The Development of Binding Theory Handout #1

Transcription

1 Sabine Iatridou EGG 2011 The Development of Binding Theory Handout #1 Chomsky 1981: Lectures on Government and Binding The Binding Conditions turn 30! We will start with a quick reminder of LGB, on the assumption that most linguists are pretty familiar with it. How do we capture the distribution of anaphors 1 and reciprocals (though most examples will be given with anaphors)? Anaphors and reciprocals need to have a DP that they can be co-indexed with/connect to/take their reference from etc : 1 a. *John saw herself in the mirror b. *Sue saw each other in the mirror c. John saw himself in the mirror d. Sue saw herself in the mirror This other DP, let s call it the intended antecedent, needs to c-command the anaphor: 2. [Sue i s sister] k saw herself *i/k in the paper 3. α c-commands β iff neither of α or β dominates the other and the first branching node dominating α also dominates β In short, what we see from (1-2) is that an anaphor must be bound : 4. α binds β iff α c-commands β and α and β are co-indexed 1 Before Chomsky 1981, the term anaphora was used for all referentially dependent items. This broader use still remains for many but there is also the convention for others to restrict the term to items like myself, oneself etc 1

2 But requiring just that an anaphor be bound is not sufficient, as can be seen below, where the intended antecedent clearly binds the anaphor: 5. *Sue thinks that Fred saw herself in the mirror So the question is what the locality constraints are on the relationship between the anaphor and the intended antecedent. Let us introduce the term Binding Domain, which refers to the domain (part of the clause) in which an anaphor must be bound. That is, we have the following condition, which incorporates the fact that an anaphor must be bound and that there are locality constraints on this binding: 6. An anaphor must be bound in its BD So the big question is, of course, what the right description of the anaphor s BD is. There were several attempts at defining the BD that led up to LGB (clausemate condition, Tensed S condition and others). We will go directly to the definition of LGB. LGB s BD is also called Governing Category (i.e. when you see this term, you know you are within the LGB framework): 7. The (BD) /governing category of an anaphor is the smallest XP that contains -the anaphor -a governor for the anaphor -an accessible subject/subject Let us look at each of the ingredients in turn. The reference to governor was deemed necessary to capture the ECM cases. Without it, the following would be predicted to be bad, as the BD of the anaphor would be the embedded infinitival clause: 8. Sue [ VP considers [ IP herself to be the best candidate] However, with the reference to governor, we can extend the BD to the matrix clause. Infinitival to is not considered a governor, so we need to extend the domain to find a governor for the anaphor, in this case the verb consider. So we go one clause up, and now the BD is the entire clause. In this BD, Sue binds the anaphor. Needless to say, we can go only one clause up. Once we find the governor, we cannot go to the next higher clause: 9. *Sue thinks that Bill considers [ IP herself to be the best candidate] Strictly speaking, the extension of the BD in (8) to include the higher verb does not end up with a BD that includes the subject Sue. The smallest XP containing the governor is 2

3 the VP, not the higher IP. With VP-internal subjects, one could say that it is the VPinternal trace that binds the anaphor. Before the days of VP-internal subject (including LGB days), one could get the desired result by saying that the BD has to be an IP (or DP) instead of just saying it has to be an XP. This way the VP headed by consider is not a possible candidate for a BD. Later yet (in Chomsky s 1986 Knowledge of Language; more on this soon) the BD had to be a Complete Functional Complex: 10. A Complete Functional Complex is the domain in which all grammatical functions associated with a head are realized. With reference to a CFC we get the desired result in (8) because once we include the head consider (the governor of the anaphor), we have to include all the GFs of that head. This will give us a BD that includes the agent of consider. What about the reference to subject/subject in (7)? Things here get a bit more complicated. What is a subject : a DP in specifier position. That some reference to subject is necessary can easily be seen: 11. Jane i saw [pictures of herself i] in the paper 12. Jane i saw [Mary k s pictures of herself *i/k ] in the paper In (11), Jane can be the antecedent of the anaphor but in (12) it cannot. We derive this by reference to the smallest XP containing a subject. In (12), the smallest XP containing a subject is the DP Mary s pictures of herself and the anaphor is bound in it. Note that in order for the DP to be the BD, it is not enough for it to have just any sort of specifier. Only a subject (i.e. a DP) will do, as can be seen in the contrast between (12) and (13), where Jane can be the antecedent of the anaphor. 13. Jane i saw [5/these/many/embarassing pictures of herself] i in the paper Going back to (7), what is SUBJECT (pronounced big subject ) and what does accessible mean? SUBJECT is AGR under the INFL node. Having AGR contributing to the determination of the BD means that the ungrammaticality of (14) can be said to be due either to the anaphor itself determining the BD as the embedded clause, or AGR in the embedded INFL doing this. In this BD, the anaphor is not bound. 3

4 14. a. *They think that each other left on time b. *Bill thinks that himself left on time However, the difference between subject and SUBJECT can be best appreciated once we bring the idea of accessible into the mix. This notion was introduced to take care of anaphors inside the subject of finite clauses. The fact that there is an issue with anaphors inside subjects of finite clauses can be seen in the contrast between (14) and the sentences in (15): 15 a. Bill thinks that [[pictures of himself] were published in the paper] b. They think that [[pictures of each other] were published in the paper] c. They think that [[each other s pictures] were published in the paper] The descriptive generalization seems to be the following: when the anaphor is the subject of a finite clause, the BD is the finite clause (as in 14). However, when the anaphor is embedded inside the subject of a finite clause, we are allowed to extend the BD domain to the matrix sentence (as in 15). What justifies this extension? This is where the notion of accessible comes in. 16. A subject S is accessible to an element α (in our case the anaphor) if co-indexation of S and α does not violate any grammatical principle. Specifically, the grammatical principle that Chomsky had in mind was the i-within-i filter: 17. The i-within-i filter. *[.X i..] Yi Let s look over (14-15) now. In (14), represented as (18) below, the SUBJECT (AGR of embedded INFL node) can be co-indexed with the anaphor without the i-within-i filter being violated. As a result, the BD must be the embedded clause (it also contains a governor for the anaphor): 18. *They think that [each other i AGR i left on time] But things are different with the sentences in (15c), represented as (19) below. Suppose that the first subject we are trying for accessibility to the anaphor is the subject of the embedded finite clause: 19. They think that [[each other i s pictures] i were published in the paper] 4

5 This indexing violates the i-within-i filter. This means that the DP [each other s pictures], the subject of the embedded clause, is not an accessible subject and therefore cannot be used to determine the BD of the anaphor. Next we try the SUBJECT, namely the AGR of the embedded clause: 20. They think that [[each other s i pictures] AGR i were published in the paper] So far, so good. However, AGR is also co-indexed with the DP [each other s pictures] because of subject-verb agreement. If we compute this indexing in, we end up with an i- within-i violation: 21. They think that [each other i s pictures] i AGR i were published in the paper] This means that the SUBJECT/AGR is not accessible either. So we need to go a clause up to find an accessible subject/subject in either They or the matrix AGR. As a result, the BD of the anaphor in the subject position is the matrix clause. The narrative for (15a,b) is similar. Note: Chomsky does not assert that the co-indexing involved in subject-verb agreement is on a par with the indexing associated with referential dependencies. Neither does he assert that the anaphor and the accessible subject/subject need to end up co-indexed. You can think of it as a game of virtual co-indexing. Something that was not discussed in LGB is that (14-15) differ in that in (14) the anaphor is impossible and the pronoun required, but in (15), both the anaphor and pronoun (with the same indexing) are an option. Along with (15), we can have (22): 22. a. Bill i thinks that [[pictures of him i ] were published in the paper] b. They i think that [[pictures of them i ] were published in the paper] c. They i think that [[their i pictures] were published in the paper] The option to have either a pronoun or an anaphor (i.e. the lack of complementary distribution between the two) is not something discussed in LGB but it was picked up later by Chomsky in his 1986 Knowledge of Language. So now we have gone over all the ingredients of (7), repeated here: 7. The (BD) /governing category ofan anaphor is the smallest XP that contains -the anaphor -a governor for the anaphor -an accessible subject/subject 5

6 And what was our (6) can now be restated as LGB s famous Condition A. 6. An anaphor must be bound in its BD 23. Binding Condition A: An anaphor must be bound in its Governing Category. Next we turn to pronouns. In many cases, pronouns and anaphors with the same index are in complementary distribution. Here are some of the sentences that we have seen already that make this point: 24 a. Sue i saw herself i /*her i b. Sue i thinks that Fred saw her i /*herself i in the mirror c. Sue i considers herself i /*her i to be the best candidate] d. Bill i thinks that *himself i /he i left on time e. Jane i saw Mary k s pictures of her i / *herself i in the paper LGB then generalized that the pronoun should be free (i.e. not bound) in the domain in which an anaphor should be bound. This became Condition B: 25. Binding Condition B: A pronoun must be free in its Governing Category. Note that Condition B does not say that a pronoun should be bound outside its GC; just that it should be free in it. Once the pronoun is free in its GC, LGB has nothing further to say about what it can or cannot felicitously refer to. In other words, while Condition B dictates where a dependency should hold, Condition B is an anti-dependency or prohibition clause. For completeness, before we conclude this section, we should also mention LGB s Condition C: 26. Binding Condition C: An R-expression must be free everywhere. 27. a. *He i thinks that Bill i is smart b. The woman who he i married think that Bill i is smart c. When he i is sick, Bill i always refuses to go to the doctor However, Condition C has a life and literature of its own and we will not be focusing on it very much in this class. 6

7 The Binding Conditions of LGB received a lot of attention. The proposal was extremely influential, and the literature commenting on it huge, even if subsequent cross-linguistic and even English-specific research showed that it didn t get things quite right. For example, several languages have long distance (or SE ) anaphors whose distribution Condition A could not capture. Moreover, LGB did not get the relationship between anaphors and pronouns quite right, and Chomsky sought to improve on this with his 1986 book. We turn to this next. Chomsky 1986: Knowledge of Language LGB defined identical BDs for pronouns and anaphors and therefore predicted complementary distribution between them: the pronoun should be free in the domain in which anaphors should be bound. However, this is not the right prediction, as we saw above and can also be seen in the following: 1. Mary i saw a snake in front of her i 2. Mary i saw a snake in front of herself i If the Binding Domain of the anaphor is the entire sentence, how can the pronoun be free in it? KoL proposes to address this problem by defining the BDs in such a way that the BD of a pronoun is different from that of an anaphor. What we want to achieve is that the BD of the anaphor in (1) is large enough to contain the subject, so that Condition A can be satisfied. At the same time, the BD of the pronoun in the same position should be smaller so that it does not contain the subject and Condition B can be satisfied. In short, what we want to derive is that the domain of the anaphor is larger than that of the pronoun. However, it is not the case that we always want the BDs of pronouns and anaphors to be different because there is, after all, complementary distribution in other environments: 3. Mary i saw herself i / *her i in the mirror 4. Mary i heard that Fred likes her i /*herself i So the challenge is to make the BDs of the anaphor sometimes be the same as and sometimes different from that of the pronoun, and when they are different, the BD of the anaphor should be larger than that of the pronoun. Several people have tried to achieve this (e.g. Huang 1983, LI). KoL proposes to meet this challenge with the notion Binding Theory compatibility. The basic idea: the BD of the anaphor/pronoun should give the anaphor/pronoun a chance to satisfy its Binding Condition. For an anaphor to have a chance to satisfy Condition A, 7

8 there should be one c-commanding DP in its BD. NOTE: the condition is not that there be a c-commanding co-indexed DP in the BD. Any c-commanding DP will do. For a pronoun to satisfy Condition B, there need not be a c-commanding DP in its BD (though there can be, of course, as in 3,4). As a result, when the BDs differ, the BD of the anaphor is larger than that of a pronoun, as it will contain a subject. Let s go through some definitions and examples. BT-compatibility Here are some definitions: 5. An anaphor or pronoun must meet its binding requirement in the smallest binding domain with a BT-compatible indexing. 6. An indexing I is BT-compatible for an anaphor if the anaphor is bound under I, and BT-compatible for a pronoun if the pronoun is free under I. 7. A binding domain D for α is the minimal CFC containing α, a governor for α and a subject. 8. A Complete Functional Complex is the domain in which all grammatical functions associated with a head are realized. For basic cases like (3,4), BT-compatibility does not make any different predictions than LGB. Consider, though, the following cases: 9. They i saw [ DP their i pictures] 10. They i saw [ DP each other i s pictures] Sentences (9, 10) are both fine, so this pair constitutes a case where the complementary distribution of pronouns and anaphors breaks down. Exactly what BT-compatibility was designed to take care of. Here is how it works: For the pronoun in (9), the BD is the DP because: -it is a CFC (in fact, the smallest CFC containing the pronoun) -it contains a governor and subject (the pronoun itself) -there is a BT-compatible indexing that permits the pronoun to satisfy its Binding Condition. (in fact, any indexing will keep the pronoun free in this BD) For the anaphor in (10), the BD cannot be the same as for the pronoun in (9) because there is no potential antecedent in it. i.e. there is no c-commanding DP in it. That is, the DP cannot be the BD for the anaphor because there is no BT-compatible indexing that 8

9 would permit the anaphor to in principle satisfy its Binding Condition. So we go to the next largest CFC, which is the entire clause. The matrix clause is the BD because it is the smallest CFC containing - the anaphor, a governor and a subject and - for which there is a BT-compatible indexing We should note that the extension of the BD licensed by the search for a BT-compatible indexing can only go exactly one clause up, that is, in search of a DP c-commanding the anaphor. This is exactly what we want: 11. *[ IP1 They i heard that [ IP2 Mary k saw [ DP each other i s pictures]]] The BD of the anaphor is IP2, not IP1. The reason is that it is the smallest CFC containing a BT-compatible indexing, namely the indexing in which i=k. The fact that there are featural differences between Mary and they that would make such an equation of indices impossible, is not computed into the equation. The BD is IP2 and in there the anaphor is, unfortunately, not bound. So this is basically how the KoL system works 2. But what about the original cases in (1,2): 1. Mary i saw a snake in front of her i 2. Mary i saw a snake in front of herself i In these sentences, the BD of the pronoun and anaphor still end up coming out the same. The smallest CFC that satisfies the relevant requirements is the entire sentence: this CFC contains the anaphor, a governor, a subject and there is a BT-compatible indexing for the anaphor. But then it is wrongly predicted that the pronoun should be unacceptable in this context, as it is bound within its BD. (If we say that the PP contains a subject and therefore it is the PP that is the BD. The same problem arises.) 2 And you can see how it works on the sentences that we saw in the LGB section where the complementary distribution breaks down: i. Bill i thinks that [[pictures of himself/him / ] were published in the paper] b. They i think that [[pictures of themselves/them i ] were published in the paper] c. They i think that [[each other s/their i pictures] were published in the paper] In the case of the anaphor, the BD must extend to the higher clause to include the subject DP to give the anaphor a potential binder/ chance to satisfy its Binding Condition. In the case of the pronoun, the BD can remain the subject of the embedded clause. 9

10 If we manage to make the BD for the anaphor the entire clause and the BD of the pronoun the PP, we get the correct results. But how can we do this? Let s take another look at (7): 7. A binding domain D for α is the minimal CFC containing α, a governor for α and a subject. Why do we need the extra specifications that the CFC has to contain a governor and a subject? The governor is necessary to get the ECM cases correctly, as we have already seen in the LGB section: An anaphor in ECM subject position would anyway warrant the extension of BD for BTcompatibility reasons: 12. They expect [each other to win] But without the reference to governor, the wrong prediction would be made for pronouns. Consider (13): 13. *They i expect [ IP them i to win] Without reference to a governor, the BD for the pronoun would be the embedded IP, as it is a CFC, it contains a subject, and there is a BT-compatible indexing that would permit the pronoun to satisfy its Binding Condition (any indexing will do, in fact, just as for (9)). But this would wrongly predict that (13) is good. If we say that the CFC should contain a governor, we are forced to extend the BD to include the ECM verb expect. Since the BD has to be a CFC, we have to extend the BD to include all GFs of expect, that is the subject as well. For this reason, it doesn t seem possible to remove reference to governor at this stage. But what about reference to the subject in (7)? Do we really need it? In the case of anaphors, reference to a subject for the BD is redundant. The search for a potential binder/c-commanding DP (i.e. the search for a BT-compatible indexing) will ensure that there is a subject. So wrt anaphors, we can remove reference to subject. What about in the case of pronouns? With pronouns, removing reference to subjects does not have any empirical consequences for cases like (9), repeated here: 10

11 9. They i saw [ DP their i pictures] That is, there will be no consequence for pronouns that are subjects themselves. When the pronoun is not the subject of the DP, then we can say that PPs do not have a subject, and the prediction is that the BD for the pronoun in (1) is the PP (The PP is anyway a CFC). 1. Mary i saw a snake [in front of her i ] Whereas with the anaphor, the BD is the entire clause because of the search for a BTcompatible indexing: 2. Mary i saw a snake in front of herself i In other words, we get the right results for (1,2) if we remove reference to subject when we define BD. That is, if instead of (7), we have (7 ): 7. A binding domain D for α is the minimal CFC containing α, a governor for α and a subject Arild 1991 (NLLT) pursues the consequences of the working hypothesis that there should be no reference made to subject in the definition of BD and discusses a number of PPs, among other constructions, with an emphasis on Norwegian. Giorgi 1987 (LI squib) discusses similar issues with a focus on Italian. As with LGB, the KoL proposal influenced a lot of research, too large to mention here. BT history seen through the eyes of the (non-) complementary distribution of pronouns and anaphors. LGB predicted perfect complementary distribution between pronouns and anaphors, as the domain in which anaphors have to be bound is defined identically to the domain in which pronouns have to be free. In KoL, the lack of complementary distribution between pronouns and anaphors drives the reformulation of the binding domains so that the domain in which an anaphor must be bound is sometimes larger than the domain in which a pronoun must be free. Since the domains are not identically defined, it is possible to sometimes have a choice between an anaphor and a pronoun in the same position. In Reinhart & Reuland 1993, the next work we will be discussing, the complementary distribution of pronouns and anaphors falls outside of the purview of reflexivity as a 11

12 strictly grammatical phenomenon, but instead belongs to the area of logophoricity, a discourse phenomenon. For R&R, the Binding Conditions should not be rewritten so as to accommodate the lack of complementary distribution between anaphors and pronouns; the BT conditions should predict exactly such a complementary distribution. In other words, whereas LGB s Binding Conditions predicted complementary distribution out of ignorance of the facts, R&R s Binding Conditions consciously predict complementary distribution, as the environments where we observe lack of complementary distribution have been exiled to discourse accessibility/licensing. In addition, R&R make a claim about where the complementary distribution will fail, i.e. where the anaphor will be licensed logophorically: when the anaphor/pronoun is not in the argument position of a syntactic predicate 3. In other words, anaphors that are not in an argument position of a syntactic predicate are exempt from BT proper 4. How can we empirically determine whether an anaphor is used logophorically? -A pronoun can be used instead of the anaphor (!!!!!!) -The anaphor s antecedent can be far -The anaphor s antecedent does not have to c-command the anaphor -The anaphor s antecedent may be altogether absent. 1 a. Max saw a snake near himself/him. b. Lucie counted five tourists in the room apart from herself/her c. Max saw a picture of him/himself in the paper d. Max thinks that Susie saw a snake near himself/him e. The picture of himself that John saw in the post office was ugly. f. Her pleasant smile gives most pictures of herself an air of confidence. f. People like myself/me are hard to find Logophors are mostly of the point-of-view variety (refer to the center of communication or consciousness (1)) or focus anaphors. Here are some examples of the latter: 2. a. This letter was addressed only to myself. b. Why should the state always take precedence over myself? 3 The syntactic predicate formed of (a head) P is P, all its syntactic arguments, and an external argument of P (subject).the syntactic arguments of P are the projections assigned theta-role or Case by P 4 As we will see, verbs always form syntactic predicates because they have a subject, Ps and Ns not so. There is an interesting discussion about Ps that form part of the predicate (Max looks after himself/him) and Ps that do not (Max saw a snake near himself/him), but we will skip it here. 12

13 c. "Bismarck's impulsiveness has, as so often, rebounded against himself." (quoted in Zribi-Hertz 1989) d. himself [Bismarck's impulsiveness has, as so often, rebounded against] Anaphors in argument position can be logophoric only if they are foci. R&R say a bit more about logophoric licensing but basically they intend to remove them from the focus of the paper, which is about grammatical, not discourse-, licensing of anaphors. Reinhart and Reuland 1993: Reflexivity (LI) Binding Conditions A and B R&R have a Condition A and a Condition B. Here they are: Condition A: A reflexive-marked syntactic predicate is reflexive. Condition B: A reflexive semantic predicate is reflexive-marked Here is a definition of some of the terms (ignore the notions of syntactic and semantic predicates for now): A predicate is reflexive iff two of its arguments are co-indexed. A predicate P is reflexive-marked iff either P is lexically reflexive 5 or one of P's arguments is a SELF anaphor. You will notice that the two Binding Conditions look nothing like those of LGB or KoL. Crucially, they do not even make reference to DPs or to structural relations between DPs; they are stated solely as conditions on predicates. 5 Intrinsic or lexical Reflexivization is an operation on the theta-grid of the verb, that results in the absorption of one of the theta-roles of the verb. Intrinsically reflexivemarked verbs can only be used reflexively. In some languages the argument must still be overtly realized (Dutch) in other it can be absent (English): i. Max wast *(zich). Max washes SE Max washed ii. Max schaamt zich. Max shames SE 'Max is ashamed.' 13

14 If they look so different, why, then, are they still called Conditions A and B? Because R&R s Conditions A and B account for the status of some of the core cases taken care of by LGB s Conditions A and B. Specifically, Condition A: 3. *Katie i saw myself m Sentence (3) is out because the predicate is -reflexive-marked (one of its arguments is a SELF anaphor), -yet it is not reflexive (its two arguments are not coindexed). Condition B: 4. *Katie i saw her i Sentence (4) is out because the predicate is -reflexive (it s two arguments are co-indexed) -yet, it is not reflexive-marked (it is not lexically reflexive, nor is one of its arguments a SELF anaphor) But there are also cases that would fall under different Binding Conditions in LGB versus in R&R. Here is a sentence that would fall only under LGB s Condition A but in R&R is relevant for both Conditions A and B: 5. Katie i saw herself i Sentence (5) is fine because -the predicate is reflexive-marked by the SELF anaphor and should therefore be reflexive -the predicate is reflexive -So Condition A is satisfied -the predicate is reflexive, so should be reflexive-marked -the predicate is reflexive marked -So Condition B is satisfied. Also, since there is no reference to types of DPs in R&R s conditions, Condition B also rules out certain cases that would fall under LGB s Condition C: 6. * He i criticized Max i 14

15 In (6), the predicate is reflexive, but it is not reflexive-marked, which is an R&R Condition B violation. More details on how the R&R system works: One SELF anaphor suffices to reflexive-mark a predicate for more than one argument. For the examples below, you need to know that SE anaphors, Dutch zich here, cannot reflexive-mark a predicate on their own, as evidenced e.g. 7a,b. These sentences are out as Condition B violations: the predicate is reflexive but it is not reflexive-marked. 7 a. *Henk i wees mij aan zich i toe. Henk assigned me to SE b. *Henk i wees zich i aan mij toe. Henk assigned SE to me However, in (8) the SE anaphor can be bound within the predicate. The SELF anaphor reflexive-marks the predicate and the binding of SE is compatible with Binding Condition B. 8. a. Henk i wees zichzelf i aan zich i toe. Henk assigned himself to SE b.. Henk i wees zich i aan zichzelf i toe. Henk assigned SE to himself These sentences also tell us that it doesn t matter in which argument position the reflexive-marking SELF anaphor is 67. It s all about reflexive (or not) predicates 6 Question: Wouldn t this wrongly predict that the following should be OK: i. *Jan i wees mijzelf aan zich i toe J assigned myself to SE This prediction is made as follows: the SELF anaphor reflexivizes the predicate and now the SE anaphor can appear without violating Condition B. R&R have taken care of this (663): The condition is relativized to an index to guarantee that a given SELF anaphor, indexed j, would not count as licensing the coindexation of two arguments with a different index, i. Their definitions are actually as follows: a. A predicate is i-reflexive iff (at least) two of its arguments are i-coindexed (that is, are indexed i). b. A predicate (formed of P) is i-reflexive-marked iff either P is lexically reflexive with respect to an i-indexed argument, or one of P's i-indexed arguments is a SELF anaphor. 15

16 So a SELF anaphor can reflexive-mark a predicate from any position, including subject position 8. BUT..This means that the following sentences do not suffer from a Condition A nor a Condition B violation: 9. *Himself i, saw him i Since the predicate is reflexive-marked by the SELF-anaphor, and it is reflexive, this sentence satisfies both Conditions A and B. What rules it out then? 7 Though economy considerations are said to be behind the degraded status of sentences with a SELF anaphor in more than one argument: i.?henk i wees zichzelf i aan zichzelf i toe. Henk assigned himself to himself 8 An empirical argument that R&R give that reflexive marking from a subject position to be possible comes in the very last section of their paper: Recall that SE anaphors cannot appear bound inside their predicates because of Condition B, since they cannot reflexive-mark the predicate: i. a. *Jan critiseerde zich. Jan criticized SE b. *Jij hoorde [Jan zich critiseren]. you heard [Jan SE criticize] However: ii. Jan horde a. *[zich zich critiseren]. b. [zich zichzelf critiseren]. c. [zichzelf zich critiseren]. d. *[zichzelf hem critiseren]. iia is out for the same reason as ib. iib,c are fine because the SELF anaphor reflexivemarks the predicate, and the SE anaphor can now be bound within its predicate Note the contrast between iic and iie. R&R say that iid is out for a reason that we will see (the Chain condition) but the fact that there is a contrast between them means that it is not the case that zichzelf is unable to reflexive-mark the predicate. However, in a discussion of similar sentences in 7.1 they say that the embedded verb moves onto the matrix verb and that the external argument of the embedded infinitival becomes an argument of the new complex verb. A SELF anaphor in the embedded subject position reflexive-marks both verbal elements of the complex verb. This means that the cases in ii don t really provide an argument for reflexive-marking from the subject position. 16

17 Enter the Chain Condition. Before we discuss the Chain Condition in detail, a few words about the typology of anaphors and pronouns that R&R assume. There are two types of anaphors: -SELF anaphors: complex expressions, bound locally -SE anaphors: simplex, bound long distance 9, English only has the SELF-type anaphor, Dutch has both, as we saw: 10. Jan haat zichzelf/*zich/*hem Jan hates SELF/*SE/*him Though both SE and SELF are anaphors, they differ in that SE anaphors cannot reflexivemark a predicate. And because SE-anaphors do not reflexive-mark their predicate, they behave like pronouns with respect to Condition B (=if a predicate is reflexive, it has to be reflexive marked): 11 a. *Max haat zich. Max hates SE b. *Max praat met zich. Max speaks with SE 12 a. Max legt het boek achter zich/hem Max puts the book behind SE/him b. Jan zag [jou achter zich/hem staan]. Jan saw [you behind SE/him stand] 'Jan saw you stand behind SE/him.' However, because SE anaphors always lack number and gender features (though not person), they cannot refer independently, unlike pronouns (P 697: An NP is + R iff it carries a full specification for [number and gender] features and structural Case ). So here is what we have: 9 Though sometimes they can be bound within their predicate as we saw. 17

18 Reflexivizing function R(eferential independence) SELF SE Pronoun The Reflexivizing function is subject to Binding Theory, as we can see from the formulation of their Binding Conditions. The R-property is subject to the Chain Condition. 10 So now let s now go to R&R s Chain Condition. R&R follow the working hypothesis of Chomsky 1973 that A-movement/chains and A- binding/anaphora are related phenomena More specifically, R&R: binding of +/-R NPs is regulated by conditions on A-chains. 13a. A maximal A-chain (a i.a n ) has exactly one link (a i ) which is both + R and Casemarked 13. A basic sentence like (13b) satisfies the Chain condition because only its head is both +R and Case- assigned. The tail is Case- assigned 14, but it is R: 13 b. Sue saw herself. On the other hand, (13c) violates the Chain Condition because the tail is both Case- assigned and +R: 10 SE anaphors are taken to move and adjoin to an INFL to acquire NG features. For this reason, part of the distribution of SE anaphors falls under Movement theory as well. 11 The locality that holds between an A-moved DP and its trace is the locality between the antecedent and the anaphor (also the domain in which a pronoun must be free). Following examples are R&R s 77a-c: a. He i is believed [t i to be smart]. b. He i believes [himself i to be smart]. c. *He i believes [him i to be smart]. 12 Chomsky s 1986 Chain Condition: An A-chain has exactly one Case, in its head position, and exactly one theta-role. As R&R point out in their fn 33, this should not be seen as a condition on wellformedness, or it would rule out John saw himself as the anaphor and its antecedent each have a theta-role. At any rate, R&R dispense with the reference to theta-roles after some discussion. 13 Should not apply to single-membered chains or logophoric anaphors would violate it because the head of the chain would be _R. 14 On page 697, they discuss the possibility that being fully specified for Case is a prerequisite for being +R. 18

19 c. *Suei saw heri The chain condition has something to say about all of SELF, SE and pronouns, as it dictates certain well-formedness conditions that are imposed because of their +/-R status. +R elements can only head chains. R elements can only be tails (or intermediate links). With the Chain condition in hand, we can now rule out the following: 14. a. *Himself i saw him i. b. *Himself i saw Max i As we said, these sentences do not suffer from a Binding Condition violation because the predicate is reflexive, as well as reflexive-marked by himself. But the sentences do violate the Chain Condition: the head of the chain is R and the tail +R. On the other hand, the following are compliant with the Chain Condition, as the head of the chains are +R and the tail R: 15. a. He saw himself b. Max saw himself. Some sentences are ruled out both by Condition B and the chain condition: 16. *Max i saw him i In (16), the predicate is reflexive without being reflexive marked (Condition B violation). In addition, the chain has a +R tail. And here is an interesting interaction between the Chain condition and SE anaphors, bringing out a difference between them and pronouns: The predicates in (17) are intrinsically reflexive, so 17a is fine wrt both Binding Conditions. It is also fine with respect to the Chain condition, as SE is a fine R tail. 17 a. Willem i. schaamt zich i Willem shames SE b. *Willem i. schaamt hem i Willem shames him (17b) on the other hand, satisfies Condition B for the same reason as (17a), that is, the predicate is reflexive and it is reflexive-marked. However, it is out because of the Chain condition, as the pronoun is a +R tail. 19

20 Similarly, consider also the following: 18 a. *Henk i wees hem i aan zichzelf i toe. Henk assigned him to himself b. Henk i wees zich i aan zichzelf i toe. Henk assigned SE to himself The predicate is reflexive-marked by the SELF anaphor. As a result, the SE anaphor in (b) can be bound by Henk. This sentence doesn t violate the Chain condition, as SE is R. However, the sentence in (a) violates the Chain condition, as the pronoun is +R. This contrast shows, according to R&R, that the issue is the Chain condition and not Condition B. If it were just Condition B, there should be no contrast between SE and the pronoun. Back to the Binding Conditions Recall Conditions A and B: Condition A: A reflexive-marked syntactic predicate is reflexive. Condition B: A reflexive semantic predicate is reflexive-marked At first we had ignored reference to syntactic and semantic in R&R s Binding Conditions. Now it is time to look at this issue in more detail. Condition A holds of syntactic predicates and Condition B of semantic ones. Here are the relevant definitions: The syntactic predicate formed of (a head) P is P, all its syntactic arguments, and an external argument of P (subject). The syntactic arguments of P are the projections assigned theta-role or Case by P. The semantic predicate formed of P is P and all its arguments at the relevant semantic level P : The definition of semantic predicates is rather trivial. It is determined solely by logical syntax, provided that the compositional rules mapping structures into functionargument configurations are defined. Deciding what counts as a syntactic predicate, however, is based on grammatical structure 20

21 Verbs form both syntactic and semantic predicates. However, subjectless predicates (such as nouns and prepositions sometimes) can only form semantic ones. Since their Condition A applies to syntactic predicates and their Condition B to semantic ones, R&R claim that Condition A effects can be found only with verbal predicates but Condition B effects with predicates headed by V or P or N. Here is an example for why Condition B should apply to semantic predicates: Consider the following: 19. a. *The queen invited myself for tea. b. The queen invited both Max and myself/me for tea. c. Max said that the queen invited both Lucie and himself/him for tea. Sentence (a) is out by Condition A: the predicate is reflexive-marked but not reflexive. In (b,c) the predicates are not reflexive marked. The anaphors are used logophorically. This is evidenced by the fact that the pronoun can optionally appear instead of the anaphor and that the antecedent can be absent (19b) or be long distance (19c). Now compare the above to the following: 20. a. The queen i invited both Max and herself i to our party. b. *The queen i invited both Max and her i to our party. In (20) there is no complementary distribution between the pronoun and the anaphor, hence it is difficult to claim that herself is a logophor in (20a). So what is the difference between (19b) and (20a) so that the anaphor can be logophoric in (19b) but not in (20a)? The answer lies in that a semantic reflexive predicate is formed in (20a). The conjunction in (20a) is interpreted as follows: 21. the queen (λx (x invited Max & x invited x)) The representation in (21) contains a reflexive predicate: x invited x. If Condition B applies on semantic representations, it finds out that in [(20a)] one of the arguments of this new semantic predicate that is about to be formed is, appropriately, realized in the syntax (LF) as a SELF anaphor; hence, the translation yielding a reflexive predicate is allowed. (p. 676). On the other hand, in (19b) there is no reflexive predicate and therefore the pronoun is allowed. But this means that Condition B can see (21) and rule out (20b). Hence it applies to semantic predicates Question: so do we consider the anaphor in (20a) a logophor or not? After all it is not in argument position of a syntactic predicate. R&R p. 675: By the same reasoning, the 21

22 Another argument that Condition B looks at semantic predicates: 22. *We voted for me. 23. We elected me In (22) the predicate receives a distributive interpretation; in (23) a collective. The same in (24, 25): 24. *Both Max i and Lucie talked about him i (distributivity forced by both; there are two separate acts of talking) 25. Max i and Lucie talked about him i Under the distributive interpretation, but not under the collective interpretation, there is a reflexive (semantic) predicate. 26. a. Max (λx (x talked about x)) & Lucie (λx (x talked about him)) (distributive) b. Max and Lucie (λx (x talked about him)) (collective) So in the distributive interpretation, there is a semantic reflexive predicate that is not reflexive-marked, so Condition B is violated in (24) but not (25). But this point can be made (and the contrasts in 22/23, 24/5 accounted for) only if Condition B can see the reflexive predicates in (26a), i.e. only if it can see semantic predicates anaphor in [20a] is logophoric as well. Then they proceed to show that the anaphor is not, however, in complementary distribution with the pronoun (our 20b). The other anaphora-diagnosing tests (lack of c-command by antecedent; absence of antecedent) will not do here because as soon as we apply them, there ceases to be semantic predicate formation. So, my [si] understanding is that they do not consider this a logophor, and that the quote is part of a rhetorical move. 17 Condition B can only see semantic predicates. That is, it cannot see syntactic ones. This discussion is spread out throughout the paper, but here are two arguments: In (i), the SE anaphor is part of the matrix syntactic reflexive predicate, given it s Case relation to the higher verb: i. Jan hoorde [zich, zingen]. Jan heard [SE sing] If Condition B could see this reflexive predicate, the sentence would be out because the predicate is not reflexive-marked. And here is one argument from their footnote 26: ii Max i was fired e i. 22

23 Condition A does not look at semantic predicates: Consider the queen sentences again: 27 a. The queen invited both Max and myself for tea. b. The queen (λx (x invited Max & x invited myself)) If Condition A applied to syntactic representations, it would rule out (27b) because the second conjunct is reflexive-marked but not reflexive. Since the sentence is grammatical, Condition A does not look at semantic predicates. The above are some arguments for the claim that the two Binding Conditions hold for different types of predicates. However, one of the biggest arguments for this distinction comes from their discussion of ECM. We turn to that next. ECM Pronouns and SELF anaphors: Look at the ECM subjects below: This syntactic predicate contains two co-indexed arguments, i.e. it is reflexive, but the semantic predicate contains only one argument, the internal one; the agent is existentially bound. If Condition B saw the syntactic reflexive predicate, this sentence would be ruled out, as the predicate is not reflexive-marked. 18 Question: What about sentences like (i), where the verb requires a plural object? The anaphor is good, the pronoun bad, so the explanation should be the same as for (20a). Yet, there is no reflexive predicate like (ii) formed: i. Max i combine Mary and himself i /*him i in a team ii. *Max i combined himself i in a team Possible answer: R&R are not precise about what they mean by semantic predicate. If we take it to mean entailment, that is, if we take it that Condition B can see reflexive predicates in entailments, then we can say that a sentence like (iii) has the entailments in (iv, v). Then we can say that it is the entailment of (i) in (vi) that Condition B sees and demands reflexive-marking on the predicate: iii. A combines B and C into a group iv. A combines B in a group with C v. A combines C in a group with B. vi. Max λx. x combines x in a group/team with Mary' 23

24 28. a. Lucie expects [herself to entertain herself]. b. *Lucie expects [myself to entertain myself]. c. *Lucie i expects [her i to entertain herself]. In (28a), the SELF anaphor is a Case-argument of the matrix verb and so it can reflexivemark it. Condition A is satisfied because the (matrix) predicate is also reflexive. (28b) is meant to show that this position is correctly predicted to not be able to support logophorically used anaphors. Condition B, however, does not rule out (28c). The reason is that Condition B only looks at semantic predicates and in (28c), the ECMed subject is not part of the matrix semantic predicate. What rules out (28c) is the Chain condition, which holds of syntactic chains. A contrast between SE and pronouns shows up in an interesting way in ECM environments. While we saw that SE anaphors and pronouns have certain similarities that follow from their SELF status (namely, they are both subject to Condition B), several differences show up as well. One case in question is ECM: 29. a. *Jan i hoorde zich ii /hem i. Jan heard SE/him b. Jan hoorde [zich zingen]. Jan heard [SE sing] c. *Jan, hoorde [hem zingen]. Jan1 heard [him1 sing] d. Jan, hoorde [zichzelf, zingen]. Jan heard [himself sing] Condition B blocks (29a) for both the pronoun and the SE anaphor. But condition B cannot rule out (29b) or (29c) as the SE anaphor or pronoun is not part of the matrix semantic predicate. What distinguishes (29b,c) is that the Chain condition is respected in (29b) but not (29c). Sentence (29d) satisfies Condition A and the Chain Condition. The above reasoning also predicts/captures the fact that SE and SELF anaphors are in complementary distribution in the complement position of a verb, but not in ECM subject position. The reason is that in the complement position of a verb, both Conditions A and B are relevant. In an ECM subject position, though, only Condition A has a say. Condition B does not apply because there is no semantic predicate (or Condition B is vacuously satisfied). This last conclusion brings us to an interesting contrast between ECM subjects and Object Control cases (and therefore a way to diagnose when we have which): 24

25 In ECM, SE and SELF are not in complementary distribution, as we just saw above, and below in (30a,b), but in object control cases (31a,b), they are, because they are in the complement of a verb. ECM: 30. a. b. Jan hoorde [zich zingen]. Jan heard [SE sing] b. Jan, hoorde [zichzelf, zingen]. Jan heard [himself sing] Object Control: 31. a. *Henk 1 overreedde zich 1 [PRO 1 te zingen]. Henk persuaded SE [PRO to sing] b. Henk 1 overreedde zichzelf 1 [PRO 1 te zingen]. Henk persuaded himself [PRO to sing] And the reasoning can be extended to anaphors in a conjunction in ECM and Objectcontrol environments: 32 a. Max 1 expects [both Lucie and him 1 to leave early]. b. *Max 1 convinced both Lucie and him 1 [PRO to leave early]. The case in (32b) is ruled out by Condition B because there is a reflexive (semantic) predicate but no reflexive marking of the predicate. On the other hand, Condition B does not apply to non-semantic predicates and an ECM subject is not part of the semantic predicate of the higher predicate. Hence the good status of (31a). (Why is (32a) not ruled out by the Chain condition? Because the Chain condition looks only at syntactic configurations, just like Condition A.) But several threads still remain to be tied, here is one. 33. a. Max heard [himself criticize Lucie]. b. Max expects [himself to pass the exam]. In (33), the SELF anaphor reflexivizes the matrix predicate, as we already saw, but it also reflexivizes the embedded predicate. This is a problem because the embedded predicate is not reflexive. These sentences should therefore be out as a Condition A violation. 25

26 The same issue arises in Dutch: dat [ IP Max [ IP zichzelf, Lucie t i ] [ hoorde critiseren i ]] that [Max [himself Lucie t] heard criticize] But in Dutch there is V-raising: the lower verb raises and adjoins to the matrix one. In (34) the SELF anaphor is not the external argument of critiseren, but is a syntactic argument of the new complex verb. As such, it reflexive-marks it and Condition A is satisfied. This V-raising is in principle optional, but it is obligatory when the ECM subject is an anaphor bound by the higher subject. Similar movement happens at LF in English (33). Such V-raising must also apply with certain Raising cases, so that Condition A can be satisfied (in 35, the SELF anaphor reflexivizes both believe and appear; does V-raising happen in Dutch in such cases? Hedde: not really): 35 a. Lucie, believes [herself, to appear [t, to be smart]]. b. Lucie, appear i -believes [herself, to t i [t, to be smart]] Tweaks of R&R Over the years, there has been a fair amount of adaptations, tweaks and changes of the R&R framework. We will see a few over the course of the semester. Here we discuss one positive (i.e. accepting of their basic ideas) follow-up of R&R. First some background. Iatridou 1988: Clitics, Anaphors and a problem of coindexation (LI squib) Greek clitic pronouns must be free in their domain: 1. [O Kostas] j ton i/*k agapai. the Kostas CL-loves 'Kostas loves him.' The Greek anaphor is complex morphologically, and subject to the known c-command and locality requirements of SELF anaphors (not shown here). 2. [O Kostas] agapai [ton eafto tu] the Kostas loves [the self his] ' Kostas loves himself.' 26

27 In addition, Greek is a clitic doubling language. However, the anaphor can be doubled by the pronoun (Greek has no reflexive clitics). 3. a. [O Kostas] j ton k agapai [ton eafto tu j] k the Kostas CL-loves the self his 'Kostas loves himself.' b. [I jinekes] j ton k idhan [ton eafto tus j] k the women CL-m.s. saw [the self their-f.p] the women saw themselves In other clitic-doubling languages, anaphors can be doubled as well but with the reflexive clitic (Greek has no reflexive clitic): 4. Queste cose Mauro se le nega a se stesso. these things Mauro Refl Cl denies to himself 'Mauro denies himself these things.' Solution: the indexing in (3). But how does the anaphor work as a whole? That is, how does the head eafton determine that the possessive pronoun must be co-indexed with something higher? After all, this does not happen with other nouns: 5. [O Kostas] j ton k agapai [ton skilo tu j/m] k the Kostas CL-loves the dog his 'Kostas loves his dog.' not in Iatridou 1988: 6. A: Did the women see themselves? B: yes, they did. B: ne, ton idhan him-m.s. saw] Anagnostopoulou and Everaert 1999: Towards a more complete typology of anaphoric expressions (LI) Pronouns: -SELF, +R SELF anaphors: +SELF, -R SE anaphors:-self, -R Greek anaphor: +SELF, +R 27