to a greater understanding of the trigger(s) and pathway(s) of language change. We might expect instances of current change to be subtle in comparison to changes measured over a longer period. While recent changes in the language 1. We do not include in this study the reduced form gotta or the form got to not preceded by an auxiliary. We exclude these from this paper because the phonologically reduced gotta does not appear in the corpus and got to occurs only once. 2. For Quirk et al. (1985: 141–146) have to is a modal semi-auxiliary, while have got to is a modal idiom. These combinations are not identical in their syntactic behaviour: have to takes do-support, while inversion is not possible for most speakers. It can occur as a non-finite form and be preceded by an auxiliary verb. By contrast, have got (to) is always finite, and therefore cannot be preceded by auxiliary verbs. It can invert with a subject.



Current change in the modal system of English 

may not appear as dramatic as some of the earlier changes such as the levelling of the English inflectional system from Old to Middle English, the shift from OV to VO or the appearance of do‑support, they could be equally significant. There exist instances of historical language changes which involve a change in the underlying grammar of the language (see Kroch 1989; Lightfoot 1979; Warner 1993; Pintzuk 1995, among others). Current change has been linked to grammar change only very recently in this way. According to Leech (2003: 223), “if we understand such [grammatical] changes to include changes of frequency, significant grammatical changes do take place within a generation.” The nature of current change is such that it is more difficult to observe than a change which has occurred over a longer period of time, e.g. several centuries. In the study of a long-term change, the end-point of the change has typically been reached, and it is often possible to track the change through the period in which it occurred. A change measured over a shorter period of thirty years or so, however, will not necessarily be completed in the time period, and it is impossible to know how, when, or even if a particular change will complete. The danger of attempting to predict these is illustrated by the subjunctive which was believed to be decreasing to the point of extinction (Fowler 1965), but has been shown more recently to be undergoing a revival (Övergaard 1995; Hundt 1998; Leech et al. 2009). It is, of course, possible to track a change as it occurs, but it will often be necessary to observe the phenomenon at a later period (see Bauer 2002; Mair 2008). 2.2

Core modal frequency in Present-Day English

One of the observed changes in the modal system of Present-Day English is a decrease in frequency of the core modals (see Krug 2000; Leech 2003; Smith 2003; Leech et al. 2009). Leech (2003) investigates modal usage in the Brown quartet of corpora. The term ‘Brown quartet’ refers to four comparable corpora: the Brown Corpus and Frown (Freiburg-Brown) Corpus from 1961 and 1991–2, respectively, and two matching corpora of British English, LOB (LancasterOslo/Bergen) and FLOB (Freiburg-Lancaster-Oslo/Bergen), again from 1961 and 1991–2, respectively. These corpora each contain a million words in fifteen written text types. He observes that the decrease in modal usage is apparent in 13 out of 15 categories ranging from a decline of 31.2% (D, Religion) to a decline of 0.5% (H, Miscellaneous). In the remaining two categories, there is a small and non-significant increase; in J (Learned) of 2.4% and L (Mystery Fiction) of 8.4%. Furthermore, the British and American corpora show a decrease over time (from 1961 to 1991) in the case of every modal (individual modals differ in frequency), except can and could which show an increase of 2.2% and 2.4%, respectively, in British English. Leech’s results also show that the infrequent modals

 Joanne Close & Bas Aarts

shall, ought to, and (in British English) need have decreased drastically, and a similar decline has taken place in the mid-frequency modals may and must, while would and should have decreased less drastically. Finally, will, can, could and might have not significantly changed. The results presented by Leech et al. (2009) also confirm that the modals which show the steepest decline are the less common modals. 2.3

Core modals vs. semi-modals

Mair & Leech (2006) point out that although many changes involve a competing construction or constructions, there is not necessarily a correlation between the form(s) losing ground and those gaining. On discovering a decrease in the frequency of the core modals in the Brown family of corpora, Leech (2003:  229) suggests that perhaps the semi-modals are “gradually usurping” the functions of the core modals, but concludes that this is not the case. The reason for his conclusion is that, although semi-modal usage is increasing overall, some semi-modals are actually declining (for example be to, (had) better, (have) got to and be going to in FLOB), and most semi-modals are much less frequent than the core modals (for example, will is about ten times more frequent than be going to). These results are supported by Krug (2000), and also by Smith (2003: 249) who states that “the rise of [have to] ... by no means makes up for the shortfall in must.” Mair (2006: 105) also investigates modal frequency using the Santa Barbara Corpus of Spoken American English and the spoken component of the British component of the International Corpus of English (ICE-GB). His results indicate that have to is the most common form in both varieties, followed by have got to and must in British English and need to and must in American English. On the possibility of have to taking over the functions of must, he comments that “epistemic must is secure in spoken British English, as have to in this function has a marginal status at best” (Mair 2006: 105). It seems, then, that if there is competition between must and have to/have got to this is in the expression of obligation or necessity, i.e. root meaning. In their investigation of modality in British dialects, Tagliamonte & Smith (2006) find exactly this; their findings indicate that “[m]ust is obsolescent and there is an unanticipated resurgence of have to alongside pan-dialectal grammatical reorganization” (Tagliamonte & Smith 2006: 341). 2.4

Monosemy of modals

Leech (2003) carries out a semantic analysis of should and every third instance of may and must found in the Brown corpora to discover whether a decline in



Current change in the modal system of English 

frequency can be attributed to one particular sense of a modal. He provides some evidence that the decline in frequency of the modals is linked to the tendency for modals to become “more monosemous”, but points out that this is not always the case. In fact, a different pattern can be observed for each of the three modals investigated. Must, the modal we are interested in here, shows a decline in both root (event modality; obligation) and epistemic (logical necessity) senses between the 1960s and the 1990s, particularly in the SEU-mini and ICE-GB-mini corpora of spoken English,3 but the root sense has shown the largest decrease in frequency.4 Leech warns that the results of the study as a whole are tentative because of uncertainties in sampling and coding procedures. It is also important to bear in mind that comparing the results of semantic analyses across studies is difficult, due to the challenging nature of the task of semantic coding. 3. Corpus and data collection 3.1

Current change and the Diachronic Corpus of Present-Day Spoken English

There are a number of “pitfalls of anecdotal observation” as far as current change is concerned, most notably “[t]he spread of salient new uses is exaggerated, while the less salient persistence of older forms is not noted or [...] a diachronic trend is read into a situation which merely shows variable or fluctuating usage” (Mair 2008: 1111). For this reason, studies on current change are carried out more reliably using corpora. One of the strengths of the methodology we have adopted is that it uses a corpus of spoken English. This is particularly important for the study of the semimodals as “these forms, being typically colloquial, are not likely to show up in their true colours in the written language” (Leech 2003: 230). The Diachronic Corpus of Present-Day Spoken English (DCPSE) contains 421,362 words from the British component of the International Corpus of English (ICE-GB) collected in the early 1990s and 464,074 words from the London-Lund 3. The SEU-mini corpus used by Leech (2003) contains 80,000 words (16 texts) from conversation, BBC discussions and news, sports and other commentaries, broadcast talks, etc., all dated between 1959–65. The ICE-GB-mini corpus contains 80,000 words from text categories chosen to match those of the SEU-mini corpus, the only difference being the date: texts are dated 1990–92. 4. Leech’s (2003) results show that the epistemic reading of may has increased in the written corpora, but all other meanings have shown a decline. With should the root sense of “weak obligation” is shown to be gaining at the expense of other senses, particularly the epistemic sense of “weak inference.”

 Joanne Close & Bas Aarts

Corpus (LLC) collected between the late 1960s and early 1980s (for more information about the LLC and ICE-GB see Svartvik (1990) and Nelson et al. (2002), respectively). The period of time between LLC and ICE-GB is within the range of what is considered ideal for studying current change. In designing the Brown family of corpora a decision was made to have an interval of approximately thirty years between the Brown and LOB corpora (1960s) and the Frown and FLOB corpora (1990s) because “the interval of one generation or so [...] is usually considered the minimum sufficient to clearly identify and document linguistic change in real time” (Mair 1998: 140). With over 800,000 words of spoken English, DCPSE is the largest available single collection of parsed and checked orthographically transcribed spoken material. All of the sentences in DCPSE have been grammatically analysed and have been given detailed parse trees which contain functional information (subject, direct object, noun phrase head, etc.), categorial information (part-of-speech, phrase, clause, etc.), and optional additional features (intransitive (verb), common (noun), main (clause), etc.). The corpus can be explored by using the International Corpus of English Corpus Utility Program (ICECUP) software developed at the Survey of English Usage. ICECUP offers a facility called Fuzzy Tree Fragments (FTFs; see Aarts et al. 1998 and Nelson et al. 2002).5 An example FTF which searches for all verb phrases (VPs) in the corpus is shown in Figure 1.1. This FTF is a partial tree diagram, which can be constructed by users. FTFs are “fuzzy” in the sense that users can specify for themselves the level of detail that they are interested in. For example, if the user adds “intr” to the lower section of the FTF in Figure 1.1, the software will search for all intransitive verb phrases in the corpus. This is shown in Figure 1.2. More complex FTFs can contain grammatical structures, words, lexical wild cards, etc. In the context of the present research, the FTF in Figure 2 searches the corpus for the string got to with the position for have left unspecified (indicated by the symbol ‘¤’). This FTF will find cases of ’ve/’d got to as well as have/has/had VP

VP intr

Figure 1.1.  An FTF which instructs the search software to find all the verb phrases in the corpus

Figure 1.2.  An FTF which instructs the search software to find all the intransitive verb phrases in the corpus

5. See also www.ucl.ac.uk/english-usage/resources/ftfs for further information on conducting experiments using FTFs.



Current change in the modal system of English 

VB

VP

AUX

AUX

AUX

¤

got

to

Figure 2.  FTF for (have) got to

got to, and even cases where an adverb intervenes between have and got.6 This is because in Figure 2 the arrows between the nodes are set so that something may intervene between the nodes. This is an example of user choice. Users can change this setting if they wish to specify that the nodes should follow each other immediately. 3.2

Data

In order to ensure reliability of results, it is important to identify contexts in which the forms being investigated may alternate. In the case of have to, have got to, and must there are a number of syntactic environments in which there is only one option, and these have been excluded from the study (see Tagliamonte & Smith 2006; Depraetere & Verhulst 2008). All exclusions are listed below, and are not included in the statistical tests in Section 4. First, negated forms were excluded because the meanings are not equivalent:7

(1) a. You mustn’t go (the modal is outside the scope of the negation: ‘It is necessary for you not to go’). b. You don’t have to go (the modal is inside the scope of the negation: ‘It is not necessary for you to go’). 6. This FTF will only retrieve instances of got to which are preceded by an auxiliary verb (always have); cases of got to (gotta) will not be captured with this FTF. Cases in which got is a main verb followed by the preposition to would also not be retrieved using this FTF. 7. Depraetere and Verhulst (2008: 16) also exclude negative uses of must and have to based on the observation that “the scope of negation is different for have to (absence of necessity, ‘not necessarily’) and must (prohibition, ‘necessarily not’).”

 Joanne Close & Bas Aarts

c. You haven’t got to go (the modal is inside the scope of the negation: ‘It is not necessary for you to go’). Secondly, interrogative contexts were excluded because interrogative forms with must did not occur at all in the later period, and only one example occurred in the early period: 

(2) Must Wordsworth speak for us in his intimations of immortality? (DCPSE:DL-J01 #0129: 2:B)

Finally, also excluded were infinitival have to (including future forms) because equivalent forms of must or have got to do not exist; and the past form had to because must cannot refer to past time, so would not be an option here. It is important to point out that tokens of must have been or must have V-ed have not been excluded. Although there are no tokens of have (got) to have been or have (got) to have V-ed in the corpus, a search of the web using WebCorp illustrates that both of these forms are possible.8 An example of have to have been is given in (3):

3.3

(3) Just my tuppence worth, but 2007 has to have been the absolute worst X Factor ever.9 Semantic coding

A number of claims have been made in the literature about the frequency of the modals and semi-modals relating to their meaning(s). For instance, Coates (1983) states that have to and have got to are infrequent as epistemic modals and Leech (2003) claims that must is one of a few modals not becoming monosemous. In order to test these claims using DCPSE, each instance of must, have got to and have to was manually classified according to the type of modality that they express. Modal meaning has been dealt with in a number of ways. The present study uses the coding system advocated by Leech & Coates (1980) and Coates (1983), in which modals are divided into two semantic classes, root and epistemic. In this approach modal meaning is dealt with in a gradience model where each class has both core and peripheral members. This can account for the fuzziness in modal meaning without the need for postulating additional classes. 8. There are two plausible reasons why these forms are not present in DCPSE: (i) they are very infrequent, or (ii) they are a post 1992 innovation. It is impossible to determine which, if either, of these is correct. 9. Example retrieved using WebCorp on 22/09/08. Web address: http://www.webcorp.org. u k /cgi‑bin/v iew.nm?url=http://w w w.unrea lit y t v.co.u k /x‑factor/t he‑worst‑x‑factor‑ever/&term=has%20to%20have%20been&case=case.



Current change in the modal system of English 

According to Palmer (2001: 9) “with deontic modality the conditioning factors are external to the relevant individual, whereas with dynamic modality they are internal”. As outlined below, we do not classify the source of the obligation for the instances of must, have to and have got to. For this reason, we prefer the term root over deontic, as this does not imply a contrast with dynamic. Our approach to semantic coding is described in more detail in Sections 3.3.1–3.3.4. 3.3.1 Root modality Coates argues that root modality is fuzzy with a cline from strong obligation (‘It is imperative/obligatory that x’) to weak obligation (‘It is important that x’), although the basic meaning of root must is ‘It is necessary for ...’. Examples of root meaning are given in (4): (4) a. My boss had said you must read those books.  (DCPSE:DI-B12 #0171:1:E) b. I’ve got to go, I’m afraid, in an hour. (DCPSE:DI-B37 #0216:2:B) c. Actually you have to be much more honest about what you’re doing.  (DCPSE:DI-A02 #0124:2:B) 3.3.2 Epistemic modality Epistemic must “conveys the speaker’s confidence in the truth of what he is saying, based on a logical process of deduction from facts known to him (which may or may not be specified)” (Coates 1983: 41). Coates divides epistemic must into core (confident inference) and peripheral (logical necessity) meanings: i. confident inference: ‘I confidently infer that x’ ii. logical necessity: ‘In light of what is known, it is necessarily the case that x’ Epistemic examples of must, have got to and have to are given in (5) below: (5) a. I am coming to conclude that it must be because I get too much water in it or something. (DCPSE:DI-B47 #0166:1:B) b. Loose shirts over jeans has got to be a sort of temporary prejudice.  (DCPSE:DI-B44 #0161:1:B) c. I mean good guidebooks really are full of impressions, aren’t they, whereas imaginative works or works of fiction often do blend facts in a distorted form. If you think of Hardy’s use of landscape as a curious blend of the identifiable, which people who research in that find interesting, and something which has to be an impression there, wouldn’t you agree? (DCPSE:DL-A06 #0310:1:E)

 Joanne Close & Bas Aarts

3.3.3 Performative modality Performative or speech act modals occur where the speaker is carrying out the action denoted by the verb. These occur frequently with verbs like admit, say, confess, as shown in (6). (6) a. There’s a piece here called Spring Fire which takes my interest, I must say. (DCPSE:DI-D12 #0141:1:A) b. When you switched to the emphasis being on Architecture <,,> did you initially think that you wanted to go into that as a career or were you doing it just as a degree because you enjoyed the subject  (DCPSE:DI-A07 #0093:1:A) Pretty much yeah I’ve got to admit. (DCPSE:DI-A07 #0094:1:B) c. I have to confess an often irking thought of am I really really two pounds less than Kate Hamilton. (DCPSE:DI-B07 #0224:2:B)



Although performative modality is considered to be a type of root modality, these examples were coded and counted separately because of their specialized meaning.10 3.3.4 Ambiguity Coates (1983: 47) asserts that “there is no overlap between the two fuzzy sets representing Root and Epistemic must” and “[c]ases where it is not possible to decide which meaning is intended are therefore ambiguous”, as illustrated in (7) taken from Coates (1983: 47, her example (35)):11

(7) And anyway, I think mental health is a very relative thing. I mean, mental health must be related to the sort of general mentality, or whatever other word you use of the community you’re living in. (S.2.14.73)

The root meaning here is ‘it’s essential that mental health is related to ...’, while the epistemic reading is ‘it’s inevitably the case that mental health is related to ...’. It might be expected that ambiguous cases would be relatively frequent. However, from the Lancaster and Survey corpora, Coates classes only 11 examples out of a total of 436 as ambiguous.12 In the ambiguous cases “syntactic features which could disambiguate such [ambiguous] utterances are either not present or are themselves ambiguous” (Coates 1983: 47). 10. Tagliamonte & Smith (2006) exclude these cases from their data based on the fact that they occurred exclusively with must, but in DCPSE have got to and have to can have performative meaning. 11. Coates’s original mark-up has been removed from the example as it is not relevant here. 12. This could be viewed as indirect support for the coding system chosen by Coates; we might expect that the better the coding system is, the fewer indeterminate examples there will be.



Current change in the modal system of English 

In coding the data from DCPSE, it was assumed that the speaker had an intended meaning when producing the utterance, and an attempt was made to classify each example, keeping ambiguous cases to a minimum. Despite this, there exist a number of examples which are impossible to disambiguate for two main reasons: (i) the utterance is compatible with both root and epistemic meanings and/or (ii) the modal is followed by ellipsis or unclear words. The numbers for ambiguous cases are shown in Table 2. 4. Results and discussion 4.1

Overall frequencies

In order to examine a possible trend for declining core modal usage and to investigate the theory of competition between core modals and semi-modals, the overall frequency of core modal must is compared with the frequencies of the semimodal forms in DCPSE. As shown in Table 1, during the thirty-year period the frequency of must has declined by over 55% while the frequency of have to has significantly increased by nearly 32%. Table 1.  Overall frequencies of must, have got to and have to in DCPSE (figures in bold are significant at p < 0.01) (Semi-) modal

LLC frequency raw per 100,000 words

must have got to have to Total

427 187 188 802

  92.01   40.30   40.51 172.82

ICE-GB frequency Change in frequency raw per 100,000 % χ2 score words 172 156 225 553

  40.82   37.02   53.40 131.24

–55.64   –8.14 +31.82 –24.06

36.29   3.10 31.94 71.32

Table 2.  Distribution of semantic types of must in DCPSE Source corpus LLC ICE-GB Total

Epistemic N % 43.96 21.12 65.08

47.78 51.74 48.99

Root N

%

36.63 16.14 52.77

39.81 39.53 39.73

Performative N %

Ambiguous N %

Total N

  9.48   2.37 11.85

1.94 1.19 3.13

  92.01   40.82 132.83

10.30   5.81 8.92

2.11 2.91 2.35

 Joanne Close & Bas Aarts

Somewhat surprisingly, have got to decreases in frequency by 8%. It is impossible for us to be sure whether have got to has reached its peak and is showing a decline in the 1990s data prior to “levelling out” in the future, or whether the decrease is simply fluctuating usage in the corpus. To be certain, it will be necessary to measure the frequency of have got to at a point later than 1992. Unfortunately, the lack of available corpora makes this impossible at this point in time. 4.2

Root and epistemic modality

One of the aims of this study is to determine whether must is becoming monosemous. Using the semantic coding carried out on the data (see Section 3.3 above for details), the frequencies of epistemic, root and performative uses were calculated. The results are presented in Table 2. To discover if the decline in root and/or epistemic must is related to the use of the semi-modals, root and epistemic uses of the three forms were compared. The results for root and epistemic uses are presented in Tables 3 and 4, respectively. Table 3 illustrates that as expressions of root obligation, must, have got to and have to combined have decreased by 12.98%. Again, the decrease in must and the rise in have to are found to be significant, and have got to shows a decline. Table 3.  Frequencies of root must, have got to and have to in DCPSE (figures in bold are significant at p < 0.01) (Semi-) modal

must have got to have to Total

LLC frequency raw per 100,000 words 174 184 185 543

37.49 39.65 39.86 117

ICE-GB frequency Change in frequency raw per 100,000 % χ2 score words   70 151 208 429

  16.61   35.84   49.36 101.81

–55.69   –9.61 +23.83 –12.98

23.61   0.12 12.32 36.05

Results for epistemic use in Table 4 show that, although epistemic must has decreased by over 50% must is the still the preferred form for expressing epistemic necessity; epistemic have got to and have to are very rare in DCPSE.13 The general view in the literature is that the decline of the core modals cannot be attributed directly to an increase in the semi-modals (Krug 2000; Leech 2003; 13. The frequencies for epistemic have to and have got to are low which may lead to questions about the reliability of the chi-square test here, but log likelihood calculations confirm low significance scores: must 0.19, have got to 3.05, have to 1.67 (total 4.92).



Current change in the modal system of English 

Table 4.  Frequencies of epistemic must, have got to and have to in DCPSE (Semi-) modal

LLC frequency raw per 100,000 words

must have got to have to Total

206    2    2 210

44.39   0.43   0.43 45.25

ICE-GB frequency Change in frequency raw per 100,000 % χ2 score words 91  4  3 98

21.60   0.95   0.71 23.26

  –51.34 +120.93   +65.12   –48.60

0.19 3.36 1.83 5.38

Smith 2003; Mair & Leech 2006). This conclusion was reached by studying frequencies in written language where, on the whole, core modals are more frequent than semi-modals. The situation with respect to the spoken data is quite different. In root contexts (see Table 3), have to is already more frequent than must in the 1960s (frequencies are 39.86 per 100,000 words for have to and 37.49 for must). By the 1990s, the frequency of have to is almost three times that of must (the frequencies are 16.61 and 49.36, respectively). This suggests that there may be a link between core modal decline and semi-modal increase. 4.3

Motivating factors

The frequency shift in the expression of root modality suggests a possible correlation between the decline of must and the rise of have to. The question, of course, is what is the motivation for the shift away from must and towards have to in root meaning? A number of suggestions to explain the decline in must have been made in the literature, many of which refer to the “authoritarian” nature of must. Myhill (1995: 202), for instance, attributes the decline in must (= ‘obligation’) and the rise of should (= ‘weak obligation’) around the time of the American Civil War to a growing tendency to avoid overt claims to authority by the speaker/writer. He calls this “democratisation”. He claims that the “old” modals were linked with people controlling the actions of other people while the “new” modals are more personal, and are used to give advice to an equal, and the like. Smith (2003) shares a similar view, and makes the following remark: It seems probable that must is a casualty of a changing society where increasing emphasis is being placed on equality of power, or at least the appearance of equality of power, and the informality of discourse found in private conversation is becoming more acceptable, even usual, in official types of discourse. Just as these conditions are likely to disfavour the use of must, they should correspondingly favour other forms which express obligation less directly (Smith 2003: 259).

 Joanne Close & Bas Aarts

The problem with the suppositions by Myhill and Smith, however, is that epistemic must has also decreased, and this verb is not related to power and authority. Leech et al. (2009: 88) suggest that the “partial decline [of epistemic must] could be due to contamination by the dramatic fall of deontic must”. This is perhaps a questionable explanation because the root and epistemic meanings of must do not intersect, so it is not clear how a fall in one would cause a fall in the other (cf. Coates’s (1983: 170) treatment of the meanings of will which do intersect). In their data, root must declines more drastically than epistemic must, which Leech et al. suggest is because “neither have to nor any other form has become widely adopted as an alternative expression of strong epistemic necessity”. In the spoken data from DCPSE, however, the fall of epistemic must is almost equal to that of root must, although it is true that there is no rise in epistemic have to or have got to. It is, of course, possible that there are independent explanations for the fall of root must on the one hand, and epistemic must on the other. However, we suggest that the cause is the same. The data from DCPSE cannot tell us which occurred first, the fall of root must or the rise of root have to. The figures for epistemic must suggest the former, otherwise it is difficult to explain why epistemic must has declined, but epistemic have (got) to has not increased. As an explanation for the decrease in must, we appeal to what Huddleston & Pullum et al. (2002: 175) refer to as “strength of commitment” where “necessity involves a strong commitment”. We suggest that the decline in must is a result of a decline in forms expressing strong commitment. In the root sense, have to is favoured over must. The difference between the two forms is illustrated in (8) where the conjoined clause is not possible with must, because the commitment is stronger and does not permit resistance (example from Sweetser (1988: 54), quoted in Hopper & Traugott (1993: 79)):

(8) I have to/??must get this paper in, but I guess I’ll go to the movies instead.

This explanation can be extended to epistemic must because this also expresses necessity (and hence involves strong commitment), although, as pointed out in Section 4.2, have to is not a rival to epistemic must. If there is a move away from forms which express a stronger commitment, then we would also expect other forms which express necessity to show a decrease. This is in fact the case: should and ought to, both classified by Coates (1983) as modals of necessity and obligation, show a decrease in frequency in DCPSE and in written corpora (see Leech (2003) and Leech et al. (2009) for results from the Brown family of corpora).



Current change in the modal system of English 

5. Conclusion This study has investigated variation between the core modal must and the semimodals have to and have got to using the Diachronic Corpus of Present-Day Spoken English. By considering only syntactic environments in which the forms are variable, any possible competition between must and the “rival” semi-modal alternates can be reliably observed. The results from the study show that even in the 1960s data, the frequency of root have to has surpassed that of root must. The continued decline of must and the rise in have to in the 1990s data suggests that have to is being used in contexts where previously we would have found must. Surprisingly, have got to, the most recent marker of root modality, has decreased. This does not appear to be a reflection of the DCPSE sampling, as Tagliamonte (2006) and Tagliamonte & Smith (2006) found that this was also the case in a number of English dialects. The semantic coding carried out on must, have to and have got to in DCPSE has led to a number of interesting results. Firstly, although both epistemic and root senses of must have declined, as proportions of the total number of instances of must they have remained fairly constant. This supports Leech’s (2003) claim that must is not becoming monosemous. Further study of the semantic senses of other modals in DCPSE is necessary in order to determine if this fact is peculiar to must. Secondly, have to is the most frequent marker of root modality, which is remarkable because we might expect the newer form have got to to be more frequent. And finally, the use of epistemic must has decreased, but the use of epistemic have to has not increased. This highlights the importance of using corpora to investigate language change, as simple “anecdotal observation” might have led to a prediction that this form is on the increase. References Aarts, Bas & Sean A. Wallis. 2006. The Diachronic Corpus of Present-Day Spoken English (DCPSE). [Software] Aarts, Bas, Gerald Nelson & Sean A. Wallis. 1998. “Using Fuzzy Tree Fragments to Explore English Grammar”. English Today 14:3.52–56. Bauer, Laurie. 2002. “Hitting a Moving Target: An Interim Report on Language Change across Generations”. English Today 18:4.55–59. Biber, Douglas, Susan Conrad & Randi Reppen. 1998. Corpus Linguistics: Investigating Language Structure and Use. Cambridge: Cambridge University Press. Coates, Jennifer. 1983. The Semantics of the Modal Auxiliaries. London: Croom Helm. Depraetere, Ilse & An Verhulst. 2008. “Source of Modality: A Reassessment”. English Language and Linguistics 12:1.1–25.

 Joanne Close & Bas Aarts Fowler, Henry Watson. 1965. A Dictionary of Modern English Usage. Second edition. Revised by Ernest Gowers. Oxford: Clarendon Press. Hopper, Paul J. & Elizabeth Closs Traugott. 1993. Grammaticalization. (Cambridge textbooks in linguistics). Cambridge: Cambridge University Press. Huddleston, Rodney & Geoffrey K. Pullum. 2002. The Cambridge Grammar of the English Language. Cambridge: Cambridge University Press. Hundt, Marianne. 1998. “It is Important that this Study (should) be Based on the Analysis of Parallel Corpora: On the Use of the Mandative Subjunctive in four Major Varieties of English.” The Major Varieties of English (Papers from MAVEN 97) ed. by Hans Lindquist, Staffan Klintborg, Magnus Levin & Maria Estling, 159–175. Växjö: Acta Wexionensia. Kroch, Anthony S. 1989. “Reflexes of Grammar in Patterns of Language Change”. Language Variation and Change 1.199–244. Krug, Manfred G. 1998. “Gotta – The tenth Central Modal in English? Social, Stylistic and Regional Variation in the British National Corpus as Evidence of Ongoing Grammaticalization”. The Major Varieties of English (Papers from MAVEN 97) ed. by Hans Lindquist, Staffan Klintborg, Magnus Levin & Maria Estling, 177–191. Växjö: Acta Wexionensia. Krug, Manfred G. 2000. Emerging English Modals: A Corpus-based Study of Grammaticalization. Berlin & New York: Mouton de Gruyter. Leech, Geoffrey. 2003. “Modality on the Move: The English Modal Auxiliaries 1961 1992.” Modality in Contemporary English ed. by Roberta Facchinetti, Manfred Krug & Frank Palmer, 223–240. Berlin & New York: Mouton de Gruyter. Leech, Geoffrey & Jennifer Coates. 1980. “Semantic Indeterminacy and the Modals.” Studies in English Linguistics ed. by Sidney Greenbaum, Geoffrey Leech & Jan Svartvik, 79–90. London: Longman. Leech, Geoffrey, Marianne Hundt, Christian Mair, & Nicholas Smith. 2009. Change in Contemporary English: A Grammatical Study. Cambridge: Cambridge University Press. Lightfoot, David W. 1979. Principles of Diachronic Syntax. Cambridge: Cambridge University Press. Mair, Christian. 1997. “Parallel Corpora: A Real-time Approach to the Study of Language Change in Progress.” Corpus-based Studies in English. Papers from the seventeenth International Conference on English Language Research on Computerized Corpora (ICAME 17), Stockholm, May 15–19 1996, ed. by Magnus Ljung, 195–209. Amsterdam: Rodopi. Mair, Christian. 1998.  “Corpora and the Study of the Major Varieties of English: Issues and Results.” The Major Varieties of English (Papers from MAVEN 97) ed. by Hans Lindquist, Staffan Klintborg, Magnus Levin & Maria Estling, 139–157. Växjö: Acta Wexionensia. Mair, Christian. 2006. Twentieth-Century English: History, Variation and Standardization. (= Studies in English Language) Cambridge: Cambridge University Press. Mair, Christian. 2008. “Corpora and the Study of Recent Change”. Corpus Linguistics: An International Handbook ed. by Anke Lüdeling & Merja Kytö, 1109–1125. Berlin & New York: Mouton de Gruyter. Mair, Christian & Geoffrey Leech. 2006. “Current Changes in English Syntax”. The Handbook of English Linguistics ed. by Bas Aarts & April McMahon, 318–342. Oxford: Blackwell. Myhill, John. 1995. “Change and continuity in the functions of theAmerican English modals.” Linguistics 33.157–211. Nelson, Gerald, Sean Wallis & Bas Aarts. 2002. Exploring Natural Language: Working with the British Component of the International Corpus of English. Amsterdam & Philadelphia: John Benjamins.



Current change in the modal system of English  Övergaard, Gerd. 1995. The Mandative Subjunctive in American and British English in the 20th Century (= Studia Anglistica Upsaliensia, 94). Uppsala: Acta Universitatis Upsaliensis. Stockholm: Almqvist and Wiksell International. Palmer, Frank R. 2001. Mood and Modality. Second edition. Cambridge: Cambridge University Press. Pintzuk, Susan. 1995. “Variation and Change in Old English Clause Structure.” Language Variation and Change 7.229–260. Smith, Nicholas. 2003. “Changes in Modals and Semi-modals of Strong Obligation and Epistemic Necessity in Recent British English”. Modality in Contemporary English ed. by Roberta Facchinetti, Manfred Krug & Frank Palmer, 241–266. Berlin & New York: Mouton de Gruyter. Svartvik, Jan, ed. 1990. The London-Lund Corpus of Spoken English: Description and Research (= Lund Studies in English, 82). Lund: Lund University Press. Sweetser, Eve E. 1988. “Grammaticalization and Semantic Bleaching”. Berkeley Linguistics Society: General Session and Parasession on Grammaticalization ed. by Shelley Axmaker, Annie Jaisser & Helen Singmaster. Tagliamonte, Sali. 2006. “Historical Change in Synchronic Perspective: The Legacy of British Dialects”. The Handbook of the History of English ed. by Ans van Kemenade & Bettelou Los, 477–506. Oxford: Blackwell. Tagliamonte, Sali & Jennifer Smith. 2006. “Layering, Competition and a Twist of Fate: Deontic Modality in Dialects of English”. Diachronica 23:2.341–380. Traugott, Elizabeth Closs. 1989. “On the Rise of Epistemic Meanings in English: An Example of Subjectification in Semantic Change”. Language 66:1.31–55. Wallis, Sean & Gerald Nelson. 2000. “Exploiting Fuzzy Tree Fragments in the Investigation of Parsed Corpora”. Literary and Linguistic Computing 15:3.339–361. Warner, Anthony R. 1993. English Auxiliaries: Structure and History. (= Cambridge Studies in Linguistics, 66). Cambridge: Cambridge University Press.

part iii

Developments in the English noun phrase

Discontinuous quantificational structures in Old English Artur Bartnik*

John Paul II Catholic University of Lublin The paper deals with discontinuous quantificational structures in Old English. It concludes that quantificational constructions in Old English are heterogeneous because they can result from movement, that is separation of the quantifier and the rest of the phrase, or base-generation when the separated nominal elements are generated in places in which they occur. Essentially, structures which display full morphological agreement result from movement. The movement types included in the paper are topicalization, scrambling, extraposition and NP (Noun Phrase) shift. By contrast, base-generated constructions do not show morphological agreement. Quantifiers in such structures are adverbial in nature.

1. Introduction This paper deals with discontinuous quantificational structures in Old English. The term “quantificational” suggests that the construction involves a quantifier, such as ealle ‘all’, and the word “discontinuous” indicates that the quantifier that qualifies the noun or pronoun is not adjacent to it. A classic example of such a structure is the so-called Quantifier Floating (QF) structure, well documented in the syntactic literature (Sportiche 1988; Bobaljik 1995; cf. also Benmamoun 1999; Bošković 2004). An example of QF in Old English is given in (1):

* I am deeply indebted to three anonymous reviewers and the editors whose insightful and critical remarks helped me avoid some mistakes. I would also like to thank the participants of the 15th International Conference on English Historical Linguistics (Munich, Germany, August 2008) for their constructive comments and advice. My heartfelt thanks are also due to Prof. Adam Pasicki, who kindly commented on an earlier draft of this paper and helped me with the historical data. However, all the people mentioned above are not to blame for the remaining errors.

 Artur Bartnik

(1) Þonne læweþ broþer oþerne hæþnum on deaþ & Then will-betray brother another heathen-dat to death and sunu se læweþ his fæder & þa gingran arisaþ son he will-betray his father and the youngers will-rise wiþ þam yldrum & hie mid deaþe geswencaþ & against the elders, and them to death will-torture and hie beoþ on feounge ealle for minum naman they-nom will-be in hatred all-nom because my-dat name-dat  (coblick, LS_32_[PeterandPaul[BlHom_15]]:171.17.2167)1 ‘Then the brother will betray the brother to the heathen to death, and the son will betray his father, and the younger ones will rise against the elders, and will torture them to death; and they will all be in hatred because of my name’ Although the pronoun hie is separated from ealle, both these elements must form a constituent at some point of the syntactic derivation because they agree morphologically. Old English offers a larger spectrum of floated quantificational forms than Modern English, where besides all, only both and each can be found. Consider the following examples, where quantifiers such as ælc ‘each’, æghwylc ‘each one’ or sum ‘some’ seem to float: (2) and hi beoð geendebyrde ælc be his geearnungum, and and they are set-in-order each by his merit and eac gewuldrode on þam micelan wurðscipe, be ðam ðe hi also glorified in the great worship according to they on life lufodon heora Scyppend in life loved their Lord (coaelhom, ÆHom_11:545.1771) ‘and each of them is set in order by his merit and glorified in the great worship, according to how they loved their Lord in life’ (3) Nu com tid & cymð þæt ge tofaron æghwylc to Now has come time, and comes that you scatter, each to his agenon & forlæton me anne his own and leave me alone (cowsgosp, Jn_[WSCp]:16.32.7121) ‘there comes the time, and (indeed) it has come, when you will be scattered, each (one) to his own, and you will leave me alone’

1. All the examples taken from The York-Toronto-Helsinki Parsed Corpus of Old English Prose (YCOE) are quoted by the title and page number of the electronic text. This is followed by information containing the original coding conventions used in YCOE.



Discontinuous quantificational structures in Old English 

(4) Ac hit is sum swa gerad þæt his nis nan nedþearf, but it is some so arranged, that his is-not no necessary, & þeah ne dereð no þeah hit geweorðe and yet not hurt no if it should-happen  (coboeth, Bo:41.144.17.2880) ‘but some of it is so arranged that it is not necessary and yet it does not hurt if it should happen’ (5) Hi habbað sume synderlice gyfe fram heora scyppende They have some separate gift from their Lord & þeahhwæþere heora wurðscipe him bið eallum gemæne and yet their worship them is all common  (cocathom1, ÆCHom_I,_24:376.169.4768) ‘and some of them have a separate gift from their Lord, and yet their worship is shared by all of them’ In the relevant literature, such examples are either derived by movement2 (cf. Sportiche 1988, for example), that is separation of the bold elements, or they are base-generated in their surface positions. In the former case, the pronoun raises, stranding the quantifier (Q) down in the clause. In the latter case, both the constituents are generated in the positions in which they occur. On this view, quantifiers resemble adverbials, as evidenced by the data from other languages (cf. Bobaljik 1995 and example (17) below). Such divergent approaches point to heterogeneity of quantifier properties crosslinguistically. In what follows I will take a closer look at other combinations of discontinuous quantifier structures in Old English and further support the hypothesis about heterogeneous properties of these elements. I will show that the split in the derivation of QF in other languages is also reflected in the discontinuous structures in Old English. In particular, I will argue that some of these structures are the result of movement (separation) of relevant elements when they display overt or covert morphological agreement, i.e. the presence or absence of overt inflectional markings. Other separated constructions where there is no full morphological agreement between the split elements are base-generated. In such cases the quantifiers resemble adverbials. The paper is organized as follows. In Section 2, I will analyse quantificational structures that involve movement. It will be seen that there is a great syntactic variety in these sequences. In Section 3, I will examine constructions that cannot 2. In this paper we assume that agreement is subject to locality. In other words, the agreeing quantifiers and their nominal associates must form a constituent at some point of the syntactic derivation.

 Artur Bartnik

involve movement despite the fact that they are separated. Section 4 briefly considers other structures that are base-generated. Section 5 concludes the article. 2. Movement structures Apart from classic examples of QF illustrated in (1) to (5), Old English offers more structures that involve discontinuous quantifiers. The first type of discontinuous quantificational structures is the structure in which the floated element (the separated element) is higher than the personal pronoun in the subject position. Pronominal subjects are important because they mark the left periphery for the subject position.3 Therefore, anything higher than the personal pronoun must be fronted (Haeberli 2000; van Kemenade 2002; Pintzuk & Traugott 2005; cf. also Fischer et al. 2000). This fronted element can be a noun, a pronoun or the quantifier itself. In the literature, this type of structure is labelled topicalization. This is illustrated in (6) to (8): (6) Wið ealle wundela genim þas wyrte zamalentition wel Against all wounds take the plant zamalentition well mid rysle gecnucude butan sealte, lege to ðam wundum. with fat pounded without salt, lay to the wound. Ealle heo hy gehæleþ All-acc she them-acc cures (coherbar, Lch_I_[Herb]:167.1.2435) ‘For all wounds, take the zamalentition plant powdered thoroughly in unsalted lard, put (it) to the wound. She [the plant] cures all of them [wounds]’ (7)

Đas niwan spel ic þe ealle in cartan This-acc new-acc story-acc I you all-acc in document awrite (coalex, Alex:6.4.26)4 write ‘All this new story I will write for you in a document’

(8) Ac þonne þa fif þing swa we ær cwædon eall But when these five things, as we before said, all gegadorede bioð þonne bið hit eall an þing & þæt an collected are, then is it all one thing and that one 3. Linguists working on adverbs in the Old English clause generally agree that a particular class of “high” adverbs split the subject position into two slots: the lower slot is occupied by nominal subjects while pronominal subjects reside in the higher slot of the subject position (cf. Kemenade 2002; Haeberli 2000). 4. The Latin equivalent reads has nouas explicaturas historias omnia cartis commendabo.



Discontinuous quantificational structures in Old English 

ðing bið God & he bið anfeald untodæled þeah he thing is God and he is single undivided though they

ær on mænig tonemned wære ða andsworede ic & before in many, named were then answered I and cwæð: þisses ic eom ealles geþafa said this-gen I am all-gen supporter (coboeth, Bo:33.76.10.1422) ‘But when these five things, as we before said, are all collected together, then is it all one thing, and that one thing is God; and he is single and undivided, though they before were in many, separately named. Then I answered and said, all of this I approve’

The reason for this split might be textual cohesion in the above examples: the topicalized elements appear to be thematic as they contain familiar information in a given context. The second type of discontinuous quantificational construction is the structure in which the floated element is lower than the subject personal pronoun. Again, this element can be a noun, a pronoun or a quantifier. In the literature, this type of structure is dubbed scrambling. This is exemplified in (9) to (11): (9) Wolde swa cepan þære cristenra lare, forðan þe heo Wished so to-keep the Christians’ teaching because she næfde on ðære byrig nænne geleaffulne mann þe did-not-have in the town not-any faithful man who hi læren cuþe, forðan Philippus aflygde þa her teach could because Philip expelled the-acc-pl cristenan of Alexandrian ealle on ær. Christian-acc-pl from Alexandria all-acc-pl before  (coaelive, ÆLS_[Eugenia]:30.208) ‘so [she] wished to keep the teaching of Christians because she did not have in the town any faithful person who could teach her because Philip had expelled all the Christian people from Alexandria before’ (10) mid his micclan cyste he forgeaf us gemænelice with his great generosity he gave us-dat generally eallum, ricum & heanum, þe heora hentan magon all-dat rich-dat and poor-dat who them catch can  (coaelhom, ÆHom_1:204.120) ‘with his great generosity/goodness, he gave [that] in general to us all, both to the rich and to the poor, who can catch/pursue them [animals]’

 Artur Bartnik

(11) Hy ahlyttriað þone stut of heora liðe mid seohhann and they purify the gnat of their drink with strainer and hy ealne forswelgað þone olfend gehalne. they all-acc swallow the-acc camel-acc whole-acc  (coaelhom, ÆHom_14:162.2091)5 ‘they use a sieve to strain a gnat out of their drink, but they swallow up the whole camel’ Examples (6) to (11) show a striking distributional similarity: topicalized and scrambled noun phrases/pronouns float away from quantifiers (examples (7)–(8) and (9)–(10), respectively), and, conversely, quantifiers float away from pronouns/noun phrases (examples (6) and (11)). The only difference lies in the landing site of these elements. Topicalized phrases are higher than subject personal pronouns, while scrambling requires separated constituents to land lower than personal pronouns. Moreover, both types of movement can appear in the same sentence. Consider: (12) & his suna swa ilce ælcne hi hæfdon for god and his sons-gen likewise each-acc they had for Gods  (coboeth, Bo:38.115.27.2300) ‘and likewise each of his sons they held for Gods’ In sentence (12), the constituent ælcne his suna is topicalized as it lands higher than the personal pronoun hi. Then the NP his suna scrambles over the adverb swa ilce leaving the quantifier below. Notice also that examples (6) to (11) must involve movement as the separated constituents display morphological agreement: ambiguous, as in (7), where quantifiers are syncretic forms or unambiguous, as in (11), where they display the lack of inflectional syncretism. Example (12) does not involve agreement between the two bold constituents because his suna is plural while ælcne is clearly singular. Still, it supports the movement analysis since ælcne selects the genitive represented by suna. Thus, the resulting split constructions must be the effect of separation. The reason for this split in scrambled structures might be syntactic. Some linguists claim that scrambling is triggered by some kind of specificity features. Other scholars argue that scrambling is needed to license appositive relations (cf. Bošković & Takahashi 1998; Vikner 1994; Hinterhölzl 2006). There are also less clear-cut structures with separated quantifiers that can be subsumed neither under topicalization nor scrambling. Quantifiers can modify relative clauses (example 13) or nominal phrases modified by relative clauses (example 14). Consider: 5. The Latin version reads Liquantes culicem et glutientes camelum.



Discontinuous quantificational structures in Old English 

(13) Nis þis na gesæd be manna sawlum, ac be Not-is this not said about men-gen souls-dat but about manna lichaman þe formolsniað to duste, and eft men-gen bodies which crumble to dust and again sceolan on domes dæg ðurh ures Drihtnes mihte shall on doom-gen day through our-gen Lord-gen power ealle of eorðan arisan þe æfre cuce wæron, swa swa all-nom from earth arise, who ever alive were just as ealle treowa cuciað æfre on Lenctenes timan all trees come-alive always in Lent-gen time  (coaelive, ÆLS[Ash_Wed]:27.2717) ‘this is not said about men’s souls but about men’s bodies which crumble to dust and again shall on doomsday through our Lord’s power all from earth arise, all [those] who ever were alive, just as all trees always come alive in the spring time’ (14) þæt seo ea, seo þe floweþ be þam burhwealle, that the river which flows around the city-wall, forhergiende wære his land, & oft gedonre yþunge se ravaging were his land and often caused fluctuation the stream mid his rynum & mid his uppgange gewunode þæt stream with its flow and with its surge continued that he togoten wæs geond his æceras, & þa it spread was through its fields and the-acc geplantedan wingeardes, þa þe he gemette, ealle he planted-acc vineyards-acc which it met all-acc it þa oferfleow then overflowed (cogregdC, GDPref_and_3_[C]:9.192.14.2448) ‘[one of the wonders that was regarded as the most famous by all those who dwelt there was] that the river which flows around the city walls, ravaged his land and often in an inundation the stream with its flow and its surge would overflow his fields and all the planted vineyards which it met – all (that) it overflowed’ In example (13), the quantifier is separated from the relative clause, a heavy element that is extraposed to the end of the clause. Relative clauses do not have to be extraposed: as shown by (14), they can appear at the front edge of the clause. Technically, ealle and the nominal phrase modified by the relative clause are not separated in (14). However, since quantifiers normally precede nominal phrases,

 Artur Bartnik

the structure might involve movement (cf. Benmamoun 1999 on such constructions in Arabic). An anonymous reviewer remarks that quantifiers in (13) and (14) could be read as adverbs modifying verbs in view of example (15) below, where the adverbial eall(e) also modifies the participle: (15) Æfter þysum worde he wearð eall gehæled After these words he was all cured  (coaelive, ÆLS_[Sebastian]:299.1391) ‘after these words he was all (=completely) cured’ Note, however, that eall(e) in (13) and (14) is construed with the plural nominal phrase: in (13) ealle refers back to manna lichaman ‘men’s bodies’ while in (14) the quantifier modifies þa geplantedan wingeardes ‘the planted vineyards’. This is quite distinct from (15), where eall is preceded by the singular subject he. Hence the adverbial meaning is more likely (cf. Bobaljik 1995, who introduces so-called completive all meaning ‘entirely, wholly’ with its singular associate). Let us finally consider example (16): (16) Hit gelamp þa sona æfter litlum fyrstum, þæt se halga It happened then soon after some time that the holy deacon gewat of worlde to Gode mid sigefæstum deacon departed from world to God with victorious martirdome, oferswiðdum deofle, to þam ecan wuldre mid martyrdom, overcoming devil, to the eternal glory with þam welwillendan Drihtene, þæt þæt he þam behet the benevolent Lord that that he them-dat promised eallum þe hine lufiað and þam þe his geleafan all-dat who him love and them-dat who his faith healdaþ oð ende hold until end (coaelive, ÆLS_[Vincent]:226.7946) ‘it happened then soon after some time that the holy deacon departed from the world to God in victorious martyrdom, overcoming the devil, to the eternal glory with the benevolent Lord which he promised all those who love him and those who stay faithful to him until the end’ In (16), eallum seems to refer to both þam þe hine lufiað and to þam þe his geleafan healdaþ oð ende. Because the italicized structure is a subordinate clause, I assume an underlying OV order that shifts the quantifier part of the heavy NP over the verb. This is just followed by the topicalization of the first þam. To sum up this section, I have shown that apart from regular QF structures presented in Section 1, discontinuous quantificational structures can be accounted for by



Discontinuous quantificational structures in Old English 

movement, either topicalization or scrambling or both. In such examples, the separated constituents display morphological agreement. I have also considered other structures with separated quantifiers that involve movement. In such cases, quantifiers usually modify relative clauses or nominal phrases modified by relative clauses. 3. Base-generated constructions This section shows discontinuous quantificational structures that cannot result from movement. One of the requirements of separation by movement is morphological agreement in the separated phrases. Thus in (11), both the quantifier and the rest of the noun phrase are unambiguously specified for accusative, masculine, singular. However, in some cases, the separated elements do not agree with each other at all. Consider (17), where eall agrees neither with the plural pronoun nor with the italicized Prepositional Phrase: (17) Ge furðum þæt ic cweðe sie sio soðe gesælð þe And also that I say is the true happiness which an ðissa fifa mæg fullice forgifan, forðæm þe on any these-gen five-gen can fully bestow because in ælcum anum hi sint eall each-dat one-dat they are all (coboeth, Bo:33.78.13.1467) ‘and moreover, I say that is the true happiness which any of these five can fully bestow: because in each one they all are’ Lack of agreement is indicative of adverbial nature of quantifiers. For example, some adverbial (uninflected) quantifiers such as eall in (18), which modifies the participle, do not display agreement (example (15) is repeated here as (18) for convenience). The translation shows that the adverbial reading of eall is possible here. Since adverbs are base-generated it seems that adverbial quantifiers should be base-generated as well. (18) Æfter þysum worde he wearð eall gehæled After these words he was all cured  (coaelive, ÆLS_[Sebastian]:299.1391) ‘after these words he was all (=completely) cured’ Examples such as (18) show that some occurrences of eall are paradigmatically the same as adverbs in that they occupy the same positions. Interestingly, quantifiers such as eall can be syntagmatically related to some adverbs. For example, in (19) below, the uninflected eall, which can be rendered as ‘exactly’, is syntactically related to swa. Although eall and swa are separated, the movement analysis is not possible because no agreement obtains between the elements in question:

 Artur Bartnik

(19) ða he cild wæs, eall hine man fedde swa man oðre when he child was all him man fed as man other cild fedeð. children feeds (cowulf, WHom_6:164.347) ‘when he was a child he was fed exactly as other children are fed’ In sum, it seems that the lack of inflection is crucial for determining the adverbial nature of quantifiers, though inflected quantifiers can be adverbial when they replace adverbials. 4. Remnant constructions In this section, I will present two other cases of discontinuous quantificational structures, in which the movement analysis is less appealing. First, consider the following sentence: (20) And feale þe ic hæfde to me atogen, bisne and healte, And many who I had to me drawn, blind and lame, lame and reoflen, ealle he heo fram me ateah. weak and leprous, all he them from me drew  (conicodC, Nic_[C]:235.234) ‘and many of those whom I had drawn to me, blind and lame, weak and leprous, all he them drew away from me’ In example (20), there appears an obligatory pronoun heo that repeats the information expressed by the associate (in our case, italicized) noun phrase. Such recapitulatory or resumptive pronouns make the movement analysis problematic, since the apparent gap left by movement of the complex nominal phrase is refilled by the pronoun, which is banned in syntax. In effect, the quantifier is base-generated rather than moved to the left in example (20) (cf. Benmamoun 1999 for a similar analysis in Arabic). Second, some discontinuous quantificational structures superficially resemble so-called split genitives. Consider examples (21) and (22): (21) & þæs oþres eorles broþor Ohteres and the-gen other-gen earl-gen brother-gen Ohter-gen  (ASC(A) 914.13) ‘and the brother of Ohter, the other earl’ (22) & Theodosius hæfde þone wind mid him þæt his fultum and Theodosius had the wind with him that his help mehte mæstra ælcne heora flana on hiora could most-gen each-acc their-gen arrows-gen on their



Discontinuous quantificational structures in Old English 

feondum afæstnian. enemies fix (coorosiu, Or_6:36.155.7.3293)6 ‘Theodosius had the wind with him so that his help could fix most of their arrows on their enemies’ Allen (2002: 58) argues that discontinuous possessive structures such as (21) are split around the head noun. She marshals some evidence that the head noun broþor does not have to case-mark the post-head position of the split genitive, that is Ohteres. Instead, she suggests that this position is an adjunct within the noun phrase, although she does not attempt a more detailed analysis of these constructions. In other words, Allen analyses split genitives as appositives, that is base-generated elements. Our example is superficially similar to the split genitive structure: the intervening head in the accusative is flanked by the split genitive. However, there are two differences between (21) and (22). The first obvious difference lies in the presence of the quantificational element in example (22) and its absence in the split genitive construction. Second, Ohteres being an adjunct is an optional element in (21) and it is not case-marked by the head. By contrast, heora flana in (22) is obligatory and, consequently, seems to be a complement case-marked by the quantificational head. This contrast, however, does not appear to affect the analysis in both cases, which favours base-generation. Hence, no movement is involved. In sum, this section provides further evidence for base-generated discontinuous structures with quantifiers. In particular, resumptive constructions, and split genitival phrases indicate that quantifiers, though separated from their associates occur in their surface positions in the clause. 5. Conclusion In this article, I discussed discontinuous quantificational structures in Old English. The catalogue of structures in Sections 2 to 4 shows that quantifier structures are not a homogeneous group. More precisely, there are structures clearly derived by movement (Section 2). The constituents in such structures can be topicalized, when they occur above the personal pronoun, or scrambled, when they land below the personal pronoun. Strikingly, the elements affected by these two movement operations are the same, namely quantifiers, noun phrases and pronouns. Other constructions that involve movement from the quantifier include relative clauses, preposed or extraposed, and nominal elements that are NP-shifted. All the structures mentioned in Section 2 display morphological agreement, 6. Alternatively, mæstra could be read as an adverb with a comparative ending, as suggested by a reviewer.

 Artur Bartnik

either ambiguous or unambiguous, which furnishes further support for the movement analysis. Sections 3 and 4 present constructions that cannot be derived by movement. They are discontinuous from the very beginning and no separation takes place. Section  3 discusses uninflected adverbial quantifiers. As shown in this section, quantifiers have both paradigmatic and syntagmatic relations with adverbs. Finally, I briefly mention resumptive pronouns and discontinuous genitival phrases resembling split genitives. In both cases, the data show that the movement analysis is problematic and the base-generation option is favoured. References Allen, Cynthia. 2002. “Case and Middle English Genitive Noun Phrases”. Syntactic Effects of Morphological Change ed. by David Lightfoot, 57–80. Oxford: Oxford University Press. Bartnik, Artur. 2008. Noun Phrase Structure in Old English: Quantifiers and Other Functional Categories. Ph.D. dissertation, John Paul II Catholic University of Lublin. Benmamoun, Elabbas. 1999. “The Syntax of Quantifiers and Quantifier Float”. Linguistic Inquiry 30.621–642. Bobaljik, Jonathan D. 1995. Morphosyntax: The Syntax of Verbal Inflection. Cambridge, Mass.: MIT Press. Bošković, Željko. 2004. “Be Careful Where You Float Your Quantifiers”. Natural Language and Linguistic Theory 22.681–742. Bošković, Željko & Daiko Takahashi. 1998. “Scrambling and Last Resort”. Linguistic Inquiry 29.347–366. Haeberli, Eric. 2000. “Adjuncts and the Syntax of Subjects in Old and Middle English”. Diachronic Syntax: Models and Mechanisms ed. by George Tsoulas, Susan Pintzuk & Anthony Warner, 109–121. Oxford: Oxford University Press. Hinterhölzl, Roland. 2006. Scrambling, Remnant Movement, and Restructuring in West Germanic. Oxford: Oxford University Press. Kemenade, Ans van. 2002. “Word Order in Old English Prose and Poetry: The Position of Finite Verbs and Adverbs”. Studies in the History of the English Language: A Millennial Perspective ed. by Donka Minkova & Robert Stockwell, 355–373. Berlin & New York: Mouton de Gruyter. Fischer, Olga, Ans van Kemenade, Willem Koopman & Wim van der Wurff. 2000. The Syntax of Early English. Cambridge: Cambridge University Press. Pintzuk, Susan & Elizabeth Closs Traugott. 2005. “The Discourse Function of Object Topicalization in English”. Paper presented at the 4th Holland-York Symposium on the History of English Syntax, Leiden, May 2005. Sportiche, Dominique. 1988. “A Theory of Floating Quantifiers and its Corollaries for Constituent Structure”. Linguistic Inquiry 19.425–449. Taylor, Ann, Anthony Warner, Susan Pintzuk & Frank Beths, eds. 2003. The York-TorontoHelsinki Parsed Corpus of Old English Prose. Oxford: Oxford Text Archive. Vikner, Sten. 1994. “Scandinavian Object Shift and West Germanic Scrambling”. Studies on Scrambling: Movement and Non-Movement Approaches to Free Word Order Phenomena ed. by Norbert Corver & Henk van Riemsdijk, 487–517. Berlin & New York: Mouton de Gruyter.

Genitive variation in letters, history writing and sermons in Late Middle and Early Modern English Teo Juvonen

University of Helsinki This paper analyzes genitive variation from about 1420 to 1640 in three genres, letters, history writing and sermons. The corpus material is selected from the Penn-Helsinki Parsed Corpus of Middle English, second edition (PPCME2), the Penn-Helsinki Parsed Corpus of Early Modern English (PPCEME) and the Corpus of Early English Correspondence (CEEC). The occurrences of the genitive variants, the s‑genitive and the of‑genitive, are quantified and categorized according to the factors of possessor weight, topicality and genitive relation. The corpus analysis shows that these factors have a significant impact on genitive variation in all the genres. The analysis also shows that the genres differ significantly from each other. However, the study shows no clear change over time, neither within the genres nor over the period as a whole.

1. Introduction This study analyzes genitive variation, the relative frequency of the s‑genitive versus the of‑genitive, in Late Middle and Early Modern English with the focus on genre. Previous research has suggested that the s‑genitive increased in relation to the of‑genitive from the early fifteenth century to the seventeenth century. However, the corpus analysis presented here shows that there does not seem to be any patterned change over time from 1420 to 1640. There are, however, significant differences among the genres. The s‑genitive is most frequent in the letter genre, which here has been selected for informal, natural text. The of‑genitive, on the other hand, is most frequent in sermons, particularly those in a highly stylized, formal register. The corpus material is selected from the Penn-Helsinki Parsed Corpus of Middle English, second edition (PPCME2), the Penn-Helsinki Parsed Corpus of Early Modern English (PPCEME) and the Corpus of Early English Correspondence (CEEC), and consists of three genres: letters, history writing and sermons. The

 Teo Juvonen

main argument is that the s‑genitive is the natural, unmarked genitive variant with human possessors throughout this period. The of‑genitive is used where the s‑genitive is least favoured, or to achieve stylistic effect. The main factor influencing genitive variation is genre. The effects of the linguistic factors of possessor weight, topicality, givenness and genitive function are also analyzed to see if there is any clear difference in the way they operate between genres. The analysis shows that there are some minor differences, mostly between letters and sermons. 2. Previous research Before the paper by Rosenbach, Stein & Vezzosi (2000) the only quantified data on the variation between the s‑genitive and the of‑genitive in earlier varieties of English were the study by Russel in 1931 (described in Mustanoja 1960: 75) and Altenberg’s (1982) study on genitive variation in the seventeenth century. The general consensus seems to have been that the relative genitive distribution reached the level of what it is today by about the beginning of the Early Modern period (cf. Rissanen 1999: 201). Rosenbach, Stein & Vezzosi (2000) argue in their corpus study on genitive variation that there was a significant increase in the use of the s‑genitive relative to the of‑genitive from 1400 to 1630. Their study shows that the most important factor in the selection of the genitive variant was animacy. Throughout the period the s‑genitive rarely occurred with inanimate possessors. The increase in the use of the s‑genitive happened with animate, human possessors. The next most important factor was topicality, and then the prototypicality of possession (see Rosenbach 2002: 193–201 for a more detailed analysis of the interaction of topicality and genitive function). The results of their study show that in the early 1400s the s‑genitive mostly occurred in its most favourable context, i.e., with highly topical possessors and in the prototypical genitive functions (see below). Even there it was quite infrequent compared to the of‑genitive. In fact, they suggest that the s‑genitive in this period was almost a relic, used with only a limited set of possessors, such as “god”, “king” and “man”. By the later sixteenth century the s‑genitive had superseded the of‑genitive as the more typical genitive variant in the possessive and subjective relations with human possessors. This study contrasts with Rosenbach, Stein & Vezzosi’s (2000) to show that the most significant factor affecting genitive variation is genre. Additionally, it aims to show that the s‑genitive was not a relic, or a marginalized structure, but rather the natural, unmarked genitive variant with human possessors. The operation of the linguistic factors of topicality and genitive function are looked at in detail to show that they affect, as Rosenbach et al. (2000) suggest, genitive variation in all the genres.



Genitive variation in late Middle and Early Modern English 

3. Factors influencing genitive variation Many factors have been shown to affect genitive variation, such as possessor length, a final sibilant ending, economy of text, style, genre, topicality and genitive function (Altenberg 1982; Anschutz 1997; Rosenbach 2002; Hinrichs & Szmrecsanyi 2007 among others). This study will focus particularly on topicality, genitive function and genre. Topicality is a primary aspect of the function of the s‑genitive. From the perspective of Present-Day English, the s‑genitive can be described as a special type of determiner, according to Taylor (1996). In Taylor’s analysis it is a grammatical tool enabling definite reference, i.e., it is a way of leading the hearer from a known entity to an unknown, or less prominent, entity. In Present‑Day English this occurs naturally with the known entity preceding the unknown entity, as is the case with the s‑genitive, with the word order of possessor (known) – possessum (unknown). In contrast, the of‑genitive has the opposite order. The studies cited above have looked at the effect of topicality on genitive variation to show that the s‑genitive is favored with more topical possessors. The other linguistic factor governing genitive variation that is looked at here is the genitive function. It is in terms of genitive function that the s‑genitive has become most constrained over time from Old English to Present‑Day English. By the Late Middle English period the s‑genitive retained mostly only those genitive relations that are available to it today. For the purposes of the corpus analysis the genitive function has been divided into three: the possessive, subjective and objective functions. The subjective and objective functions are those where the possessum is an action and the possessor is the agent or the object of the action, respectively, as in examples (1) and (2).

(1) the king’s return [subjective] (2) the king’s murder [objective, if the king is being murdered]

The so-called possessive function seems to allow for a wide range of possible semantic relations between the possessor and the possessum. The possessive relations of kinship, part-whole and concrete possession are usually seen as most basic or prototypical (cf. Taylor 1996; Heine 1997; Rosenbach 2002). This is discussed further in Section 4.3. Altenberg’s (1982) study on genitive variation in the seventeenth century shows that the variation is very sensitive to genre. Genitive variation has also been linked to levels of formality, although indirectly through genre or text type evidence (Altenberg 1982: 254; Biber et al. 1999: 302; Rosenbach 2002: 38–40). What is very important for the following corpus analysis is that also the absolute frequency of total genitive constructions (s‑genitives and of‑genitives together) is

 Teo Juvonen

tied to genre, as is shown in Tables 1 to 3.1 For this reason, in the corpus analysis presented below, the selection of occurrences is based on quota sampling, i.e., selecting the first 300 tokens from each genre within each period, to enable a more balanced comparison of genres and overall change over time. 4. Genitive variation in the letter, history and sermon genres 4.1

Selection of corpus

The corpus texts were selected with the main aim of showing the differences among the genres.2 This corpus selection is not a comprehensive representation of total genre differences, but it covers the range from informal, speech-like texts, i.e., letters, to highly stylized texts, i.e., sermons (cf. Biber 1988; Biber & Finegan 1992; Nevalainen & Raumolin-Brunberg 2003 and others). The texts are categorized into three periods: Period I (1420–1500), Period II (1500–1570) and Period III (1570–1640). The letter sub-corpus (see Appendix A) was selected from the Corpus of Early English Correspondence (CEEC). The selection of material aimed for continuity in the background of the writers in terms of region, social rank and occupation. The emphasis was on informal letters. London and East Anglia were chosen since both regions had a reasonable number of writers with a comparatively large word count in the fifteenth and the sixteenth centuries. The genres of history and sermons were chosen as representative of more formal writing (see Appendix A). 4.2

Selection of genitive variants

All s‑genitives3 and of‑genitives were searched for in the corpus texts. These occurrences were then analyzed to see where there was indeed a possibility for variation. Identifying the contexts in which the s‑genitive and the of‑genitive are truth-propositionally equivalent and where there is real choice between the 1. The data in Altenberg (1982: 254) can also be re-calculated to show a similar distribution among the genres he analyses. 2. There are significant differences among the texts within the genres. This aspect is, however, beyond the scope of this article. 3. The s‑genitives, the so-called his‑genitives, e.g., “the king his daughter”, and the s‑less genitives, e.g., “the erle of Sowthefolke brothyr” (CHRLOND, p. 164.855) were initially classified separately, but on closer examination they appear to be socio-regionally conditioned orthographic variants of the same morpho-syntactic construction (cf. Allen 2008 for the hisgenitives and their relation to the s-genitives).



Genitive variation in late Middle and Early Modern English 

variants was problematic. The problems had to do with both theoretical concerns and with the nature of the historical material where native speaker intuition cannot be relied on and where stylistic rules are different from today. Only specifying genitives (see examples (3) and (4)) were included, as they are the only s‑genitives that allow for variation at all or in sufficient frequency for quantitative analysis. In these the possessor can be a proper or common noun, animate or inanimate, but it needs to be referential, i.e., it needs to refer to a specific entity or a set of specific entities. Both possessor and possessum need to be full lexical NPs:

(3) the king’s daughter (4) the daughter of the king

Thus all classifying genitives (e.g., an iron ram’s horn), double genitives (e.g., a servant of the king’s), absolute genitives (e.g., the book that was my brother’s) and all elliptic occurrences were excluded. Subjective gerunds (e.g., the king’s coming) were included, but not objective gerunds.4 Locative s‑genitives were also excluded (e.g., at the Johnsons’). With of‑genitives, only those with definite possessums were included. In the following two examples, the referent of daughter is different, and thus they cannot be considered equivalent:

(5) the king’s daughter (6) a daughter of the king

The problem with the two examples above is that in example (5) the daughter is definite and specified, while in (6) the daughter can be any one of the possible daughters of the king. There is no s‑genitive construction that is equivalent to (6). The of‑genitive equivalent to (5) would be the daughter of the king. A similar problem exists with determiners such as some and no, and these instances were also excluded. In the corpus material there were a number of instances where the possessum was not overtly marked for definiteness, in a way that would be ungrammatical in Present-Day English.

(7) þis was the laste yere of raygne of the fadyr, and the fyrste yere of the (CHRLOND, p. 107.287)6 raygne of the sone, Kyng Harry the v.5

4. There were no s-genitive objective gerunds in the corpus texts. 5. The possessor and genitive marker of the genitive construction, either the s-genitive or the of‑genitive, is in bold type in all the quotes taken from the corpus analysis. 6. The sources for the quotes from the corpus analysis are in Appendix A. The first word identifies the book the quote is from. In letters, the name of the author and the year the letter was written or sent is also given.

 Teo Juvonen

In example (7) the first reign is unmarked for definiteness. However, from the context and the second use of reign it becomes apparent that overt marking was not required. Attestations like these are thus included. There are also a number of instances in which one or the other genitive variant occurs very rarely. These cases are excluded and are discussed below. S‑genitives with inanimate possessors were rare (about two to four percent of all s‑genitives depending on the writer and text). Because they only occur in a limited set of relations (see examples (8) to (10)) and the factors governing their variation seem to be different from those with human possessors, all inanimate possessors were left out from the corpus analysis. (8) for scant ij yere’s disbursing ther monney (Johnson, p. 1182; John Johnson, 1551) (9) the schypys namys (Cely, p. 81; Richard Cely Jr, 1480) (10) at deaths doore (Chamberlain, part II, p. 241; John Chamberlain, 1619)



Likewise excluded are animals and collective nouns as possessors. Animal possessors were extremely rare, and collective nouns as possessors in the s‑genitive mostly occurred with one writer, John Chamberlain: (11) the counsailes letters  (Chamberlain, part I, p. 139; John Chamberlain, 1602) Genitive constructions with long and complex possessors are included. This is contrary to the practice of some of the other studies because, on the whole, s‑genitives rarely occur with long possessors. This is also the case with this corpus, where there were few examples of the s‑genitive with long or complex possessors, with the exception of titles in the form of: (12) as for the matter of my Lord of Canterbery ys cortte  (Cely, p. 141; William Maryon, 1482) (13) sent by John the Abottys man of Seynt Benet.  (Paston, part I, p. 584; John III Paston, 1472?) Example (12) is a group genitive, while (13) is a split genitive. Both forms occur throughout the period, although split genitives become increasingly rare after the fifteenth century. Compare with Present-Day English where, for example, it is possible to split the s‑genitive, although it is stylistically marked: (14) has allayed people’s fears who’ve been used for those residential home agreements (quoted in Denison et al. 2008: 6)



Genitive variation in late Middle and Early Modern English 

Similar constructions were used, rarely, by one of the writers in Letters Period I (1420–1500) and one in History Period I. Example (15) is from the history genre: (15) and in that furynys they wente, as they sayde, for the comyn wele of the realme of Ingelonde, evyn strayght unto a marchaunte ys place i-namyd Phylyppe Malpas of London. (CHRLOND, p. 191.1463) Due to the possibility for long possessors as in example (15) it was seen better to be more inclusive than not. Since the focus of this study is on the change and variation in the use of genitive variants, only instances of productive use are included. Thus fixed expressions, titles, place names, feast days and names of institutions, such as the King’s Bench were excluded since they form unvarying lexical units. Likewise, the expressions god’s grace and the grace of god were excluded. Although these show variation, the variation is not affected by topicality or genitive relation in the same sense as with productive variation. Similarly, in the Bacon letter collection, there are s‑genitives that function as a polite equivalent of the second person singular or plural pronoun, which are excluded. Example (16) shows a typical use, where the s‑genitive is functionally equivalent to the pronoun your: (16) your lordship’s well doinge (Bacon, part I, p. 41; Nathaniel Bacon, 1572) The genitive variants were chosen according to these criteria from the corpus texts. The overall results are displayed in Tables 1 to 3. Each period for each genre contains a total of 300 tokens. The normalized frequency (occurrences/1,000 words) is based on this combined genitive count. Table 1 shows that letters contain a high relative frequency of s‑genitives. The normalized frequency of total genitive constructions is lower than in the history or sermons genre: Table 1.  Genitive constructions in the letter genre Period Period I (1420–1500) Period II (1500–1570) Period III (1570–1640) Letters

Word count

Normalized frequency

  77,991   72,077   50,515 200,583

3.8 4.2 5.9 4.5

s‑genitive N % 236 265 242 743

79% 88% 81% 83%

of‑genitive N % 64 35 58 157

21% 12% 19% 17%

 Teo Juvonen

Table 2.  Genitive constructions in the history genre Period

Word count

Normalized frequency

Period I (1420–1500) Period II (1500–1570) Period III (1570–1640) History

  38600   34717   41240 114557

7.8 8.6 7.3 7.9

s‑genitive N % 141 167 116 424

47% 56% 39% 47%

of‑genitive N % 159 133 184 476

53% 44% 61% 53%

Table 3.  Genitive constructions in the sermons genre Period

Word count

Normalized frequency

Period I (1420–1500) Period II (1500–1570) Period III (1570–1640) Sermons

25081 30850 29460 85391

12.0   9.7 10.2 10.5

s‑genitive N % 121 110 102 333

40% 37% 34% 37%

of‑genitive N % 179 190 198 567

60% 63% 66% 63%

Table 2 shows that history writing has proportionally more of‑genitives than the letter genre and also a higher normalized frequency of total genitive constructions. Table 3 shows that the sermons genre has the highest use of the of‑genitive and the most total genitive constructions. Tables 1 to 3 show that the absolute frequency of genitive constructions varies according to genre. This is the reason for controlling genre by token count rather than word count: If word count were used, when the genres were combined to look at overall change over time, sermons would dominate over the other genres, in favour of the of‑genitive. This is also possibly the major reason why the results of this study differ so markedly from those of Rosenbach et al. (2000), since in their selection sermons and other texts from religious registers predominate in the earlier periods. These texts are also the ones in which the s‑genitive appears most fossilized. 4.3

Classification of genitive variants

The genitive variants were classified according to length of possessor, topicality, givenness and genitive function. Long possessors are those that have four or more words or are compounds or appositives (e.g., his brother John’s) or contain a prepositional phrase or relative clause. Topicality, here, is based on the idea of how prominent a possessor is. More topical information is that which can be seen as being part of global knowledge,



Genitive variation in late Middle and Early Modern English 

while less topical information is that which is contextually derived. In practice, proper nouns and titles of prominent people, such as the king or bishop are classified as more topical, or “global”, while other definite common nouns are classified as less topical, or “local”. This practice follows that of Rosenbach et al. (2000). Indefinite and generic (e.g., man in man’s law) possessors were classified as “indefinite”. Global and local possessors were further divided into “given” and “new” based on whether the same referent had appeared in the text within 50 words or before a text boundary (cf. Hinrichs and Szmrecsanyi 2007 for a similar method). Genitive relations were divided into the subjective gerund, possessive, subjective and objective relations. Of these, the objective relation is excluded from the corpus analysis, because in practice it is extremely rare in the s‑genitive (seven clear objective uses out of 1500 tokens)7. The possessive relation was divided into six types (following Rosenbach et al. 2000 and using their terminology): 1. Kinship 2. Body parts 3. 4. 5. 6.

Ownership Social relations States Abstract possession

(17) the kinges doughter of Spain (Moreric, p. 60.181) (18) Crysten mennes handes (Paston, part I, p. 446; John II Paston 1472) (19) poore mennes goodes (Moreric, p. 31.111) (20) the Kings souldiers (Stow, p. 555.88) (21) saynt Payles authoryte (Fisher, p. 314.11) (22) Kyng Henryes marryage (Hayward, p. 4.18)

Kinship, body parts and ownership are most often seen in the literature as “prototypical” of the possessive function. To this it was decided to add social relations, because in the cultural context of this period servants and other similar referents were part of the household (cf. Juvonen 2009). All the other categories, i.e. states, abstract possession, subjective and subjective gerund, were categorized as “non-prototypical”. This differs from the categorization of Rosenbach et al. (2000), who divide prototypical patterns into kinship, body parts and ownership and non-prototypical ones into social relations, states and abstract possession and analyze subjective relations separately. The reason for treating subjective relations together with non-prototypical relations was to simplify the analysis into a two-way rather than three-way distinction. The effect of my decisions in categorization is to heighten the difference between prototypical and non-prototypical genitive function.8 7. In the initial selection of 300 first tokens (Tables 1 to 3) occurrences with the objective genitive function were included. However, since it became apparent that the objective genitive function was a virtual knockout factor, they are excluded from the corpus analysis (Tables 4 to 10). 8. The variation of each individual genitive function was looked at separately, but no clear pattern emerged that differed from the general prototypical/non-prototypical pattern. The

 Teo Juvonen

4.4 Corpus analysis The data from the three genres were combined to provide an overall diachronic picture from 1420 to 1640. Table 4 shows that no obvious change happened with the relative frequency of the s‑genitive from 1420 to 1640. The differences between genres, on the other hand, are striking. Table 4.  Genitive variation 1420–1640 in three periods: share of the s‑genitive out of all s‑ and of‑genitives s‑genitive % N Period I (1420–1500) Period II (1500–1570) Period III (1570–1640)

56% 62% 53%

491 542 460

Total N 879 879 862

Table 5 shows that genre has a clear impact on genitive variation. The differences among all the genres are statistically significant (Yates’ chi-square p < 0.0019 between all). Table 5.  Genitive variation in the letter, history and sermon genres 1420–1640: share of the s‑genitive out of all s‑ and of‑genitives s‑genitive % N Letter History Sermon

84% 49% 38%

743 422 328

Total N 887 861 872

4.4.1 Comparison of genres The linguistic factors of possessor length, genitive function, topicality and givenness are all significant in genitive variation in all the genres.

divisions are of necessity somewhat arbitrary, and fail to capture many of the nuances of the writers. The way the word soul is used, for example, would make an interesting case study. 9. All the following calculations of statistical significance use the chi-square test with the Yates’ correction.



Genitive variation in late Middle and Early Modern English 

Table 6 shows that the length of the possessor is a highly significant factor in genitive variation in all the genres.10 Long possessors tend to favor the of‑genitive, even in the letter genre. Table 6.  Genitive variation in the letter, history and sermon genres according to length of possessor: share of the s‑genitive out of all s‑ and of‑genitives Short possessors s‑genitive % N Letter History Sermon

88% 57% 40%

703 381 327

Total N 801 663 811

Long possessors s‑genitive Total % N N 47% 21% 2%

40 41 1

86 198 61

The prototypicality of the genitive function is also a significant factor, as shown by Table 7. The difference between prototypical and non-prototypical possessors is statistically significant within all the genres (p < 0.001, except p < 0.01 in sermons).11 Table 7.  Genitive variation in the letter, history and sermon genres according to prototypicality of the genitive function: share of the s‑genitive out of all s‑ and of‑genitives Prototypical

Letter History Sermon

s‑genitive %

N

Total N

95% 67% 44%

391 245 145

412 364 327

Non-prototypical s‑genitive Total % N N 74% 36% 34%

352 177 183

475 497 545

Table 8 shows while there is clear difference among the genres according to topicality, the differences within genres are small.12 Most of the occurrences are 10. The difference in the relative frequency of the s‑genitive between long and short possessors is statistically significant (p < 0.001) within all the genres. The differences between genres are also significant (p < 0.001) in both long and short possessors. 11. With the prototypical relation the difference is significant both between letters and history and history and sermons (p < 0.001). In the case of the non-prototypical relation the difference is significant between letters and history and letters and sermons (p < 0.001), but not between history and sermons (p > 0.20). 12. Letters and history show a significant difference between global and local topicality (p < 0.01 and p < 0.001 respectively). Sermons shows a significant difference between global and indefinite topicality (p < 0.01). The letter genre differs significantly from history and sermons

 Teo Juvonen

Table 8.  Genitive variation in the letter, history and sermon genres according to topicality: share of the s‑genitive out of all s‑ and of‑genitives Global s‑genitive % N Letter History Sermon

86% 52% 35%

612 325 219

Total N 714 623 633

Local s‑genitive % N 74% 35% 43%

94 53 35

Total N 127 150   82

Indefinite s‑genitive Total % N N 80% 50% 47%

37 44 74

  46   88 157

of global topicality, and thus it is this category which has the most impact on the overall results. Table 9 shows that givenness is a significant factor in genitive variation in all the genres.13 All genres show a significant preference for the s‑genitive with short, given possessors in the prototypical genitive function. The only clear difference among genres is in topicality. Here letters and history favour global topicality with the s‑genitive, while sermons favour indefinite topicality with the s‑genitive. All the genres have a high proportion of globally topical possessors. This is especially the case in letters, where possessors in genitive constructions usually refer to a person who is of interest to both the sender and the recipient. The content is often news or gossip, with little extended narrative. In sermons, the most common possessors are God, Christ and man. The possessor man is mostly used generically (e.g., man’s law). A Table 9.  Genitive variation in the letter, history and sermon genres according to givenness: share of the s‑genitive out of all s‑ and of‑genitives14 Given s‑genitive % N Letter History Sermon

90% 64% 43%

141 164 137

Total N 156 256 316

New s‑genitive % N 82% 41% 29%

565 214 117

Total N 685 517 399

in global, local and indefinite topicality (p < 0.001 in all except between letters and history in indefinite, where p < 0.01). History differs from sermons only in global topicality (p < 0.001). 13. The difference between prototypical and non-prototypical possessors is statistically significant within all the genres (p < 0.001, except p < 0.025 in letters). With both given and new possessors the difference is significant among all the genres (p < 0.001). 14. Indefinite possessors are not included.



Genitive variation in late Middle and Early Modern English 

prominent feature of sermons is the use of generic possessors for rhetorical purposes and the limited use of the s‑genitive compared with the other two genres. The interaction of prototypicality of the genitive relation and givenness is analyzed further to show the most and least favoured contexts for the s‑genitive. Table 10 shows that in each genre the prototypical genitive function with given possessors is the favoured linguistic context for the s‑genitive, while the non-prototypical genitive function with new possessors is the least favoured (all differences p < 0.001). In letters and history prototypicality is the determining factor, followed by givenness, while in sermons givenness is the determining factor.15 Table 10 also shows that the s‑genitive is the clearly preferred variant, and thus the unmarked choice, in all the linguistic contexts in the letter genre. In history, the factor of prototypicality seems to be the dividing line, while in sermons the of‑genitive seems to be more the unmarked variant. All of these factors were further analyzed in interaction with each other to see if any discernible change over time could be found, but there were no clear and significant changes that formed a continuous pattern. Table 10.  Genitive variation in the letter, history and sermon genres according to interaction of prototypicality of the genitive relation and givenness: share of the s‑genitive out of all s‑ and of‑genitives Prototypical Given s‑gen. Total

Letter History Sermon

% (N) 99% (67) 79% (85) 50% (59)

N   68 108 119

Prototypical New s‑gen. Total % (N) 94% (306) 62% (139) 34% (53)

N 324 225 155

Non-Prototypical Non-Prototypical Given New s‑gen. Total s‑gen. Total % (N) 84% (74) 53% (79) 40% (78)

N   88 148 197

% (N) 72% (259) 26% (75) 26% (64)

N 361 292 244

15. In letters, the differences between prototypical new and non-prototypical given and new are significant (p < 0.01 and p < 0.001 respectively). In the prototypical given context there is only one of‑genitive, so in letters this is almost exclusively s‑genitive territory. In history, only the difference between prototypical new and non-prototypical given is not significant (prototypical given to new is p < 0.01, others are p < 0.001). In sermons, the differences between prototypical given and new (p < 0.025) and non-prototypical given and new (p < 0.01) are significant.

 Teo Juvonen

5. Conclusion All the genres show a significant preference for the s‑genitive with short, given possessors in the prototypical genitive function. Furthermore, in all the genres, the prototypical genitive function with given possessors is the most favoured context for the s‑genitive, while the non-prototypical genitive function with new possessors is the least favoured. However, genre does affect the relative weight of the linguistic factors. In letters and history writing it is the prototypicality of the genitive relation that is the determining factor, while in sermons it is givenness. Particularly in letters, the s‑genitive is the natural, unmarked variant. In sermons, on the other hand, the of‑genitive is more common. Overall, this study shows that while genre has a significant impact on genitive variation, there appears to be no patterned change over time within the period 1420 to 1640. References Allen, Cynthia. 2008. Genitives in Early English: Typology and Evidence. Oxford: Oxford University Press. Altenberg, Bengt. 1982. The Genitive v. the of‑Construction: A Study of Syntactic Variation in 17th Century English. Lund: Gleerup. Anschutz, Arlea. 1997. “How to Choose a Possessive Noun Phrase Construction in Four Easy Steps”. Studies in Language 21:1.1–35. Biber, Douglas. 1988. Variation across Speech and Writing. Cambridge: Cambridge University Press. Biber, Douglas & Edward Finegan 1992. “The Linguistic Evolution of Five Written and Speechbased English Genres from the 17th to the 20th Centuries”. History of Englishes: New Methods and Interpretations in Historical Linguistics ed. by Matti Rissanen, Ossi Ihalainen, Terttu Nevalainen & Irma Taavitsainen, 688–704. Berlin & New York: Mouton de Gruyter. Biber, Douglas, Stig Johansson, Geoffrey Leech, Susan Conrad & Edward Finegan. 1999. Longman Grammar of Spoken and Written English. London: Longman. Denison, David, Alan Scott & Kersti Börjars. 2008. “What’s Wrong with Possessive ‘s?” Paper presented at the First Triennial Conference of the International Society for the Linguistics of English (ISLE1), University of Freiburg, 8–11 October 2008. . Heine, Bernd. 1997. Possession: Cognitive Sources, Forces, and Grammaticalization. Cambridge: Cambridge University Press. Hinrichs, Lars & Benedikt Szmrecsanyi. 2007. “Recent Changes in the Function and Frequency of Standard English Genitive Constructions: A Multivariate Analysis of Tagged Corpora”. English Language and Linguistics 11:3.473–474. Juvonen, Teo. 2009. “Everyday Possessions: Family and Identity in the Correspondence of John Paston II”. The Language of Daily Life in England (1400–1800) ed. by Arja Nurmi, Minna Nevala & Minna Palander-Collin, 253–277. Amsterdam & Philadelphia: Benjamins.



Genitive variation in late Middle and Early Modern English  Kytö, Merja. 1996. Manual to the Diachronic Part of the Helsinki Corpus of English Texts: Coding Conventions and Lists of Source Texts. (3rd Edition). Helsinki: Helsinki University Printing House. Lambrecht, Knud. 1994. Information Structure and Sentence Form: Topic, Focus and the Mental Representations of Discourse Referents. Cambridge: Cambridge University Press. Mustanoja, Tauno. 1960. A Middle English Syntax. Part I. Parts of Speech. (= Memoires de la Société Néophilologique de Helsinki, XXIII.) Helsinki: Société Néophilologique. Nevalainen, Terttu & Helena Raumolin-Brunberg. 2003. Historical Sociolinguistics: Language Change in Tudor and Stuart England. London: Longman. Rissanen, Matti. 1999. “Syntax”. The Cambridge History of the English Language. Vol. III: Early Modern English 1476–1776 ed. by Roger Lass, 187–331. Cambridge: Cambridge University Press. Rosenbach, Anette, Dieter Stein & Letizia Vezzosi. 2000. “On the History of the s‑genitive”. Generative Theory and Corpus Studies (A Dialogue from 10 ICEHL) ed. by Ricardo Bermúdez-Otero, David Denison, Richard M. Hogg & C. B. McCully, 183–210. Berlin & New York: Mouton de Gruyter. Rosenbach, Anette. 2002. Genitive Variation in English: Conceptual Factors in Synchronic and Diachronic Studies. Berlin & New York: Mouton de Gruyter. Taylor, John. 1996. Possessives in English: An Exploration in Cognitive Grammar. Oxford: Clarendon. Vezzosi, Letizia. 2000. “S‑genitive and of‑genitive: Competitors or Complementary Strategies? A Diachronic Analysis”. English Diachronic Pragmatics ed. by Gabriella Di Martino & Maria Lima, 399–432. Napoli: CUEN.

Appendix A: Sources for the genre corpus material 1. Letters (From CEEC) (http://www.helsinki.fi/varieng/CoRD/corpora/CEEC/index.html) 1.1 Period I Cely: Hanham, Alison, ed. 1975. The Cely Letters 1472–1488. (= Early English Text Society, 273.) London, New York and Toronto: Oxford University Press. Paston: Davis, Norman, ed. 1971 & 1976. Paston Letters and Papers of the Fifteenth Century. Parts I-II. Oxford: Clarendon Press. 1.2 Period II Johnson: Winchester, Barbara, ed. 1953. The Johnson Letters, 1542–1552. Unpublished doctoral dissertation, University of London. 1.3 Period III Chamberlain: MacClure, Norman Egbert, ed. 1939. The Letters of John Chamberlain. Parts I-II. (= American Philosophical Society, Memoirs, 12.) Philadelphia: American Philosophical Society.

 Teo Juvonen

Bacon: Smith, A. Hassel, Gillian M. Baker & R.W. Kenny, eds. 1978, 1979, 1982, 1983, 1987 & 1988. The Papers of Nathaniel Bacon of Stiffkey. (= Norfolk Record Society, 46, 49 & 53.) Norwich: Norfolk Record Society. 2. History (From PPCME2 and PPCEME) (http://www.ling.upenn.edu/hist-corpora/)16 2.1 Period I CAPCHR (pp. 32–160, 209–217, 238–249), CHRLOND (pp. 95.21–128, 138–143, 148–149, 156–239) 2.2 Period II FABYAN (pp. 556.46–583.8), MORERIC (pp. 15.26–65.12, 76.12–80.24) 2.3 Period III HAYWARD (pp. 1.1–47.30, 60.11–64.9, 86.28–90.29), STOW (pp. 543.1–549.28, 549.35–585.3, 589.8–594.4) 2.4 Extras Hayward: Hayward, John. 1599. The First Part of the Life and Raigne of King Henrie the IIII. Extending to the End of the First Yeare of his Raigne. Written by I.H. London: E. Allde and T. Judson (pp. 1–10). Stow: Stow, John. 1580. The Chronicles of England from Brute unto this Present Yeare of Christ. London: Ralphe Newberie. (pp. 518–543) http://eebo.chadwyck.com/home 3. Sermons (From PPCME2 and PPCEME) (http://www.ling.upenn.edu/hist-corpora/) 3.1 Period I MIRK (pp. 1.4–5.36, 82.9–85.10), FITZJ (pp. A2R.1–C1R.20), CAPSER (pp. 143.1– 148.28), INNOC (pp. 1.1–13.4), GAYTR (pp. 1–15), ROYAL (pp. 9.3–12.17, 16.2– 19.34, 251.22–261.21) 3.2 Period II FISHER (pp. 314.4–348.5, 388.1–403.6), LATIMER (pp. 17.1–38.37, 22.1–43.12, 55.1–57.32) 3.3 Period III HOOKER (pp. 1.14–29.10, 31.13–35.30, 36.1–56.31), SMITH (pp. 1.1–47.19, B1R.1– F4V.13) 16. The sources for the history and sermon texts are presented in their abbreviated titles as printed in the Helsinki Corpus Manual (Kytö 1996). The full bibliographical information is available online at http://khnt.hit.uib.no/icame/manuals/HC/KHWTITLE.HTM.



Genitive variation in late Middle and Early Modern English 

3.4 Extras Fisher: Mayor, John  E.B., ed. 1876. The English works of John Fisher, Bishop of Rochester, Part I. (= EETS ES, 27.) Berlin, New York & Philadelphia: Asher, Scribner, Lippincott. (pp. 268–277) Smith: Smith, Henry. 1591. A Preparatiue to Mariage The Summe whereof was Spoken at a Contract, and Inlarged After. Whereunto is Annexed a Treatise of the Lords Supper, and another of Vsurie. By Henrie Smith. London: R. Field. (pp. 47– 52)

part iv

Syntactic variation and change through contact

On the use of beon and wesan in Old English Ilse Wischer

Potsdam University This paper commences by characterizing the exceptionality of the Old English double paradigm of s‑ and b‑forms of the verb ‘to be’ in the present tense among the (West) Germanic languages, followed by an analysis of the use of beon and wesan in the Old English section of the Helsinki Corpus with a focus on the linguistic contexts of these verbs in their indicative, subjunctive and imperative forms and on their dialectal distribution. A comparison of my findings with parallel forms and their uses in the Celtic languages supports the assumption that the emergence of the double paradigm in Old English, with wesan expressing current relevance and beon referring to habitual or future events, can only be assigned to Insular Celtic influence.

1. Introduction In Old English there existed two parallel forms of the verb to be in the present tense. One of them derived from an IE s‑root (eom, eart, is, sindon), and the other from a b‑root (beo, bist, biþ, beoþ), the latter originally meaning ‘become’. Additionally there were parallel forms in the infinitive and imperative, one having also derived from the b‑root (beon, beo/beoþ) and the other from an IE w‑root (wesan, wes/wesaþ). According to the general tradition, in the following the verbs originating in b‑roots shall be called beon, and the verbs derived from s‑roots and w‑roots will be referred to as wesan, even though the term is not fully correct. Although beon‑forms also entered the paradigms of other West Germanic languages, only in Old English did they develop a parallel paradigm in relation to wesan, both being semantically distinct. The standard/handbook treatments of Old English morphology and syntax (see below) agree that beon refers to future events and statements of general truth, while wesan is used for immediate present relevance (cf. (1)), although it is admitted that this is just an affinity.

 Ilse Wischer

(1) þu byst æfter fæce þæt ic nu eom. thou beon–2sg.prs after while that I now wesan–1sg.prs  (Blickling Homilies, p. 113) ‘you will be after a while what I am now.’ Towards the end of the Old English period this distinction blurred and a new mixed paradigm emerged in Middle English. This functional distinction in the use of beon and wesan in Old English seems to be taken for granted by most scholars. The descriptions in the handbooks are rather vague, though. Brunner (1951: 258) argues that the b‑forms are often, but not always, used with a future sense. He also refers to some dialectal differences. Campbell (1991 [1959]: 351) also mentions dialectal variation, especially with regard to the imperative and the subjunctive. Hogg (1992) and Traugott (1992) in the Cambridge History of the English Language only briefly refer to the present topic, Mitchell (1985) mainly agrees with Campbell’s position. The only detailed analyses of the uses of beon and wesan are some early twentieth century studies, such as Karl Jost’s (1909) or Otto Exter’s (1911) dissertations, and more recently Kilpiö’s (1992) entry for beon in the Dictionary of Old English and his analysis of the present indicative forms of beon in the Old English part of the Helsinki Corpus (1993) as well as his (1997) investigation of the development of b‑ and non‑b‑forms from Old to Modern English, which rely on empirical data analyses. The question of a possible Celtic origin of the b‑paradigm in Old English was already addressed by Keller in 1925 and has more recently been discussed at length by Lutz (2009). In my contribution I will present results from my detailed corpus investigation of the use of beon and wesan in the Old English section of the Helsinki Corpus. I will focus on the linguistic contexts of beon and wesan in their indicative, subjunctive and imperative forms as well as on their dialectal distribution. From this I will try to draw some conclusions about a possible origin of this unique formal and functional pattern of Old English. 2. The Old English double paradigm The b‑forms in the present tense paradigms of the verb to be have their origin in an Indo-European (IE) root *bheu-,*bhu- with the meaning ‘become’ (Schumacher 2007: 193). They can be found in other Old Germanic languages as well, but only in West Germanic. Lass (1994:  171) argues that the “incorporation of the b‑root into this group seems to be a West Germanic innovation”. However, it is only in Old English that we find a complete double paradigm of s‑ and b‑forms in the present tense, see Figure 1.

The use of beon and wesan in Old English 



Indicative Sg.

1 2 3

Pl.

ēom eart is sind(on)/āron

Subjunctive bēo bist biþ bēoþ

sīe sīe sīe sīen

bēo bēo bēo bēon

Imperative

Infinitive wesan

sīe/wes

bēo

wesaþ

bēoþ

bēon

Figure 1.  Parallel Paradigms of beon and wesan in Old English

The other West Germanic languages have mixed paradigms, see Figures 2 and 3. This situation has been known for a long time. As early as in 1900 Bethge (1900: 390) claims that in West Germanic there appear forms of a root *bheu‑, but concedes that a complete present of *bheu‑ exists only in Old English. Similarly Dieter (1900) maintains that b‑forms exist neither in Gothic nor Old Norse. Old English, on the other hand, has complete parallel paradigms in present indicative and subjunctive, imperative and infinitive.

Sg.

1 2 3

Pl.

Indicative

Subjunctive

bium, biun, bion bis, bist ist, is sind(un), sindon

sī sīs sī sīn

Imperative

Infinitive/Gerund sīn, uuesan/uuesanne

uuis, uues uuesađ

Figure 2.  Mixed Paradigms of beon and wesan in Old Saxon* Indicative

Subjunctive

Sg.

1

bim, bin

sī, uuese

Pl.

2 3 1 2 3

bist, bis ist birum, birun birut, birent sint, sindun

sīs(t), uuesēs(t) sī, uuese sīm, sīn, uuesēm sī, sīnt, uuesēt sīn, uuesēn

Imperative

Infinitive/Gerund uuesan, sīn/sīnne, uuesanne

uuis, bis

uueset

Figure 3.  Mixed Paradigms of beon and wesan in Old High German†

* The data have been compiled from Streitberg (1974 [1896]: 316), Holthausen (1900: 199), Dieter (1900), Robinson (1992: 115).

† The data have been compiled from Streitberg (1974 [1896]: 316), Wright (1906: 88), Dieter (1900), Robinson (1992: 115), Besch et al. (2000: 1192).

 Ilse Wischer

In the other West Germanic languages, the paradigms are not only mixed in the sense of being suppletive, but they may even be called contaminated1 (cf. forms like bim in Old Saxon or bium in Old High German). Schumacher (2007: 194–195) relates the existence of these mixed and contaminated paradigms in a convincing way to a very early language contact between West Germanic and Celtic tribes on the continent. His hypothesis, however, that the Old English double paradigm is an archaic feature2 that was inherited from earlier West-Germanic and preserved through the contact situation with the Celtic languages is rightly criticized by Lutz (2009). Lutz is certainly right in stating that the West Germanic dialects rather had mixed paradigms from the beginning and that the Old English double paradigm is a result of Insular Celtic influence after the Germanic settlement on the British Isles. This influence is also confirmed by Schumacher (2007: 201–202), who provides evidence that in the Celtic languages, especially in the earlier attested languages on the British Isles, there is a clear dichotomy between the use of an unmarked s‑stem and a marked b‑stem for the verb to be in the present indicative (for the occurrence of subjunctive, imperative and verbal noun b‑forms in the Celtic languages, see below). The functional distinction between s‑ and b‑forms in Old English is described by Campbell (1991: 350–351) in the following way: “[eom refers to a present state] provided its continuance is not especially regarded”; beo is used to express an “invariable fact”, a future event, or an “iterative extension into the future”. Traugott (1992: 182) observes that “there seems to be no absolute distinction between ‘beon’ ... and ‘wesan’ when time reference to the future is concerned.” She also stresses the fact that beon “seems to be preferred for reference to habitual, repetitive, and therefore pluralized, situations”, while wesan “is favoured for singular situations ... or situations regarded as eternal, and therefore singular ...” (1992: 183). As evidence, she refers to Ælfric, who, in his grammar, describes the use of sum/ic eom in a similar way as a word of existence that in particular belongs to God, since God is eternal without beginning and end: “Sum i c e o m is edwistlic word and gebyrað to gode anum synderlice, forðan þe god is æfre unbegunnen and ungeendod on him sylfum and ðurh hine sylfne wunigende.” (Zupitza 1880: 201, ‘Sum i c e o m is a substantive word and belongs especially to God alone because God is always without beginning and end, existing in himself and by himself.’). However, Ælfric does not distinguish in that respect between 1. In Schumacher‘s (2007:  202) words: “... und die beiden Paradigmen wurden zu einem einzigen kontaminiert.” – ‘... and the two paradigms were contaminated to a single one.’ 2. “... der Gegensatz zweier distinktiver Verbalstämme im Altenglischen [ist] ein Archaismus und keine Neuerung ...” (Schumacher 2007: 197f) – ‘...the opposition of two distinctive verbal stems in Old English is an archaism and not an innovation ...’



The use of beon and wesan in Old English 

wesan and beon. He only lists the forms of beon as translations of Latin future, imperative and optative/hortative forms (ero/ic beo, sis/beo ðu, sit/beo he, simus/ beon we, estote/beon ge, sunto/beon hi) (1880: 201f.).3 Hogg (1992: 164) only very vaguely refers to the divided usage claiming that “[t]he Anglo–Saxons appear to have distinguished in meaning between the two sets of forms more often than not (but not, alas, always)”. In one of the early more detailed investigations, Jost (1909), who studied the use of beon and wesan mainly in Beowulf and the Cura Pastoralis, comes to the conclusion that the future use of beon is only attested for a small number of examples. This is supported by Exter’s (1911) findings, who analyzed the Alfredian translation of Boethius and of St. Augustine’s Soliloquies. Both authors, instead, maintain that wesan predominates in so‑called “concrete” sentences, where it denotes a present state or a general state without focusing on the duration of the state. On the other hand, they argue that beon is instead used in so‑called “abstract” sentences referring to a state which from the perspective of the speaker will happen in the future or which already exists in the present, but whose continuation in the future is especially emphasized, or which takes place repeatedly. Although their division into concrete and abstract is somewhat arbitrary and theoretically not very well founded, the affinity to the statements by Traugott, Schumacher and other scholars on a distinction between actual and habitual events becomes clearly obvious. Kilpiö (1993) provides the most detailed and empirically founded analysis of the two forms in the present indicative. His results do not disagree with the previous observations, yet they are more exact and based on quantitative data. He concludes that although there is no sharp demarcation between the uses of the b‑ and s‑forms, there are clear tendencies for the use of beon to be preferred with future reference, in contexts where the verb is given a durative interpretation, in generic contexts and in actional passive constructions. Furthermore it is the only form to be used when the verb is given an iterative interpretation. Wesan, on the other hand, is preferred with reference to the present moment or situation as well as with statal passives. We will see that all these interpretations are in line with Schumacher’s (2007: 186) observation for the present indicative paradigms of the verb to be in the Celtic languages. He claims that the unmarked s‑stem in the Insular Celtic languages, especially in the earlier attested languages Irish and Welsh, is used with reference to the actual present, while the marked Celtic b‑stem has primarily a habitual meaning, and sometimes refers to future events. 3. Cf. Laing (this volume), who also refers to Ælfric’s grammar and the role that a contemporary grammarian may play for our understanding of the future tense in Old English.

 Ilse Wischer

3. Analysis of the examples 3.1

Indicative

There were altogether 6,551 be‑forms in the Old English part of the Helsinki Corpus, of which 65 percent occurred in the s‑ and 35 percent in the b‑forms, which corresponds – not surprisingly – to Kilpiö’s (1993) data, since they are based on the same corpus. The distribution of s‑ and b‑forms according to person and number distinctions can be seen in Figure 4:

s‑form b‑form Σ

1st Pers Sg

2nd Pers Sg

3rd Pers Sg

Plural

229 (81%)   55 (19%) 284 (100%)

176 (85%)   32 (15%) 208 (100%)

3,032 (65%) 1,630 (35%)   4,662 (100%)

  810 (58%)   587 (42%) 1,397 (100%)

Figure 4.  Distribution of indicative s‑ and b‑forms according to person and number distinctions in the Old English part of the Helsinki Corpus

3.1.1 1st Person Singular In my database, there were 284 examples of be‑verbs in the 1st person singular indicative. 229 (81 percent) belonged to the s‑paradigm (eom/eam) and 55 examples (19 percent) were b‑forms. Among the latter there are 31 (56 percent) contaminated forms (biom/beom), which appear almost exclusively in Mercian (Vespasian Psalter, Rushworth Gospels) and Northumbrian (Durham Ritual, Lindisfarne Gospels) texts, i.e. in the Anglian dialects. A few examples are recorded in Old English poetic texts, particularly in some of the Riddles, which are translations of earlier, probably Anglian, versions. The contaminated forms closely correspond to the Old Saxon forms, with Old Saxon being a language which had derived from one of the source dialects of Old English. The main difference between the use of the s‑ and the b‑form in the 1st person singular indicative is that between present tense and future. The majority of the b‑forms (67 percent) occur in translations from Latin, usually rendering Latin futures, as in example (2): (2) L:

respondens autem iesus dixit o generatio infidelis et peruersa usque quo ero [1 Ps Sg. Fut.] apud uos et patiar uos ... OE: se hælend cuoeð la cneoreso ungeleafull & wohfull hu longe ic biom [1 Ps Sg. Prs. b‑form] mið iuh & ic ðola iuih ...  (Lindisfarne Gospels, Luke 9.41) ‘And Jesus, answering, said: O faithless and perverse generation, How long shall I be with you and suffer you? ...’

The use of beon and wesan in Old English 



Latin present tense, sum, is generally translated by eom/eam, as in (3): nam et ego homo sum [1 Ps Sg Prs.] sub potestate constitutus habens sub me milites ... OE: wiotudlice & ic monn eam [1 Ps Sg. Prs. s‑form] under mæhti geseted hæbbende under me cempa ...  (Rushworth Gospels, Matthew 8.9) ‘For I am a man under authority; having soldiers under me ...’ (3) L:

However, not in every case was the glossator absolutely sure how to translate the Latin form, as examples such as (4) illustrate, where both s‑ and b‑form, eam and beom, are alternatively used (see Latin vel ‘or’) to render the Latin future ero: dicebat enim intra se si tetigero tantum uestimentum eius salua ero [1 Ps Sg. Fut.] OE: heo cwæþ forþon in innan hire gif ic gehrine efne vel swa micel hrægl his hal ic eam [1 Ps Sg. Prs. s‑form] vel ic beom [1 Ps Sg. Prs. b‑form]. (Rushworth Gospels, Matthew 9.21) ‘For she said within herself, if I may but touch his garment, I shall be whole.’ (4) L:

As mentioned before, in addition to the present–future distinction, Old English makes use of the same stems to refer to different aspectual types of present states. The s‑stem, like in (3), expresses a general or current state, often combined with a past participle to indicate a stative passive (cf. also Kilpiö 1993). The b‑form instead focuses on a temporary state, a duration which may extend into the future, or can be understood as repetitive or habitual, see example (5).

(5) Saga hwæt ic hatte, oþþe hwa mec rære, þonne ic restan ne mot, oþþe hwa mec stæðþe, þonne ic stille beom [1 Ps Sg. Prs. b‑form]. (Riddles 3.72) ‘Tell me my name, and who it is rouses me when I may not rest or who restrains me when I remain silent.’ Sentence (3) is an example of what Jost (1909: 13) and Exter (1911: 20) characterize as a “concrete” sentence. The first person is a concrete being and the situation is expressed as a real fact. Sentence (5) would be considered as “abstract” according to their theory since the event is only a hypothetical one. It is interesting to note that in the 1st person singular the percentage of b‑forms is considerably lower (19 percent) than that of s‑forms (81 percent) in Old English, although in Old Saxon, which is so closely related to Old English, there are only contaminated b‑forms attested. Such contaminated b‑forms do appear in Old English as well, but are almost exclusively restricted to Northumbrian and

 Ilse Wischer

Mercian texts. They seem to represent the oldest layer of 1st person singular be‑verbs. In contrast, the s‑form (eom/eam) must be an innovation, which has its origin most likely in the South. In the earliest attested Old English texts, it is already the dominant form. Nevertheless, the b‑form is still in use and a functional distinction, especially between present tense (s‑form) and future reference (b‑form), is clearly observable. Laing (this volume) provides a detailed account on how b-forms as markers of future reference continue to be used in different dialects of early Middle English. 3.1.2 2nd person singular In my database there were 208 examples of 2nd person singular constructions with the verb to be in the present tense indicative. Of these examples 85 percent (176) appeared in the s‑form (eart/earð), and 15 percent (32) in the b‑form (bist). In their distribution, I could not observe any particular preferences with regard to the different dialects, genres or historical subperiods. Like with the 1st person singular indicative, the s‑form is used to refer to a present or general state, as for instance a characteristic property or an identification, such as in example (6). Here Christ is directly addressed, and he is identified as a particular being, which exists now, at the time of speaking, and in general: (6) ðu eart [2 Ps Sg. Prs. s‑form] cyninga cyningc cwicera gehwilces, ðu eart [2 Ps Sg. Prs. s‑form] sigefest sunu and soð helend ofer ealle gesceft angla and manna. (The Kentish Hymn, 87.15) ‘you are the king of all living kings, you are the victorious son and true saviour of all creatures of angels and men.’



In example (7), both s‑ and b‑forms appear in the same sentence: (7) ðu eart [2 Ps Sg. Prs. s‑form] soðlice simle halig, and ðu eart [2 Ps Sg. Prs. s‑form] ana æce dryhten, and ðu ana bist [2 Ps Sg. Prs. b‑form] eallra dema cwucra ge deadra, ... (The Kentish Hymn, 88.36) ‘you are truly always holy, and you alone are the Lord forever, and you alone will be the judge over the living and the dead ...’

Eart assigns a property to the Lord that holds now and forever (simle ‘always’, æce ‘eternally’), while bist may refer to the role of the Lord in future, when the Last Judgement is supposed to take place. The general function of bist in contrast to eart is clearly to mark future tense. This becomes most obvious in translations from Latin, where eart regularly translates Latin es, but bist translates Latin eris or future passives, as in (8) and (9):

The use of beon and wesan in Old English 



Ecce tu pulchra es [2 Ps Sg. Prs.] amica mea ecce tu pulchra occuli tui columbarum OE: heono ðu feger arð [2 Ps Sg. Prs. s‑form] mego min heonu ðu feger ego ðino culfra. (The Durham Ritual, 3.6) ‘How beautiful you are, my maid, how beautiful you are with your dove eyes.’ (8) L:

(9) L: Et ecce eris [2 Ps Sg. Fut.] tacens et non poteris loqui usque in diem ... OE: & heono ðu bist [2 Ps Sg. Prs. b‑form] suigendæ & ne mæge ðu gesprece oðð on doege ... (Lindisfarne Gospels, Luke 1.20) ‘And behold, thou shalt be dumb and shalt not be able to speak until the day ...’ The 2nd person singular forms show a similar distribution as the 1st person singular forms. The s‑forms (eart) clearly predominate with 85 percent over the b‑forms (bist) with only 15 percent, although in Old Saxon the only attested form is the b‑form. The origin of the 2nd person s‑form (eart), like that of the 1st person (eam) may be attributed to a very early North Germanic influence, maybe already on the Old Anglian and Jutish dialects. Krause (1948: 100) lists the forms am (1sg) and est/ert (2sg) as the only forms for Old West Norse. However, a functional distinction between s‑ and b‑forms can be attested for all Old English dialects and all subperiods. 3.1.3 3rd person singular The 3rd person singular is by far the most frequent type of construction in the data. There are altogether 4,662 examples with a form of the verb to be in the present tense indicative. 65 percent of these (3,032 examples) are attested with an s‑form and 35 percent (1,630 examples) with a b‑form. Both forms occur in almost all of the texts and there is no particular preference with regard to the dialect or the time period, nor does the distribution across genres (as distinguished in the Helsinki Corpus) appear to follow a clear pattern of usage, see Figure 5 (next page). The preponderance of bið‑forms in medical texts may be due to frequent statements about future or general consequences of particular medical treatments. It can be observed that the b‑form (bið), which Schumacher (2007: 201– 202) calls the “marked” form, is regularly used when the author predicts a consequence of something which may have future or habitual implications, as in (10) and (11):

 Ilse Wischer

Genre Documents Verse Law Medicine Philosophy Religion Preface History Bible Astronomy Homily Rule Geography Travelogue Biography Fiction ‡ §

850–950

950–1050

1050–1150

is 0‡ bið bið is bið is is is – – – – – – –

is is is bið – is – bið is is is is is is is is

is –§ is – bið is is is – bið is is – – is –

No 3rd person singular indicative present tense used. No texts of that genre included.

Figure 5.  Dominance of bið‑ or is‑forms in the various genres

(10) Wifgemanan to donne, nim drige fearres sceallan, wyrc to duste, oððe elcor gnid on win, & drince gelome. He bið [3 Ps Sg. Prs. b‑form] þy gearwra to wifþingum. (Quadrupedibus, 752) ‘To have intercourse, take dry testicles of a bull, crumble them into dust, or else crumble on wine, and drink it frequently. Consequently, he will be better prepared for intercourse. (11) Gif XXX, he bið [3 Ps Sg. Prs. b‑form] lange seoc, þæhweðere ariseð.  (Prognostications, 30) ‘If 30 (the 30th night of the month), he (consequently) will be sick for a long time, yet nevertheless he will get well.’ In (12) the state of remembering something is not a current state, but is expressed as depending on a particular situation in future or generally:



The use of beon and wesan in Old English 

(12) He sorgað ymb ða, & bið [3 Ps Sg. Prs. b‑form] ðara suiðe gemyndig, & forgiett his selfes, ðonne he suiðor his mod gebint to ðam unnyttran weorcum ðonne he ðyrfe. (Alfred’s Cura Pastoralis, 4.37.19) ‘He cares for them, and remembers them a lot, and forgets himself, when he directs his mood to the useless works more than he should.’ Furthermore, the b‑form is often used in the protases of conditional clauses, which also have a future or repetitive implication. Such constructions are typical in law texts. Since the situation described is a hypothetical one, bið can vary here with the subjunctive sie, see examples (13) and (14): (13) Gif se nægl bið [3 Ps Sg. Ind. b‑form] ofaslegen, ðam sculon V scillinga to bote. (Laws (Alfred), 56.1) ‘If the nail is cut off, he shall be paid 5 shillings as compensation.’ (14) Gif se midlesta finger sie [3 Ps Sg. Subj. s‑form] ofaslegen, sio bot bið [3 Ps Sg. Ind. b‑form] XII scillinga. (Laws (Alfred), 58) ‘If the middle finger is cut of, the compensation shall be 12 shillings.’ In (14) the subjunctive sie is probably preferred to avoid a repetition of bið, which occurs in the apodosis of the same sentence. The “unmarked” s‑form (is) is found in simple statements where a term or a concept is explained, as in (15) and (16): (15) þy læs cild sy hreosende, þæt is [3 Ps Sg. Ind. s‑form] fylleseoc, ...  (Quadrupedibus, 434) ‘lest a child be falling, that is epileptic, ...’ (16) æt ærestan we lærað, þæt mæst ðearf is [3 Ps Sg. Ind. s‑form], þæt æghwelc mon his að & his wed wærlice healde. (Laws (Alfred), 1) ‘first we teach that the most important necessity is that each person truly keeps his oath and pledge.’ The b‑form in the 3rd person singular is clearly an innovation in Old English. It has no parallel in any other West‑ or North Germanic language. There is, however, a similar form in Cymric, which is used in habitual contexts (cf. Keller 1925: 58). Thus, it is possible that this form was borrowed from Insular Celtic and integrated into the Old English b‑paradigm, or, it was formed in analogy to the inherited 1st and 2nd person singular b‑forms. Considering that this form is an innovation in Old English, it is even more remarkable that the percentage of its use (35 percent) in relation to that of the s‑form (65 percent) is even higher than that of the inherited 1st (19 percent) and 2nd person singular b‑forms (15 percent). This means that the borrowing or creation of the form bið together with the formation of the double paradigm must have happened very early after the settlement on the British Isles.

 Ilse Wischer

3.1.4 Plural The highest percentage of b‑forms compared to s‑forms could be found among the plural forms.4 Of all 1,397 examples, 58 percent (810 items) occurred in the s‑form and 42 percent (587 items) in the b‑form. Most of the s‑forms are variants of sind/sindon (781 examples). The form aron is comparatively rare (only 29 examples) and is restricted to Northumbrian texts. Northumbrian texts also contain the form biðon instead of beoð. Otherwise the distinction between the use of sind and beoð seems not to depend on the dialect or subperiod. The s‑form is used for the intransitive verb meaning ‘exist’, as in (17), but in the same contexts, even in the same text, we can alternatively find beoð, as in (18). (17) ðonne syndan [3 Ps Pl. Ind. s‑form] oþere wif þa habbað eoferes tuxas & feax oð helan side, & oxan tægl on lendunum. (Marvels, 27.1) ‘Then there are/exist other women, who have boar’s teeth and hair to their heels, and an ox tail at their loins.’ (18) þonne syndon [3 Ps Pl. Ind. s‑form] oþere ealond suð from brixonte on þon beoð [3 Ps Pl. Ind. b‑form] men buton heafdum. þa habbað on hyra breostum heora eagan & muð. Hy seondon [3 Ps Pl. Ind. s‑form] eahta fota lange & eahta fota brade. ðar beoð [3 Ps Pl. Ind. b‑form] dracan cende, þa beoð [3 Ps Pl. Ind. b‑form] on lenge hundteontiges fotmæla lange & fiftiges. Hy beoð [3 Ps Pl. Ind. b‑form] greate swa stænene sweras micle. (Marvels, 15.1–16.2) ‘Then there are / exist other islands south of the Brixonte in which there are / exist men without heads. They have their eyes and mouth in their chests. They are eight feet tall and eight feet wide. There are (usually) dragons born which are 150 feet long. They are as big as huge stone pillars.’ In (18) it is difficult, if at all possible, to identify a functional or contextual difference between the uses of the b‑ and s‑forms.5 The first two forms, syndon and beoð, both mean ‘exist’ and are used in almost the same contexts. Then seondon and beoð are similarly used as copula to link the subject with a complement describing the size of the subject. Only in the construction ðar beoð dracan cende the use of the b‑form can be accounted for in terms of a dynamic passive, which again has a repetitive or habitual sense. 4. Due to lack of time I treated the three persons in plural together. For data on the individual persons see Kilpiö (1993: 98). 5. A comparison with the Latin original is not very helpful either (‘Est etiam in Brixonte insula, in qua nascuntur homines sine capitibus qui in pectore oculos et ora habent. altitudine novem pedum. et latitudine octo. hos epifagos vocamus. Nascuntur autem ibi et dracones longitudine centum quinquaginta pedum grossitudine columnarum’ (Graff 1827: 196)).

The use of beon and wesan in Old English 



The plural form beoð is, like the 3rd person singular bið, an innovation in Old English. Again, there is no parallel in the other North Sea Germanic dialects. And still its use was firmly established when the first Old English texts appeared. With 42 percent of all be‑verbs in plural its percentage frequency is even higher than that of bið. As in the case of bið, we can again assume direct borrowing from Insular Celtic, since in Cymric we can find parallel forms to the one attested in the Northumbrian dialect, biðon. On the other hand, analogical formation on the basis of the inherited 1st and 2nd person forms might be even more likely to account for the phonological form beoð. The functional distinction, however, in the use of s‑ and b‑forms has clear parallels in Insular Celtic. 3.2

Subjunctive

It is interesting to note that in Old English the b‑forms also entered the subjunctive paradigm, while in all other West Germanic languages, the s‑form remained the only one for the subjunctive. In English, be later even replaced the old subjunctive form sie(n), which was still the dominant one in Old English. Of a total of 1,476 subjunctive constructions with the verb ‘to be’, 87 percent (1,277 items) are expressed with s‑forms, while only 13 percent (199 items) contain b‑forms:

s‑form b‑form Σ

1st Pers Sg

2nd Pers Sg

3rd Pers Sg

Plural

Σ

9 (56%) 7 (44%) 16 (100%)

33 (80%) 8 (20%) 41 (100%)

986 (89%) 117 (11%) 1,103 (100%)

249 (79%) 67 (21%) 316 (100%)

1,277 (87%) 199 (13%) 1,476 (100%)

Figure 6.  Distribution of subjunctive s‑ and b‑forms according to person and number distinctions in the Old English part of the Helsinki Corpus

These data support Brunner’s (1951: 259) observation on the frequency distribution of “optative” forms. He claimed that in Old West Saxon and in poetic texts beo/beon is used much more rarely than sie/sien. Brunner further points out that in Northumbrian and West Mercian the b‑form is not found at all. For Northumbrian this can be confirmed by my studies, as I could not find any single subjunctive b‑form in my Northumbrian data (cf. Figure 7). In the Mercian texts, however, which I was not able to subdivide clearly into West and East Mercian,6 I could find both s‑ and b‑forms, as in example (19): 6. It is not clear which texts Brunner refers to when he claims that in West Mercian the b‑form is not found at all.

 Ilse Wischer

Northumbrian beo(n) sie(n)    –850 850–950 950–1050 1050–1150

0 – 0 –

50 – 30 –

Mercian beo(n) sie(n) 0 1 5 3

86 21 17 20

Kentish beo(n) sie(n) – 0 – –

– 25 – –

West Saxon beo(n) sie(n) – 2 5 4

– 44 21 16

Figure 7.  Average use of subjunctive beo(n)/sie(n) according to dialect and time (normalized figures, per 10,000 words)

(19) & ðis sio [3 Ps Sg. Subj. s‑form] gelæst to sancte Michaelæs tide. & bio [3 Ps Sg. Subj. b‑form] he ælces wites wyrðe. & gif hwilc forwyrht man hiowan gesæce bio [3 Ps Sg. Subj. b‑form] se ðingað swa hit medlic sia [3 Ps Sg. Subj. s‑form] be ðæs geltes meðe.  (Documents 3 (Robertson) 60, 8/9) ‘& this (shall) be done to St. Michael’s time. & he (shall) deserve any punishment. & if any ruined man comes to visit people of a religious house, the agreement (shall) be as it may be befitting according to the size of the crime.’ For East Mercian, with reference to the Rushworth Gloss, Brunner (1951:  259) states that subjunctive sie does not occur at all. This again could not be verified by my data. I found both forms in the Rushworth Gloss (28 examples with sie and 4 with beo), as in examples (20) and (21): (20) gif þu godes sunu siæ [3 Ps Sg. Subj. s‑form] gecwæþ þæt þas stanes hlafes beon [infinitive b‑form] vel gewærþe. (Rushworth Gospels, Matthew 4.3) ‘if thou be the Son of God, command that these stones will be or become bread.’ (21) (\sic ergo uos orabitis pater noster qui es in caelis sanctificetur nomen tuum\) þus ge þonne eow gebiddað fæder ure þu þe in heofunum earð beo [3 Ps Sg. Subj. b‑form] gehalgad þin noma. (Rushworth Gospels, Matthew 6.9) ‘after this manner therefore pray ye: Our Father which art in heaven, Hallowed be thy name.’ The difference seems to be that while the s‑form has a present relevance here, the b‑form refers to a future event, usually with the original sense of this verb, namely ‘become’. So in the Rushworth Gospels, beon is in all other occurrences, just like in (20), equated with the verb weorþan ‘become’ (beon vel ‘or’ gewærþe).

The use of beon and wesan in Old English 



As we can see, there is no dialect in which subjunctive beo(n) clearly dominates. Its use, however, is very limited in Northumbrian and Kentish.7 Thus, it will probably not have had its origin in these dialects. The texts in which subjunctive b‑forms predominate over subjunctive s‑forms can almost all be assigned to Ælfric: Ælfric’s Catholic Homilies, Ælfric’s Preface to Catholic Homilies I, Ælfric’s Preface to Genesis, Ælfric’s First and Second Letters to Wulfstan, The Old Testament (translated partly by Ælfric); additionally there is Byrhtferth’s Manual (on astronomy) (950–1050), and finally The Life of St. Chad (1050–1150), whose author is unknown. All texts are Westsaxon8 and date from the late tenth to early eleventh century. Nevertheless, if we consider all texts in our corpus, we must admit that sie is still the prevailing form of the subjunctive at the end of the Old English period. It is not clear on what evidence Brunner (1951: 259) based his conclusion when he says that in late West Saxon beo/beon advances more and more, and that it had almost completely displaced sie/sien from the eleventh century on. My corpus findings do not corroborate this development. 3.3

Imperative and infinitive

Other than with the subjunctive, the dominant form for the imperative and the infinitive in Old English is the b‑form. Here, Old English again behaves in a unique way compared to the other West Germanic languages.9 Only in English the imperative and infinitive is formed with the b‑stem. And this has its origin early in Old English. Although there are only 60 imperatives in our corpus, 68 percent of the singular and 94 percent of the plural constructions are formed with beo. The infinitive beon is used in 91 percent of the 358 examples. In the rest of the examples we find mainly forms of wesan, like in Old Saxon, and a few instances of the imperative with sie, like in Old High German, see Figure 8:

s‑form b‑form Σ

Imp. Sg.

Imp. Pl.

Infinitive

14 (32%) 30 (68%) 44 (100%)

1 (6%) 15 (94%) 16 (100%)

32 (9%) 326 (91%) 358 (100%)

Figure 8.  s‑forms and b‑forms in the imperative and infinitive 7. There is, unfortunately, only one Kentish text in the Helsinki Corpus, but Campbell also lists Kentish as a dialect for which sie was typical. 8. The Life of St Chad is probably a West Saxon copy of an earlier Mercian text, although this issue is much disputed (cf. Vleeskruyer 1953). 9. Lucia Kornexl (personal correspondence) made me aware of the fact that there exists an imperative form bi in Bavarian. But this is mainly used in one collocation: Bi schdaad! “Be quiet” (cf. Merkle 1996). Renn & König (2006: 77) assume an origin of this form in analogy to the 2nd person singular indicative du bist.

 Ilse Wischer

As there appear no imperatives in the Northumbrian texts of our data, I cannot verify the general assumption that in Northumbrian there are only wes‑imperatives (cf. Brunner 1951: 259; Campbell 1991: 351). It would, however, make sense, if we consider that in this dialect beo‑forms are not used for the subjunctive either. The same is true for the infinitive. Here I could confirm Brunner’s findings that in Northumbrian the infinitive is only wesa or wosa. So we might assume that there is the same conservatism with the imperative in Northumbrian. Brunner (1951:  259) further claims that the wes‑imperative, as well as the wesan‑infinitive, on the other hand, were missing in Older West Saxon. If this is true, we must assume a very early replacement of wes/wesan, which had been the regular imperative and infinitive in Old Saxon, by beo/beon under Celtic influence. In the Celtic languages, the verb to be in the subjunctive, imperative and the verbal noun is only based on one stem, so that the forms all have an initial b‑ and are etymologically related to the b‑forms of the indicative. See the forms in Middle Welsh as given in Figure 9:

Sg.

Pl.

1 2 3 1 2 3

Subjunctive

Imperative

bwyf bych bo bom boch bont

byd bit bydwn bydwch bint

Verbal noun

bot

Figure 9.  The forms of the subjunctive, imperative and verbal noun in Middle Welsh10

In our data the 11 wes(aþ)‑imperatives occur mainly in Old English verse, which are all West Saxon translations from earlier versions in other dialects, mainly Anglian. The only plural wesaþ‑imperative could be identified in the Rushworth Gospel, which is a Mercian text. In this dialect probably both wes‑ and beo‑imperatives were in use (cf. also Campbell 1959: 351), since in example (22) the glossator translates Latin estote with the two alternating forms.

10. I would like to thank Stefan Schumacher (personal correspondence) for the information on the subjunctive, imperative and verbal noun forms of the verb to be in the Celtic languages, and particularly in Middle Welsh.

The use of beon and wesan in Old English 



Ecce ego mitto uos sicut oues in medio luporum estote [pl. imperative] ergo prudentes sicut serpentes et simplices sicut columbe. OE: henu ic sende eow swa swa scep in midde uulfum bioþ [pl. imperative b‑form] vel wesaþ [pl. imperative s‑form] forþon snottre swa swa nedra & bilwite swa swa culfra. (Rushworth Gospels, Matthew 10.16) ‘Look, I send you forth as sheep in the midst of wolves: be[beon] or be[wesan] therefore wise as serpents, and harmless as doves.’

(22) L:

A similar alternation can be found in the same text between beon and wesan as infinitive forms, as in example (23): prohibebat autem eum iohannis dicens ego a te debeo baptizari [passive infinitive] et tu uenis ad me? OE: iohannes þonne werede him cweþende ic sceal from þe beon [infinitive b‑form] vel wesa [infinitive s‑form] deped vel fullwihted & ðu cymest to me? (Rushworth Gospels, Matthew 3.14) ‘but John didn’t allow it, saying, I shall be[beon] or be[wesan] dipped or baptized by you & and you are coming to me?’ (23) L:

So Brunner is not right when he argues that in Mercian the only infinitive is beon. In my data wesa(n) is found in all dialects except for the Kentish one, although it has to be admitted that there is only one single Kentish document in the Old English part of the Helsinki Corpus. Beon, on the other hand, is found in all dialects except Northumbrian. The b‑form for the infinitive as well as for the imperative is quite obviously a very early innovation, which did not originate in the Northern dialect area. 4. Conclusion In the light of the findings presented above, it can be claimed that the 1st and 2nd person singular b‑forms in the Old English present indicative paradigm are inherited from Proto‑West Germanic. While in the other West Germanic languages they form a mixed paradigm with s‑forms, in Old English alone a double paradigm emerged, which can only be assigned to Celtic influence. The 3rd person singular form bið and the plural form beoð, which have no parallels in the other North Sea Germanic dialects, had been firmly established as parts of the Old English double paradigm by the time of the earliest Old English texts. The difference in use between beon and wesan is one between future and present, and partly, parallel to the Celtic languages, one between habitual present and current present. Further investigations will be needed to confirm these observations.

 Ilse Wischer

In Old English, as the only West Germanic language, b‑forms also penetrated into the subjunctive paradigm and eventually even displaced the inherited sie(n)‑forms. This process originated most likely in the South (West), not in the North, again as a contact phenomenon, based on similar b‑verbs in the Celtic languages. The same is true for beon in the imperative and infinitive. The displacement of the original wes‑forms must have taken place at a very early time, possibly right after the settlement, which again did not affect the North. References Besch, Werner, Anne Betten, Oskar Reichmann & Stefan Sonderegger, eds. 2000. Sprachgeschichte: ein Handbuch zur Geschichte der deutschen Sprache und ihrer Erforschung. (= Handbücher zur Sprach- und Kommunikationswissenschaft, 2) 2nd edition. Vol. 2. Berlin & New York: Mouton de Gruyter. Bethge, Richard. 1900. “Konjugation des Urgermanischen”. Laut– und Formenlehre der altgermanischen Dialekte ed by. Ferdinand Dieter, 345–391. Leipzig: Reisland. Brunner, Karl. 1951. Die englische Sprache: ihre geschichtliche Entwicklung, vol. 2. Halle: Niemeyer. Campbell, A. 1991 [1959]. Old English Grammar. Repr. with corrections. Oxford: Clarendon. Dieter, Ferdinand, ed. 1900. Laut– und Formenlehre der altgermanischen Dialekte. Leipzig: Reisland. Exter, Otto. 1911. Beon und wesan in Alfreds Übersetzung des Boethius, der Metra und der Soliloquien. Kiel: Fiencke. Graff, Eberhard Gottlieb, ed. 1827. Diutiska: Denkmäler deutscher Sprache und Literatur, aus alten Handschriften zum ersten Male theils herausgegeben, theils nachgewiesen und beschrieben. 2nd Volume. Stuttgart & Tübingen: Cotta‘sche Buchhandlung. Holthausen, Ferdinand. 1900. Altsächsisches Elementarbuch. Heidelberg: Winter. Hutterer, Claus Jürgen. 1999. Die Germanischen Sprachen. Ihre Geschichte in Grundzügen. Wiesbaden: Albus. Jost, Karl. 1909. Beon und wesan, eine syntaktische Untersuchung. (= Anglistische Forschungen, 26) Heidelberg: Winter. Hogg, Richard. 1992. “Phonology and Morphology”. The Cambridge History of the English Language. Volume 1: The Beginnings to 1066 ed. by Richard Hogg, I, 67–167. Cambridge: Cambridge University Press. Keller, Wolfgang. 1925. “Keltisches im englischen Verbum”. Anglica: Untersuchungen zur Englischen Philologie, Alois Brandl zum Siebzigsten Geburtstage überreicht, I, Sprache und Kulturgeschichte. (= Palaestra, 147), 55–66. Leipzig: Mayer & Müller. Kilpiö, Matti. 1992. “BEON. With Attested Spellings by Robert Millar, Using Materials Assembled by Haruko Momma”. Dictionary of Old English. Toronto: Pontifical Institute of Mediaeval Studies. Kilpiö, Matti. 1993. “Syntactic and Semantic Properties of the Present Indicative Forms of the Verb to be in Old English”. Early English in the Computer Age: Explorations through the Helsinki Corpus ed. by Matti Rissanen, Merja Kytö & Minna Palander-Collin, 97–116. Berlin & New York: Mouton de Gruyter.



The use of beon and wesan in Old English  Kilpiö, Matti. 1997. “On the Forms and Functions of the Verb to be from Old to Modern English”. English in Transition. Corpus-Based Studies in Linguistic Variation and Genre Styles ed. by Matti Rissanen, Merja Kytö & Kirsi Heikkonen (= Topics in English Linguistics, 23), 87–120. Berlin & New York: Mouton de Gruyter. Krause, Wolfgang. 1948. Abriss der altwestnordischen Grammatik. Halle: Niemeyer. Lass, Roger. 1994. Old English. A Historical Linguistic Companion. Cambridge: Cambridge University Press. Lutz, Angelika. 2009. “Celtic Influence on Old English and West Germanic”. English Language and Linguistics 13.2: 227–249. Merkle, Ludwig. 1996. Bairische Grammatik, 6th edition. München: Hugendubel. Mitchell, Bruce. 1985. Old English Syntax, vol. I. Oxford: Clarendon. Prokosch, Eduard. 1939. A Comparative Germanic Grammar. Philadelphia: Linguistic Society of America. Renn, Manfred & Werner König. 2006. Kleiner Bayerischer Sprachatlas. München: dtv. Robinson, Orrin W. 1992. Old English and its Closest Relatives: A Survey of the Earliest Germanic Languages. London: Routledge. Schumacher, Stefan. 2007. “Die Deutschen und die Nachbarstämme: Lexikalische und strukturelle Sprachkontaktphänomene entlang der keltisch–germanischen Übergangszone”. Johann Kaspar Zeuß im kultur- und sprachwissenschaftlichen Kontext (19. bis 21. Jahrhundert), Kronach 21.7. –23.7.2006 ed. by Hans Hablitzel & David Stifter unter redaktioneller Mitarbeit von Hannes Tauber (= Keltische Forschungen, 2), 167–207. Wien: Praesens. Streitberg, Wilhem. 1974 [1896]. Urgermanische Grammatik: Einführung in das vergleichende Studium der altgermanischen Dialekte. 4th edition. Heidelberg: Winter. Traugott, Elizabeth Closs. 1992. “Syntax”. The Cambridge History of the English Language. Volume 1: The Beginnings to 1066 ed. by Richard Hogg, 168–289. Cambridge: Cambridge University Press. Vleeskruyer, Rudolf, ed. 1953. The Life of St. Chad. An Old English Homily. Amsterdam: North–Holland Publishing Company. Wright, Joseph. 1906. An Old High German Primer. 2nd edition. Oxford: Clarendon. Zupitza, Julius, ed. 1880. Ælfrics Grammatik und Glossar. Berlin: Weidmannsche Buchhandlung.

The reflexes of OE beon as a marker of futurity in early Middle English Margaret Laing*

University of Edinburgh In Old English the finite forms of the b‑root for ‘be’ (beo, bist, bið, etc.) were more likely to appear in contexts involving futurity than the s‑root (eom, eart, is, etc). The use of the b‑root for future continues into Middle English. During the compilation of LAEME, we have observed that the complex and variable Old English distinction can become simplified and systematized. In early Middle English the use of b‑forms in the present indicative singular is in some text languages1 restricted entirely to future senses. In the areas where the b‑root is the norm for present indicative plural, this system is confined to the singular. But in the North and to a certain extent the North Midlands, where ar-/er‑forms are available, the system is extended into the plural. Ilse Wischer’s contribution to this volume offers fascinating and detailed insights into the different forms of the verb ‘to be’ in Old English and their distinctive functions. This paper looks mainly at subsequent developments. It therefore only briefly summarizes the Old English distinctions as background to a micro-dialectal study of three subsystems that emerge during early Middle English. Their identification gives rise to further questions that might reward investigation in the future. * These observations arise from work on early Middle English manuscript texts undertaken at the Institute for Historical Dialectology, Linguistics and English Language, School of Philosophy, Psychology and Language Sciences, University of Edinburgh towards the compilation of A Linguistic Atlas of Early Middle English (LAEME). This research project was supported from 2000-2006 by AHRC for which gratitude is here expressed. I also thank Roger Lass, Keith Williamson, the anonymous reviewers and the volume editors for very helpful comments on earlier drafts. 1. The term “text language” is adopted from Suzanne Fleischman (2000: 34), who used it to refer in general to historical languages for which our only informants are written texts. In this paper the term is used as it is in Laing & Lass (2008: see esp. Introduction, Chapter 1) as the surviving written English of some particular scribe. A text language may represent the written dialect of the scribe who perpetuates it; or (in the case of a literatim copyist), that of the exemplar from which he copies. In the first case, a single text language can comprise the usage of any number of surviving texts, either composed by a particular scribe, or translated by him

 Margaret Laing

1. Background 1.1.

Futurity in Old English

As in Present-Day English, morphological tense markers in Old English are confined to past and present (or non-past). Temporal relations may be signed more fully by context and/or by temporal adverbs and conjunctions. Present (or nonpast) can also express the future: (1) & ic arise of deaðe on þam þriddan dæge and I will-arise from death on the third day  (quoted from Traugott 1992: 180-182) The verb to be appears to be at least a partial exception to this rule. It is one of the socalled “anomalous verbs” in Old English, being both irregular and defective. Its full conjugation in English (from Old English to Present-Day English) is made up from a union of the surviving forms of three originally distinct and independent verbs: a. the s‑root, i.e. the original IE substantive verb with stem *h1es‑, Skr. as‑, ′s‑, Gr. εσ‑, L. es‑, ′s‑, PGmc. *es‑, *′s‑. This has no surviving past tense in IndoEuropean languages. b. the b‑root, i.e. IE *bheu‑ Skr. bhū‑, bhaw‑, Gr. ϕυ‑, L. fu‑, PGmc. *βeu‑, *βeo‑, OE bēon ‘to become, come to be’. This also has no surviving past tense. c. the w‑root, i.e. the verb with stem *wes‑, Skr. vas‑ ‘to remain’, PGmc. *wes‑ Gothic wis‑an ‘to remain, stay, continue to be’, OS, OE, OHG wesan, OFris. wes‑a, ON ver‑a. This provides the past tense in English. Other parts of wesan fell out of use during the Old English period when it was a defective strong verb of Class V, subject to Verner’s Law: ind. sg. 1st and 3rd wæs, 2nd w󰀳xre, pl. w󰀳xron, subj. w󰀳xre(n) (Campbell 1959: §768). Only roots (a) and (b) are relevant to the present discussion.

from an exemplar or exemplars whose language he translates into his own dialect. In the second case, a single literatim copyist may provide us with more than one text language. A text language may be homogeneous dialectally or mixed.

Reflexes of OE beon as a marker of futurity 



1.2

The OE bēon paradigm (Taken from Campbell 1959: §768 (d)) the s- root

Indicative Present Sg. 1 eom I am 2 eart 3 is Pl. sindon, sint [Anglian erun, arun]

the b- root

bēo I shall be2 bist biþ bēoþ

Subjunctive Present Sg. sīe Pl. sīen

bēo bēon

Imperative Sg. Pl.

bēo3 bēoþ

Infinitive

bēon

The present forms of weorþan are also sometimes used in Old English and in early Middle English to express the future.4 The verb weorþan is not discussed in this paper, but a more detailed study of futurity in Old and early Middle English would certainly need to take account of it.5

2. These are Campbell’s glosses. See further the quotation from Mitchell (1985) in §1.3 below, qualifying the implication that there was a clear-cut present/future distinction. 3. Wischer (this volume) also lists the rare s-root and w-root imperatives sie and wes, wesaþ. 4. Weorþan is a strong verb of Class III, which also has the senses ‘become, happen’. It is also used, in combination with forms of wesan, to provide the passive voice. 5. Cf. e.g. Hwa wæs æfre, oþþe is nu, oððe hwa wyrþ get æfter us ‘Who always was, or is now, or who shall be yet after us’ (Ælfred’s translation of Boethius’ Consolation of Philosophy 11.I.) and Ich shal mid one bare worde Do þat þi speche wrht forworþe ‘I shall with one single word make it that thy speech shall be destroyed’ (London, British Library, Cotton Caligula A ix, The Owl and the Nightingale lines 547–48, Language 1, Worcs, C13b2).

 Margaret Laing

1.3

OE beon and the expression of futurity

In Old English the finite forms of the b‑root for ‘be’ were more likely to appear in contexts involving futurity than the s‑root e.g. eom, is. Mitchell (1985:§659ff) notes considerable semantic complexity, but essentially follows Campbell’s (1959:§768) conclusions:6 [...] bēo expresses what is (a) an invariable fact, e.g. ne bið swylc cwenlic þeaw ‘such is not a queenly custom’, or (b) the future, e.g. ne bið þe wilna gad ‘you will have no lack of pleasures’, or (c) iterative extension into the future, e.g. biþ storma gehwylc aswefed ‘every storm is always allayed’ [...]; eom expresses a present state provided its continuance is not especially regarded, e.g. wlitig is se wong ‘the plain is beautiful’.7

1.4

The grammarian’s view

Amongst his numerous other works in English, the great homilist and hagiographer Ælfric of Eynsham also produced a Grammar. It is a grammar of the Latin language, and is rich in Latin examples, but the text itself is couched in Ælfric’s late-tenth-century West-Saxon English. This work appears to have had considerable contemporary popularity, since it still survives in ten Old English manuscripts (often accompanied by Ælfric’s Latin-English Glossary). There are also a number of fragments of the text surviving in other manuscripts, as well as some later copies.8 It is clearly of great interest for our perception of the “future tense” in Old English to see how its expression was understood by a contemporary grammarian. Ælfric’s examples indicate that the simple future in Latin is expressed in Old English by present indicative forms, with or without adverbial support, e.g. to merjen ‘tomorrow’ (quoted from Zupitza 1880 [2001]: 131): 6. Cf. a similar summary in Traugott (1992: 182-183). Traugott follows Mustanoja (1960: 583; cf. Jost 1909) in using wesan (non-historically) as the citation form for the s‑root as well as the w‑root. 7. These basic handbook summaries are perhaps sufficient to provide the background for the present study of early Middle English data. There has been a great deal of more recent work on the double paradigm for ‘be’ in Old English. See for instance Kilpiö (1992, 1993, 1997). Kilpiö (1997: 89) observes the general tendency for b‑forms rather than s‑forms to be used with the future, and that “conversely, deictic locatives or temporals linking the state or action to the present moment or situation are more common with non‑b‑forms than b‑forms”. For a convincing account of ‘the Celtic hypothesis’, that the double paradigm and its contrasting functions is the result of sub-stratal Celtic influence on Old English, see Lutz (2008) and works there cited. Cf. Wischer (this volume), who also offers a syntactic and dialectal survey. 8. For details of all the manuscript copies see Zupitza (1880 [2001]: iv-ix) and references there to Ker (1957).

Reflexes of OE beon as a marker of futurity 



(2) futuro tempore on toweardre tide amabo ic lufige gyt to dæg oððe to merjen, amabis þu lufast, amabit he lufað et pluraliter amabimus we lufjað, amabitis ge lufjað, amabunt hi lufjað. ‘ futuro tempore in future tense amabo “I love yet today or tomorrow”, amabis “thou lovest”, amabit “he loveth”; and in the plural amabimus “we love”, amabitis “ye love”, amabunt “they love”.’

1.5

Old English forms of the verb to be as a marker of futurity

For the verb to be, we find in Ælfric’s Grammar clear illustration of the relationships of the s‑root and the b‑root to the expression of futurity in Old English under the section headed De Verbo Passivo (‘Concerning the Passive Verb’). Compare the s‑root usage (3) with the b‑root usage (4), where adverbial support is nevertheless still supplied (quoted from Zupitza 1880 [2001]: 139-140): (3) Amor ic eom gelufod ys passivvum, swa swa we ær cwædon, amaris þu eart gelufod, amatur he ys gelufod; et pluraliter amamur we synt gelufode, amamini ge synd, amantur hi synd. ‘Amor “I am loved” is passive as we said before, amaris “thou art loved”, amatur “he is loved”; and in the plural amamur “we are loved”, amamini “ye are”, amantur “they are”.’



(4) tempore fvtvro amabor ic beo gelufod gyt, amaberis ðu bist, amabitur he byð; et pluraliter amabimur we beoð gelufode gyt, amabimini ge beoð, amabuntur hi beoð gelufode. ‘in future tense amabor “I [shall] be loved yet”, amaberis “thou [shall] be”, amabitur “he [shall] be”; and in the plural amabimur “we [shall] be loved yet”, amabimini “ye [shall] be”, amabuntur “they [shall] be loved”.’9

2. Late Middle English 2.1

Survival of ‘be’ as a marker of futurity

Mustanoja (1960: 583) summarizes the Old English division between the use of the s‑root10 and the b‑root, and further observes: “Traces of this old use of the 9. In these glosses I use shall throughout the paradigm to express simple future, as was most commonly the practice in late Middle English with all verbs, and also in early Middle English with verbs other than be. No deontic sense is here implied. The practice of using shall in the first person and will in second and third persons to imply simple future, with the reverse usage in each case implying obligation, is a tradition that emerged later in the history of English. In Scots, the future versus deontic paradigms of shall and will are commonly the reverse of those in Standard English. 10. Which he refers to as wesan, cf. note 6 above.

 Margaret Laing

b‑forms to express futurity occur in early ME and to some extent even later in the period” (cf. also Fischer 1992: 241). 2.2

Late Middle English grammars

Ælfric’s Grammar was designed to explain the morphology and syntax of Latin by means of examples translated into Old English. Unfortunately, we have nothing of this kind for early Middle English. Post 1066, until the late fourteenth century, French and not English was the medium of instruction in the schools: After the Norman Conquest, English fell out of use as the language of elementary instruction in Latin grammar, and no grammatical texts in Middle English survive from before the closing years of the fourteenth century. Ranulf Higden, who died c. 1363, was still able to complain in his Polychronicon that “pueri in scholis contra morem caeterarum nationum a primo Normannorum adventu, derelicto proprio vulgari, construere Gallice compelluntur.” [‘Since the first arrival of the Normans, contrary to the custom of other nations, boys in schools are obliged to abandon their own language and to construe in French. i(Thomson 1984: xi).]

As we know from John Trevisa’s interpolation into his translation of Higden’s Polychronicon, by 1385 (the time Trevisa was writing): “in al þe gramerscoles of Englelond childern leueþ Frensch and construeþ and lurneþ an Englysch” (quoted from Sisam 1970: 149). Once English began again to be used as the language of instruction and of the construal of Latin in schools, Latin grammars couched in English also started to reappear. Compared with Ælfric’s splendidly full treatment, the Middle English grammatical texts that emerged in the late fourteenth and fifteenth century (Thomson 1984) are very slight, and tend to be hardly more than checklists. Those that deal fully with the verbal conjugations agree that Latin simple future tense is expressed in English by the periphrastic use of shall: e.g. “Qwerby knowyst þe future tens? For it spekyth of tyme þat is to come, and hath þis Englysch wurd ‘schal’, as amabo ‘I schal louyn’” (Thomson 1984: 26 – Acedence text C line 429–31, from Cambridge, St. John’s College, MS F. 26 (163), fols. 1r–12r). There is rarely mention in these grammars of esse, sum or the other anomalous verbs in Latin, and therefore no Middle English examples useful for our present discussion on the forms of the verb to be are given. How, then, was the simple future expressed in English between these two periods for which we have evidence from contemporary grammarians, and how are we to interpret the early Middle English evidence for the forms of the verb to be?

Reflexes of OE beon as a marker of futurity 



3. Early Middle English The grammarian’s view

3.1

Although there are no new grammars in Middle English much before 1400, the early Middle English Tremulous Scribe of Worcester made a copy of Ælfric’s Grammar in the early thirteenth century (Worcester Cathedral, Chapter Library F 174, fols. 1r–63r). He updated the spellings of the English parts of the text, including the exemplary material, to those of his own thirteenth-century Worcestershire language. He often curtailed the English examples where there was deducible repetition, and also truncated the Latin, giving just the endings when the repeated root could be inferred. The passages from Ælfric quoted in (2) to (4) above appear in the Tremulous Scribe’s version as (5) to (7) below (transcribed from a microfilm of the manuscript):



(5) futuro tempore . amabo . ic lufie get to-dai . oþer tomorwen \ amabis . amabit . & pluraliter amabimus bitis . bunt (6) [A]mor . ic am ilufod is passiuum so we ær cweþon . amaris . þu ert ilufod . amatur . he is ilufod . & pluraliter amamur . we beoþ ilufod . amamini . amantur . (7) tempore . futuro amabor ic beo ilufed get . amaberis . þu bist amabitur he biþ . & pluraliter amabimur we beoþ amabimini . amabuntur .

It can be seen that the Tremulous Scribe copied the structure of Ælfric’s English with minimal formal updating. Did he make his copy of the Grammar only from antiquarian interest, or did the content still hold for thirteenth-century English usage? 3.2

A Linguistic Atlas of Early Middle English (LAEME)

The evidence from the LAEME corpus of tagged texts (CTT) is that the Old English practice of using the s‑root for “a present state” and the b‑root for (among other things) futurity continues in at least some dialects of early Middle English. Of the 167 text languages in the CTT, 49 show at least some examples of be-future. The process of tagging itself turned out to be a powerful heuristic. 3.2.1 Tagging of ‘be’ with future sense Consider the following from the version of Ancrene Riwle in Cambridge, Gonville and Caius 234/120:

 Margaret Laing

(8) O þen an half adomesdai shulen ure swarte sunnen On the one side on doomsday shall our black sins strongliche biclepien us of ure sawle murþre. O þer oþer strongly accuse us of our soul’s murder. On the other half stont rich-wisnesse þe nan rewþe is wid. dredful & side stands righteousness whom no mercy is with, dreadful & grislich. & grure-ful to bi-holden. Buuen us þe orre deme. grisly & gruesome to behold. Above us the angry judge. for ase softe as he is her; ase hard he bid þer. Ase For as soft as he is here, as hard he shall be there; as milde ase he is nu; ase sturne þenne. Lomb her; leon þar mild as he is now, as stern then; lamb here, lion there. The Ancrene Riwle11 was probably written in the first quarter of the thirteenth century (C13a1) in South Salop or North Herefords. The Gonville and Caius manuscript (G) dates from the third quarter of the century (C13b1) and belongs in N Worcs. It follows closely the structure of the original here as attested by a combination of the Cleopatra text (C) and the author’s revised version represented by the Corpus text (A). The Nero (N) and Titus (T) versions are also similar. All four of these copies probably date from C13a2. The example above is a paraphrase and expansion of the first part of a Latin quotation from “seint Anselme” which immediately precedes it in the text: Hinc erunt peccata accusancia . Illinc terrens iusticia supra iratus Iudex (‘On this side shall be the accusing sins. On the other side [shall be] terrifying justice. Above us the angry judge.’). The English version freely adapts the Latin, but it does show a strict response to the Latin tense structure. The overtly expressed simple future erunt ‘shall be’ combined with present participle accusancia ‘accusing’ is here expressed with periphrastic shall plus infinitive: shulen biclepien ‘shall accuse’. The Latin parallel construction has present participle terrens ‘terrifying’ presumably with erit ‘shall be’ understood. Here the Middle English does not supply the necessary future shal form that would create strict parallelism. Instead the Latin is paraphrased, the missing finite verb and the present participle terrens being rendered respectively by two present indicatives: stont rich-wisnesse þe nan rewþe is wid ‘stands righteousness whom no mercy is with’. Stont here could be taken as present indicative implying 11. The surviving early Middle English versions are found in Cambridge, Corpus Christi College 402 (A), London, British Library, Cotton Cleopatra C vi (C), London, British Library, Cotton Nero A xiv (N), London, British Library, Cotton Titus D xviii (T), and Cambridge, Gonville and Caius College 234/120 (G). The passage illustrated here does not form part of the tagged sample in the LAEME CTT for any of the texts except G.



Reflexes of OE beon as a marker of futurity 

future tense, or possibly as a state that exists at the moment of writing, while is expresses a general truth. With the next two instances of the verb to be (for which corresponding Latin is lacking) there is a contrast between present indicative expressing a state that exists at the moment of writing: is her ‘is here’ and what is, judging from the context, a clear expression of future time: bid þer ‘shall be there’. The A, C and N texts (all like G from the South-West Midland area) have the same syntactic structure as G, contrasting is and bið, but the N text also repeats bið before ‘then’ reinforcing the distinction between ‘is now’ and ‘shall be then’. The T text, which belongs further north, in Cheshire, preserves the distinction between the s‑root and b‑root but has the northerly form beos rather than the bid/ bið of the other texts. The sample from the G version of Ancrene Riwle used in the CTT was tagged for LAEME following our usual tagging procedures (LAEME, Introduction, Chapter 4). Parts of the above example are illustrated below, (a) in the tagged text format and (b) in the format (including tags) that is retrievable using the concordancing programme on the LAEME TASKS page. (a) Tagged text format (b) Concordance format12 $for/cj_FOR as/av>=soft/aj as/av /P13NM be/vps13 here/av $as/av>=_ASE ASE SOFTE AS HE IS HER $soft/aj_SOFTE $as/av_AS so/cj<= hard/aj /P13NM $/P13NM_HE ASE HARD HE $be/vps13_IS $here/av_HER BID [be/v-fut13 (C13b1) # 276 caiusart.tag] {.’} $so/cj<=_ASE there/av as/av>= mild/aj as/av /P13NM $hard/aj_HARD {\} yER *ASE MILDE ASE HE $/P13NM_HE $be/v-fut13_BID be/vps13 now/av so/cj<= stern/aj $there/av_yER IS NU ASE STURNE From these illustrations it can be seen that the tags assigned to the finite verb forms is and bid from our short example are different: $be/vps13 and $be/v-fut13. The beginning of a tag is signalled by $. The lexical element (lexel) appears between $ and / and is here the modern English citation form be. The grammatical element (grammel) comes between / and _ after which follows the manuscript 12. To save space, examples from now on will be given in “normal” rather than internal LAEME format, and without the tagging being included.

 Margaret Laing

form, here illustrated in “internal format” in capitals, viz IS and BID. The grammel for IS is vps13: v = verb; ps = present indicative; 1 = singular; 3 = third person. The grammel for BID substitutes ‑fut for ps. The ‑fut tag label is used only for forms of the verb to be and I have tried to confine it to examples where clear future is intended; though the dividing line between this and Campbell’s (see §1.2 above) “iterative extension into the future” is fuzzy. Often cases could be argued either way, though sometimes there are rather strong clues when Latin exempla in a text are (more or less) translated into Middle English, as in this example from the same text as above: (9) Quid confusionis. quid ignominie erit quando dissipatis What of confusion, what of ignominy shall be when, scattered foliis & dispersis. Vniuersa nudabitur turpitude leaves and dispersed, all shall be laid bare foulness sanies apparebit. hwuch schendlac. & hwuch sorhe filth shall appear. what ignominy and what sorrow bið þer hwenne alle þe leaues schulen beon towarpled. shall be there when all the leaves shall be scattered, & al þat fulþe schawen him & wringed and all that filth [shall] show itself and wrings/shall wring ut þat wursum biuoren alle wide world out that corruption before all [the] wide world Here the singular bið directly translates Latin erit ‘shall be’. Thereafter, the Middle English paraphrases the Latin text. The Latin ablative absolute dissipatis foliis et dispersis ‘the leaves having been scattered’ and singular future tense nudabitur and apparebit (‘shall be laid bare’ and ‘shall appear’) are combined in the Middle English periphrastic expression in the plural, schulen beon plus a past participle towarpled ‘shall be scattered’, followed by an infinitive schawen ‘show itself’.13 The text then continues with an expansion, for which there is no equivalent Latin text, with a present tense wringed ut ‘wrings/shall wring [itself] out’.14 In this text 13. The G version (which is a much shortened and reordered text of Ancrene Riwle) differs here syntactically from the other early Middle English texts of Ancrene Riwle. This seems to be the only plausible interpretation of its syntax in this context. In G’s language, schawen cannot be plural indicative (which is always expressed by the ‘–eth’-type ending). Formally it can only be infinitive or plural subjunctive. I take it to be infinitive depending on schulen in a zeugma construction. 14. The A, C, N and T texts (after whatever form of towarpled they have) all proceed with a version of the structure: & al $ fulþe schaweð him . & wringeð ut $ wursum biuoren alle þe wide world. The T version has the NWML -es variants for 3sg pres ind: scheawes and wringes. The G version’s



Reflexes of OE beon as a marker of futurity 

language there is no distinctive simple verbal form available to differentiate present from future sense of ‘be’ in the plural, b‑root beoð‑type spellings being the norm for the present plural indicative ‘are’. Using periphrastic schulen beon was therefore the only way for the scribe to translate the Latin simple future with an English equivalent that was distinctively future in expression. 3.2.2 Shall as a marker of tense or of obligation? By the early Middle English period shall has begun to emerge as a regular marker for the expression of futurity (Fischer 1992: 241, 250), especially with verbs other than to be, because for those verbs there is no formal distinction available. However, alongside this emergent function for future expression, shall in the present tense can still be used deontically in Middle English, in general statements of what is right or becoming, to mean ‘ought, should’. Sometimes therefore the use of shall can be ambiguous. Consider the following example from British Library, Egerton 613, Poema Morale (e text - C13a2-b1): (10) Ac drihte ne demð nanne man æfter his But the Lord not judges no man according to his bi-ginning ac al his lif sceal beo sich se buð his endinge beginning, but all his life shall/must be such as is/shall be his ending When tagging, we have to decide whether the periphrastic expression sceal beo implies merely futurity (future ‘shall’) or whether sceal here retains at least some of the sense of obligation it had in Old English (deontic ‘shall’). If sceal here represents future ‘shall’, we would have to supply an underlying ‘judged’ to make sense of the expression: a man’s life will be judged not according to how he begins it but according to how he ends it. If sceal beo implied future sense only, this might suggest that the finite simplex buð was not normally used for this purpose in this text language, and that its appearance in the second clause should be tagged as present indicative: $be/vps13 meaning ‘is’. If buð were taken to be a possible variant for the temporal expression of ‘shall be’ and tagged $be/v-fut13, it would imply in the writer a conscious avoidance of a strictly parallel expression of what would appear to be parallel (though inverted) senses: ‘his life shall be’ and ‘shall be his ending’. Of course the demands of metre have to be taken into account as well: it may be that the periphrastic variant was chosen merely to fill up the line. That aside, if we take sceal beo here to represent deontic ‘shall be’, both the semantic context and its formal contrast suggests that buð may be analysed as implying future ‘shall be’. wringed is clearly also intended to be 3sg pres ind. In this text language and <ð> are frequently interchangeable: cf. bid for bið in example (8) and see further Lass & Laing (2009).

 Margaret Laing

3.2.3 The importance of comparison In practice of course we rarely have to treat such a short example in isolation. The text of the Egerton (e) version of Poema Morale ends imperfectly, but it still runs to 368 lines of verse. It is vital when making analytical decisions, whether for tagging or other purposes, to look at variant usages within and across text languages as whole systems. There are six other surviving copies of Poema Morale, originating from different parts of the country and dating from between the late twelfth century to about 1300. The two verse lines above and the two lines following are set out here from all seven versions for (admittedly limited) comparison, and also to illustrate some of the variation in use of be and shall in different places and at different times. T = Cambridge, Trinity College B.14.52, fols. 2r–9v. C12b2. W Essex. L = London, Lambeth Palace Library 487, fols. 59v–65r. ca 1200. NW Worcs. D = Oxford, Bodleian Library, Digby 4, fols. 97r–110v. C13a1. W Kent. e = London, British Library, Egerton 613, fols. 64r–70v. C13a2-b1. SW Worcs. E = London, British Library, Egerton 613, fols. 7r–12v. C13a2-b1. SW Worcs. J = Oxford, Jesus College 29, fols. 169r–174v. C13b2. E Herefords. M = Cambridge, Fitzwilliam Museum, McClean 123, fols. 115r–120r. ca 1300. Gloucs. (11)

T: L: D: e: E: J: M:

Ac drihte ne demeð noman after his Ah drihten ne demeð nenne .M. efter his Drihte ne demeð nenne man. bi his Ac drihte ne demð nanne man æfter his Ac drithte ne demeð nanne man after his Ne schal nomon beon ydemed. after his Ac crist ne demeþ nanne man after his

(12)

T: L: D: e: E: J: M:

Ac al his lif Ah al his lif Al his lif ac al his lif ac his lif Ah dom Ac al

(13)

T: gief þe endinge is god al hit is god. & euel gief euel is þe ende L: gef $ is uuel al hit is uuel. & God gefe god his ende D: Ef se ende is euel hit is al euel & god ef god is se ende e: Ac gif þe ende is uuel eal hit is uuel. & gód gif gód is þenne E: Ac õif þe ende is euel al it is uuel & al god õif god is ende J: If þe ende is vuel. al hit is vuel. god yef vs god ende. M: Õef his ende is euel al hit is euel & god õef god beoþ his ende

sal scal sel sceal scal schal scal

ben teald after his bon suilch boð his ben iteald bi his beo sich se buð his beo swulc se bued his þolyen vych mon. after his beo his lif iteld suich is his

biginninge biginnigge biginninge bi-ginning bi-ginninge bigynnynge. ginnigge endinge endinge endinge endinge endinge endinge . endigge

Reflexes of OE beon as a marker of futurity 



(14)

T: L: D: e: E: J: M:

God gieue þat God geue $ God geue $ god gyue $ god õuue þat God yef vs Iesucrist leue þat

ure ende be ure ende bo ure ende bi ure ende beo ure ende beo vre ende ure ende beo

god & gieue god . & wite god . & wite gód . & wit god . & wite god . hwider god & witie

þat $ hwet þét þet þat þat

he us lende he us lende he us lende hé us lenne he us lende he vs lende he us lende

Apart from J, which has a free paraphrase very different from the rest, there are a few minor differences of lexis and word order between the texts that give minor differences also in sense. But it is clear from the examples in (13) (confirmed from examination of the text dictionaries derived from the complete LAEME samples for these text languages) that is is the normal form for the present indicative 3rd person singular ‘is’. Confirming this gives us more licence to interpret the b-root forms in L, e and E as implying future. T, D and M illustrate a variant textual tradition, and have the past participle ‘told, judged’ that we supplied as understood for the e version above; they therefore lack the context for the b-root future form. The M version, however, displays an interesting variation in its deployment of s-root and b-root forms. It has is where L, e and E have b-forms and beoþ for the third example of ‘is’ in (13). It looks as if in the M scribe’s system the s-root and b-root forms are simply variants for ‘is’ and he could write either to represent what was probably is in his exemplar.15 This variation would also mean that he could read exemplar bið-type forms as simple present indicative and substitute his is variant for them if he wished. Examination of his practice across the whole of his text confirms this: he uses is, beoþ and beþ for the 3rd singular present indicative where present sense is certainly implied, and there are no clear contexts in his output where b-forms must imply future sense. Clearly the v-fut13 tag cannot be sensibly applied to this scribe’s forms for ‘is’. The free paraphrase of the J scribe shows that he has opted to express the future with periphrastic ‘shall’, which here has no implication of obligation. Note too the survival in all seven texts of subjunctive b(e)o, bi for the optative expression ‘grant that our end be good’. 3.2.4 Classifying the variation During the tagging process there emerged three early Middle English b‑root “systems” for expression of futurity. These systems are possible because of the availability of both s‑root and b‑root forms in the present tense. In early Middle English this is true for the singular in all dialects and also for the plural in non-southerly dialects. 15. The order of the lines of the Poema Morale in M differs greatly from that in the other versions. Paues (1907: 225) conjectures that M’s text may therefore have been written down from memory. The point here holds whether or not the ‘exemplar’ was in front of the M scribe or in his head.

 Margaret Laing

TYPE 1: Southerly mixed — the Old English system continued? Text languages of this type show s‑root spellings of the am art is type for the present indicative singular of ‘be’ and b‑root spellings of the be(o)þ type for present indicative plural. Subjunctive is expressed by be(o) singular and be(o)n plural. b‑root spellings are also found to express the future, more commonly in the singular (biþ type) than in the plural be(o)þ type. Biþ-type spellings may also be used in the singular for present tense without future sense. More work would be needed to look at the contexts of b‑root spellings that do not imply futurity to see if they are all or mostly of the “general truth” type that could imply a continuation of the Old English system. The following examples are from London, Lambeth Palace Library 487, Lambeth Homilies, language 1 (ca 1200, North-West Worcestershire). Compare the following similar constructions: (15) gif þet ege ablindað ne bið naut þe hond wel lokinde If the eye becomes blind the hand shall not be well-seeing (16) Gif god bið his ifulsta ne bið his mehte nowher forsegen If God is/shall be his helper his might nowhere shall be despised The first example illustrates a present single occurrence followed by a future occurrence and is expressed by the 3rd person singular indicative ablindað followed by a b-root form of ‘be’ to express future. The second example has bið in both halves of the expression. The second bið is used, as in the first example, to express a future occurrence. The first bið may be interpreted as an invariable fact or as an example of the present continuous or as implying future. Any of these interpretations would show a continuation of Old English usage. TYPE 2: Southerly discrete Text languages of this type show s‑root spellings of the am art is type for the present indicative singular of ‘be’ and b‑root spellings of the be(o)þ type for present indicative plural. Subjunctive is expressed by be(o) singular and be(o)n plural. This leaves the singular be, bist, biþ type spellings “free” to express future sense. If be(o)þ types are used to express the future in the plural they are formally indistinguishable from the present forms, making the system defective. Recourse may then be had to ‘shall’ periphrasis as in the G Ancrene Riwle example in section 3.2.1 above. The illustrations of the Southerly discrete system below are from Oxford, Bodleian Library, Bodley 34: Hali Meiðhad (S Salop C13a2).

Reflexes of OE beon as a marker of futurity 



(17) Singular is = present; beo = subjunctive; bið = future: Meiðhad is $ tresor . $ beo hit eanes forloren; ne Maidenhood is that treasure, that be it once lost, not bið hit neauer funden shall be it never found (18) Plural beoð = present (parallel to singular is): Hwen þus is of riche; hwet wenest tu of þe poure. When thus it is of the rich, what thinkest thou of the poor, þe beoð wacliche iõeuen who are unworthily given (19) Plural beoð = formally identical to present, but perhaps implying future, parallel to ‘shall sing’ in the second part of the sentence. If so, the system is formally defective: For õef ha þus beoð acwiket [..] ha beoð For if they thus are/shall be revived [..] they are/shall be in widewene reng & schulen in widewene reng biuore þe in widows’ rank and shall in widows’ rank before the iweddede. singen in heouene wedded, sing in heaven TYPE 3: Midland system As with TYPE 1, text languages of this type show s‑forms of the am, art, is type for the present indicative singular, leaving be, bist, biþ or bes type spellings available to express future sense. Subjunctive singular is be(o) (or less common si(e), se). Present indicative plural is s‑root sinden and/or arn/are type with be(o)n for the plural subjunctive. Inflected be(o)þ type spellings are available to be reserved for plural future sense. The example below is from Oxford, Bodleian Library, Laud Misc 108, Havelok (West Norfolk C14a1). Bernard reports to Ubba how Havelok fought and slew 60 ruffians single-handed and is a bit worse for wear. (20) Singular is = present; bes = future: But it is of him mikel scaþe \ I woth þat he bes But he is badly wounded \ I believe that he shall be ded ful raþe dead very soon The second example is from London, British Library, Cotton Titus D.xviii, Ancrene Riwle (S Cheshire C13a2).

 Margaret Laing

(21) Plural arn = present (parallel to singular is); beoð = future; beon = subjunctive: Þis is a swiðe dredliche word to wepman & to wimmen $ swiðe sone scheawen ham to hwase wile . õe arn bitacnet þe vnwrihene put . $ is owre feire neb owre hwite swire . owre lihte ehe . owre honde õif õe ham scheawen . õet beoð owre wordes put . bote ho beon þe betere biset . ‘This is a very dreadful saying to men and to women who are quick to show themselves to whoever desires. Ye are a symbol of the uncovered pit: that is, your fair face, your white neck, your bright eye, your hands, if ye show them. Even your words shall be a pit unless they be better directed.’ TYPE 4: Northern system Text languages of this type are confined to the northern counties. They have for the present indicative singular am, er(t)/es, es/is and for the plural er/es/ar; subjunctive singular and plural be. The inflected b‑root forms are available to express futurity and are the same for singular and plural: the bes-type. For early Middle English very few texts survive from the North and the LAEME time-span had to be expanded beyond 1300 in order to have any coverage there at all. So the texts in LAEME that show this system are from the first quarter of the fourteenth century. They are therefore nearer in both time and space than the rest of the LAEME corpus to the Older Scots materials displayed in LAOS, whose finite forms for ‘be’ are also of great interest. The example below is from Edinburgh, Royal College of Physicians, MS of Cursor Mundi, Hand A (Yorks East Riding C14a1). Much of the context is future in sense, being about what is to occur at doomsday at the end of the world. As the numerous examples in the text show, the simple future is expressed here with periphrastic sal ‘shall’. Contrast, however, er present indicative plural (line 1, cf. singular es line 5) with bes future plural (line 12). (22) 1

Mani wenis þat er vnwis

5

Þat tat fleis hal suld neuer ris Nou I sal te resun rede And out of mistrouning you lede Wit ye wel it es na rihtt For-to mistroun in godes miht ... Sua halli sal tai risin þar Þaim sal noht want an hefdis har Ne noht a nail of fot or hand

Many think, who are unwise, That the flesh whole should never rise. Now I shall thee reason read, And out of misbelieving you lead. Know ye well, it is not right To misbelieve in God’s might ... So wholly shall they rise there, To them shall not want a head’s hair, Nor not a nail of foot or hand.

Reflexes of OE beon as a marker of futurity 



10 Þohqueþir we sal wnderstand Þat nail and har þat hauis ben schorn Bes noth al quar þai war biforn Bot als potter wit pottes dos Quen his new vessel fordos 15 He castis it al in a bal A better forto mak wit-al Of noht he lokes quilk was quilk Bot makes anoþer of þat ilk Wel fairer þan þe first was wroth 20 Riht sua sal crist ne dout þou noht

However, we shall understand That nail and hair that have been shorn Shall be not all where they were before. But as a potter with pots does, When [he] his new vessel destroys, He makes it all into a ball To make a better [one] with it. Of none [of it] he considers which [bit] was which, But makes another of the same stuff, Much fairer than the first was made; Right so shall Christ [do], doubt thou not.

4. Prospect This paper has drawn attention to the continuation into early Middle English of a variable Old English grammatical distinction, and its apparent exaptation for use in different subsystems in early Middle English. More work is needed before we can assess the extent and duration of these temporary subsystems. Questions about ‘be’ and futurity in Middle English that would reward investigation are: 1. Does TYPE 1 above simply continue an Old English system? 2. What part does the verb weorþan play in the story of ‘be’ futurity in Old and Middle English? 3. How does the rise of the expression of futurity with ‘shall’ (especially the ‘shall be’ type) interact with TYPES 1-4 above? References Campbell, A. 1959. Old English Grammar. Oxford: Clarendon Press. Fischer, Olga. 1992. “Syntax”. The Cambridge History of the English Language. Volume II: 1066-1476 ed. by Norman F. Blake, 207-408. Cambridge: Cambridge University Press. Fleischman, Suzanne. 2000. “Methodologies and Ideologies in Historical Linguistics: On Working with Older Languages”. Textual Parameters in Older Languages ed. by Susan C. Herring, Pieter van Reenen & Lene Schøsler, 33-58. Amsterdam & Philadelphia: John Benjamins.

 Margaret Laing Jost, Karl. 1909. Beon und Wesan: eine syntaktische Untersuchung. (= Anglistische Forschungen, 26). Heidelberg: C. Winter. Ker, N. R. 1957. Catalogue of Manuscripts containing Anglo-Saxon. Oxford: Clarendon Press. Kilpiö, Matti. 1992. “BEON. With Attested Spellings by Robert Millar, Using Materials Assembled by Haruko Momma”. Dictionary of Old English. Toronto: Pontifical Institute of Mediaeval Studies. Kilpiö, Matti. 1993. “Syntactic and Semantic Properties of the Present Indicative Forms of the verb to be in Old English”. Early English in the Computer Age: Explorations through the Helsinki Corpus ed. by Matti Rissanen, Merja Kytö & Minna Palander-Collin, 97-116. Berlin & New York: Mouton de Gruyter. Kilpiö, Matti. 1997. “On the Forms and Functions of the Verb to be from Old to Modern English”. English in Transition. Corpus-Based Studies in Linguistic Variation and Genre Styles (= Topics in English Linguistics, 23) ed. by Matti Rissanen, Merja Kytö & Kirsi Heikkonen, 87-120. Berlin & New York: Mouton de Gruyter. LAEME = Laing, Margaret & Roger Lass. 2008– . Laing, Margaret & Roger Lass. 2008– . A Linguistic Atlas of Early Middle English 1150-1325 [http://www.lel.ed.ac.uk/ihd/laeme1/laeme1.html] compiled by Margaret Laing & Roger Lass. Edinburgh: © The University of Edinburgh. Lass, Roger & Margaret Laing. 2009. “Databases, Dictionaries and Dialectology. Dental Instability in Early Middle English: A Case Study”. Studies in English and European Historical Dialectology ed. by Marina Dossena & Roger Lass, 91-131. Bern: Peter Lang. LAOS = Williamson, K. 2008– . Lutz, Angelika. 2008. “Celtic influence on Old English and West Germanic”. English Language and Linguistics 13:2.227–249. Mitchell, Bruce. 1985. Old English Syntax. Vol. 1. Oxford: Clarendon Press. Mustanoja, Tauno F. 1960. A Middle English Syntax. Part 1. Helsinki: Société Néophilologique. Paues, Anna C. 1907. “A Newly Discovered Manuscript of the Poema Morale”. Anglia 30.217-237. Sisam, Kenneth, ed. 1975. Fourteenth Century Verse and Prose. Reprinted with corrections. Oxford: Clarendon Press. Thomson, David, ed. 1984. An Edition of the Middle English Grammatical Texts. New York & London: Garland. Traugott, Elizabeth Closs. 1992. “Syntax”. The Cambridge History of the English Language. Volume 1: The Beginnings to 1066 ed. by Richard M. Hogg, 168-289. Cambridge: Cambridge University Press. Williamson, Keith. 2008– . A Linguistic Atlas of Older Scots Phase 1 1380-1500 [http://www.lel.ed.ac.uk/ihd/laos1/laos1.html] compiled by Keith Williamson. © Edinburgh: The University of Edinburgh. Zupitza, Julius, ed. 1880 [2001]. Ælfrics Grammatik und Glossar: Text und Varianten. Third reprint with an updated introduction by Helmut Gneuss. Hildesheim: Olms. First published Berlin: Weidmann 1880.

Stylistic fronting in the history of English Masayuki Ohkado* Chubu University

Stylistic fronting is an operation which moves elements generally occupying the position to the right of the finite verb such as adjectives, past participles, and adverbs to the position immediately preceding it in clauses with a subject gap. The operation is typically observed in Modern Icelandic and in earlier stages of the Scandinavian languages. In this article I will extensively examine Old and Middle English texts and show that word order patterns arguably attributed to stylistic fronting are widely observed both in Old and Middle English and are not confined to texts which are likely to have been heavily influenced by Old Norse speakers. This is contrary to Trips (2002), who attributes the presence of the stylistic fronting patterns in the Ormulum to the Scandinavian invasions. Taking into consideration the wide distribution of relevant examples, I will conclude that the process of stylistic fronting was a genuine property of Old and Middle English.

1. Introduction The aim of this article is to show that stylistic fronting, which is typically observed in Modern Icelandic and earlier stages of the Scandinavian languages in general, is also extensively attested in Old and Middle English. This analysis enables us to account for word order differences between clauses with a full NP subject on the one hand and those with a personal pronoun subject or those with a subject gap on the other. An interesting consequence of the present study is that we cannot attribute the origin of stylistic fronting to the Scandinavian influences as claimed in Trips (2002). * This research was supported by a Grant-in-Aid for Scientific Research, Grant No. 18320076, from the Japan Society for the Promotion of Science. I am deeply grateful to the editors and anonymous reviewers for their invaluable comments on an earlier version of this paper. I would also like to express my sincere gratitude to Dr. Richard Ingham for helping me at the conference, giving helpful comments on and correcting stylistic errors of the final version of this paper. All remaining inadequacies are of course mine.

 Masayuki Ohkado

The organization of the present article is as follows. In Section 2 I will briefly introduce the characteristics of stylistic fronting. In Section 3 and Section 4, Old and Middle English data are analyzed. In Section 5, I will discuss the origin of stylistic fronting in the English language. 2. Stylistic fronting in Icelandic Stylistic fronting is an operation typically observed in Modern Icelandic subordinate clauses, which moves elements (past participles, adjectives, some adverbs, particles, etc.) generally occupying the position to the right of the finite verb to the position immediately preceding it. Examples involving stylistic fronting are given in (1) to (3), which Maling (1990: 74) cites from a story by Gestur Pålsson (1970). (1) Clauses with Fronted Past Participles a. Honum mætti standa á sama, hvað sagt væri um him.dat might stand on same what said was about

hann him (“Tilhugalíf”, Chapter 5) ‘It might be all the same to him what was said about him.’

b. og hann léti þá vita, að hann ... kæmi and he would-let them know that he ... would-come

ekki heim, eins og ráðgert hafði verið not home as planned had been (“Tilhugalíf”, Chapter 1) ‘and he would let them know that he wouldn’t be coming home as had been planned’

(2) Clauses with Fronted Adjectives a. Hann raulaði þá vísuna sína nokkuð hærra en he hummed then tune his somewhat louder than

venjulegt var. usual was (“Hans Vöggur”) ‘Then he hummed his tune somewhat louder than usual.’

b. Engum datt í hug, að vert væari að reyna til að No‑one.dat fell to mind that worth was to try P to

kynnast honum. know him (“Hans Vöggur”) ‘It didn’t occur to anyone that it was worth trying to get to know him.’



Stylistic fronting in the history of English 

(3) Clauses with Fronted Adverbs1 a. og nú eigum við að vita ... hvort ekki finnst meira and now ought we to know ... whether not finds more

þýfi hjá piltinum. stolen-goods on the-boy (“Tilhugalíf”, Chapter 4) ‘and now we are supposed to find out whether any more stolen goods can be found where the boy lives.’

b. Hún ]horfði á þennan stóra sterka mann, sem aldrei hafði She looked at this big strong man that never had

beygt lund sína fyrir nokkrum kvenmanni. bent will his for any woman (“Tilhugalíf”, Chapter 1) ‘She looked at the big, strong man who had never changed his mind because of a woman.’

Notable properties of stylistic fronting pointed out in Maling (1990) are that its application observes two constraints: (i) the Subject Gap Condition and (ii) the Accessibility Hierarchy. The Subject Gap Condition requires that stylistic fronting applies only to clauses with a subject gap, and Maling (1990) assumes that the fronted elements occupy the empty subject position, as illustrated in (4). (4) Clauses with Fronted Past Participles a. Honum mætti standa á sama, hvað [ ] væri sagt um hann him might stand on same what was said about him => Honum mætti standa á sama, hvað [sagt] væri um hann b. og hann léti þá vita, að hann ... kæmi and he would-let them know that he would-come

ekki heim, eins og [ ] hafði ráðgert verið not home as had planned been => eins og [ráðgert] hafði verið

Subject gaps are created (i) through extraction of the subject NP by unbounded transformational rules such as relativization, question formation, comparative

1. An anonymous reviewer points out that the function of ekki is often not equivalent to that of other adverbs. Here I simply follow Maling’s (1990) treatment of ekki, leaving aside its possible differences from other fronted elements.

 Masayuki Ohkado

clause formation, topicalization, and clefting; (ii) through use of impersonal passives or lexically impersonal predicates; and (iii) through indefinite-NP postposing.2 The Accessibility Hierarchy is given in (5), which states that if a given clause has more than one potential candidate for stylistic fronting, only the leftmost element in the hierarchy can undergo the process. (5) Accessibility Hierarchy ekki ‘not’ > predicate adjective > past participle/verbal particle For instance, a predicate adjective cannot be fronted in clauses with ekki ‘not’ as the contrast between (6a) and (6b) illustrates. (6) a. Þetta er nokkuð, sem ekki er hægt að gera við. this is something that not is possible to fix PRT b. *Þetta er nokkuð, sem hægt er ekki að gera við. this is something that possible is not to fix PRT In this paper, due to the limited number of relevant examples in our corpus, I will focus on the first property of stylistic fronting: the Subject Gap Condition. Although stylistic fronting is not observed in contemporary Mainland Scandinavian languages, it is attested in their older stages, as illustrated in (7) to (9). (7) Old Danish (1200–1325) vatn hvært, ær æi ær mæþ damme fæst water every which not is with dam closed  (Falk & Torp (1900: 296)/Hrafnbjargarson (2004: 182)) ‘every water that is not closed off with a dam’ (8) Middle Danish (1325–1550) som sagd er ved Propheten as said is with prophet-the  (Falk & Torp (1900: 296)/Hrafnbjargarson (2004: 182)) ‘as is told by the prophet’

2. Reflecting the recent development of the structural analysis, Maling’s (1990) “subject position” where the fronted elements occupy is interpreted in three ways. In the analysis of Platzak (1987) and Rögnvaldsson & Thráinsson (1990), the position is the specifier of IP. In Cardinaletti & Roberts (2002), fronted elements are analyzed as being moved into [Spec, Agr1P] via [Spec, Agr2P]. In Jónsson (1991), and Poole (1996), they are analyzed as head-adjoined to I0. For details of these analyses and their advantages and disadvantages, see Trips (2002: Section 7.3) and also Hrafnbjargarson (2004).



Stylistic fronting in the history of English 

(9)

Old Swedish (1250–1500) En ... som likir war enom hofman one ... that alike was a courtier (Falk (1993: 178)/Trips (2002: 276)) ‘one ... that resembled a courtier’

Stylistic fronting is also found in Old Norse, as illustrated in (10). (10) Old Norse (early ninth century to late fourteenth century) lágu hestarnir á kafi annat skeið, svá at draga varð upp lay horses-the on snow other time so that pulled became upw  (1330 Egils saga Skalla-Grímssonar 260.23/Faarund (2004: 237)) ‘The horses sank into the drifts now and again, so that they had to be pulled out.’ 3. Stylistic fronting in Old English In this section I will examine Old English data and demonstrate that phenomena which will best be analyzed as the result of stylistic fronting are observable in Old English. 3.1

Theoretical assumptions

I assume that the underlying structure of Old English is double-based in the sense that both head-initial and head-final structures can be base-generated (Double-Base Hypothesis). 3 I also assume that the ratio of head-final structures to head-initial structures of a given construction is constant across contexts (Constant Ratio Hypothesis). The underlying word order patterns are affected by various operations, notably the following two classes of operations: (i) topic movement and V-to-C movement that apply in main clauses and (ii) extraposition or rightward movement of heavy elements that applies both in main and subordinate clauses.4 Under this analysis both head-initial and head-final clauses are basically ambiguous in structural interpretation: They can be either base-generated or derived by movement. For instance, subordinate clauses 3. For discussions about the underlying structure of Old English and how various word order patterns are derived, see, for instance, Fischer et al. (2000), Ohkado (2005), and Pintzuk (1999). 4. Other notable operations proposed in the literature are Verb Raising and Verb Projection Raising. Ohkado (2005: Chapter 1) proposes that the latter should be interpreted as an instance of extraposition.

 Masayuki Ohkado

illustrated in (11a) may reflect the underlying SVO pattern or they may have been derived from the underlying SOV structure, which is illustrated in (11b), by extraposition.5 (11) a. þæt God lufode Iacob that God loved Jacob ‘that God loved Jacob’ b. buton Iohannes attor drunce unless John poison drank ‘unless John drank poison’

(ÆCHom I. 7: 236)

(ÆCHom I. 4: 213.212.818)

It is impossible to say whether the word order of a given case reflects the underlying structure or is derived by movement. However, we can estimate the ratio of head-initial and head-final structures of a given construction by examining the frequencies of the two competing patterns in a case that is unlikely to be affected by movement operations. For explanatory purposes, let us call such constructions “Genuine Constructions” in the sense that they are likely to reflect the underlying word order patterns. Under the assumption that elements undergoing movement operations are typically finite verbs and heavy elements, one candidate for Genuine Constructions is a VP headed by a nonfinite verb with a non-heavy NP object, as illustrated in (12). (12) Đa het se cyngc scipa gegearcian then ordered the king ships prepare ‘Then the king ordered that the ships be prepared.’

(ApT: 8.1.118)

Heavy NP objects may undergo extraposition so that VPs involving them might not reflect the underlying order. Likewise, VPs with personal pronoun objects, quantified objects, and negative objects might not show the underlying order since they may undergo leftward movement (cf. Pintzuk & Taylor 2006). Let us examine the situation by searching the York-Toronto-Helsinki Parsed Corpus of Old English. The result of the search for VPs containing only a nonfinite verb and a oneword object NP is given in Table 1. The search was conducted by using the program CorpusSearch 2, developed by Beth Randal.6 The command lines used for the present study (for Old English

5. The reference system is the one adopted in the York-Toronto-Helsinki Parsed Corpus of Old English (YCOE), the corpus used for the present research. 6. CorpusSearch 2 can be downloaded at: http://corpussearch.sourceforge.net/.



Stylistic fronting in the history of English 

Table 1.  OV and VO patterns in VPs headed by nonfinite verbs with a single-word full noun object OV pattern 13 (65%)

VO pattern

total

7 (35%)

20

data) are given in the appendix.7 The figures in Table 1 show that VPs with an accusative object in Old English are 65 percent OV and 35 percent VO. Needless to say, this is a broad approximation since in Table 1 the figures of all the texts are put together. Although separate treatment of individual texts is ideal, the restrictions imposed by the database significantly reduce the number of relevant examples. In the present case, there are 314 examples of VPs headed by infinitives with accusative NPs. The restrictions on the number of constituents reduce the relevant occurrences to 20, as shown in Table 1. Now suppose that we have constructions comparable to Genuine Constructions (e.g., VPs headed by an infinitive with a single-word NP object). If the ratio of the two competing patterns in the former is significantly different from that in the Genuine Constructions, the difference will be accounted for in terms of (an) additional operation(s) that affect(s) the underlying ratio of the two patterns. As an example, consider a case where VPs headed by an infinitive contain an object composed of three or more words, as illustrated in (13), which is comparable to the Genuine Construction we have just seen in Table 1. (13) Ða het he for þi acwellan ealle þa hysecild then ordered he for that kill all the male-children þære burhscire fram twywintrum cilde oð anre nihte of-the city from two-years children till one night  (ÆCHom I, 5: 219.72.957) ‘Then, because of this, he ordered that all the male children in the city, from two years to one night old be killed.’ The result of the search for such examples in the York-Toronto-Helsinki Parsed Corpus of Old English is given in Table 2.

7. As argued in Ohkado (2005: 12–14), for the “true development” of historical linguistics and other related fields, it is essential for researchers using corpora to spell out the exact procedures used to retrieve the relevant data so that interested readers may examine the relevant data. The command lines in the appendix are given in this spirit.

 Masayuki Ohkado

Table 2.  OV and VO patterns in VPs headed by nonfinite verbs with an NP object containing three or more words OV pattern 0 (0%)

VO pattern

total

10 (100%)

10

The figures in Table 2 are in sharp contrast with those in Table 1: While more than half of the examples are OV in Table 1, there is not a single instance of OV patterns in Table 2. The difference is accounted for by assuming that in the case where VPs contain accusative NP objects with more than three words, extraposition applied and all the underlying OV examples have become VO. The basic assumptions discussed above are summarized in (14). (14) Basic assumptions a. Both head-initial and head-final patterns can be base-generated in Old English. b. The ratio of head-initial and head-final patterns in a given construction is constant across contexts. (The Constant Ratio Hypothesis) c. The ratio of the two competing patterns at the surface structure may differ from the underlying ratio due to movement operations. 3.2

Database and the constructions analyzed

The database for Old English is the York-Toronto-Helsinki Parsed Corpus of Old English, which we have already used in the preceding subsection. The constructions analyzed in this paper are the subordinate clauses listed in (15). (15)

a. b. c. d.

adjective + BE (AB)/BE + adjective (BA) past participle + BE (VB)/BE + past participle (BV) infinitive + modal (VM)/modal + infinitive (MV) NEG + BE/HAVE/modal verb/verb, and BE/HAVE/modal verb/verb + NEG8 e. adverb (non NEG) + BE/HAVE/modal verb/verb, and BE/HAVE/modal verb/verb + adverb (non NEG) If stylistic fronting was available in Old English, we would expect that the first patterns given in (15a) to (15e) (AB, VB, VM, NEG-BE, etc. and adverb-BE, etc.) are attested with higher frequencies in clauses with a subject gap than in clauses without a subject gap. This is because the nonfinite predicates in (15a) to (15c) and 8. The negating particle ne, which is regarded as cliticized to the finite verb, is excluded.



Stylistic fronting in the history of English 

adverbs in (15d) to (15e) may undergo stylistic fronting and move into the subject gap, leading to the higher frequencies in the two cases. Note that in clauses with a personal pronoun subject, the subject is analyzed as being cliticized to the C position, creating a subject gap as schematically illustrated in (16). (16) complementizer-subject [ ] ... Therefore, clauses with a personal pronoun subject are expected to behave in parallel fashion to clauses with an empty subject. Although infinitives in Modern Icelandic do not undergo stylistic fronting, they do in Old Icelandic, as illustrated in (17). (17) Ertu Gísli sá er finna vildir Grettir Ásmundarson? are-you Gísli that who find wanted Grettir Ásmundarson  (Willson (2001: 134)) ‘Are you the Gísli that was looking for Grettir Ásmundarson?’ To examine whether Old English is of the Old Icelandic or Modern Icelandic type, clauses with infinitives accompanied by modals are included in the present search. In analyzing the constructions in (15a) to (15c) we should take into account that, as demonstrated in Ohkado (2001), in constructions with multiple predicates, the presence or absence of extra elements affects the choice between headinitial and head-final patterns. For instance, clauses with modal verbs and infinitives tend to show higher frequencies of head-initial MV patterns if they involve object NPs or PPs.9 (18) a.

Clauses with No Extra Element (Subject) Adjective BE (AB) (Subject) [Past Participle] BE (VB) (Subject) Verb Modal (VM)

b. Clauses with Extra Elements (Subject) O/PP Adjective BE → Adjective (BA) (Subject) O/PP [Past Participle] BE → Participle] (BV) (Subject) O/PP Verb Modal →

(Subject) BE (Subject) BE [Past (Subject) Modal Verb (MV)

9. In Ohkado (2001) the tendency is accounted for in terms of the well-known tendency of Old English that heavy elements are likely to undergo rightward movement: VPs involving extra elements such as object NPs and PPs are heavy in comparison with VPs with no extra elements and they tend to undergo extraposition, leading to the higher frequencies of head-initial patterns.

 Masayuki Ohkado

To avoid these possibilities, the first three constructions in (15) analyzed here are restricted to cases with (i) the subject (including a subject gap), (ii) the finite verb (BE, HAVE, or a modal), and (iii) the nonfinite predicate (an adjective, past participle, or infinitive) and clauses involving other elements are excluded from the data. As an illustration, examples with BE and past participle are given in (19). (19) Clauses with BE and past participle a. “past participle-BE” order with a full NP subject ... gif heora lar onfongen wære, ... if their teaching accepted were (Bede_1: 13.56.6.521) ‘... if their teaching were accepted ...’ b. “past participle-BE” order with a personal pronoun subject ... ær þan þe he acenned wære: ... before he born was (ÆHom_2: 95.295) ‘... before he was born ...’ c. “past participle-BE” order with an empty subject ... þæt gedon is, ... that done is (ÆLS_[Christmas]: 43.29) ‘... that which is done ...’ d. “BE-past participle” order with a full NP subject ... syððan se dolh wæs geopenod ... since the wound was opened (ÆLS_[Æthelthryth]: 65.4181) ‘... since the wound was opened ...’ e. “BE-past participle” order with a personal pronoun subject ... þeah þe he beo gehæled, ... although he be healed (ÆHom_2: 105.300) ‘... although he will be healed ...’ f. “BE-past participle” order with an empty subject ... and se þe beo tosliten, ... and those who will-be torn (ÆHom_21: 324.3241) ‘... and those who will be torn, ...’ The examples in (19a) to (19c) show head-final order and those in (19d)-(19f) show head-initial order. The NEG in (15d) is either a negative adverb na/no, or næfre or nawiht ‘not.’ (20) NEG in (15d) na/no, næfre, nawiht Note that the finite verb in (15d) can be without ne as shown in (21a), or it can be with ne as in (21b) and (21c).



Stylistic fronting in the history of English 

(21) a. NEG + BE/HAVE/modal verb/verb without ne ... þæt hi na mihton incumen. that they neg might in-come  (ChronE_[Plummer]:1070.20.2585) ‘... that they might not come in.’ b. NEG + BE/HAVE/modal verb/verb with the contracted ne ... þæt hit na nære on ðæm dagum mid Romanum that it neg neg.were in the days with Romans

buton gewinne, ... without conflict (coorosiu, Or_3: 10.75.15.1476) ‘... that in those days it was not without conflict with the Romans, ...’

c. NEG + BE/HAVE/modal verb/verb with the independent ne ... ða ðe hi na ne cuðon, ... those that they NEG NEG knew (ÆHom_11: 229.1611) ‘... those that they did not know, ...’ 3.3

The results

The results of the search are given in Tables 3 to 9. In order to see whether there is a statistically significant difference between the figures concerning examples with full NP subjects and those concerning examples with personal pronoun subjects or empty subjects, Fisher’s exact probability test was conducted.10 The results of the test are given below each Table. The italicized parts indicate that there is a statistically significant difference between the relevant figures. Table 3, for example, shows that there is a statistically significant difference (i) between the figures concerning examples with full NP subjects and those concerning personal pronoun subjects and (ii) between the figures concerning examples with full NP subjects and those concerning examples with empty subjects. Table 3.  Adjective-BE (AB) and BE-Adjective (BA) patterns AB 62 (37%)

full NP subject BA total 105 (63%)

167

personal pronoun subject AB BA total 347 (68%)

167 (33%)

514

empty subject AB BA total 152 (70%)

Full NP vs. personal pronoun: P value < 0.0001, extremely significant Full NP vs. empty: P value < 0.0001, extremely significant 10. For the calculation, InStat 3 for Macintosh was used.

65 (30%)

217

 Masayuki Ohkado

Table 4.  Past participle-BE (VB) and BE-past participle (BV) patterns VB 206 (60%)

full NP subject BV total 137 (40%)

343

personal pronoun subject VB BV total 195 (71%)

79 (29%)

VB

274

empty subject BV total

56 (72%)

22 (28%)

78

Full NP vs. personal pronoun: P value = 0.0050, very significant Full NP vs. empty: P value = 0.0696, NOT QUITE significant

Table 5.  Infinitive-modal (VM) and modal-infinitive (MV) patterns VM 13 (59%)

full NP subject MV total 9 (41%)

22

personal pronoun subject VM MV total 101 (62%)

63 (38%)

empty subject VM MV total

164

35 (78%)

10 (22%)

45

Full NP vs. personal pronoun: P value = 0.8197, NOT significant Full NP vs. empty: P value = 0.1506, NOT significant

Table 6.  NEG-BE/HAVE/modal verb/verb and BE/HAVE/modal verb/verb-NEG (NEG = na/no) NEG-V 1 (3%)

full NP subject V-NEG total 32 (97%)

33

personal pronoun subject NEG-V V-NEG total 8 (14%)

50 (86%)

58

empty subject NEG-V V-NEG total 3 (12%)

23 (89%)

26

Full NP vs. personal pronoun: P value = 0.1480, NOT significant Full NP vs. empty: P value = 0.3113, NOT significant

Table 7.  NEG-BE/HAVE/modal verb/verb and BE/HAVE/modal verb/verb-NEG (NEG = næfre) NEG-V 2 (20%)

full NP subject V-NEG total 8 (80%)

10

personal pronoun subject NEG-V V-NEG total 30 (67%)

15 (33%)

45

Full NP vs. personal pronoun: P value = 0.0115, significant Full NP vs. empty: P value = 0.0502, NOT significant

empty subject NEG-V V-NEG total 12 (63%)

7 (37%)

19



Stylistic fronting in the history of English 

Table 8.  NEG-BE/HAVE/modal verb/verb and BE/HAVE/modal verb/verb-NEG (NEG = nawiht) NEG-V 0 (0%)

full NP subject V-NEG total 27 (100%)

27

personal pronoun subject NEG-V V-NEG total 22 (24%)

68 (76%)

empty subject NEG-V V-NEG total

90

7 (19%)

30 (81%)

37

Full NP vs. personal pronoun: P value = 0.0019, very significant Full NP vs. empty: P value = 0.0180, significant

Table 9.  Adverb (non NEG)-BE/HAVE/modal verb/verb and BE/HAVE/modal verb/ verb-adverb (non NEG) full NP subject adverb-V V-adverb total 134 (22%)

488 (79%)

622

personal pronoun subject adverb-V V-adverb total 810 (40%)

1207 (60%)

2017

empty subject adverb-V V-adverb total 508 (42%)

696 (58%)

1204

Full NP vs. personal pronoun: P value < 0.0001, extremely significant Full NP vs. empty: P value < 0.0001, extremely significant

The pattern observed in Tables 3, 8, and 9 is clear: Clauses with a personal pronoun subject or those with an empty subject consistently show higher frequencies of head-final patterns than clauses with a full NP subject. The differences are statistically significant. In these cases we can safely conclude that the higher frequencies of head-final patterns are due to stylistic fronting which fronted adjectives, nawiht, and adverbs to the empty subject slot. In Tables 4 and 7 statistical significance is observed only between the figures concerning examples with full NP subjects and those concerning examples with personal pronoun subjects. Yet, the percentages of head-final patterns are higher in examples with empty subjects than those with full NP subjects. Thus, we tentatively conclude that the account in terms of stylistic fronting applies to these cases as well. In Tables 5 and 6 statistical significance is not observed in both pairs, so we should conclude that stylistic fronting is irrelevant in these cases. Note that inapplicability of stylistic fronting to infinitives is identical with the situation in Modern Icelandic and different from that in Old Icelandic (cf. Willson 2001: 134). These considerations suggest that in clauses with a subject gap, stylistic fronting was able to apply to adjectives, past participles, NEG (nawiht), and adverbs and led to the higher frequencies of head-final patterns.

 Masayuki Ohkado a. [adjective]

BE

b. [past participle]

BE

c. [NEG (nawiht)]

BE/HAVE/modal verb/verb

d. [adverb]

BE/HAVE/modal verb/verb

4. Stylistic fronting in Middle English 4.1

Theoretical assumptions

In Middle English the underlying order is predominantly VO, as the analysis of Genuine Constructions (VPs containing only a nonfinite verb and a one-word object NP) reveals, as illustrated in Table 10. Table 10.  OV and VO patterns in VPs headed by nonfinite verbs with a single-word NP object OV pattern 2 (12%)

VO pattern

total

15 (88%)

17

An example of each pattern is given in (23). (23) a. OV pattern ... he ne ssel naŠt ariŠt conne god louie ... he not shall not rightly can god love  (CMAYENBI, 117.2249) ‘... he shall not know how to love god ...’ b.

VO pattern oðer is bodien fulcninge other is preach baptism ‘The second is to preach baptism’

(CMTRINIT, 139.1874)

The situation shown in Table 10 is in accordance with the traditional analysis of word order change in English within the framework of generative grammar: Canale (1978), Lightfoot (1979) and Van Kemenade (1987) among others assume that English underwent underlying word order change from SOV to SVO in the Early Middle English period.



Stylistic fronting in the history of English 

4.2

Database and the constructions analyzed

The database used for the present study is the Penn-Helsinki Parsed Corpus of Middle English, second edition (PPCME2) (Kroch & Taylor 2000). Examples examined are subordinate clauses showing the patterns given in (24). (24) a. b. c. d. e.

adjective + BE (AB)/BE + adjective (BA) past participle + BE (VB)/BE + past participle (BV) infinitive + modal (VM)/modal + infinitive (MV) NEG + BE/HAVE/modal verb/verb, and BE/HAVE/modal verb/verb + NEG adverb (non NEG) + BE/HAVE/modal verb/verb, and BE/HAVE/modal verb/verb + adverb (non NEG)

A sample of each pattern is given in (25) to (29). (25) a. 

adjective + BE (AB) Wif wolde þat hire louerd dead ware woman wanted that her lord dead were (CMTRINIT, 165.2246) ‘The woman wanted her lord to be dead’

b. BE + adjective (BA) ... gief þe licame beð euel ... if the body is evil ‘... if the body is evil ...’

(CMTRINIT, 183.2521)

(26) a. past participle + BE (VB) And whan this folk togidre assembled weren, ... and when this folk together assembled were  (CMCTMELI, 218.C2.55) ‘And when this folk was assembled together, ...’ b. BE + past participle (BV) His wyf and eek his doghter hath he left inwith his his wife and also his daughter has he left inside his hous, of which the dores weren faste yshette house of which the doors were fast shut  (CMCTMELI, 217.C1b.7) ‘He has left his wife and his daughter within his house, the doors of which were shut fast.’

 Masayuki Ohkado

(27) a.

infinitive + modal (VM) Sey what men seye wil say what men say will ‘Say what men want to say.’

(CMCLOUD, 81.405)

b. modal + infinitive (MV) ... it is þe schortest werke of alle þat man may ymagyn it is the shortest work of all that man can imagine  (CMCLOUD, 17.86) ‘... it is the shortest work of all that man can imagine’ (28) a. NEG + BE/HAVE/modal verb/verb ... þat Seint Ruffus not be-gan þis ordr, ... that Saint Ruffus not began this order  (CMCAPSER, 147.42) ‘... that Saint Ruffus did not begin this order, ...’ b. BE/HAVE/modal verb/verb + NEG ... he asked why sche came not to soper he asked why she came not to supper (CMCAPCHR, 49.534) ‘... he asked why she did not come to supper’ (29) a. adverb (non NEG) + BE/HAVE/modal verb/verb ... þat Coryn pryuyly shuld gon out ... that Coryn privily should go out (CMBRUT3, 10.241) ‘... that Coryn should go out secretly ...’ b. BE/HAVE/modal verb/verb + adverb (non NEG) ... þat Richard, þe son of Wiliam, schuld frely holde that Richard the son of William should freely hold al Normandy (CMCAPCHR, 93.1852) all Normandy ‘... that Richard, the son of William should hold all Normandy freely’ If stylistic fronting is involved, it is expected that the head-initial patterns or patterns with fronted negatives or adverbs are observed with higher frequencies in clauses with a subject gap than in clauses without a subject gap. 4.3

The results

The results are given in Tables 11 to 15. Taking into account possible dialectal differences, the figures in the Tables are broken down by dialects. Note that the Ormulum text, which is contained in the corpus, is excluded from the data because it is a poetic text. Just as in the analysis of Old English data, Fisher’s exact probability test was conducted to see whether there is a statistically significant difference



Stylistic fronting in the history of English 

Table 11.  Adjective-BE (AB) and BE-Adjective (BA) patterns AB

full NP subject BA total

personal pronoun subject AB BA total

AB

empty subject BA total

EAST MIDLANDS 16 285 301 (5.3%) (94.7%)

12 (3.7%)

313 (96.3%)

325

KENTISH   46 (97.9%)

 5 (8.8%)

  52 (91.2%)

  57

4 (1.7%)

238 (98.3%)

242

0 (0.0%)

  28 (100.0%)

  28

 1 (2.1%)

0 (0.0%)

  28 (100.0%)

  28

 0 (0.0%)

NORTHERN 285 285 (100.0%)

11 (3.4%)

313 (96.6%)

324

0 (0.0%)

  77 (100.0%)

  77

 0 (0.0%)

SOUTHERN   75   75 (100.0%)

 3 (3.8%)

  77 (96.3%)

  80

0 (0.0%)

174 (100.0%)

174

WEST MIDLANDS 12 238 250 (4.8%) (95.2%)

16 (6.3%)

239 (93.7%)

255

  47

Table 12.  Past participle-BE (VB) and BE-past participle (BV) patterns VB

full NP subject BV total

personal pronoun subject VB BV total

VB

empty subject BV total

EAST MIDLANDS  6 420 426 (1.4%) (98.6%)

3 (0.7%)

418 (99.3%)

421

KENTISH   32 (86.5%)

1 (3.2%)

  30 (96.8%)

  31

2 (0.5%)

434 (99.5%)

436

1 (4.8%)

  20 (95.2%)

  21

 5 (13.5%)

0 (0.0%)

  79 (100.0%)

  79

 0 (0.0%)

NORTHERN 113 113 (100.0%)

0 (0.0%)

  72 (100.0%)

  72

0 (0.0%)

138 (100.0%)

138

 0 (0.0%)

SOUTHERN 111 111 (100.0%)

0 (0.0%)

198 (100.0%)

198

0 (0.0%)

258 (100.0%)

258

WEST MIDLANDS 12 309 321 (3.7%) (96.3%)

7 (2.2%)

306 (97.8%)

313

  37

 Masayuki Ohkado

Table 13.  Infinitive-modal (VM) and modal-infinitive (MV) patterns VM

full NP subject MV total

personal pronoun subject VM MV total

empty subject VM MV total

EAST MIDLANDS 4 (1.0%)

393 (99.0%)

397

26 (2.8%)

890 (97.2%)

916

12 (4.9%)

231 (95.1%)

243

  67

 0 (0.0%)

  37 (100.0%)

  37

242

 0 (0.0%)

  80 (100.0%)

  80

SOUTHERN 176 176 (100.0%)

 1 (1.7%)

  58 (98.3%)

  59

WEST MIDLANDS 12 548 560 (2.1%) (97.9%)

 3 (1.8%)

166 (98.2%)

169

KENTISH 0 (0.0%)

  17 (100.0%)

  17

 2 (3.0%)

  65 (97.0%) NORTHERN

0 (0.0%)

  60 (100.0%)

  60

 1 (0.4%)

0 (0.0%)

  92 (100.0%)

  92

 0 (0.0%)

0 (0.0%)

180 (100.0%)

180

241 (99.6%)

Table 14.  NEG-BE/HAVE/modal verb/verb and BE/HAVE/modal verb/verb-NEG full NP subject NEG-X X-NEG total

personal pronoun subject NEG-X X-NEG total

empty subject NEG-X X-NEG total

1 (0.6%)

181

EAST MIDLANDS 9 406 415 (2.2%) (94.7%)

5 (2.7%)

179 (97.3%)

184

   5

KENTISH 0   27 (0.0%) (100.0%)

6 (16.7%)

  30 (83.3%)

  36

  26

NORTHERN 1 153 154 (0.6%) (99.4%)

2 (2.8%)

  69 (97.2%)

  71

  56

SOUTHERN 0   88   88 (0.0%) (100.0%)

0 (0.0%)

  48 (100.0%)

  48

  64

WEST MIDLANDS 8 154 162 (4.9%) (95.1%)

4 (5.3%)

  72 (94.7%)

  76

1 (20.0%) 1 (3.8%) 0 (0.0%) 0 (0.0%)

180 (99.4%)    4 (80.0%)   25 (96.2%)   56 (100.0%)   64 (100.0%)

  27



Stylistic fronting in the history of English 

Table 15.  Adverb (non NEG)-BE/HAVE/modal verb/verb and BE/HAVE/modal verb/ verb-adverb (non NEG) full NP subject ADV-X X-ADV total

personal pronoun subject ADV-X X-ADV total

empty subject ADV-X X-ADV total

EAST MIDLANDS 26 (9.8%)

240 (90.2%)

266

59 (10.7%)

492 (89.3%)

551

151 (40.6%)

221 (59.4%)

372

  60

  37 (38.9%)

  58 (61.1%)

  95

161

  27 (29.3%)

  65 (70.7%)

  92

  78

  12 (14.3%)

  72 (85.7%)

  84

255 (34.1%)

132 (65.9%)

387

KENTISH 6 (23.1%)

  20 (76.9%)

  26

 6 (10.0%)

  54 (90.0%)

NORTHERN 5 (10.4%)

  43 (89.6%)

  48

11 (6.8%)

150 (93.2%) SOUTHERN

3 (6.3%)

  45 (93.8%)

  48

 2 (2.6%)

  76 (97.4%)

WEST MIDLANDS 33 (17.4%)

157 (82.6%)

190

91 (20.7%)

349 (79.4%)

440

between the figures concerning examples with full NP subjects and those concerning examples with personal pronoun subjects or empty subjects.11 The italicized parts indicate that there is a statistically significant difference between the relevant figures. Table 11, for instance, indicates that there is such a statistically significant difference between the figures concerning the following examples: (i) examples with full NP subjects and those with personal pronoun subjects in the East Midlands dialect; (ii) examples with full NP subjects and those with personal pronoun subjects in the West Midlands dialect; (iii) examples with full NP subjects and those with empty subjects in the West Midlands dialect. The situation observed in Tables 11 to 15 is summarized in Table 16. Table 16 suggests that stylistic fronting was operative in all the dialects except the Southern dialect in Middle English. It should be noted that, although the result is not statistically significant, the Southern dialect also shows the “stylistic fronting pattern”, that is, the pattern in which examples with empty subjects shows higher frequencies of head-final order than those with full NP subjects in three of the five constructions (cf. Tables 11, 13, and 15). This seems to indicate 11. Again, the calculation was done by using InStat 3 for Macintosh.

 Masayuki Ohkado

Table 16.  Number of constructions showing the “stylistic fronting” patterns with statistical significance Number of constructions showing the “stylistic fronting” patterns Dialect East Midlands Kentish Northern Southern West Midlands

Clauses with a personal pronouns subject 2 of the 5 constructions 1 of the 5 constructions 0 of the 5 constructions 0 of the 5 constructions 3 of the 5 constructions

Clauses with a subject gap 2 of the 5 constructions 0 of the 5 constructions 1 of the 5 constructions 0 of the 5 constructions 3 of the 5 constructions

that we might be able to draw an even stronger conclusion that stylistic fronting was operative in all the dialects of Middle English. 5. Scandinavian influences? Examining the Middle English text of Ormulum, Trips (2002) argues that the “stylistic fronting pattern” is observable in this text and that the “existence” of the pattern is due to Scandinavian influence. Likewise, demonstrating that the similar phenomenon is found in the Old English interlinear glosses to the Lindisfarne Gospels, which are written in the Northern variety of Old English, Ohkado (2006) suggests that Scandinavian influence on syntax is already visible in the Old English period. If the presence of stylistic fronting in Old and Middle English is due to language contact with Old Norse as these authors suggest, it is expected that the “stylistic fronting pattern” is attested in the dialects heavily influenced by the Scandinavian invaders (Northern and East Midlands) and not attested in the other dialects (Southern, Kentish, and West Midlands) (cf. Thomason and Kaufman 1988: 263–342). The data presented in Section 3 and 4 show that the “stylistic fronting pattern” is widely observed in texts which are unlikely to have been influenced by the Scandinavian invaders so that the existence of stylistic fronting in Old and Middle English cannot solely attributed to the influence of Old Norse. Therefore, I conclude that the process of stylistic fronting was a genuine property of Old and Middle English.



Stylistic fronting in the history of English 

References Canale, Michael. 1978. Word Order Changes in OE: Base Re-analysis in Generative Grammar. Ph.D dissertation, McGill University. Faarlund, Jan Terje. 2004. The Syntax of Old Norse: With a Survey of the Inflectional Morphology and a Complete Bibliography. Oxford: Oxford University Press. Fischer, Olga, Ans van Kemenade, Willem Koopman & Wim van der Wurff. 2000. The Syntax of Early English. Cambridge: Cambridge University Press. Hrafnbjargarson, Gunnar Hrafn. 2004. Oblique Subjects and Stylistic Fronting in the History of Scandinavian and English: The role of IP-spec. Ph.D dissertation, University of Aarhus. Kroch, Anthony & Ann Taylor. 2000. The Penn-Helsinki Parsed Corpus of Middle English, second edition (PPCME2). Philadelphia: Department of Linguistics, University of Pennsylvania. Lightfoot, David. 1979. Principles of Diachronic Syntax. Cambridge: Cambridge University Press. Maling, Joan. 1990. “Inversion in Embedded Clauses in Modern Icelandic”. Modern Icelandic syntax. (= Syntax and semantics, 24) ed. by Joan Maling & Annie Zaenen, 71–91. New York: Academic Press. Ohkado, Masayuki. 2001. Old English Constructions with Multiple Predicates. Tokyo: Hituzi Syobo. Ohkado, Masayuki. 2005. Clause Structure in Old English, Ph.D dissertation, University of Amsterdam. Ohkado, Masayuki. 2006. “On Word Order in Constructions with Two Predicates in Old English Interlinear Glosses”. Textual and Contextual Studies in Medieval English: Towards the Reunion of Linguistics and Philology ed. by Michiko Ogura, 127–145. Frankfurt am Main: Peter Lang. Pintzuk, Susan. 1999. Phrase Structures in Competition: Variation and Change in Old English Word Order. New York: Garland. Pintzuk, Susan & Ann Taylor. 2006. “The Loss of OV Order in the History of English”. The Handbook of the History of English ed. by Ans van Kemenade & Bettelou Los, 249–278. Oxford: Blackwell. Taylor, Ann, Anthony Warner, Susan Pintzuk & Frank Beths. 2003. The York-Toronto-Helsinki Parsed Corpus of Old English Prose. York: University of York. Thomason, Sarah Grey & Terrence Kaufman. 1988. Language Contact, Creolization, and Genetic Linguistics. Berkeley: University of California Press. Trips, Carola. 2002. From OV to VO in Early Middle English. Amsterdam & Philadelphia: John Benjamins. Van Kemenade, Ans. 1987. Syntactic Case and Morphological Case in the History of English. Dordrecht: Foris. Willson, Kendra. 2001. “Old Icelandic Topicalization and the Emergence of Stylistic Fronting”. Proceedings of the 18th Scandinavian Conference of Linguistics, Vol. 2 ed. by Arthur Holmer, Jan-Olof Svantesson, & Åke Viberg, 127–137. Lund: Lund University.

 Masayuki Ohkado

Appendix: Command lines used for the present study This appendix gives the command lines for CorpusSearch 2 used for obtaining the Old English data in the present study. To obtain the Middle English data, the command lines should be slightly modified because of the differences in the annotation systems in the two corpora. For the relevant information, see the “The major ways in which the YCOE differs from the PPCME2” file which is available at: http:// www-users.york.ac.uk/~lang22/YCOE/doc/annotation/ppcme2-to-ycoe.htm The command used for searching OV patterns in Table 1 is given in (i). (i)

Command for searching OV patterns node: IP-INF query: (IP-INF iDomsTotal 2) AND (IP-INF iDominates NP-ACC) AND (NP-ACC iDomsOnly N^A|NR^A) AND (IP-INF iDominates VB) AND (NP-ACC Precedes VB) AND (VB iDominates![1]¥**)

For examples with VO patterns, the sixth line in (i) is replaced by “AND (VB Precedes NP-ACC).” The command in (ii) is for OV patterns in VPs headed by nonfinite verbs with an NP object containing three or more words in Table 2. (ii) Command for searching OV patterns node: IP-INF query: (IP-INF iDomsTotal 2) AND (IP-INF iDominates NP-ACC) AND (NP-ACC iDomsTotal> 3) AND (NP-ACC iDominates N^A|NR^A) AND (IP-INF iDominates VB) AND (NP-ACC Precedes VB) AND (VB iDominates![1]¥**) For corresponding examples with VO patterns, the seventh line in (ii) is replaced by “AND (VB Precedes NP-ACC).” The command in (iii) is for clauses with “adjective + BE (AB)” patterns with full NP subjects in Table 3.



Stylistic fronting in the history of English 

(iii) Command for searching AB patterns with full NP subjects node: IP-SUB* query: (IP-SUB* iDomsTotal 3) AND (IP-SUB* iDominates NP-NOM) AND (NP-NOM iDominates N^N|NR^N) AND (IP-SUB* iDominates *BED*|*BEP*) AND (IP-SUB* iDominates ADJP-NOM-PRD) AND (ADJP-NOM-PRD iDomsOnly ADJ^N) AND (NP-NOM Precedes ADJP-NOM-PRD) AND (ADJP-NOM-PRD Precedes *BED*|*BEP*) AND (*BED*|*BEP* iDominates![1]\**) AND (ADJP-NOM-PRD iDominates ![2]\**) For clauses with personal pronoun subjects, the fourth line “AND (NP-NOM iDominates N^N|NR^N)” is replaced by “AND (NP-NOM iDominates PRO^N).” Likewise, for clauses with empty subjects the fourth line is replaced by “AND (NP-NOM iDominates \*con\*|\*exp\*|\*pro\*|\*T\**).” For clauses with BE and past participles in Table 4, “ADJP-NOM-PRD” in (iii) is replaced by “*VBN.” For clauses with modals and infinitives in Table 5, “*BED*|*BEP*” and “ADJP-NOM-PRD” in (iii) are replaced by “*MDD*|*MDP*” and “*VB.” For clauses with NEG in Tables 6 to 8, the basic command is the one given in (iv). (iv) Command for searching clauses with NEG (for clauses with full NP subject with NEG preceding the finite verb) node: IP-SUB* query: (IP-SUB* iDominates NP-NOM) AND (NP-NOM iDominates N^N|NR^N) AND (IP-SUB* iDominates *BED*|*BEP*|*HVD*|*HVP*|* MDD*|*MDP*) AND (IP-SUB* iDominates ADVP) AND (ADVP iDominates NEG + ADV) AND (NP-NOM iPrecedes ADVP) AND (ADVP iPrecedes *BED*|*BEP*|*HVD*|*HVP*|*MDD *|*MDP*) AND (*BED*|*BEP*|*HVD*|*HVP*|*MDD*|*MDP* iDominates![1]\**) For clauses with NEG preceding the finite verb, the command in (v) is also necessary to retrieve examples with the negating particle ne not contracted with the finite verb.

 Masayuki Ohkado

(v) Command for searching clauses with NEG with ne not contracted (for clauses with full NP subject with NEG preceding finite verb) node: IP-SUB* query: (IP-SUB* iDominates NP-NOM) AND (NP-NOM iDominates N^N|NR^N) AND (IP-SUB* iDominates *BED*|*BEP*|*HVD*|*HVP*|* MDD*|*MDP*) AND (IP-SUB* iDominates ADVP) AND (ADVP iDominates NEG + ADV) AND (IP-SUB* iDominates NEG) AND (NEG iDominates ne) AND (NP-NOM iPrecedes ADVP) AND (ADVP iPrecedes NEG) AND (NEG iPrecedes *BED*|*BEP*|*HVD*|*HVP*|*MDD *|*MDP*) AND (*BED*|*BEP*|*HVD*|*HVP*|*MDD*|*MDP* iDominates![1]\**) The third line in (iv) and (v) is replaced by “AND (NP-NOM iDominates PRO^N)” for clauses with personal pronoun subjects and by “AND (NP-NOM iDominates \*con\*|\*exp\*|\*pro\*|\*T\**)” for clauses with empty subjects. For clauses with NEG following the finite verb, the seventh line in (iv) is replaced with “AND (NP-NOM iPrecedes *BED*|*BEP*|*HVD*|*HVP*|*MDD*|*MDP*)” and the order of the elements sandwiching “iPrecedes” in the eighth line in (iv) is reversed. After obtaining examples with negative adverbs, I searched for na/no, næfre, and nawiht separately to retrieve relevant examples in Tables 6 to 8. For clauses with non NEG adverbs in Table 9, the sixth line in (iv) and (v) is replaced by “ AND (ADVP* iDominates !NEG*).”

Subject and Word index A A Representative Corpus of Historical English Registers (ARCHER)  79–93 accessibility hierarchy  257, 258 accusative  4, 11, 12, 13, 15, 17–26, 80, 193, 195, 260, 261 accustomed  4, 29–43 ACE; see Australian Corpus of English active  13, 15n., 48, 65, 66 adjective  5, 7, 13, 29–43, 75, 95–119, 146, 155, 255–274 adverb  7, 55, 143–161, 171, 188, 192, 193, 195n., 196, 238, 255–274 adverbials  17, 70, 74, 185, 187, 192–194, 240, 241 Ælfric’s Grammar  240–243 agreement  6, 13, 14n., 15, 18, 26n., 185, 187, 190, 193, 195 American English  29–43, 45–48, 82n., 130, 168 analogy  95, 97, 115, 119, 148n., 227, 229, 231n. Ancrene Riwle  243, 244–246, 250, 251 ARCHER; see A Representative Corpus of Historical English Registers article  87, 88, 91 Australian Corpus of English (ACE)  123–140 B be  63–77, 217–234, 237–253 benefactive  80 bleaching  89, 92, 136, 137, 139 BNC; see British National Corpus British National Corpus (BNC)  45–60, 79n., 123–140 Brown Corpus  45–47, 123–140, 167–170, 178

C CB; see Collins COBUILD Corpus CEEC; see Corpus of Early English Correspondence Celtic  3, 6, 7, 217–234, 240n. cigan (OE)  19, 20 clefting  258 CLMET; see Corpus of Late Modern English texts Collins COBUILD Corpus (CB)  46, 50, 95–119 collocation  1, 2, 5, 6, 74, 89n., 143–161, 231n. collostructional analysis  98, 99, 106, 111, 118 commitment  165, 178 complementation of adjective  4, 5, 29–43, 95–119 of verb  4, 11–28, 45–60, 124–126, 137, 138, 148 Constant Ratio Hypothesis  259, 262 construction  1, 3–5, 79–92, 95–119 copularity  4, 11–28 Corpus of Early English Correspondence (CEEC)  124, 197–210 Corpus of Late Modern English texts (CLMET)  98, 111–115, 123–140 Cursor Mundi  252 D dative  11, 12, 15, 17, 18, 20, 21, 25, 80, 91 DCPSE; see Diachronic Corpus of Present-Day Spoken English deixis/deictic  101, 132, 240 deontic; see also modality; shall, deontic; root (uses)  5, 6, 95–119, 144, 145, 147, 148, 150, 153, 157, 161, 173, 178, 241n., 247

descriptivism  63, 64, 66, 76 Diachronic Corpus of Present-Day Spoken English (DCPSE)  165–179 ditransitive  15n., 17, 19–21, 23, 89 Double Base Hypothesis  259 E Early Modern English  5, 6, 85, 97, 98, 114, 118, 119, 197–210 eaþe (OE)/eaþelice (OE)/esili (ME)  143–161 Emergent Grammar  118 entrenchment  2, 41, 43, 49, 89, 99, 118 epistemic; see also modality  6, 101, 143–161, 165–179 Extraction Principle  29, 32, 34, 35, 42 extraposition  5, 95–119, 185, 191, 195, 259, 260, 262, 263n. F fail  5, 123–140 Fisher’s Exact Test  99, 103, 104, 107n., 114, 265, 270 FLOB; see The Freiburg-LOB Corpus of British English The Freiburg-Brown Corpus of American English (Frown)  45–47, 123–140, 167–170, 178 The Freiburg-LOB Corpus of British English (FLOB)  45–47, 80, 82, 92, 123–140, 167–170, 178 French  3, 91, 92, 124n., 146, 242 Frown; see The Freiburg-Brown Corpus of American English future (tense)/futurity; see also shall, future  6, 7, 217–234, 237–253

 English Historical Linguistics 2008 G genitive  6, 12n., 15, 16n., 17, 18, 20, 21, 46, 190, 194, 195, 196, 197–210 split  194–196, 202 genre(s); see also letters; history writing  6, 68, 197–210, 224–226 give  79–92 good  5, 95–119 grammars, late Middle English  242 grammaticalization  2, 3, 5, 89n., 123, 125, 136–140 Great Complement Shift  29, 32, 42 H habitual  6, 217, 220, 221, 223, 225, 227, 228, 233 Hali Meiðhad  250 hatan (OE)  13, 19, 20 have  63–76 have got to  165–179 have to  165–179 Havelok  251 HC; see Helsinki Corpus of English Texts Helsinki Corpus of English Texts (HC)  79–93, 124, 157, 158, 217–234 history writing; see also genre(s)  197, 204, 210 I Icelandic  255–258, 263, 267 idiomatization  79–92, 159, 161 imperative  7, 38, 40, 217–221, 231–234, 239 inference, invited  148, 150, 157, 160 infinitive  7, 34, 70, 113n., 124n., 217, 219, 231–234, 261‑272 to-infinitive  4, 5, 29–43, 73, 98–101, 104–108, 110–113 K KAK pattern/knowledge/acquisition of knowledge  95–119 L LAEME; see Linguistic Atlas of Early Middle English

læran (OE)  20, 23, 25 Lambeth Homilies  250 The Lancaster-Oslo-Bergen Corpus (LOB)  45–47, 123–140, 167– 170, 178 LAOS; see Linguistic Atlas of Older Scots Late Modern English  63–76, 82n., 95, 97, 98, 114, 115, 118, 127, 128 Latin  3, 26, 91, 150, 154, 157, 161, 221–224, 232, 240, 242–247 letters; see also genre(s)  4, 6, 63–76, 197, 198, 200, 203, 207n., 208–210 lexicalization  5, 89, 143, 150, 159–161 Linguistic Atlas of Early Middle English (LAEME)  237, 243–252 Linguistic Atlas of Older Scots (LAOS)  252 LOB; see The Lancaster-OsloBergen Corpus locative/locative pattern  5, 73, 95–119, 201, 240n. M mæg (OE)  5, 6, 143–161 mandative  5, 95–119 Mercian  222, 224, 229–233 Middle English  4–7, 64, 80, 82, 83, 90–92, 98, 112, 113n., 115, 143, 146, 148–150, 153, 155n., 157–161, 167, 197–210, 218, 237–253, 255, 256, 268–274 modal/modality; see also deontic; epistemic; modal verb; verb, pre-modal  5, 6, 143–161, 165–179 modal verb; see also modal/modality  5, 6, 101, 116n., 143–161, 165–179, 262–273 multiple distinctive collexeme analysis  98–100, 105–107, 110, 111 must  165–179 N na (OE)  264, 266 ne (OE)  262n., 264, 265 negation/negative  5, 7, 70, 74, 123–140, 160, 171, 172, 260, 262n., 264, 270

nemnan (OE)  19, 20 no (OE)  264, 266 nominative  4, 11, 12, 13, 15, 16n., 17, 18, 20, 21, 24, 25, 26 Northumbrian  222, 223, 228–233 noun, deverbal  85, 86n., 89n., 92 nouniness  34, 79, 86–88, 92 NP shift  195 O object  5, 11, 13, 15–18, 23–26, 38, 45, 46, 49, 50, 54, 55, 57, 58, 60, 79–92, 260–263, 268 eventive  84, 85, 92 obligation  125, 126, 168, 169, 173, 176–178, 241n., 247, 249 oblique  11, 17, 18, 19 Old English  4–7, 11–28, 63, 80, 89n., 90, 98, 112, 115, 118, 143–157, 159–161, 185–196, 199, 217–234, 237–243, 259–268, 274 Old Norse; see also Scandinavian  3, 7, 146, 219, 225, 255–274 Older Scots  252 Ormulum  255, 270, 274 ought  116n., 165, 168, 178 OV; see also SOV; word order  167, 192, 261, 262, 268 P partially filled construction  5, 95–119 participle  7, 108, 111, 114, 117, 244 ing-participle  50, 54, 57, 58, 60, 109 past participle  12n., 25n., 63–76, 146, 191, 193, 223, 246, 255–258, 262–271, 276 particle  256, 258, 262, 277 passive  12, 15, 16n., 18, 23–28, 38, 39, 47, 48, 64–66, 221–224, 239n., 241, 257 Penn-Helsinki Parsed Corpus of Middle English (PPCME)  98, 111–115, 197–210, 269–274 Penn-Helsinki Parsed Corpus of Early Modern English (PPCEME)  98, 111–115, 197–210 perception  105–111, 114–131, 133 Poema Morale  247–249 possessive  46, 87, 88, 109, 195, 198, 199, 205



Subject and Word index  possibility  6, 143–161 PPCEME; see Penn-Helsinki Parsed Corpus of Early Modern English PPCME; see Penn-Helsinki Parsed Corpus of Middle English prescriptivism  60, 63, 64, 67, 75 prevent  45–60 Priestley, Joseph  4, 5, 63–76 productivity  80, 86, 89, 92, 143, 159–161, 203 pronoun personal  12, 50, 83, 91, 185–190, 193–195, 203, 255, 260, 263–267, 271–274 resumptive  194, 196 Q quantifier  6, 13, 185–196, 260 quantifier floating  185–196 R root (uses); see also deontic  165, 168, 169, 172–179 S Scandinavian; see also Old Norse  67, 255, 258, 259, 274 scrambling  6, 185, 189, 190, 193, 195 semi-modal; see also modal/ modality  165–179 sermons; see also genre(s)  6, 197–210

shall  143, 168, 241n., 242, 244, 245, 247–250, 252, 253 deontic  247, 249 future  244, 245, 247, 250, 252, 253 should  101, 116n., 165, 168, 169n., 177, 178 SOV; see also OV; word order  260, 268 status  4, 11, 17, 19 structures, discontinuous  6, 185–196 stylistic fronting  7, 255–274 subject derived  15, 18 empty  7, 257, 263–274 inserted  17, 28 raised  18, 28 subject control  29–33, 36, 39, 42, 43 subject gap  7, 255–274 subjunctive  7, 101, 153, 167, 217–220, 227, 229–232, 234, 239, 249–252 subordination  155, 256, 259, 262, 269 SVO; see also VO; word order  260, 268 T text language  237, 243, 247–252 that-clause  95–119 TIME corpus  2, 4, 29–43 to -ing complement  29–43 to-clause  95–119

to-infinitive; see infinitive topicality  6, 197–199, 203–208 topicalization  6, 14n., 34–36, 185, 188–190, 192,193, 195, 258 transitivity; see also ditransitive  4, 11, 17, 19, 65 þyncan (OE)  12, 13, 17 V variation, syntactic  6, 7, 28 verb, mutative intransitive  63–76 verb, pre-modal; see also modal/ modality  143–161 VO; see also SVO; word order  167, 261, 262, 268 W WebCorp  123–140, 172 Wellington Corpus  123–140 weorþan (OE)  12, 17, 230, 239, 253 Worcester Tremulous Scribe  243 word order  7, 83, 199, 249, 255, 259, 260, 268 Y YCOE; see York-TorontoHelsinki Parsed Corpus of Old English York-Toronto-Helsinki Parsed Corpus of Old English (YCOE)  98, 111–115, 185–196, 260–268

CURRENT ISSUES IN LINGUISTIC THEORY

E. F. K. Koerner, Editor

Zentrum für Allgemeine Sprachwissenschaft, Typologie und Universalienforschung, Berlin [email protected] Current Issues in Linguistic Theory (CILT) is a theory-oriented series which welcomes contributions from scholars who have significant proposals to make towards the advancement of our understanding of language, its structure, functioning and development. CILT has been established in order to provide a forum for the presentation and discussion of linguistic opinions of scholars who do not necessarily accept the prevailing mode of thought in linguistic science. It offers an outlet for meaningful contributions to the current linguistic debate, and furnishes the diversity of opinion which a healthy discipline must have. A complete list of titles in this series can be found on the publishers’ website, www.benjamins.com 315 COLINA, Sonia, Antxon OLARREA and Ana Maria CARVALHO (eds.): Romance Linguistics 2009. Selected papers from the 39th Linguistic Symposium on Romance Languages (LSRL), Tucson, Arizona, March 2009. xiv, 421 pp. + index. Expected November 2010 314 LENKER, Ursula, Judith HUBER and Robert MAILHAMMER (eds.): English Historical Linguistics 2008. Selected papers from the fifteenth International Conference on English Historical Linguistics (ICEHL 15), Munich, 24-30 August 2008.. Volume I: The history of English verbal and nominal constructions. 2010. vii, 282 pp. 313 ARREGI, Karlos, Zsuzsanna FAGYAL, Silvina A. MONTRUL and Annie TREMBLAY (eds.): Romance Linguistics 2008. Interactions in Romance. Selected papers from the 38th Linguistic Symposium on Romance Languages (LSRL), Urbana-Champaign, April 2008. 2010. vii, 266 pp. 312 OPERSTEIN, Natalie: Consonant Structure and Prevocalization. 2010. x, 234 pp. 311 SCALISE, Sergio and Irene VOGEL (eds.): Cross-Disciplinary Issues in Compounding. 2010. viii, 382 pp. 310 RAINER, Franz, Wolfgang U. DRESSLER, Dieter KASTOVSKY and Hans Christian LUSCHÜTZKY (eds.): Variation and Change in Morphology. Selected papers from the 13th International Morphology Meeting, Vienna, February 2008. With the assistance of Elisabeth Peters. 2010. vii, 249 pp. 309 NICOLOV, Nicolas, Galia ANGELOVA and Ruslan MITKOV (eds.): Recent Advances in Natural Language Processing V. Selected papers from RANLP 2007. 2009. x, 338 pp. 308 DUFRESNE, Monique, Fernande DUPUIS and Etleva VOCAJ (eds.): Historical Linguistics 2007. Selected papers from the 18th International Conference on Historical Linguistics, Montreal, 6–11 August 2007. 2009. x, 311 pp. 307 CALABRESE, Andrea and W. Leo WETZELS (eds.): Loan Phonology. 2009. vii, 273 pp. 306 VIGÁRIO, Marina, Sónia FROTA and M. João FREITAS (eds.): Phonetics and Phonology. Interactions and interrelations. 2009. vi, 290 pp. 305 BUBENIK, Vit, John HEWSON and Sarah ROSE (eds.): Grammatical Change in Indo-European Languages. Papers presented at the workshop on Indo-European Linguistics at the XVIIIth International Conference on Historical Linguistics, Montreal, 2007. 2009. xx, 262 pp. 304 MASULLO, Pascual José, Erin O'ROURKE and Chia-Hui HUANG (eds.): Romance Linguistics 2007. Selected papers from the 37th Linguistic Symposium on Romance Languages (LSRL), Pittsburgh, 15–18 March 2007. 2009. vii, 361 pp. 303 TORCK, Danièle and W. Leo WETZELS (eds.): Romance Languages and Linguistic Theory 2006. Selected papers from ‘Going Romance’, Amsterdam, 7–9 December 2006. 2009. viii, 262 pp. 302 FERRARESI, Gisella and Maria GOLDBACH (eds.): Principles of Syntactic Reconstruction. 2008. xvii, 219 pp. 301 PARKINSON, Dilworth B. (ed.): Perspectives on Arabic Linguistics. Papers from the annual symposium on Arabic linguistics. Volume XXI: Provo, Utah, March 2007. 2008. x, 206 pp. 300 VAJDA, Edward J. (ed.): Subordination and Coordination Strategies in North Asian Languages. 2008. xii, 218 pp. 299 GONZÁLEZ-DÍAZ, Victorina: English Adjective Comparison. A historical perspective. 2008. xix, 252 pp. 298 BOWERN, Claire, Bethwyn EVANS and Luisa MICELI (eds.): Morphology and Language History. In honour of Harold Koch. 2008. x, 364 pp. 297 DOSSENA, Marina, Richard DURY and Maurizio GOTTI (eds.): English Historical Linguistics 2006. Selected papers from the fourteenth International Conference on English Historical Linguistics (ICEHL 14), Bergamo, 21–25 August 2006. Volume III: Geo-Historical Variation in English. 2008. xiii, 197 pp.

296 DURY, Richard, Maurizio GOTTI and Marina DOSSENA (eds.): English Historical Linguistics 2006. Selected papers from the fourteenth International Conference on English Historical Linguistics (ICEHL 14), Bergamo, 21–25 August 2006. Volume II: Lexical and Semantic Change. 2008. xiii, 264 pp. 295 GOTTI, Maurizio, Marina DOSSENA and Richard DURY (eds.): English Historical Linguistics 2006. Selected papers from the fourteenth International Conference on English Historical Linguistics (ICEHL 14), Bergamo, 21–25 August 2006. Volume I: Syntax and Morphology. 2008. xiv, 259 pp. 294 FRELLESVIG, Bjarke and John WHITMAN (eds.): Proto-Japanese. Issues and Prospects. 2008. vii, 229 pp. 293 DETGES, Ulrich and Richard WALTEREIT (eds.): The Paradox of Grammatical Change. Perspectives from Romance. 2008. vi, 252 pp. 292 NICOLOV, Nicolas, Kalina BONTCHEVA, Galia ANGELOVA and Ruslan MITKOV (eds.): Recent Advances in Natural Language Processing IV. Selected papers from RANLP 2005. 2007. xii, 307 pp. 291 BAAUW, Sergio, Frank DRIJKONINGEN and Manuela PINTO (eds.): Romance Languages and Linguistic Theory 2005. Selected papers from ‘Going Romance’, Utrecht, 8–10 December 2005. 2007. viii, 338 pp. 290 MUGHAZY, Mustafa A. (ed.): Perspectives on Arabic Linguistics. Papers from the annual symposium on Arabic linguistics. Volume XX: Kalamazoo, Michigan, March 2006. 2007. xii, 247 pp. 289 BENMAMOUN, Elabbas (ed.): Perspectives on Arabic Linguistics. Papers from the annual symposium on Arabic Linguistics. Volume XIX: Urbana, Illinois, April 2005. 2007. xiv, 304 pp. 288 TOIVONEN, Ida and Diane NELSON (eds.): Saami Linguistics. 2007. viii, 321 pp. 287 CAMACHO, José, Nydia FLORES-FERRÁN, Liliana SÁNCHEZ, Viviane DÉPREZ and María José CABRERA (eds.): Romance Linguistics 2006. Selected papers from the 36th Linguistic Symposium on Romance Languages (LSRL), New Brunswick, March-April 2006. 2007. viii, 340 pp. 286 WEIJER, Jeroen van de and Erik Jan van der TORRE (eds.): Voicing in Dutch. (De)voicing – phonology, phonetics, and psycholinguistics. 2007. x, 186 pp. 285 SACKMANN, Robin (ed.): Explorations in Integrational Linguistics. Four essays on German, French, and Guaraní. 2008. ix, 239 pp. 284 SALMONS, Joseph C. and Shannon DUBENION-SMITH (eds.): Historical Linguistics 2005. Selected papers from the 17th International Conference on Historical Linguistics, Madison, Wisconsin, 31 July - 5 August 2005. 2007. viii, 413 pp. 283 LENKER, Ursula and Anneli MEURMAN-SOLIN (eds.): Connectives in the History of English. 2007. viii, 318 pp. 282 PRIETO, Pilar, Joan MASCARÓ and Maria-Josep SOLÉ (eds.): Segmental and prosodic issues in Romance phonology. 2007. xvi, 262 pp. 281 VERMEERBERGEN, Myriam, Lorraine LEESON and Onno CRASBORN (eds.): Simultaneity in Signed Languages. Form and function. 2007. viii, 360 pp. (incl. CD-Rom). 280 HEWSON, John and Vit BUBENIK: From Case to Adposition. The development of configurational syntax in Indo-European languages. 2006. xxx, 420 pp. 279 NEDERGAARD THOMSEN, Ole (ed.): Competing Models of Linguistic Change. Evolution and beyond. 2006. vi, 344 pp. 278 DOETJES, Jenny and Paz GONZÁLEZ (eds.): Romance Languages and Linguistic Theory 2004. Selected papers from ‘Going Romance’, Leiden, 9–11 December 2004. 2006. viii, 320 pp. 277 HELASVUO, Marja-Liisa and Lyle CAMPBELL (eds.): Grammar from the Human Perspective. Case, space and person in Finnish. 2006. x, 280 pp. 276 MONTREUIL, Jean-Pierre Y. (ed.): New Perspectives on Romance Linguistics. Vol. II: Phonetics, Phonology and Dialectology. Selected papers from the 35th Linguistic Symposium on Romance Languages (LSRL), Austin, Texas, February 2005. 2006. x, 213 pp. 275 NISHIDA, Chiyo and Jean-Pierre Y. MONTREUIL (eds.): New Perspectives on Romance Linguistics. Vol. I: Morphology, Syntax, Semantics, and Pragmatics. Selected papers from the 35th Linguistic Symposium on Romance Languages (LSRL), Austin, Texas, February 2005. 2006. xiv, 288 pp. 274 GESS, Randall S. and Deborah ARTEAGA (eds.): Historical Romance Linguistics. Retrospective and perspectives. 2006. viii, 393 pp. 273 FILPPULA, Markku, Juhani KLEMOLA, Marjatta PALANDER and Esa PENTTILÄ (eds.): Dialects Across Borders. Selected papers from the 11th International Conference on Methods in Dialectology (Methods XI), Joensuu, August 2002. 2005. xii, 291 pp. 272 GESS, Randall S. and Edward J. RUBIN (eds.): Theoretical and Experimental Approaches to Romance Linguistics. Selected papers from the 34th Linguistic Symposium on Romance Languages (LSRL), Salt Lake City, March 2004. 2005. viii, 367 pp. 271 BRANNER, David Prager (ed.): The Chinese Rime Tables. Linguistic philosophy and historicalcomparative phonology. 2006. viii, 358 pp.

270 GEERTS, Twan, Ivo van GINNEKEN and Haike JACOBS (eds.): Romance Languages and Linguistic Theory 2003. Selected papers from ‘Going Romance’ 2003, Nijmegen, 20–22 November. 2005. viii, 369 pp. 269 HARGUS, Sharon and Keren RICE (eds.): Athabaskan Prosody. 2005. xii, 432 pp. 268 CRAVENS, Thomas D. (ed.): Variation and Reconstruction. 2006. viii, 223 pp. 267 ALHAWARY, Mohammad T. and Elabbas BENMAMOUN (eds.): Perspectives on Arabic Linguistics. Papers from the annual symposium on Arabic linguistics. Volume XVII–XVIII: Alexandria, 2003 and Norman, Oklahoma 2004. 2005. xvi, 315 pp. 266 BOUDELAA, Sami (ed.): Perspectives on Arabic Linguistics. Papers from the annual symposium on Arabic linguistics. Volume XVI: , Cambridge, March 2002. 2006. xii, 181 pp. 265 CORNIPS, Leonie and Karen P. CORRIGAN (eds.): Syntax and Variation. Reconciling the Biological and the Social. 2005. vi, 312 pp. 264 DRESSLER, Wolfgang U., Dieter KASTOVSKY, Oskar E. PFEIFFER and Franz RAINER (eds.): Morphology and its demarcations. Selected papers from the 11th Morphology meeting, Vienna, February 2004. With the assistance of Francesco Gardani and Markus A. Pöchtrager. 2005. xiv, 320 pp. 263 BRANCO, António, Tony McENERY and Ruslan MITKOV (eds.): Anaphora Processing. Linguistic, cognitive and computational modelling. 2005. x, 449 pp. 262 VAJDA, Edward J. (ed.): Languages and Prehistory of Central Siberia. 2004. x, 275 pp. 261 KAY, Christian J. and Jeremy J. SMITH (eds.): Categorization in the History of English. 2004. viii, 268 pp. 260 NICOLOV, Nicolas, Kalina BONTCHEVA, Galia ANGELOVA and Ruslan MITKOV (eds.): Recent Advances in Natural Language Processing III. Selected papers from RANLP 2003. 2004. xii, 402 pp. 259 CARR, Philip, Jacques DURAND and Colin J. EWEN (eds.): Headhood, Elements, Specification and Contrastivity. Phonological papers in honour of John Anderson. 2005. xxviii, 405 pp. 258 AUGER, Julie, J. Clancy CLEMENTS and Barbara VANCE (eds.): Contemporary Approaches to Romance Linguistics. Selected Papers from the 33rd Linguistic Symposium on Romance Languages (LSRL), Bloomington, Indiana, April 2003. With the assistance of Rachel T. Anderson. 2004. viii, 404 pp. 257 FORTESCUE, Michael, Eva Skafte JENSEN, Jens Erik MOGENSEN and Lene SCHØSLER (eds.): Historical Linguistics 2003. Selected papers from the 16th International Conference on Historical Linguistics, Copenhagen, 11–15 August 2003. 2005. x, 312 pp. 256 BOK-BENNEMA, Reineke, Bart HOLLEBRANDSE, Brigitte KAMPERS-MANHE and Petra SLEEMAN (eds.): Romance Languages and Linguistic Theory 2002. Selected papers from ‘Going Romance’, Groningen, 28–30 November 2002. 2004. viii, 273 pp. 255 MEULEN, Alice ter and Werner ABRAHAM (eds.): The Composition of Meaning. From lexeme to discourse. 2004. vi, 232 pp. 254 BALDI, Philip and Pietro U. DINI (eds.): Studies in Baltic and Indo-European Linguistics. In honor of William R. Schmalstieg. 2004. xlvi, 302 pp. 253 CAFFAREL, Alice, J.R. MARTIN and Christian M.I.M. MATTHIESSEN (eds.): Language Typology. A functional perspective. 2004. xiv, 702 pp. 252 KAY, Christian J., Carole HOUGH and Irené WOTHERSPOON (eds.): New Perspectives on English Historical Linguistics. Selected papers from 12 ICEHL, Glasgow, 21–26 August 2002. Volume II: Lexis and Transmission. 2004. xii, 273 pp. 251 KAY, Christian J., Simon HOROBIN and Jeremy J. SMITH (eds.): New Perspectives on English Historical Linguistics. Selected papers from 12 ICEHL, Glasgow, 21–26 August 2002. Volume I: Syntax and Morphology. 2004. x, 264 pp. 250 JENSEN, John T.: Principles of Generative Phonology. An introduction. 2004. xii, 324 pp. 249 BOWERN, Claire and Harold KOCH (eds.): Australian Languages. Classification and the comparative method. 2004. xii, 377 pp. (incl. CD-Rom). 248 WEIGAND, Edda (ed.): Emotion in Dialogic Interaction. Advances in the complex. 2004. xii, 284 pp. 247 PARKINSON, Dilworth B. and Samira FARWANEH (eds.): Perspectives on Arabic Linguistics. Papers from the Annual Symposium on Arabic Linguistics. Volume XV: Salt Lake City 2001. 2003. x, 214 pp. 246 HOLISKY, Dee Ann and Kevin TUITE (eds.): Current Trends in Caucasian, East European and Inner Asian Linguistics. Papers in honor of Howard I. Aronson. 2003. xxviii, 426 pp. 245 QUER, Josep, Jan SCHROTEN, Mauro SCORRETTI, Petra SLEEMAN and Els VERHEUGD (eds.): Romance Languages and Linguistic Theory 2001. Selected papers from 'Going Romance', Amsterdam, 6–8 December 2001. 2003. viii, 355 pp. 244 PÉREZ-LEROUX, Ana Teresa and Yves ROBERGE (eds.): Romance Linguistics. Theory and Acquisition. Selected papers from the 32nd Linguistic Symposium on Romance Languages (LSRL), Toronto, April 2002. 2003. viii, 388 pp. 243 CUYCKENS, Hubert, Thomas BERG, René DIRVEN and Klaus-Uwe PANTHER (eds.): Motivation in Language. Studies in honor of Günter Radden. 2003. xxvi, 403 pp.

242 SEUREN, Pieter A.M. and Gerard KEMPEN (eds.): Verb Constructions in German and Dutch. 2003. vi, 316 pp. 241 LECARME, Jacqueline (ed.): Research in Afroasiatic Grammar II. Selected papers from the Fifth Conference on Afroasiatic Languages, Paris, 2000. 2003. viii, 550 pp. 240 JANSE, Mark and Sijmen TOL (eds.): Language Death and Language Maintenance. Theoretical, practical and descriptive approaches. With the assistance of Vincent Hendriks. 2003. xviii, 244 pp. 239 ANDERSEN, Henning (ed.): Language Contacts in Prehistory. Studies in Stratigraphy. Papers from the Workshop on Linguistic Stratigraphy and Prehistory at the Fifteenth International Conference on Historical Linguistics, Melbourne, 17 August 2001. 2003. viii, 292 pp. 238 NÚÑEZ-CEDEÑO, Rafael, Luis LÓPEZ and Richard CAMERON (eds.): A Romance Perspective on Language Knowledge and Use. Selected papers from the 31st Linguistic Symposium on Romance Languages (LSRL), Chicago, 19–22 April 2001. 2003. xvi, 386 pp. 237 BLAKE, Barry J. and Kate BURRIDGE (eds.): Historical Linguistics 2001. Selected papers from the 15th International Conference on Historical Linguistics, Melbourne, 13–17 August 2001. Editorial assistance Jo Taylor. 2003. x, 444 pp. 236 SIMON-VANDENBERGEN, Anne-Marie, Miriam TAVERNIERS and Louise J. RAVELLI (eds.): Grammatical Metaphor. Views from systemic functional linguistics. 2003. vi, 453 pp. 235 LINN, Andrew R. and Nicola McLELLAND (eds.): Standardization. Studies from the Germanic languages. 2002. xii, 258 pp. 234 WEIJER, Jeroen van de, Vincent J. van HEUVEN and Harry van der HULST (eds.): The Phonological Spectrum. Volume II: Suprasegmental structure. 2003. x, 264 pp. 233 WEIJER, Jeroen van de, Vincent J. van HEUVEN and Harry van der HULST (eds.): The Phonological Spectrum. Volume I: Segmental structure. 2003. x, 308 pp. 232 BEYSSADE, Claire, Reineke BOK-BENNEMA, Frank DRIJKONINGEN and Paola MONACHESI (eds.): Romance Languages and Linguistic Theory 2000. Selected papers from ‘Going Romance’ 2000, Utrecht, 30 November–2 December. 2002. viii, 354 pp. 231 CRAVENS, Thomas D.: Comparative Historical Dialectology. Italo-Romance clues to Ibero-Romance sound change. 2002. xii, 163 pp. 230 PARKINSON, Dilworth B. and Elabbas BENMAMOUN (eds.): Perspectives on Arabic Linguistics. Papers from the Annual Symposium on Arabic Linguistics. Volume XIII-XIV: Stanford, 1999 and Berkeley, California 2000. 2002. xiv, 250 pp. 229 NEVIN, Bruce E. and Stephen B. JOHNSON (eds.): The Legacy of Zellig Harris. Language and information into the 21st century. Volume 2: Mathematics and computability of language. 2002. xx, 312 pp. 228 NEVIN, Bruce E. (ed.): The Legacy of Zellig Harris. Language and information into the 21st century. Volume 1: Philosophy of science, syntax and semantics. 2002. xxxvi, 323 pp. 227 FAVA, Elisabetta (ed.): Clinical Linguistics. Theory and applications in speech pathology and therapy. 2002. xxiv, 353 pp. 226 LEVIN, Saul: Semitic and Indo-European. Volume II: Comparative morphology, syntax and phonetics. 2002. xviii, 592 pp. 225 SHAHIN, Kimary N.: Postvelar Harmony. 2003. viii, 344 pp. 224 FANEGO, Teresa, Belén MÉNDEZ-NAYA and Elena SEOANE (eds.): Sounds, Words, Texts and Change. Selected papers from 11 ICEHL, Santiago de Compostela, 7–11 September 2000. Volume 2. 2002. x, 310 pp. 223 FANEGO, Teresa, Javier PÉREZ-GUERRA and María José LÓPEZ-COUSO (eds.): English Historical Syntax and Morphology. Selected papers from 11 ICEHL, Santiago de Compostela, 7–11 September 2000. Volume 1. 2002. x, 306 pp. 222 HERSCHENSOHN, Julia, Enrique MALLÉN and Karen ZAGONA (eds.): Features and Interfaces in Romance. Essays in honor of Heles Contreras. 2001. xiv, 302 pp. 221 D’HULST, Yves, Johan ROORYCK and Jan SCHROTEN (eds.): Romance Languages and Linguistic Theory 1999. Selected papers from ‘Going Romance’ 1999, Leiden, 9–11 December 1999. 2001. viii, 406 pp. 220 SATTERFIELD, Teresa, Christina TORTORA and Diana CRESTI (eds.): Current Issues in Romance Languages. Selected papers from the 29th Linguistic Symposium on Romance Languages (LSRL), Ann Arbor, 8–11 April 1999. 2002. viii, 412  pp. 219 ANDERSEN, Henning (ed.): Actualization. Linguistic Change in Progress. Papers from a workshop held at the 14th International Conference on Historical Linguistics, Vancouver, B.C., 14 August 1999. 2001. vii, 250 pp. 218 BENDJABALLAH, Sabrina, Wolfgang U. DRESSLER, Oskar E. PFEIFFER and Maria D. VOEIKOVA (eds.): Morphology 2000. Selected papers from the 9th Morphology Meeting, Vienna, 24–28 February 2000. 2002. viii, 317 pp.

217 WILTSHIRE, Caroline R. and Joaquim CAMPS (eds.): Romance Phonology and Variation. Selected papers from the 30th Linguistic Symposium on Romance Languages, Gainesville, Florida, February 2000. 2002. xii, 238 pp. 216 CAMPS, Joaquim and Caroline R. WILTSHIRE (eds.): Romance Syntax, Semantics and L2 Acquisition. Selected papers from the 30th Linguistic Symposium on Romance Languages, Gainesville, Florida, February 2000. 2001. xii, 246 pp. 215 BRINTON, Laurel J. (ed.): Historical Linguistics 1999. Selected papers from the 14th International Conference on Historical Linguistics, Vancouver, 9–13 August 1999. 2001. xii, 398 pp. 214 WEIGAND, Edda and Marcelo DASCAL (eds.): Negotiation and Power in Dialogic Interaction. 2001. viii, 303 pp. 213 SORNICOLA, Rosanna, Erich POPPE and Ariel SHISHA-HALEVY (eds.): Stability, Variation and Change of Word-Order Patterns over Time. With the assistance of Paola Como. 2000. xxxii, 323 pp. 212 REPETTI, Lori (ed.): Phonological Theory and the Dialects of Italy. 2000. x, 301 pp. 211 ELŠÍK, Viktor and Yaron MATRAS (eds.): Grammatical Relations in Romani. The Noun Phrase. with a Foreword by Frans Plank (Universität Konstanz). 2000. x, 244 pp. 210 DWORKIN, Steven N. and Dieter WANNER (eds.): New Approaches to Old Problems. Issues in Romance historical linguistics. 2000. xiv, 235 pp. 209 KING, Ruth: The Lexical Basis of Grammatical Borrowing. A Prince Edward Island French case study. 2000. xvi, 241 pp. 208 ROBINSON, Orrin W.: Whose German? The ach/ich alternation and related phenomena in ‘standard’ and ‘colloquial’. 2001. xii, 178 pp. 207 SANZ, Montserrat: Events and Predication. A new approach to syntactic processing in English and Spanish. 2000. xiv, 219 pp. 206 FAWCETT, Robin P.: A Theory of Syntax for Systemic Functional Linguistics. 2000. xxviii, 360 pp. 205 DIRVEN, René, Roslyn M. FRANK and Cornelia ILIE (eds.): Language and Ideology. Volume 2: descriptive cognitive approaches. 2001. vi, 264 pp. 204 DIRVEN, René, Bruce HAWKINS and Esra SANDIKCIOGLU (eds.): Language and Ideology. Volume 1: theoretical cognitive approaches. 2001. vi, 301 pp. 203 NORRICK, Neal R.: Conversational Narrative. Storytelling in everyday talk. 2000. xiv, 233 pp. 202 LECARME, Jacqueline, Jean LOWENSTAMM and Ur SHLONSKY (eds.): Research in Afroasiatic Grammar. Papers from the Third conference on Afroasiatic Languages, Sophia Antipolis, 1996. 2000. vi, 386 pp. 201 DRESSLER, Wolfgang U., Oskar E. PFEIFFER, Markus A. PÖCHTRAGER and John R. RENNISON (eds.): Morphological Analysis in Comparison. 2000. x, 261 pp. 200 ANTTILA, Raimo: Greek and Indo-European Etymology in Action. Proto-Indo-European *aǵ-. 2000. xii, 314 pp. 199 PÜTZ, Martin and Marjolijn H. VERSPOOR (eds.): Explorations in Linguistic Relativity. 2000. xvi, 369 pp. 198 NIEMEIER, Susanne and René DIRVEN (eds.): Evidence for Linguistic Relativity. 2000. xxii, 240 pp. 197 COOPMANS, Peter, Martin EVERAERT and Jane GRIMSHAW (eds.): Lexical Specification and Insertion. 2000. xviii, 476 pp. 196 HANNAHS, S.J. and Mike DAVENPORT (eds.): Issues in Phonological Structure. Papers from an International Workshop. 1999. xii, 268 pp. 195 HERRING, Susan C., Pieter van REENEN and Lene SCHØSLER (eds.): Textual Parameters in Older Languages. 2001. x, 448 pp. 194 COLEMAN, Julie and Christian J. KAY (eds.): Lexicology, Semantics and Lexicography. Selected papers from the Fourth G. L. Brook Symposium, Manchester, August 1998. 2000. xiv, 257 pp. 193 KLAUSENBURGER, Jurgen: Grammaticalization. Studies in Latin and Romance morphosyntax. 2000. xiv, 184 pp. 192 ALEXANDROVA, Galina M. and Olga ARNAUDOVA (eds.): The Minimalist Parameter. Selected papers from the Open Linguistics Forum, Ottawa, 21–23 March 1997. 2001. x, 360 pp. 191 SIHLER, Andrew L.: Language History. An introduction. 2000. xvi, 298 pp. 190 BENMAMOUN, Elabbas (ed.): Perspectives on Arabic Linguistics. Papers from the Annual Symposium on Arabic Linguistics. Volume XII: Urbana-Champaign, Illinois, 1998. 1999. viii, 204 pp. 189 NICOLOV, Nicolas and Ruslan MITKOV (eds.): Recent Advances in Natural Language Processing II. Selected papers from RANLP ’97. 2000. xi, 422 pp. 188 SIMMONS, Richard VanNess: Chinese Dialect Classification. A comparative approach to Harngjou, Old Jintarn, and Common Northern Wu. 1999. xviii, 317 pp.

187 FRANCO, Jon A., Alazne LANDA and Juan MARTÍN (eds.): Grammatical Analyses in Basque and Romance Linguistics. Papers in honor of Mario Saltarelli. 1999. viii, 306 pp. 186 MIŠESKA TOMIĆ, Olga and Milorad RADOVANOVIĆ (eds.): History and Perspectives of Language Study. Papers in honor of Ranko Bugarski. 2000. xxii, 314 pp. 185 AUTHIER, Jean-Marc, Barbara E. BULLOCK and Lisa A. REED (eds.): Formal Perspectives on Romance Linguistics. Selected papers from the 28th Linguistic Symposium on Romance Languages (LSRL XXVIII), University Park, 16–19 April 1998. 1999. xii, 334 pp. 184 SAGART, Laurent: The Roots of Old Chinese. 1999. xii, 272 pp. 183 CONTINI-MORAVA, Ellen and Yishai TOBIN (eds.): Between Grammar and Lexicon. 2000. xxxii, 365 pp. 182 KENESEI, István (ed.): Crossing Boundaries. Advances in the theory of Central and Eastern European languages. 1999. viii, 302 pp. 181 MOHAMMAD, Mohammad A.: Word Order, Agreement and Pronominalization in Standard and Palestinian Arabic. 2000. xvi, 197 pp. 180 MEREU, Lunella (ed.): Boundaries of Morphology and Syntax. 1999. viii, 314 pp. 179 RINI, Joel: Exploring the Role of Morphology in the Evolution of Spanish. 1999. xvi, 187 pp. 178 FOOLEN, Ad and Frederike van der LEEK (eds.): Constructions in Cognitive Linguistics. Selected papers from the Fifth International Cognitive Linguistics Conference, Amsterdam, 1997. 2000. xvi, 338 pp. 177 CUYCKENS, Hubert and Britta E. ZAWADA (eds.): Polysemy in Cognitive Linguistics. Selected papers from the International Cognitive Linguistics Conference, Amsterdam, 1997. 2001. xxviii, 296 pp. 176 VAN HOEK, Karen, Andrej A. KIBRIK and Leo NOORDMAN (eds.): Discourse Studies in Cognitive Linguistics. Selected papers from the 5th International Cognitive Linguistics Conference, Amsterdam, July 1997. 1999. vi, 187 pp. 175 GIBBS, JR., Raymond W. and Gerard J. STEEN (eds.): Metaphor in Cognitive Linguistics. Selected papers from the 5th International Cognitive Linguistics Conference, Amsterdam, 1997. 1999. viii, 226 pp. 174 HALL, T. Alan and Ursula KLEINHENZ (eds.): Studies on the Phonological Word. 1999. viii, 298 pp. 173 TREVIÑO, Esthela and José LEMA (eds.): Semantic Issues in Romance Syntax. 1999. viii, 309 pp. 172 DIMITROVA-VULCHANOVA, Mila and Lars HELLAN (eds.): Topics in South Slavic Syntax and Semantics. 1999. xxviii, 263 pp. 171 WEIGAND, Edda (ed.): Contrastive Lexical Semantics. 1998. x, 270 pp. 170 LAMB, Sydney M.: Pathways of the Brain. The neurocognitive basis of language. 1999. xii, 418 pp. 169 GHADESSY, Mohsen (ed.): Text and Context in Functional Linguistics. 1999. xviii, 340 pp. 168 RATCLIFFE, Robert R.: The “Broken” Plural Problem in Arabic and Comparative Semitic. Allomorphy and analogy in non-concatenative morphology. 1998. xii, 261 pp. 167 BENMAMOUN, Elabbas, Mushira EID and Niloofar HAERI (eds.): Perspectives on Arabic Linguistics. Papers from the Annual Symposium on Arabic Linguistics. Volume XI: Atlanta, Georgia, 1997. 1998. viii, 231 pp. 166 LEMMENS, Maarten: Lexical Perspectives on Transitivity and Ergativity. Causative constructions in English. 1998. xii, 268 pp. 165 BUBENIK, Vit: A Historical Syntax of Late Middle Indo-Aryan (Apabhraṃśa). 1998. xxiv, 265 pp. 164 SCHMID, Monika S., Jennifer R. AUSTIN and Dieter STEIN (eds.): Historical Linguistics 1997. Selected papers from the 13th International Conference on Historical Linguistics, Düsseldorf, 10–17 August 1997. 1998. x, 409 pp. 163 LOCKWOOD, David G., Peter H. FRIES and James E. COPELAND (eds.): Functional Approaches to Language, Culture and Cognition. Papers in honor of Sydney M. Lamb. 2000. xxxiv, 656 pp. 162 HOGG, Richard M. and Linda van BERGEN (eds.): Historical Linguistics 1995. Volume 2: Germanic linguistics.. Selected papers from the 12th International Conference on Historical Linguistics, Manchester, August 1995. 1998. x, 365 pp. 161 SMITH, John Charles and Delia BENTLEY (eds.): Historical Linguistics 1995. Volume 1: General issues and non-Germanic Languages.. Selected papers from the 12th International Conference on Historical Linguistics, Manchester, August 1995. 2000. xii, 438 pp. 160 SCHWEGLER, Armin, Bernard TRANEL and Myriam URIBE-ETXEBARRIA (eds.): Romance Linguistics: Theoretical Perspectives. Selected papers from the 27th Linguistic Symposium on Romance Languages (LSRL XXVII), Irvine, 20–22 February, 1997. 1998. vi, 349 pp. + index. 159 JOSEPH, Brian D., Geoffrey C. HORROCKS and Irene PHILIPPAKI-WARBURTON (eds.): Themes in Greek Linguistics II. 1998. x, 335 pp. 158 SÁNCHEZ-MACARRO, Antonia and Ronald CARTER (eds.): Linguistic Choice across Genres. Variation in spoken and written English. 1998. viii, 338 pp.