Reading In Different Languages

6

Reading in different languages

At the neuronal level, a person who learns to read in Chinese uses a very particular set of neuronal connections that differ in significant ways from the pathways used in English. When Chinese readers first try to read in English, their brains attempt to use Chinese-based neuronal pathways. The act of learning to read Chinese characters has literally shaped the Chinese reading brain. (Wolf, 2007: 5)

This chapter outlines some of the differences in learning to read in differing L1s. Because different languages vary in their phonological, orthographic, morphological, syntactic, and semantic systems, a given L1 and L2 could be quite different from each other linguistically. Generally speaking, the linguistic differences, or linguistic distance, between any given L1 and L2 will be a factor to consider in L2 reading development. When two languages are more distinct linguistically, there is a greater likelihood (though no guarantees) that there will be greater linguistic processing interference. An examination of reading in different L1s helps to interpret the notion of linguistic distance as a useful concept. Moreover, the exploration of reading development in different languages can help identify the extent to which there are universals of reading development and the extent to which the interaction between a given L1 and L2 will create unique issues for learning and instruction. This issue of different patterns of L1 reading development provides additional perspectives on some of the difficulties that L2 readers might face as they learn to read an L2. In exploring reading development in different languages, this chapter does not make any claim for comprehensiveness. Crosslinguistic variability in reading is already a basis for many books and major reviews (Cook & Bassetti, 2005b; Frost, 2005; Genesee et al., 2006; Harris & Hatano, 1999; Joshi & Aaron, 2006; Riches & Genesee, 2006; Schwartz & Kroll, 2006; Seymour, 2005, 2006), and several topics could be added to the discussion that follows, but they would go beyond the constraints of the present book. However, the differences that are outlined in this chapter do point out the many complexities of learning to read in a second language and many reading-instruction issues that need to be considered as well, including the transfer effects that are observable when students 109 Downloaded from https://www.cambridge.org/core. The University of Notre Dame Australia, on 03 Apr 2019 at 06:27:31, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781139150484.010

110

Patterns of variation in reading

read in their L2s. Different phonology, orthography, morphology, syntax, lexical semantics, and discourse features create countless patterns of similarities and differences across any two languages. How these patterns interact with growing L2 reading proficiency is the subject of the latter part of this chapter.

Reading in different languages In learning to read, children, in all languages, must first recognize which language elements are directly encoded in the writing system. . . . Although the linking requirement is universally imposed, how it is accomplished varies across languages. (Koda, 2007: 12)

An early model of reading development, the Psycholinguistic Guessing Game Model, made the claim that reading is a universal process. Research over the past 20 years has demonstrated that this claim is partly true but also partly false. (The larger claims concerning the Psycholinguistic Guessing Game are more certainly false; see Chapter 5.) There are aspects of reading that are very likely universal. Most of these aspects have to do with the nature of cognitive-processing mechanisms more generally and universal concepts of language knowledge. All readers make use of visual word-recognition processes while reading and engage in phonological processing in reading at the earliest possible moment that the orthography allows. All readers use syntactic information to determine text meaning and text comprehension. All readers set goals, engage in reading strategies, and apply some level of metacognitive awareness to text comprehension. On the purely cognitive side, all readers use a capacity-limited working-memory system, draw on long-term memory (background knowledge) to interpret text meaning, engage in very rapid pattern recognition, and automatize well-practiced processing skills. In terms of language knowledge, there is a long and interesting literature on language universals (Comrie, 1989; Comrie, Matthews, & Polinsky, 2003; Croft, 1990; Hawkins, 1988; Whaley, 1997). For example, all languages have nouns and verbs. All languages have vowels. All languages have stop phonemes. All languages have first- and secondperson pronouns. All languages have relative clauses or relativizationlike structures. Furthermore, all languages use recursion (e.g., embed a noun phrase within a noun phrase, embed a clause within a clause) to build more complex syntactic clauses. In all of these respects, one can say that reading requires fluent readers to draw on universal human processing skills and language knowledge. Of course, the above skills and knowledge resources do not comprise all that there is to reading comprehension. In many cases, these skills and

Downloaded from https://www.cambridge.org/core. The University of Notre Dame Australia, on 03 Apr 2019 at 06:27:31, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781139150484.010

Reading in different languages

111

resources are similar at the level of general cognitive architecture and language learning, but they are not similar in the specific details of how they work in differing languages. Thus, a major issue is determining what is general across languages and what is specific to reading development in each language (Geva & Siegal, 2000; Geva & Wang, 2001; KahnHorwitz, Shimron, & Sparks, 2005; Koda, 2007). This chapter will expand on these similarities and differences. The outcome of this exploration demonstrates that reading in different languages involves a number of specific differences that cannot be ignored. These differences influence the path of L1 reading development and patterns of L1–L2 transfer. Foremost among the differences that influence how fluent readers process a text are the diverse orthographies that visually recode the phonological and morphological systems of each language (see Comrie, Matthews, & Polinsky, 2003; Crystal, 1997, for interesting and accessible examples). It is also obvious that the vocabulary and grammar of each language can be quite distinct from one language to another and that there is a wide range of variability across languages. Even in the area of strategic processing, certain languages may privilege one set of strategies over others, depending not only on the orthography and morphology of each language but also the social and cultural uses of texts (e.g., Chinese readers place more emphasis on visual processing, Hebrew readers place more emphasis on processing consonants, Czech readers place more emphasis on initial consonant-cluster recognition). Many books describe linguistic differences across first languages and a number examine their impact on reading-skills development. (For additional information on crosslinguistic differences, many linguistics textbooks have chapters on linguistic typology; see Comrie, 1989; Comrie, Matthews, & Polinsky, 2003; Croft, 1990; Finegan, 2008). This chapter focuses primarily on word-recognition processes and the impact of differing L1 orthographies, phonologies, and morphologies on wordrecognition processes. While languages vary considerably in their lexicons and syntactic structures, little research demonstrates how these differences would impact readers from different L1s (and L2s) (cf. Cook & Bassetti, 2005a). Differences across L1s in higher-order reading processes, to the extent that they can be generalized, are less likely to be due to language differences in the orthography, phonology, morphology, and syntax of the languages involved (see Chapter 7).

Differences in orthography, phonology, and morphology Anyone who has learned both a nonalphabetic language and an alphabetic language (one as L1, the other as L2) is well aware of the large

Downloaded from https://www.cambridge.org/core. The University of Notre Dame Australia, on 03 Apr 2019 at 06:27:31, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781139150484.010

112

Patterns of variation in reading

differences that the orthography of the language can impose on visual comprehension processing. These differences include the extent to which phonology is directly reflected in the orthographic segments and the extent to which morphology is coded in the orthography. They also include whether or not the orthography is primarily phonological or syllabic, and the extent to which an alphabetic system is consonantal or identifies all vowels orthographically (Frost, 2005). Roughly, one can classify the orthographies of the world’s languages as phonological (e.g., English, Hebrew, Hindi, Turkish), syllabic (e.g., Cambodian, Cherokee, Javanese, Japanese Kana, Kannada, Sanskrit, Thai, Tibetan), or morphosyllabic (Chinese), though few languages are purely one of these three types. In a phonological language, each phoneme (or most of the phonemes) of the language has some graphic counterparts in the writing system. In a syllabic system, distinct graphic symbols reflect syllables in the languages (e.g., there could be separate symbols for ba, be, bi, bo, bu in a syllabic language). Only Chinese is classified as a morphosyllabic language. This notation reflects the fact that Chinese characters indicate both a meaning component and a phonological, or syllabic, component (80–90% of Chinese characters have both a phonetic and a semantic component; Cheung, McBride-Chang, & Chow, 2006; Frost, 2005; Hanley, Tzeng, & Huang, 1999). Moreover, most Chinese “words” are multicharacter, usually involving two characters (80% according to Hanley, Tzeng, & Huang, 1999: 174; see also Frost, 2005). The alphabetic designation itself includes several types of distinctions. Alphabetic languages may include all vowel sounds orthographically (e.g., English, Spanish) or may be primarily consonantal (e.g., Arabic, Hebrew). Alphabetic languages may be more or less reflective of morphological processes in the language. English strongly preserves morphemes in its orthography. For example, the plural suffix s in cats and dogs looks like the same morpheme, though it represents two different phonological sounds. We do not spell dogs as dogz, so the English orthographic system preserves the same morpheme rather than spelling the words exactly as they sound. Alphabetic languages are more or less transparent and more or less consistent (two aspects of the Orthographic Depth Hypothesis, see Frost, 2005). The general conclusion with respect to orthographies of differing languages is that they are quite likely to impact the reading process in multiple ways (for both L1 reading and L2 reading). The phonology and morphology of different L1s can influence reading processes. First, languages vary considerably in their phonological systems, some being quite limited, others being very extensive. Crystal (1997) notes that the fewest phonemes in a language are 11 (Mura, a Chibchan language). The largest number of phoneme sounds in a language is 141 (!Xu, a Khoisan language). English has 39 phonemes. Downloaded from https://www.cambridge.org/core. The University of Notre Dame Australia, on 03 Apr 2019 at 06:27:31, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781139150484.010

Reading in different languages

113

Vowel systems range from three vowels (Greenlandic, an Eskimo-Aleut language; Quechua, a South American Indian language) to 24 vowels (again !Xu). Indo-European languages tend to have large vowel systems in comparison to most languages of the world (e.g., English, French, and German all have 15 vowel sounds). Most languages have 5–7 vowels (e.g., Bengali, Chinese, Hawaiian, Hebrew, Japanese, Serbo-Croatian, Spanish, Persian, Tagalog, Zulu). Beyond the level of the phoneme itself, languages vary considerably in the allowable structure of syllables. A number of languages use only consonant – vowel (CV) or vowel (V) structures as allowable syllables (e.g., Hawaiian, Japanese). Many others also include CVC as an option (e.g., Chinese, Hebrew, Turkish). Some languages can be quite extreme in what they allow. English, for example, allows CCCVCCC (e.g., strengths). This extreme syllable structure combination strikes speakers of many languages as very unusual. Languages vary considerably in their use of morphology – the affixes and stem-form changes that indicate grammatical and semantic changes to the base form of the word. Languages like Chinese have very few grammatical or semantic morphemes (and are therefore called isolating languages). Chinese words basically come in only one form. English is also fairly simple in terms of its morphology (though Old English was very complex morphologically). There are many derivational prefixes and suffixes, but inflectional morphology is very simple, making English noun case and verb conjugation forms fairly easy to learn and use. In contrast, most Indo-European languages have much more complex inflectional morphologies and case markings on nouns. Turkish is a language typically labeled as an agglutinative language. Words in Turkish often have strings of multiple affixes after the stem of the word, often with each affix adding some additional information. Finnish is an even more complex agglutinative (and fusional) language, adding many affixes to the stem with many types of grammatical and semantic information. In both Finnish and Turkish, words can become quite long through affixation processes. Fusional languages are identified by complex inflectional morphology that combines multiple semantic and syntactic aspects of information into a single form. For example, Spanish verb endings combine information on person, tense, and number. Two other types of complex morphological systems lead to languages being labeled as polysynthetic languages (e.g., Eskimo, Greenlandic) and complex consonantal root languages (e.g., Hebrew). Greenlandic morphology can create full sentences by complex affixation processes on a single root word form. Hebrew builds many word variations on three-consonant root forms, such as the root G-D-L (with the meaning of “grow”). It also adds a range of external affixes and has an extensive inflectional morphology system for nouns, adjectives, and verbs. Downloaded from https://www.cambridge.org/core. The University of Notre Dame Australia, on 03 Apr 2019 at 06:27:31, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781139150484.010

114

Patterns of variation in reading

These variations across languages are reflected by numerous options in the orthography of these languages. Hebrew, for example, in line with its consonantal root emphasis, does not write out all vowels in its “adult” print version. (Diacritics indicate vowel sounds on Hebrew print for children who are learning to read Hebrew.) Chinese only reflects phonology indirectly and fairly loosely, incorporating inexact phonetic components in a large subset of Chinese characters (approximately 40% consistency between phonetic component pronunciation and character pronunciation; Koda, 2005). English has a particularly complex set of relationships between the sounds and the letter combinations for vowels. Japanese combines three script systems to reflect syllables (Hiragana, Katakana) and morphemes (Kanji). These examples and many others illustrate the impact of a language’s phonology and morphology on its orthography, the variations in orthographic systems across languages, and the possible transfer or interference effects that can arise. Many of these distinctions, and their impact on reading, are captured by the Orthographic Depth Hypothesis.

The Orthographic Depth Hypothesis There is a significant body of experimental evidence which suggests that different cognitive processes are involved in skilled reading of deep and shallow orthographies. (Frost, 2005: 286)

The Orthographic Depth Hypothesis states that orthographies (particularly alphabetic orthographies) have varying degrees of transparency between the phonological segments of the language and the orthographic symbols intended to represent the phonological segments. An orthography that closely represents the phonology of the language in clear one-to-one relationships is a shallow (or transparent) orthography (e.g., Finnish, Serbo-Croatian). An orthography that does not closely represent the phonology in a clear one-to-one relationship is a deep (or opaque) orthography (e.g., English). The orthographic depth of a language varies along a continuum rather than being an either / or determination. Most importantly for the present volume, the orthographic depth of a language has important implications for word-recognition processes in reading. The shallowness or depth of letter-sound correspondences can be examined in multiple ways. Two factors are considered basic for orthographic depth discussions: regularity and consistency. The two concepts are very similar and often overlap, but they are separable. (See Balota, Yap, & Cortese, 2006, for discussion of these two factors.) Regularity refers to the extent to which words in a language follow expected correspondences between words and word-level pronunciation (usually

Downloaded from https://www.cambridge.org/core. The University of Notre Dame Australia, on 03 Apr 2019 at 06:27:31, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781139150484.010

Reading in different languages

115

caused by specific types of letter-sound correspondences). For example, in English, the first sound in chili is pronounced /ˇc/, and this is the expected letter-sound correspondence. There are exceptions to this expectation in English: chord, chef, and yacht. There are many exceptions of this type in English, so English is a fairly irregular language. Consistency is specifically designated to refer to the rhyme segment of words (the word-final coda). For example, in English, the orthography of the word trip is completely regular. The /i/, the /p/, and the syllable rhyme (coda) ip can only be pronounced in one way. The rhyme relationship is fully regular (trip, slip, skip, ship, lip, rip, nip, tip). If you have a certain spelling, it can only be sounded out in one way. However, English has many inconsistent spelling patterns for the rhyming part of words. For example, in English, the letter sequence ‘ea’ can be pronounced in either of two common ways: first as /i/: heal, deal, peal, dream, beat; or as /ε/: head, bread, sweat, read, dealt, lead (also, uncommon options, perhaps irregular, would involve /∧/ learner, /a/ heart, or /e/ steak, break). It should be clear from this example that a learner seeing ‘ea’ will not know how to pronounce it outside of the specific word. The complexity of a language could include it being more or less regular and more or less consistent. English appears to be both less regular and less consistent (making English a deep-orthography language). The depth of an orthography can also be associated with the completeness of the orthography in representing sounds. In Hebrew, for example, certain vowel sounds are not represented in the normal adult orthography. In this way, Hebrew has a deep orthography because it is potentially ambiguous in how to sound out words. It is not a matter of irregular or inconsistent mapping between letter and sounds. It is simply the case that some sounds in normal Hebrew have no orthographic representation. In fact, the orthography of Hebrew, for the sounds that it does represent, is very shallow, or transparent (that is, highly regular and very consistent) across all letter-to-sound correspondences. In Hebrew, if you see a letter, you know how to pronounce it. It is also possible to consider orthographic depth from a spelling perspective (from sound to letter rather than from letter to sound). If a word in an L1 is heard, is there a regular and consistent correspondence so that one would know how to spell the word just from hearing it pronounced? In fact, one orthographic form may have one pronunciation, but one pronunciation may be represented by more than one orthographic form (or form sequence). In English, for example, if you saw ‘ff,’ ‘f,’ or ‘ph,’ you know that the sound is /f/, but if you hear /f/, you cannot be sure how to spell it. More generally, in French the letter-to-sound mappings are fairly consistent and regular (if you see a word, you tend to know how to pronounce it). However, the mapping

Downloaded from https://www.cambridge.org/core. The University of Notre Dame Australia, on 03 Apr 2019 at 06:27:31, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781139150484.010

116

Patterns of variation in reading

from sound-to-letter forms is very irregular, thus making spelling very difficult when just hearing a word. In English, most commonly the correspondences between sounds and graphic forms are opaque in both directions. The graphic form ‘ough’ can be pronounced in a number of different ways: though, through, thought, bough, enough (decoding, letter-to-sound direction). The sound /i/ can be represented with different graphic forms in the following words: seal, peel, ceiling, icy, recent (spelling, sound-to-letter direction) (see Finegan, 2008). In English, very often a person has to know the specific word to know how to spell it. Language can also be classified for orthographic depth in terms of simple or complex correspondences. In some languages, the correspondences between individual letters and individual sounds are very regular. However, in specific contexts in relation to other letters and other sounds, the sound of a given graphic form changes. In English, the letter ‘c’ varies in a patterned way in words such as cell and call (because a following “low” vowel changes /s/ to /k/). More generally, in English, the addition of derivational morphemes changes stress patterns and changes vowel (and some consonant) sounds. These morphological changes lead to letter-to-sound changes within words (electric to electricity, nation to national) but do not significantly disturb the spelling of English words. This preference in English to preserve morphology over phonology is a major reason why it is a deep orthography (but not the only reason). A further example helps explain why English is a deep orthography that preserves morphological information, but loses letter-to-sound regularity. Using the contrast between nation and national, the first vowel in national (/æ/) has become different from nation (/e/), but we keep the same spelling in English (‘a’). The same two vowel sounds, in other spoken English words, create very different words (e.g., bait vs. bat). But in national, despite the different first vowel sound, we know that we are referring to the same base word. In this way, English preserves the morphological similarity (and follows complex stress patterns), and sacrifices the sound similarity. If the spelling of ‘a’ in national were changed to reflect a different pronunciation, we would save the letter-sound relationship, but we would lose the information that the two words come from the same source (losing morphological information). In other languages that are much more regular and consistent, in Serbo-Croatian for example, the addition of morphemes to base words almost never changes the letter-to-sound correspondences. Considering all of the above factors, the orthographic depth of any language is not a single nor a simple factor. Languages all vary in their extent of regularity, consistency, and completeness of representation. Languages can also vary in the extent to which the additions of morphemes impact the letter-to-sound correspondences. Moreover, all

Downloaded from https://www.cambridge.org/core. The University of Notre Dame Australia, on 03 Apr 2019 at 06:27:31, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781139150484.010

Reading in different languages

117

languages vary in direction (decoding or spelling) in terms of regularity, consistency, and completeness of representation: For purposes of reading, the discussion that follows focuses on letter-tosound correspondences (reading and decoding) rather than focusing on sound-to-letter correspondences (spelling). Many alphabetic languages have been categorized in terms of orthographic depth. Finnish, Hungarian, and Serbo-Croatian are often cited as very shallow languages. Only a bit less shallow are Greek, Icelandic, Italian, Norwegian, and Spanish. Most cases of irregularities with these languages are in the direction of sounds to letters (spelling). The letter-tosound correspondences are very regular and consistent (that is, if you see the letter, you know how to pronounce it). More moderate in depth are Portuguese, Dutch, and Swedish. Fairly deep languages include French and Danish. English is the deepest orthography in terms of regularity and consistency (Seymour, 2005, 2006). Hebrew (and also Arabic) is also a deep orthography, but this categorization is because of incomplete graphic representation of sounds. Of course, languages that are nonalphabetic in nature will be yet deeper (more opaque) with respect to being able to pronounce words from the graphic forms (Chinese, Japanese).

Consequences of the Orthographic Depth Hypothesis for reading One major consequence of the Orthographic Depth Hypothesis is that a given language reader will be more or less able to rely easily on letterto-sound correspondences as the major developmental path to L1 word recognition and early L1 reading. For example, languages that have a shallow orthography (e.g., Finnish, German, Greek, Hungarian, Italian, Spanish, Turkish) allow schools to teach students letter-to-sound correspondences within months, as a relatively effortless process (Cossu, ˘ 1999; Durgunoglu, 2006; Seymour, 2006; Wimmer, Landerl, & Frith, 1999). These students benefit greatly from basic instruction in letter-tosound correspondences and are able to read and recognize words quickly and easily (Goswami, 2006; Harris & Giannouli, 1999; Landerl & Thaler, 2006; Lyytinen et al., 2006; Porpodas, 2006; Wimmer, Landerl, & Frith, 1999). A related consequence of the Orthographic Depth Hypothesis is that phonological awareness measures that are commonly used to assess beginning English readers turn out to be poor indicators of differences among learners in transparent languages. In shallow languages, children in early stages of reading development almost all perform very well on

Downloaded from https://www.cambridge.org/core. The University of Notre Dame Australia, on 03 Apr 2019 at 06:27:31, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781139150484.010

118

Patterns of variation in reading

phonological awareness tasks because even weak readers gain a quick command of letter-to-sound correspondences. As a result, these measures also do not predict later reading development with young readers of shallow languages (Seymour, 2006). The situation is quite different for deeper orthographies, where phonological awareness is a major indicator of individual differences and a predictor of later reading abilities. Students learning to read in a deeper orthography (e.g., English) experience reading difficulties because of complex letter-to-sound relationships. They need to master letter-tosound correspondences as quickly as they can, but it is a much more difficult task. In these situations, it is useful for students to build a store of simple sight words that they can recognize quickly and automatically because of the longer time it takes to develop recognition skills with letter-to-sound correspondences. Thus, L1 children reading in deeper orthographies (e.g., Danish, English, French, Portuguese) need to build two skills at the same time: a fast lexical access process for quick recognition of very common words (sight-word reading) and a more gradual process of mastering letter-to-sound correspondences that might take one to two years to control completely. In brief, greater lexical access processing (sight-word reading) is required in languages with deeper orthographies (Frost, 2005; Seymour, 2006). This second system of sight-word reading (recognizing whole words and word parts that are very common) also makes greater use of larger letter-sound units, typically the rhyme units in syllables and affix forms to assist word recognition (Frost, 2005; Gaskins, 2005; Gaskins et al., 1995; Goswami, 1999; Koda, 2007; Pressley, 2006). The need for recognition of full word forms at an early stage in reading development also holds for nonalphabetic languages (e.g., Chinese and Japanese) in which children learn two much more independent systems (i.e., phonological and logographic, or morphemic systems) (Seymour, 2006). It is worth noting that English is the most opaque alphabetic orthography, so learning to read in English is very different from most other languages in the world. English reading instruction, as a consequence of its deep orthography, places a fairly unique emphasis on early sightword reading for very-high-frequency words. These early sight words give learners examples for learning more complex letter-to-sound correspondences through analogies with the known sight words. L1 reading instruction in most languages, however, does not need to emphasize early sight-word reading to the extent that it is done in English as a way to aid learning more complex letter-to-sound correspondences. Nonetheless, phonological awareness is still very important for languages with shallow orthographies. All learners need to develop phonological awareness abilities for word recognition. But for shallow orthographies, it is easy to learn, and it does not constrain weaker

Downloaded from https://www.cambridge.org/core. The University of Notre Dame Australia, on 03 Apr 2019 at 06:27:31, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781139150484.010

Reading in different languages

119

students. Thus, it is not a good way to predict future reading abilities. Once learned, almost every learner can carry out phonological recoding well in shallow languages (Goswami, 2006). However, phonological awareness does need to be taught to young learners to help them see the connections between orthographic units and sounds. In languages with shallow orthographies, this instruction may only take six months and not prove to be difficult for learners. Just as all children benefit from phonological awareness training (Bialystok, 2001), all learners are likely to benefit from some level of training in lexical access processes, or learning sight words. Even very shallow orthographies have some complications that can be accounted for usefully by learning common sight words (Frost, 2005; Seymour, 2006) or by training on larger orthographical units above the phonological segment (e.g., the rhyme segment of words) (Goswami, 2006). An additional consequence of the Orthographic Depth Hypothesis is that learners from deeper orthographies show a much stronger wordfrequency effect in experiments. In deeper orthographies, words that are less frequent take longer to recognize than words that are frequent. Languages that are highly regular and transparent allow learners to sound out longer, morphologically complex words quickly and quite easily. Even though these more complex words are less frequent and not familiar to learners, they can pronounce the words easily and working-memory rehearsal will be more efficient. Learners from deeper orthographies, because they do not have many encounters with less frequent words, will not know how to pronounce them as accurately and their word recognition will be slower (Koda, 2005).

Other factors in word-recognition development Orthographies in several languages have additional consequences for word reading that may be related to orthographic depth but also to other features of the orthography, such as densely organized orthographic print, the typical length of words in the language, and, with nonalphabetic languages, an emphasis on visual processing skills. One such consequence of orthographic differences is word-reading time in different L1s. Learners from certain orthographies read consistently more slowly than readers in other orthographies. One strong case of this difference involves L1 readers of Chinese. They read words more slowly than readers of English. Readers pause longer at each word, their perceptual span is shorter (in Chinese, one character to the left and three to the right vs. English, five letters to the left and 12 letters to the right), and their eye movements to the next orthographic unit are shorter (Rayner, Juhasz, & Pollatsek, 2005). A likely cause for these differences, and for

Downloaded from https://www.cambridge.org/core. The University of Notre Dame Australia, on 03 Apr 2019 at 06:27:31, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781139150484.010

120

Patterns of variation in reading

slower word reading, is that these languages have greater lexical density; that is, they have more information per word (or orthographic unit) than does English. Other languages with greater informational density per word also have slower reading rates per word than English. Hebrew readers, for example, are slower readers in terms of words per minute than English readers (Share & Levin, 1999; Shimron & Sivan, 1994). Slower Hebrew reading is likely due to a combination of an extremely dense set of morphological information and a dense blocklike print shape that requires more time to decode (Rayner, Juhasz, & Pollatsek, 2005). German readers also read at a slower reading rate than English, even though it is an alphabetic language. Finnish and Turkish readers (and perhaps Norwegian and German readers) have more complex morphology and wordformation processes that create longer words. Unpackaging these words requires slower reading through phonological recoding, an important aspect of reading that leads to slower word recognition while reading ˘ texts (Durgunoglu, 2006; Lyytinen et al., 2006). These longer words also require readers “to attend to the word endings very carefully to ˘ understand the intended meaning” (Durgunoglu, 2006: 220). Word-reading time is also tied to orthographic depth because wordreading time differences are strong predictors of reading difficulties in languages with shallow orthographies. In English, phonological awareness (and especially pseudoword reading-time measures) is a powerful indicator of learners with early reading difficulties. German L1 children who experience reading difficulties often do so because of slow wordrecognition speeds and less so because of phonological processing difficulties (Landerl & Thaler, 2006; Lyytinen et al., 2006; Wimmer, Landerl, & Frith, 1999). Similar patterns of reading difficulties arise in a number of languages with transparent orthographies where slower speed of word reading is the best indicator of reading difficulties (Landerl & Thaler, 2006; cf. Caravolas, 2005). These findings do not mean that phonological processing is not a useful early indicator of reading difficulties across languages, but that it may not be the most important indicator of difficulties and may be limited to very early reading-proficiency differences in some orthographies. Other consequences of orthographic differences across languages include the roles of visual processing skills, morphological awareness, and syllable-reading effects. Languages that are nonalphabetic or that have dense and complex orthographic forms will lead readers to develop greater emphasis on visual-form processing as part of reading development. Most obviously, L1 readers in nonalphabetic languages (Chinese and Japanese) develop specific skills for processing visual units that are not strongly tied to phonological processing. Character recognition depends on both linguistic and nonlinguistic visual memory. In Hebrew,

Downloaded from https://www.cambridge.org/core. The University of Notre Dame Australia, on 03 Apr 2019 at 06:27:31, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781139150484.010

Reading in different languages

121

with its dense orthographic form, visual-spatial processing is also an important predictor ability. This may be due both to the dense blocklike print and the need to attend to morphemes carefully to unpack the multimorphemic strings into constituents (Share & Levin, 1999). Young Hebrew readers are also more sensitive to consonants than to vowels, reflecting the basic orthographic patterning of Hebrew. Morphemic awareness – a learner’s understanding of how words are formed and changed by word-internal subparts and affixation – also is impacted by different L1 orthographies. Morphological awareness appears to be a better predictor of reading abilities than phonological awareness among Chinese children (Koda, 2005; Leong et al., 2005; McBride-Chang et al., 2005; Wang, Koda, & Perfetti, 2003). Moreover, the processing of syllable structures may have some role to play in L1 reading in some languages. In Korean, syllable processing is noted along with naming speed and phonological processing as important predictors of word reading. Korean orthography in syllable block form may be a cause for the relative importance of morphological processing in Korean (Cho & McBride-Chang, 2005a, 2005b; McBride-Chang et al., 2005; Wang, Park, & Lee, 2006). Whether syllable processing is an important skill for reading in most languages is still an open question. (Rhyming abilities in English, German, Swedish, and Greek do not seem to be strong predictors of later reading ability. See Goswami, 1999; Landerl & Thaler, 2006; Lundberg, 1999; Seymour, 2006; cf. Goswami, 2006.)

From the L1 to the L2: Transfer effects of different L1s on L2 reading The patterns of variation across languages with respect to phonology, morphology, and orthography suggest that a learner from a given L1 may encounter some difficulties when processing words in an L2. To review these complexities, a learner may come from a language with few vowels and relatively few consonants. That learner needs to recognize the many more sounds in the L2 that make a meaning difference (the phonemes), as well as their recoding as letters in the orthography of the L2. The learner needs to learn new processes of affixation and word-stem changes that are productive in the L2 and that differ greatly from her or his L1. The learner needs to recognize what the acceptable options are for syllable structures in the L2. The learner needs to learn any new complications in letter-to-sound patterning, particularly if the L2 is less transparent in its orthography. These patterns of L1–L2 variation created by differences across L1s are likely to have a significant impact on the speed and accuracy of word-recognition processes in L2 reading development, particularly at lower proficiency levels.

Downloaded from https://www.cambridge.org/core. The University of Notre Dame Australia, on 03 Apr 2019 at 06:27:31, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781139150484.010

122

Patterns of variation in reading

The patterns of differences in L1 reading development reviewed in this chapter have their most obvious link to L2 reading development in terms of L1–L2 transfer effects at the word-reading level (see Chapter 7 for research on transfer more generally). Based on the above descriptions, it is evident that L2 reading development will be affected to some degree by the L1 background of the L2 readers. Before providing specific evidence and examples of these patterns of word-level L1 transfer (either as facilitation or interference), it is important to be clear first about the concept of transfer and the issue of universal aspects of reading across all languages. Koda (2007) provides a good overview of L1 transfer effect for L2 word-recognition abilities. She identifies transfer as “the ability to learn new skills by drawing on previously acquired resources (Genesee, Geva, Dressler & Kamil, 2006)” (2007: 17). For Koda, transfer is not simply reliance on L1 knowledge when there is a language or processing limitation with the L2. Instead, this view of transfer implies three important criteria: (1) For L1 effects to be active in L2 reading, the L1 resources must be well-learned and automatic (through extensive practice); otherwise these effects will not intrude without intentional effort on L2 reading; (2) transfer does not end at any specific point of improved L2 reading processes; rather, L1 resources will always be activated to some extent when these resources are closely associated with L2 reading processes; and (3) the L1 transferred abilities will continue to develop in relation to L2 reading development; useful L1 support resources will integrate with growing L2 reading abilities (Bialystok, 2001; Cook & Bassetti, 2005a; Koda, 2005). These three implications suggest that L2 reading is inevitably an integration of dual language abilities and dual language processing – thus L1 transfer effects of several types will play some role throughout L2 reading development (see also Akamatsu, 2005). What role the L1 plays depends both on universal aspects of reading abilities and language-specific aspects influencing L1–L2 transfer effects.

Universals of reading development At the beginning of this chapter, a number of very basic cognitive and linguistic universals were noted that should influence learning to read in any language (Kahn-Horwitz, Shimron, & Sparks, 2005; Koda, 2007; Perfetti, 2003; Perfetti & Liu, 2005). These universals, outlined in Table 6.1, emphasize general cognitive skills and language resources for all readers. It is possible to create a very large inventory of specific skills and resources that are likely to be universal aspects of reading abilities and are capable of fully transferring. However, understanding how universals are deployed differentially in a given language is what leads to observable

Downloaded from https://www.cambridge.org/core. The University of Notre Dame Australia, on 03 Apr 2019 at 06:27:31, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781139150484.010

Reading in different languages

123

Table 6.1. Universal aspects of cognitive and linguistic processing All readers: 1. carry out phonological processing while reading 2. use syntactic information to determine text meaning and text comprehension 3. set goals, engage in reading strategies 4. apply some level of metacognitive awareness to text comprehension 5. engage a capacity-limited working-memory system 6. draw on a long-term memory (background knowledge) to interpret text meaning 7. carry out very rapid pattern recognition and automatic processing skills

differences across languages and sources of variable language transfer. At the same time, aspects of L1 language knowledge do not always transfer well. Word-recognition abilities in the L1 do not always transfer to the L2, particularly when the writing system is different. L1 vocabulary knowledge generally does not transfer (except for cognates), and syntactic knowledge usually does not transfer positively (Verhoeven, 1994). For purposes of L1–L2 transfer, even with similarities of underlying cognitive skills, it is important to recognize the limitations of universal aspects of reading abilities. For example, one might say that all readers engage in reading strategies, but at a more specific level, a reader from a given L1 might engage in a specific combination of L1 reading strategies that are most suited to the L1. A Chinese strategy might focus on identifying the semantic component and the phonological component in a character, and phonological processing might be secondary to visual processing. A strategy in Hebrew might be to identify the morphological information that is critical to word recognition, thus emphasizing the ways that word parts and morphological information are extracted from the printed word (Frost et al., 2005). Even at higher-order levels, strategies for main-idea comprehension may vary considerably from language to language (and this is where strategy instruction enters the picture). Koda (2007) distinguishes language universals from languagespecific patterns by labeling universals as “general mapping principles” and the specific abilities as “the mapping details” (p. 12). The combined exploration of what is universal and what is specific to each language involved in L1–L2 reading combinations helps to determine just what can transfer, and in what ways (as facilitation or as interference). Among the universal mapping principles that support reading abilities in all languages are the set outlined in Table 6.2. All of these general mapping principles will have “mapping details” that distinguish one language from another. Thus, the principles may be universal but how languages operationalize the principles, and carry them out as specific language skills and abilities, are still likely to vary from language to language, though in ways constrained by the universal principles. In

Downloaded from https://www.cambridge.org/core. The University of Notre Dame Australia, on 03 Apr 2019 at 06:27:31, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781139150484.010

124

Patterns of variation in reading

Table 6.2. Universals of Reading Abilities 1. Consistent cognitive architecture: All readers integrate pattern-recognition skills, working memory, long-term memory, inferencing capacity, drive for coherence of interpretation, and general cognitive learning principles (see Chapters 1–4). 2. Print relations to speech: All orthographies activate spoken-language processing, though the levels at which each language optimally develops this relationship will vary. Moreover, phonological decoding is a universal of all languages. 3. Reading transfer facilitation: L2 reading will be facilitated when features being transferred are similar to the features of the L2. When any two languages differ in specific mapping details, facilitative transfer is more limited. 4. Metalinguistic awareness: Reading in all languages requires that readers become aware of multiple levels of language knowledge and their use in building reading skills. These levels of metalinguistic knowledge useful to reading include phonological awareness, orthographic awareness, morphological awareness, syntactic awareness, and discourse awareness (Koda, 2007; Nagy, 2007) (Chapter 11). 5. Similarities in text-interpretation principles: All languages build meaning through an elaboration of the text model of comprehension to a situation model that is strongly linked to reader background knowledge, attitudinal responses, and cultural socialization (Chapter 3).

the following discussion of transfer effects on the L2, comments will highlight universals 2 and 3.

Print relations to speech: The role of phonological decoding There is extensive evidence that phonological decoding is an important part of reading in all languages, including nonalphabetic languages (Koda, 2005), that phonological processing in the L1 influences word reading in the L2 (Cho & McBride-Chang, 2005b; Chow, McBrideChang, & Burgess, 2005; Gottardo et al., 2001; He, Wang, & Anderson, 2005; Wang, Park, & Lee, 2006), and after an alphabetic language has been learned, that the “alphabetic principle” does not need to be relearned when learning to read in additional languages (Koda, 2007). At the same time, for all alphabetic languages (including all shallow languages), L1 instruction that explicitly raises awareness of letter-to-sound correspondences is necessary (Lundberg, 1999). When there is no systematic instruction, phonological skills deficits arise and contribute to weak readers; this is true in all languages (Koda, 2005). Chinese, as a nonalphabetic language, is often assumed not to require phonological decoding skills but rather the processing of characters directly from print to meaning. However, recent research on Chinese

Downloaded from https://www.cambridge.org/core. The University of Notre Dame Australia, on 03 Apr 2019 at 06:27:31, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781139150484.010

Reading in different languages

125

reading shows that Chinese children do make use of phonological awareness for Chinese L1 reading. Phonological awareness is a predictor of Chinese reading abilities in research studies, and phonological components in Chinese characters are processed, influencing children’s reading abilities. Chinese children attend to the phonetic component in characters and make use of this information while reading, including recognizing unfamiliar characters (Chow, McBride-Chang, & Burgess, 2005; Gottardo et al., 2001; Hanley, Tzeng, & Huang, 1999; He, Wang, & Anderson, 2005). As Koda (2005) notes, “Phonological processing may be guided by [phonetic] radical activation to a far greater extent that has been suspected” (p. 81; see also Hanley, Tzeng, & Huang, 1999). From the perspective of reading difficulties among Chinese L1 learners, reading problems correlate highly with poor phonological awareness skills. One of the clearest consequences of the universal role of phonological awareness in reading is the transfer benefit of being an L1 reader from any alphabetic system. These readers (e.g., Spanish L1 readers of English) will be much better at internalizing the more opaque letter-to-sound correspondences of an L2 like English than will L1 readers of a nonalpha˘ ˘ betic orthography (Bialystok, 2001; Durgunoglu, 1998; Durgunoglu, Nagy, & Hancin-Bhatt, 1993; Genesee et al., 2006). The process of establishing consistent letter-to-sound correspondences (understanding the alphabetic principle) in the L1, and then experiencing large amounts of repetition in the L1 with the letter-to-sound correspondences, will make internalization of the more complex L2 system much less difficult and time-consuming. “To paraphrase Frith (1998), once an alphabetic code is acquired, phonological awareness is never the same” (Goswami, 2006: 464).

Reading transfer facilitation: Similarities between two languages Transfer in L2 reading is facilitated by the extent to which two languages share similar properties at any linguistic level (Bialystok, 2001). More specifically, word recognition is faster and more accurate with readers from related L1 orthographic backgrounds (Koda, 2005). Koda (1989), for example, showed that among groups learning to read Japanese, the group with prior experience of reading nonalphabetic scripts outperformed learner groups from alphabetic backgrounds. Muljani, Koda, and Moates (1998) showed that Indonesian L1 students were faster than Chinese L1 students in English L2 word recognition because of their prior alphabetic experiences. Moreover, the Indonesian students were significantly faster in word recognition when the English words conformed to the permissible syllable structures of Indonesian Bahasa. Ryan and

Downloaded from https://www.cambridge.org/core. The University of Notre Dame Australia, on 03 Apr 2019 at 06:27:31, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781139150484.010

126

Patterns of variation in reading

Meara (1991) showed that Arabic L1 readers (who are well practiced at reading a consonantal-root system) were slower and less accurate in identifying English words that were misspelled with short vowels missing. Koda (1990) showed that Japanese L1 students were better at reading non-words that were not pronounceable (Sanskrit forms) than were Arabic and Spanish L1 students, transferring their better visual processing abilities from Japanese. These studies, and many others, all show that transfer effects do commonly occur in various L1–L2 relations, and that L1–L2 similarities enhance facilitative transfer (see also Cook & Bassetti, 2005a).

Conclusion One could easily reach the conclusion that word-level transfer effects are always facilitative and are pervasive. However, many transfer effects involve interference because of the differences between two languages at any linguistic level. For example, Japanese students are likely to rely more on visual processing with difficult words than will students from L1 alphabetic languages, even though the visual processing will not facilitate English L2 word-recognition processing (Cook & Bassetti, 2005a; Koda, 1990). At the level of syntax-processing differences between the L1 and the L2, several studies show that L1 processing preferences will cause interference with L2 syntactic processing (Koda, 2005, 2007). Nonetheless, it is also true that positive L1–L2 lower-level transfer effects are real, as this chapter demonstrates for word-recognition processes. This chapter has also pointed out that more current views of transfer do not depict transfer as a temporary process, one that only lasts until the L2 becomes proficient enough. Rather, L1 processes will always be present at some level because L2 reading involves reading with a dual-language system (Bialystok, 2001; Cook, 1997; Cook & Bassetti, 2005a; Koda, 2007). In effect, transfer, when it occurs, is a permanent effect. As interference, transfer effects can be suppressed to some extent, particularly at very advanced L2 levels (Bialystok, 2001); as facilitation, transfer effects support L2 reading on an ongoing basis. The interaction of two languages as part of L2 reading processes has been termed “multicompetency” by Cook (1997; Cook & Bassetti, 2005a). It also suggests that L2 reading processes will impact L1 reading abilities, though that is not a focus of this chapter (see Cook, 1997). As we will see in the next chapter, the concept of multi-competence, or a dual-language L2 reading system, is viewed as a major theme in discussions of L1–L2 reading differences.

Downloaded from https://www.cambridge.org/core. The University of Notre Dame Australia, on 03 Apr 2019 at 06:27:31, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781139150484.010

Reading in different languages

127

In closing this chapter, a final consideration should be given to the long-term consequences of reading development in different L1s and the potential problems created when L1 readers learn to read in an L2 that has many differences and relatively few facilitative similarities. As Akamatsu (2006) states: Studies suggest that word-processing skills or strategies cultivated in a first language . . . [have] a major impact on the cognitive processes that are used in reading a second written language, L2. Nevertheless, this does not necessarily imply that L2 learners whose L1 is orthographically different from the target language encounter insurmountable difficulty in acquiring particular L2 reading skills or strategies. Research findings . . . simply suggest . . . some qualitative differences in [a student’s] ability to recognize written English words. (p. 494)

Implications for teaching This chapter has provided an overview of reading in different languages, outlined briefly some transfer effects for L2 reading, and introduced a number of very important concepts. Implications for instruction follow generally from a number of issues raised throughout the chapter. First, teachers should have some idea about the linguistic and literacy backgrounds of their students. (At the same time, it is not necessary for teachers to be language typologists, tracking down every possible difference between a student’s L1 and L2 that impacts reading.) Second, the issue of differences in learning to read in different L1s suggests that teachers should check beginning-level students’ abilities to read a set of high-frequency sight words in English with reasonable ease and check to make sure that they can read a basic list of English pseudowords without being seriously impeded or unable to complete the task. A simple check can take no more than five minutes per student at the outset. Practice in learning to read frequent sight words also has uses beyond a check for reading difficulties. All students need to develop some level of fluency in English with basic sight words, and learning a set of high-frequency sight words, in turn, should address many L1 word-recognition transfer issues. If some beginning-level students have difficulties with sight-word reading and pseudoword-reading tasks, students can learn to develop lettersound relationships over units larger than the letter-sound unit. Larger pronounceable units, such as the rhyme parts of word endings, provide English L2 learners with a more regular set of relationships from letter group to sounds and allow readers to make efficient analogies from spelling patterns of learned sight words to new words that students

Downloaded from https://www.cambridge.org/core. The University of Notre Dame Australia, on 03 Apr 2019 at 06:27:31, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781139150484.010

128

Patterns of variation in reading

encounter. These larger sets of letter and word-part patterns in English provide a more consistent set of relationships that can be reinforced through practice in reading frequent sight words and engaging in extensive reading. Third, the patterns of variability across different L1 orthographies, both with respect to the Orthographic Depth Hypothesis as well as other features of a language’s orthography, show that an important goal for L2 word recognition rests with accurate and fluent recognition and the development of automatic letter-to-sound correspondences. The distinct orthographic, phonological, and morphological patterns for a given L2 need to be practiced consistently and learned well. Students need wellpracticed L2 word-processing skills to carry out L2 word recognition efficiently. These skills are more important for students than the influence of L1 transfer on L2 word recognition. One of the best ways to build such automaticity is through consistent inclusion of extensive reading in the curriculum. Fourth, this chapter has noted the importance of metalinguistic awareness for L1 word-recognition skills and reading development. The same goal should be developed for L2 reading instruction (see also Chapters 2 and 11). Raising student awareness of how sounds change during affixation, how word parts have consistent form-sound relations, how morphologically derived words include sound and stress changes, and how morphologically complex words build from more basic words are all important instructional and learning goals in a reading curriculum. Students who have some awareness of the roles of phonology, orthography, and morphology in word recognition will be more efficient at lower-level language processing as well as at vocabulary learning. Finally, it is important to recognize that not all word-recognition difficulties are the result of L1 transfer effects. In most respects, wordrecognition development in L2 contexts needs to focus more on the complexities of L2 orthography, phonology, and morphology. L1 transfer effects will have an impact at the level of L2 word-recognition processing, and these effects will vary from L1 to L1, but they can be addressed by developing L2 processing efficiency. For teachers, the development of L2 processing efficiency is brought about through consistent practice, extended exposure to L2 reading, the development of sight-word reading of high-frequency words, and the growth of receptive L2 vocabulary knowledge.

Downloaded from https://www.cambridge.org/core. The University of Notre Dame Australia, on 03 Apr 2019 at 06:27:31, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781139150484.010