Syntax Iii

1. Chomsky’s View on Language and Linguistics Noam Chomsky revolutionized the field of Linguistics as from the 60s, when he argued that language is to be considered a COGNITIVE SYSTEM, i.e. a system of knowledge. Moreover, we do not need to learn our native language since somehow this system of knowledge is part of the human mind. Secondly, he seeks to explain how language is acquired (you learn consciously and you acquire unconsciously). According to Chomsky, language is a property of the human mind. Every human being is born with a biological basis for acquiring language. The human brain is said to be divided into modules or faculties, such as Math, Vision, Logic, Emotions, Senses, etc. Since birth, human beings are genetically endowed to acquire any language, since the language faculty is innate.It can be defined as an autonomous area of the mind devoted to language knowledge, independent of other aspects of knowledge. Chomsky proposed a theory of the language faculty: UG ( Universal Grammar).The basic idea is that any speaker of any language knows a set of principles that apply to all languages (that’s why they are universal), and parameters that account for language variation. As a result, acquiring a certain language means knowing how these pples apply to the target language and which value is appropriate for each parameter (either + or -). You are already acquainted with a good number of pples, for instance the EPP: all clauses must have a subject. It is claimed that pples are universal, so in all languages of the world clauses must have a subj. Whether the subject can be understood or tacit will depend on the setting of the pro-drop parameter (or null subj parameter). Thus, languages which allow tacit subj are said to be pro-drop languages (like Spanish), while those that do not are said to be non-pro-drop languages (like English). Question: What would happen if UG contained pples only? Let’s turn to language acquisition. The discussion goes back to Plato who put forward the famous Plato’s problem: if the child cannot create language knowledge from the data or evidence he/she gets, then the source must be within the mind itself. This would be the explanation to why children say things they have never heard. Chomsky also states that language is innate on the basis of the following pieces of evidence: - Language acquisition is fast: children complete it by the age of 2. That’s too little time to learn a whole list of sentences. - It is also uniform, since all children throughout the world go through the same stages. - It is an inherently creative process, given that children do not memorize sentences in order to repeat them later, but rather create new or novel sentences they have never encountered. They can do so because the system of rules is innate and can apply unlimitedly. Remember: children do not imitate. He concludes that language acq Is determined by a biologically endowed innate language faculty within the brain, which gets activated by input or evidence. The child unconsciously fixes the parameters for the right value rather soon and accurately. To sum up, language acquisition depends on two crucial factors: input and UG. Without input, UG does not get activated and without UG we would have to memorize everything isolatedly, instead of making generalizations. 2. STRUCTURE-DEPENDENCE Following Chomsky’s approach, Language has two main properties: a. It’s infinite: as the rules can apply indefinitely, the longest sentence does not exist. Furthermore, language cannot be conceived of as a list of sentences that the child learns and then reproduces. If it were a list, it would be finite. What is finite is the set of rules contained in the mind. In spite of that, this system of rules can produce an infinite number of sentences. b. It has structure: whenever movement takes place, it targets constituents, rather than isolated words.This is why a rule like the following will be discarded: RULE #1: To form a passive sentence, move the 4th word to the beginning. Even tough it may work (eg. a), it does not work accross-the –board (as shown by the ungrammaticality of b). The police arrested Tom. a. Tom was arrested by the police The police arrested my friend Tom. b. *My was arrested friend Tom by the police. In GG, we are interested in rues that will ALWAYS work ONLY. A rule like #1 would produce ungrammatical sentences (like b). We must seek for a rule that yields grammatical results. Consider: RULE #2: to from a passive sentence, move the patient to subject position. This rule gives raise to grammmatical results both for a. and b. That’s why it is better than Rule #1.How can we account for this?

The Structure Dependence Principle (SDP) says that all garmmatical operations in natural languages make reference to constituents of some sort and not to linear order. In other words, Rule #2 is constituent-based whereas Rule #1 is linear-order-based. Any rule that goes “the first, second, last, penultimate word...” will violate SD and therefore will not be a possible rule for GG. Instead, the terms “patient” and “subject oposition” refer to syntactic notions that are more reliable since they relate to constituency or hierarchy (the same as I, aux, modal, C, spec of VP, etc.) To sum up, Rule #2 is preferred since it follows SD. That’s why it is said to be a Structuredependent rule. By contrast, Rule #1 is said to be a structure-independent rule, given that it violates the SDP. Le’s compare the following rules of Question-Formation: RULE#3:to form a yes/no question, move the 2nd word to the beginning. RULE#4: To form a yes/no question, move I to C. The first one will work in a sentence like: Mary will go on holiday. => Will Mary go on holiday? However, it gives wrong results in the following case: Mary always goes on holiday.=> *Always Mary goes on holiday? Evidently, something’s wrong with Rule#3.Let’s see the results for Rule#4: Mary will go on holiday? => Will Mary go on holiday? Mary always goes on holiday? => Does Mary always go on holiday? Clearly, the problem with Rule#3 is that it looks at the linear arrangement of words, i.e. it is a Structure-Independent rule.Whereas Rule#4 takes into account the hierarchy within the sentence, i.e. its constituents. Therefore, it is a Structruture-Dependent rule since it complies with the SDP. Note: the SDP applies ONLY to grammatical operations.In other fields, such as Phonology, rules can make reference to a certain order, by using terms such as “penultimate”, “antepenultimate”, “previous”, etc. Suggested Practice 1. Formulate 2 rules for Raising constructions (John seems a good teacher). Prove one of them to be strucuture-dependent. Explain why the other is NOT. 2. Universal Principles work for all languages.Apply what you have seen to Spanish, on the basis ofthe following: María felicitó a Juan. Juan fue felicitado por María. 3. Decide whether the following HYPOTHETICAL rules are structure-dependent or not. Account. a.To form a sentence of type X, insert an X after the 3rd word in a sentence. b. To form a sentence of type W, move the head of an XP to the head of the WP. c. To form a sentence of type Z, move the 2nd word of a sentence to final position. 3. English Verbal Morphology The main issue of this chapter is how English expresses Tense.As you know, Tense is carried by I, which may contain features, modals (can, should, must, etc.) or auxiliaries (be-have -do).We will be concerned with the 1st option: features, which are also known as "affixes", esp. when they show 3rd person sing [-s] or past[-ed]. To start with, affixes originate at the head of IP. But we all know that they must ATTACH to a Verb sooner or later. How does this happen? Let's see: a. Mary [t often [reads detective stories.]] b.* Mary[ reads [often t detective stories.]] The 1st sentence shows I-lowering (also referred to as "AFFIX-HOPPING"), which is the process of moving the affix down to the head of VP (hop/h p/=saltar). While the 2nd sentence shows V-RAISING (the opposite), which consists in raising the V up to I so that the affix gets attached. Evidently, this is NOT an option since the results are bad. More precisely, if we raise the V,the adv will block adjacency between the object DP and the V. As a consequence, ACC Case cannot be assigned and the DP remains Caseless (violating the Case Filter). However, if we lower the affix, none of the above happens and the order is grammatical (since the DP CAN get ACC Case, as required by the Case Filter again). To sum up, the empirical evidence in a. and b. shows that A-Hopping is obligatory in sentences that contain a main verb. But why should the affix move after all? In order not to violate the STRANDED AFFIX FILTER*, which sates that we cannot have unattached affixes in the derivation. They cannot be on their own or in isolation ("stranded" means "suelto"). Remember that affixes do not have morphological/phonological realization, i.e the sentence "Mary [s] often read detective stories" is unpronounceable (apart from being ungrammatical).

Now the question is whether A-Hopping is always possible. The expected answer is "No" and the reason has to do with the ADJACENCY CONDITION*, which states that A-Hopping will not take place if the Affix and the V are not ADJACENT/ei d eisnt/, i.e the V must immediately follow the affix. In other words, Strict/linear adjacency is a requirement for A-Hopping to take place. Let's see how it works: c.*Mary [-s] not like Syntax. As you can see, Negation (not) blocks adjacency between the affix and the verb. As a consequence, A-hopping will not be possible and the affix will remain stranded.This sentence violates the Adjacency Condition and the SAF. d.*Mary likes not Syntax. Raising the V is not a solution either (remember that main V do not raise in English). When A-Hopping fails as in this case, Do-SUPPORT comes into play!!!! By inserting the aux "does" we can support/realize the features so that the affix is not stranded any more. e. Mary DOESN'T like Syntax. It is the presence of the aux that makes the affix pronounceable since it counts as a full-fleshed word. Mind you, "DO" appears at SS, not at DS. It is predicted that A-Hopping and Do-Support will never apply simultaneously since the latter applies only when A-Hopping fails, namely, in questions and negatives.Hence, thay can be said to be in complementary distribution. *Mary doesn't likes Synatx. A sentence like this implies that there are two affixes. Now let's see what happens in Questions. In the case of Yes/No Questions, I-to-C movement takes place OBLIGATORILY. Consider: *Mary [-s] like Synatx. *[-s] Mary like Synatx? It is clear that this sentence is ungrammatical, since it violates the SAF. Consequently, we need to do something about the affix. Can we possibly do Affix-Hopping? The answer is "No": I-to-C movement applies first (because it is a syntactic rule). After movement, the DP "Mary" blocks adjacency between the afix and the verb. That's why the only possibility is to apply Do-Support (which is a post-syntactic rule). Does Mary like Syntax? Tree Let's turn to the following examples: a.*What Mary [-s] like? b.*What [-s] Mary like? c.*What likes Mary? All of them are ungrammatical: in the first place, a. violates the SAF. Secondly, I-to-C movement hasn't applied, being obligatory in English. In b. I-to-C movement has applied but the affix is still stranded. In c. the V has raised to C, which is not allowed since Vs never leave the VP.The correct version is: d. What does Mary like? Do- Support has applied given that A-Hopping is impossible(due to the fact that there ¡s no adjacency): I precedes the subject and the affix has been attached. However, there is ONE verb that CAN raise in Negatives and Questions, namely the verb TO BE. Cf: a. Does Mary like Syntax? DS: ____Mary [-s] like Syntax. b. Is Mary a student of Syntax? DS:____Mary [-s] like Syntax. c.*Does Mary be a student of Syntax? a. is perfect: Do-Supp is obligatory in Questions that contain a main V. But in b. the verb has left the VP: 1st it moved to I to get Tense (V-raising) and then to C to form the Yes/No question (I-toC). Meanwhile, c. is not possible since Do-supp is not necessary when the V can raise.In other words, the affix gets attached after V-raising, so the SAF is obeyed.The conclusion is that TO BE does not behave as a main verb, given the fact tat it can undergo V-raising.You may be wondering whether adjacency is a condition for V-raising. Well, it isn't. Negation and Subjects are barriers for adjacency ONLY when we lower an element, but not when raising one. Empirical evidence that TO be doesn't behave as a main V also comes from negative sentences, as follows:

a. Mary doesn’t like Syntax. b. Mary isn’t a student of Syntax. c. *Mary doesn’t be a student of Syntax. Example a. requires do-support since A-Hopping is not possible.Whereas b. goes to show that TO BE has raised to I in order to become IS and after that it forms a contraction (ISN’T). However, if we apply do-support we obtain c. which is wrong due to the fact that BE doesn’ t behave as a main verb, as “like”. To conclude, we can say that V-Raising makes Do-support impossible. Another exception is the verb HAVE when used as a main V. 4. Case Theory Although it is difficult to define Case, we can say that it is a way of marking syntactic functions. Thus, subjects will get nominative from tensed inflection, object will get accusative case from transitive verbs and complements of prepositions will get oblique case from the head. Oblique case can also be assigned by empty prepositions, e.g., “Leave the premises”. `Leave ´ being unaccusative cannot assign accusative case to the DP "the premises". It looks like a DP, but in fact it is a PP: the head is the covert feature [+ location]. The feature assigns oblique case to the pseudo DP. Case may be classified as ABSTRACT or MORPHOLOGICAL. Abstract case does not have an impact on the morphology/spelling of the noun. If we have the DP "John" in isolation, we cannot tell wether it is the subject or the object unless we put it in context. John = ? Mary = ?

He loves her.

John loves Mary.

She loves him.

Mary loves John. Futhermore, if we replace DPs by pronouns, we can tell in which Case they are. Pronouns are said to reflect morphological case, since they have different forms depending on their function (NOM for subjects, ACC for complements of verb, OBL for complements of preposition). Another distinction drawn in the literature concerns structural versus inherent case. While the first one relies exclusively on configuration, the second one is tied to theta-role assignment. Structural Relations So, for NOM to be assigned there must be spec-head agreement. ACC is assigned under government, as well as OBL. Genitive case is assigned under spec-head agreement by an empty determiner. All these are structural relations. a. John wrote a book "John" gets nominative K under spec-head agreement, while "a book" gets accusative K under government. Additionally, it gets inherent Case since it also gets a theta-role from the V. However, structural K prevails and inherent K is redundant (we will need inherent K when there is no other K assigner!!!). It's only DPs that need Case: this is known as the Case Filter (*DP if it doesn't get K). Clauses do not need Case, neither do predicative DPs or QPs or PRO. The DPs subject of exceptional clauses poses a problem, though. Consider: b. They believe Mary to be a good friend. Who assigns Case to Mary? Evidently, it cannot get NOM from the lower INFL, since it is untensed. A solution is in order here, and it is called exceptional case marking: the verb assigns K to the subject of the embeded clause (even though it is not its complement). The lower IP will be in transparent for government. Could Mary get inherent K? Obviously not, since it does not get a theta-role from "believe" but from the nominal predicate "a good friend". This is why it's called exceptional: the DP gets a θ-role from one element (the N “friend”) and K from another one, the ECM verb "believe". This goes against the Uniformity Condition. There are more contexts for ECM. Let's see: c. For Mary to pass the exam was a surprise.

d. She cried herself to sleep. e. He laughed a scomful laugh. In c. the complementizer "for" will act as a preposition and will assign OBL to the DP "Mary" under ECM. In d. the complement of the verb "cry" is a resultative small clause. The subject "herself" gets ACC from the main verb under ECM. Although is not common that unergative verbs select complements, they can. They are able to do so because they selct an external argument and Burzio's generalization says that only verbs that select an external argument will assign ACC K. The verb "cry" does select an external argument, that's why it can assign ACC to the subject of the RSC exceptionally. In e. the same situation holds. The unergative verb "laugh" selects a cognate object, namely an object to which it is semanticaly related. BG predicts that "laugh" will be able to assign ACC to the cognate object exceptionally. On the contrary, if it were an unaccusative verb, ECM would not be possible: *The ghost dissapeared a quick disappearance. Due to the fact that it doesn't select an external argument, the unaccusative verb "disappear" cannot assign K to the cognate object “ a quick disappearance”, which will remain Caseless. Now let's turn to inherent K. The uniformity condition states that for there to be inherent Case marking there must be θ-marking as well. Given that A, P and deverbal nouns can be predicates, they will be able to assign INH, just like verbs. Consider: f. Mary's envy of Peter is a joke. g. Mary is envious of Peter. h. Mary envies Peter . All three examples share the same θ-grid, namely <experiencer, stimulous>. In f. the predicate is a deverbal noun and the experiencer is realized by a GenP (thus “Mary” will get GEN K). In g. the predicate is a deverbal adjective: “Mary” gets NOM case from tensed inflection and “Peter” gets INH case which is manifested through of-insertion. Finally, in h. "Peter" gets ACC and inherent from the transitive V "envy".

Other examples of Inherent Case are: 1. Collective nouns: the school of whales traveled south. “School” will assign IHN to “whales”. 2. Partitive nouns: a loaf of bread. “Loaf” will assign INH to “bread”. 3. Complex prepositions: out of sight. “Out” will assign INH to “sight”. Let's return to ECM. Are ECM nouns possible? Consider: i. *The belief of the students to pass the exam was deep. What's wrong here is that "the students" is Caseless. It cannot get NOM (since INFL is non-tensed), and it cannot get inherent from "belief" because it doesn't get a θ-role from it. The θ-role of proposition is assigned to the whole clause. Instead, it gets the role from the embedded predicate "pass". Of-insertion is irrelevant since no inherent K is possible. The clue is that ECM is an instance of structural K and nouns can only assign INH by definition. On the whole, ECM applies only to verbs: nouns can never assign structural K. In the case of Nouns,inherent and structural K are like water and oil, so try to keep them apart.

Last but not least, there a few cases where Structural and Inherent K do not overlap, viz. ECM verbs readily assign ACC to the subject of the embeded clause. Since the ECM verb does not θ-mark the subject ,it cannot assign it INH K. The same holds for the other instances of ECM that we have considered. Some other verbs only assign INH K. One possible example is found in Dative Shift constructions, which work as follows: j. Mary gave a present to John. k. Mary gave John a present. Let's say that j. is basic and k. is derived from it. The question is how. It's claimed that the preposition "to" incorporates into the verb "give", leaving the DP John Kless. Consequently, "John" must become adjacent to the verb (in k.) so as to get ACC. But what about "a present"? Given that it gets a θ-role from the V, it already has inherent K. So, in k. the V "give" assigns inherent to "a present" but not ACC, since each K can only be assigned once. The second example is that of Existencial Constructions, namely sentences that contain an unaccusative V and an expletive THERE in subject position. Consider: l. The money disappeared. m. There disappeared the money. As we know, unaccusative verbs of existence assign only one θ-role of theme. In l. "the money" is θ-marked by the V, and then moves to SPEC of IP to get NOM. Whereas in m. the expletive "there" is inserted: it gets NOM but it doesn't get a θ-role. "The money" is still the theme so it has inherent. It can never get ACC since the verb doesn't select an external argument (Burzio's G). Alas, unaccusative verbs are UN-ACC, i.e. they cannot assign ACC K by definition. Suggested Exercises 1. One big difference between English and Spanish is that the latter has CLITICS, which get or reflect Case.Another crucial difference is that Sp has DATIVE Case, which is assigned to indirect objects of ditransitive verbs or to the indirect object of unaccusative verbs like encantar, fascinar, gustar, molestar, etc.Try to determine what type of K each clitic has. a. Juan me dio un mensaje para que te lo diga a vos. b. Juan me lo dio para te lo dijera. c. A José le gusta mucho el pan dulce con nueces. d. Me molesta la actitud de ciertas personas. e. Nunca se lo devolví porque no lo vi más por el barrio. 2.Which of the following are instances of ECM? Underline the DP that gets ECM and circle the K assigner. Besides, one of the sentences is ungrammatical. Say which and why. a. John found Mary a job. b. Tom swam himself sober. c. John finds Mary an interesting girl. d. There happened a deadly accident. e. John proved his boss to be wrong. f. Peter has little expectations of the exam to be easy. g. For you to recover so fast is remarkable. h. I can't imagine the teacher drunk. 3. Explain the ungrammaticality of: a.* He arrived the station late. b.* John to arrive late is very unusual. c.* The chlidren description of the accident sounds unlikely. d.* The doctor persuaded to cut down coffee Jamie. e.* The project discussion by the Minister is rather convincing. f.* They reported of Peter absent. g.* The police entered to the abandoned house. h.* It appeared some problems lately. k.* The possibility of Jon to find the money is remote. j.* A book was given John. Question: Is adjacency a requirement for NOM K assignment? If not, think of another one. John really likes Syntax. 5. Binding Theory

That's what we call the module of Grammar that deals with the interpretation of NPs. In other words, BT establishes what can (or cannot) refer to what and why. For our purposes, we're going to classify NPs into 3 classes: a. ANAPHORS: reflexive (myself, herself, etc.) and reciprocal pronouns (each other). b. PRONOMINALS: objective pronouns (him, them, us, me, etc.) c. R(eferential)-Expressions: proper nouns(Tom, Mary, etc) and epithets (the idiot, the fool, etc.). For each class, there's a corresponding Principle, as follows: Principle A: anaphors must be BOUND within their MGC. What does BOUND mean? An element is BOUND when it has a proper BINDER which BINDS it. For there to be BINDING, 3 conditions (and all of them) must be met: 1. COINDEXATION The binder and the element in question (the one you're analyzing) must have the same index (If one is i and the other is j, no way). 2. ϕ-FEATURES It means that both elements must agree in person, number and gender. Thus, "herself" will never be bound by something like "John" or "my parents", but by something like "Susan". 3. C-COMMAND This refers to a particular configuration in the tree. We start counting from X and we go up once to find the 1st branching node. Everything that appears to the right will be c-commanded by X (downwards of course). TREE a. Does "cook" c-c "dinner"? If we go up, we will see that the 1st branching node (BN) is V'. "Dinner" is to right , so it IS c-commanded by the V. b. Does "John" c-c "dinner"? Yes, the 1st BN is IP and "dinner" is to its right-down. c. Does "cook" c-c "John"? The 1st BN is V' but "John" is higher up . So, it is NOT c-commanded by the V. Mind you: it's no good going up more that once: moving up twice or three times, etc. will be useless. One is all you can go up. d. Does INFL c-c "dinner" and "cook"? Yes, since the 1st BN is I' and both are to its right-down. a. Johni blames himselfi. "Himself" should be bound by "John" as required by Pple A. Let's check: a. Is there coindexation? YES: both have an i. b. Do they match in ϕ-features? YES: both are masc, sing. c. Is there c-command? YES: the subject c-c all the rest. Conclusion: Pple A is satisfied, since the anaphor "himself" is bound by "John". Now, let's see what happens if it John's father that blames himself. Cf: b. * Johni' s father blames himselfi. Is there still binding? Let's check: a. Coindexation is OK. b. phi-features match. c. C-commands fails. How? Tree Crucially, the 1st BN is DP: it c-commands D and NP. Nothing else. We cannot go up to IP because that would entail going up twice, which is not possible. As a conclusion, "John" cannot bind "himself" because it doesn't c-c it. As a result, "himself" is FREE (= not bound), violating Pple A. However, there is a way of rescuing the sentence, and that is to change the index. The binder of "himself" should be the whole DP "John's father", viz. c.[John's father]i blames himselfi. How is it different? Now we must start off from DP, and not from GenP. The 1st BN is IP which c-c "himself" allright. Conclusion: "himself" is bound by the DP " John's father", satisfying Pple A. Note: c-command can be compared to a fall-rise: you go up and then down all the way(^). Now we need to define "MGC". Let's see. MGC stands for Minimal Governing Category. Why "minimal"? Because it must cover just what's necessary. Why "governing"? Because we must find a governor as a previous step.

Government is like C-command only that it applies to heads. We said that the DP "John's father" Cc "himself". However, it cannot govern it since it's a phrase, i.e. a maximal projection. The formulation goes: A governs B iff ( A rige a B si): - A is a head - A c-commands B - No barriers intervene (CPs) More generally, the governor is the element that assigns Case. In our example, "cook" governs " dinner". Therefore, in order to determine the MGC we need: 1. the element in question (anaphor or pronoun) 2. its governor 3. a subject (anything that appears in spec position). Let's see how it works.

c.[John's fatheri blames himselfi].MGC As easy as it may seem, the MGC is the whole sentence. The anaphor is duly bound within it, as stated by Pple A. But some other examples are not as straightforward, viz. d. [John's fatheri considers himselfi to be guilty].MGC The exceptional clause cannot serve as the MGC since it does not contain a governor for the element in question "himself" (neither does it contain a subject). What we must do is expand the MGC so that it covers the whole sentence. Now that we have determined the MGC, we apply the Binding Conditions, as follows: Coind YES ϕ-features YES C-Command YES Conclusion: "himself" is bound within its MGC, satisfying Pple A. Let's consider the next example: e.* John's fatheri considers that [himselfi is guilty].MGC Why is it wrong? Tensed Infl will count as the GOV and as the ACCESIBLE subject. So the MGC will be the nominal clause. And that's that: the anaphor is FREE since it binder is outside the MGC. This leads to ungrammaticality, given Pple A. So remember: untensed INFL triggers MGC expansion, while tensed INFL does not, since it can be the ACC SUBJ and the GOV as well. Now we shall look at the behavior of Pronominals or Pronouns. Their interpretation is regulated by Pple B: Pronominals must be free within their MGC. The procedure will be: 1st we determine the MGC and 2nd we check binding conditions to see whether it's free. f. [Johni's father blames himi ].MGC Once we establish that the MGC is the whole sentence, we pass onto checking the Binding Conditions: Coind YES phi-features YES C-command NO Crucially, the 3rd condition is enough for binding to be impossible. "John" doesn't c-command "him" so the pronoun is FREE as predicted by Pple B. The sentence is OK. But: g. [John's father]i blames himi. Now the whole DP " John's father" DOES c-command "him" so it also binds it. This is forbidden by Pple B. Therefore the sentence is ungrammatical, which means that "him" cannot refer back to "John's father", i.e. they cannot be the same person in the real world. A further example is: h. Johni' s father considers that [hei is guilty].MGC

As we have said, the MGC is going to be the embedded clause. Evidently, "he" is not bound within it but from outside. In other words, "John's father" binds the pronoun from outside the MGC. To our purposes, "he" is FREE within the MGC, so Pple B is respected. Cf. i. [John's father]i considers himi to be guilty].MGC Again, the whole clause is taken to be the MGC. Now, the binder of the pronoun appears in the same clause, therefore binding it. Pple B is not obeyed. Finally, let's look at R-expressions. The good news is that they must be FREE everywhere (Pple C), which means that we don't have to look for the MGC. j. Johni's father blames Johni. This sentence is OK for binding since Pple C requires that the second "John" be FREE. How? Coind YES Phi-features YES C-command NO There ¡s no c-c so there's no binding. Coindexation and phi-features are possible as long as ccommand fails. This means that the 1st "John" and the 2nd one can perfectly well be the same person in the real world. Cf: k. John's fatheri blames Johni. This sentence will be ruled by Pple C, since the R-expression is bound given that the DP “John’s father” c-commands the R-expression (John). Same as in: l. Hei blames Johni. The pronoun "he" binds "John", violating Pple C. However, this sentence is possible, i.e. you can say this. How come? The key to this example is that "he" and "John" cannot be the same person. To show that they are different people, we are going to change the indexes. m. Hei blames Johnj. (or viceversa) Binding is no longer possible and Pple C is satisfied. The following is probably one of the most difficult examples: n. Johni thinks that [hei/j shouldn't blame himselfj].MGC Let's start by the anaphor. The MGC is the lower or embedded clause. As required by Pple A, "himself" is bound by "he", the closest antecedent. Question: Could "John" be its binder? Taking into account the Locality Requirement, "NO". We must always choose the closest binder. In this case "he" instead of "John". They refer to same individual, but syntactically speaking they are not the same element. Plus, “John" is not within the MGC. Secondly, the pronominal "he" is FREE, so Pple B is obeyed. Its binder is the DP "John" which is outside the MGC. The R-expression "John" is FREE also. Binding theory says that this sentence is grammatical. 6. CONTROL, ECM AND RAISING PREDICATES Verbs can also be classified taking into account the type of complement they select. Thus we obtain: a. Control verbs: the ones that select an infinitival clause (CP) with a PRO subject. b. ECM verbs: the ones that select an exceptional clause as complement c. Raising verbs: the ones that select an IP as complement with a trace as subject. Consider the following examples: 1. John tried [? to pass the exam]. 2. John seems [? to have passed the exam]. What's the status of the clausal complement in each example? In order to decide on the type of clause, we need to determine the nature of the embedded subject first. If it's PRO, the verb is said to be a control verb. But if it's a trace, the verb will be a raising verb. The relevant empirical evidence comes from the tests of "SUBJECTHOOD",which consist in changing the subject for an expletive or an inanimate entity. If the results are acceptable, it means that the subject is not an argument of the verb (i.e. it doesn't get a θ-role from it). In contrast if the results are bad, the claim is that the subject is an argument of the verb, since it gets the θ-role of agent. Let’s put all this into practice:

1a. *It tried to rain all day. 1b. *The novel tried to be a best seller. Evidently, there is something wrong with 1a. and 1b.. When aplying the subjecthood tests, we can see that they fail. Therefore, the verb "try" is a control verb which selects an external argument and an infinitival clause with PRO as subject.Cf: 1c. John tried [PRO to pass the exam]. Since PRO is controlled by the subject, the verb is said to be a subject-control verb. Now, let’s consider 2: 2a. It seems to have rained all night. 2b. The novel seems to be a huge success. The results are acceptable, so this is proof that the verb "seem" is a raising verb, which does not select an external argument. That's why the subject needs not be human or referencial due to the fact that it doesn't get a θ-role. Some provisions must be made: -If you use the expletive "it", you need a weather verb (rain, snow, hail). -If you use the expletive "there", you must form an existencial construction (there seems to be...) -You may need to make small adjustments so as to get a meaningful sentence (as long as you don't change the meaning of the verb). So far, we know how to distinguish control and raising verbs. Let’s turn to ECM verbs. Consider: 3. John expects Mary to be loyal. 4. John persuaded Mary to lie. The question is how we should analize the DP "Mary": as subject of the exceptional clause or as object of the main verb. Again we can seek evidence by applying the subjecthood tests. Careful: what we are going to replace is the embedded DP subject (not the main subject as before!!!). Let’s check: 3a. John expects it to stop raining. 4a. John expects the exercise to be difficult. The tests worked perfectly so we can conclude that the DP "Mary" is the subject of the exceptional clause, given that it has subject-like properties. Therefore, “expect” in this context is an ECM verb which selects two arguments, its θ-grid will be <experiencer, proposition>. Let’s pass on to 4: 4a . *John persuaded it to stop raining. 4b. *John persuaded the exersice to be difficult. The results are unacceptable which shows that “persuade” is a control verb: it requires an object (the DP Mary) and an infinitival clause complement [PRO to lie]. Given that PRO is controlled by the object rather than by the subject, “persuade” is said to be an object-control verb. Its θ-grid will be . TESTS------ POSSIBILITY 1-- good results =>raising v.(seem) Bad results =>subj-control v.(try) ------ POSSIBILITY 2-- good results=>ECM v.(consider) Bad results =>obj-control v(persuad

The behavior of PRO is regulated by the PRO Theorem which says that PRO must appear in ungoverned positions, such as spec of CP. As a consequence, it will always be Caseless (sort of a syntactic renegade...). Whereas DPs can appear in governed positions ONLY: otherwise they wouldn’t get K. Traces appear in +θ -K positions. As regards Binding Theory, PRO does not follow any of the principles since it does not have a governor or MGC. Anaphors can never appear where traces go, or they will be Kless. E.g. 5. *Mary seems herselfi to be smart.

7. A-BAR MOVEMENT Raising is an instance of A-movement because a DP moves to an A-position (argumental position). For example, SPEC of IP is an A-position. But then there is A-bar movement, that's to say movement to a "non-argumental position": a position where no θ-role can be assigned. More precisely, we will be concerned with SPEC of CP. WH- movement and relative clauses are instances of A-bar movement. A. WH-movement consists in moving a wh-phrase to the SPEC of CP. The motivation is the presence of a strong feature [+wh] at the head of CP. Strong features must be cancelled, otherwise the derivation will crash. Once spec-head agreement is obtained, the strong feature is checked away. Another property of the strong feature is that it attracts inflection giving rise to I-to-C movement. In English, both of them are obligatory in direct questions, i.e. those that end with a question mark. For example:

However, in indirect questions only wh-movement applies. Compare: 1. They wondered whether the teacher has cancelled the test. *They wondered whether has the teacher cancelled the test. Another example are reported questions: 2. They asked me what my name was. *They asked me what was my name. Wh-movement is regulated by the Subjacency condition, which states that movement applies cyclically; that's to say it must stop at every intermediate SPEC of CP and start another cycle. In each step it cannot cross more than one bounding node (IPs and NPs count as bounding nodes in English). Consider: 3. What do [you believe [t' that [they solved t]]? Does the movement of the wh-phrase take place in only one step? NO, because there is an intermediate empty SPEC of CP. The complementizer "that" is the head of CP so its SPEC is empty. An intermediate trace is postulated there. Subjacency is respected, since the first step only crosses IP2 and the second step crosses IP1. The sentence is grammatical. Let's turn to 4: 4. *What do [you believe [the fact t' [that [they solved t]]]] ? We need to determine what the bounding nodes are. The first step of the movement goes from comp of V to SPEC of CP2. It has crossed only one bounding node so it's OK. However, the second step (from SPEC of CP 2 to SPEC of CP1) crosses two bounding nodes, namely NP and IP1. Subjacency rules out this sentence (or we can say that this sentence violates the subjacency condition). This construction receives special name: NP-island or complex MP. An island is a certain type of clause that disallows extraction. As it is headed by the noun "fact" it's going to be an NP-island. There's a second type of island. Let's see: 5. *What do [you believe [how [they solved t t]]] ? As SPEC of CP2 is occupied by the wh-phrase "how", the movement of the wh-phrase "what" occurs in a single step, crossing two bounding nodes (IP1 and IP2). As a result, this sentence violates the Subjacency condition. This context is referred to as "wh island": the lower clause does not allow extraction. Whenever the clause is introduced by a wh-phrase, movement out of it will be impossible. Alternatively, we can say that there's an island constraint which is violated in 4 and 5. B. RELATIVE CLAUSES are divided into two groups depending on the presence (or the absence) of a relative pronoun.

6. I don't know the book [ which [John read t ]]. 7. *I don't know the book [ which [John made [the claim [that [he read t ]]]]]. 8. I don't know the book [ which [John claimed [that [he read t ]]]] 9. I don't know the student [ who [t read Syntactic Structures]] 10. * I don't know the student [ who [you believe [the fact [that [t read Syntactic Structures]]]]] 11. *I don't know the student [ who [you wonder [why [t read Syntactic Structures]]]] 12. *I don't know the students [ who [ you believe [that [t read Syntactic Structures]]]] 13. * I don't know the book [ which [John wonders [why [Mary read t]]]] When a relative clause does not contain a relative pronoun it looks like a "that clause". Since "that" is not a wh-phrase we cannot claim that it has moved. Rather, we are going to say that these clauses contain a null operator (op) which moves. As expected, movement of the null operator must obey Subjacency. 14. I liked the book [op that [John claimed [that [he read t ]]]] 15.*I liked the book[op that[John made[the claim [that [he read t 16.*I like the book[op that[John wonders[when [he will read t ]]]] 17. I met the person [op that [John says [he likes t]]] 18. *I met the person [op that [John says [t likes him]]]