English Nonfinite Complements. The external and internal syntax of nonfinite expressions in English

Transcription

1 English Nonfinite Complements The external and internal syntax of nonfinite expressions in English

2 Syntactic processes affecting both finite and nonfinite clauses Arguments from the external and internal syntax of nonfinites 1. Infinitival and gerundial subjects 2. Extraposed infinitival subjects 3. Infinitival and gerundial objects 4. Pseudo-clefting 5. Finite and infinitival clauses conjoined 6. Subject-oriented adverbs in object control structures 7. Bound anaphora 8. Split-antecedent phenomena 9. The structure at LF Syntactic processes affecting both finite and nonfinite clauses 1

3 1. Infinitival and gerundial subjects The argument: Infinitives are either VPs or clauses by hypothesis. Clauses but not VPs may occupy the subject position in sentences. Infinitives and gerunds may occupy the subject position in sentences. Therefore, infinitives and gerunds are clauses. Evidence: As is well known, infinitives and gerunds may occur in subject position: (1) a. [To teach Elizabeth] is a pleasure. b. [Teaching Elizabeth] is a pleasure. Conclusion (supported by empirical evidence): Therefore, infinitives and gerunds are clauses. A stylistic variant of the argument: If infinitives and gerunds are either VPs or clauses by hypothesis, and if clauses but not VPs may occupy the subject position in sentences, *[ IP VP VP], and if infinitives and gerunds may occupy the subject position in sentences, then it follows that infinitives and gerunds are clauses. Infinitival and gerundial subjects 2

4 2. Extraposition of infinitival subjects It is well known that clausal subjects may (sometimes must) be extraposed. (2) a. [That the world is round] is obvious. b. It is obvious [that the world is round]. The assumption that infinitival subjects are clauses predicts that the may be extraposed. Evidence: (3) It is a pleasure [to teach Elizabeth]. As this example shows, the prediction is borne out by the facts, providing additional support (i.e. empirical evidence) for the conclusion that infinitives are clauses. Extraposition of infinitival subjects 3

5 3. Infinitival and gerundial objects The argument: Infinitives and gerunds are either VPs or clauses by hypothesis. Clauses but not VPs may occupy the object position in sentences. Infinitives and gerunds may occupy the object position in sentences. Therefore, infinitives and gerunds are clauses. Evidence: (4) Mary wants [to stay at home]. (5) John quit [smoking]. Conclusion (supported by empirical evidence): Therefore, infinitives and gerunds are clauses. A stylistic variant of the argument: If infinitives and gerunds are either VPs or clauses by hypothesis, and if clauses but not VPs may occupy the object position in sentences, *[ VP V VP], and if infinitives and gerunds may occupy the subject position in sentences, then it follows that infinitives and gerunds are clauses. Infinitival and gerundial objects 4

6 4. Pseudo-clefting Introduction: the pseudo-cleft structure (6) a. I need a long cool drink. b. What I need is a long cool drink. cleft clause background focus foreground As is well known, only constituents may occur in the focus of a pseudo-cleft. Very importantly, this restriction implies that constituents may not be split between the cleft clause and the focus. This is clear from the data below: Pseudo-clefting 5

7 (6) a. I need a long cool drink. b. What I need is [a long cool drink] c. * What I need [a] is [long cool drink] d. * What I need [a long] is [cool drink] e. * What I need [a long cool] is [drink] The expression [a long cool drink] may occupy the focus position (cf. (b)). This shows that this expression is a constituent of (6a). Fragments of this constituent in the focus (e.g., long cool drink) with the rest of the constituent (e.g., a) remaining in the cleft clause result in ungrammaticality (cf. (c-e)). The data above show that the string a long cool drink cannot be torn apart by a syntactic operation. No part of this string may be moved. What may be moved is the entire string. From this we may infer that the string is a constituent (and that substrings like long cool drink etc. of this string are not). Pseudo-clefting 6

8 The argument: If infinitives are either VPs or clauses by hypothesis, and if clauses but not VPs can be pseudo-clefted, and if infinitives can be pseudoclefted, then it follows that infinitives are clauses. (7) a. He suspected that Bill saw Monument Valley. b. What he suspected was [ CLAUSE that Bill saw Monument Valley]. c. * What he suspected that [Bill] was [ VP saw Monument Valley]. Contrast (7b) and (7c): a clause may but the VP of a clause may not occur in the focus of a pseudo-cleft. As (7c) shows, the VP of a sentence may not be separated from the subject of that sentence by pseudo-clefting. The empirical question: Do infinitives occur in the focus of pseudo-clefts? (If they do, they are clauses.) This is an important difference between VPs and clauses in general. Note that a VP is always the VP of a sentence, because VPs occur only as predicates within sentences. i.e., by moving the VP only, without its subject Pseudo-clefting 7

9 Evidence: (8) a. He wanted to visit Monument Valley. b. What he wanted was TO VISIT MONUMENT VALLEY. Conclusion (supported by empirical evidence): Therefore, infinitives like to visit Monument Valley in sentences like (8a-b) are clauses. So, the structure of (8a) is something like this: (8) c. He wanted [ CLAUSE to visit Monument Valley] Now consider (9) a. He wanted (for) Bill to see Monument Valley. b. What he wanted was FOR BILL TO SEE MONUMENT VALLEY. c. * What he wanted for Bill was TO SEE MONUMENT VALLEY. Precisely the same kind of reasoning applies to (9a-c), and exactly the same conclusion follows: Infinitives like for Bill to see Monument Valley in sentences like (9a-b) are clauses. As (8b) shows, infinitives do occur in the focus of pseudo-clefts. This raises a problem. Do you notice what? If not, bear with me. We shall return to the problem directly. (If to visit MV is a clause, where is its subject?) Pseudo-clefting 8

10 So, the structure of (9a) is something like this: (9) d. He wanted [ CLAUSE for Bill to see Monument Valley] Now compare (7a-c) and (9a-c), repeated: (7) a. He suspected that Bill saw Monument Valley. b. What he suspected was [ CLAUSE that Bill saw Monument Valley]. c. * What he suspected that Bill was [ VP saw Monument Valley]. (9) a. He wanted (for) Bill to see Monument Valley. b. What he wanted was [ CLAUSE FOR BILL TO SEE MONUMENT VALLEY]. c. * What he wanted for Bill was [ VP TO SEE MONUMENT VALLEY]. The (b) examples and the (c) examples in (7) and (9) are parallel in structure as well as grammaticality. (This, of course, shows that the (a) examples are also parallel in structure.) These facts are, again, evidence that infinitives are clauses. This conclusion raises another problem. Do you see what? If not bear with me. We shall return to it directly. (If the bracketed expression is a clause, what is for in it? If it is a P, how does a P introduce a clause? If it is not a P, what is it? Pseudo-clefting 9

11 Summary We are led to conclude that infinitives like to visit Monument Valley in sentences like (8) and infinitival expressions like for Bill to see Monument Valley in sentences like (9) are clauses. The good news We have seen how pseudo-clefting reveals the constituent structure of sentences. We have derived some nice conclusions about the syntactic category of infinitives. Some of these conclusions nicely account for the intuition (often observed but never satisfactorily explained in traditional grammar) that sentences like (7) and (9) are so strikingly similar in both form and meaning. The bad news These same conclusions raise some new problems. (Wait to see whether this bad news is indeed bad news, or maybe good news.) It often happens in research that what seems to be bad news first later turns out to be good news. Pseudo-clefting 10

12 Problems raised 1. The subject of the infinitival clause in (8) The object of want in (8a) is an apparently subjectless infinitive clause. (8) a. He wanted [ CLAUSE to visit Monument Valley]. Why is this a problem? Because of the obligatory subject condition we have adopted from TradG (cf. Lecture 1 Data and Sample of Problems), which requires that every clause/sentence have a subject. This condition is technically known as the Extended Projection Principle, EPP (cf. Chomsky 1981). What is the subject of the apparently subjectless infinitival clause in (8)? Even worse: If the infinitive in (8) has no subject, it is not a sentence. To make things even worse: the infinitive in (8) has an understood subject: the wanter (the matrix subject) and the visitor are the same person, he. This would mean that at least some infinitives are not sentences, which would seriously undermine the hypothesis that claims that all are. The only way to avoid this conclusion would be to abandon the EPP, as Huddleston and Pullum (2002) do. But that runs up against even more serious problems. For a discussion of those serious problems, see Czeglédi (2007). How do speakers of English figure that out? What is it that they know? These issues will be taken up directly. Until then, note the following Pseudo-clefting 11

13 1. If we cannot show that infinitives like to visit Monument Valley in (8) have a subject, we are in trouble: the conclusion we have just drawn must be revised. Note that this must be shown independently. 2. On the other hand, if we CAN independently show that infinitives must be assumed to have subjects anyway, then we will have found strong additional support for the conclusion just reached. The conclusion will be corroborated, because if independent evidence suggests that infinitives always have subjects, then, by the EPP, it follows that infinitives are sentences. 2. The category of for in (9) (9) a. He wanted [ CLAUSE for Bill to see Monument Valley]. What is for? P? No! A (new) Complementizer we have just discovered? Very likely! We will soon consider some facts and arguments that address the issue of apparently missing subjects of infinitives and gerunds directly. It cannot be P forming a PP with Bill, because then the infinitive would have a PP subject [for Bill], but sentences never have a PP subject. It cannot be a dangling P (not in construction with Bill in a PP) either, because then it would introduce a clause/sentence, and sentences are never introduced by a P. Clauses are never introduced by a P. what may introduce a clause is a C. Note the structural parallelism between (7) and (9). So, for is very likely a new C, a clausal subordinator (traditionally called a subordinating conjunction), introducing infinitival clauses, which we have just discovered. There is independent evidence that this for here is a C, but we cannot discuss the issue any further. Pseudo-clefting 12

14 3. The notions sentence and clause (10) a. John knows [that the world is round]. b. John knows [the world to be round]. On traditional assumptions, the complement of know in (10a) is a clause, but the bracketed expression in (10b) is not. We have been led to conclude that strings of words like the world to be round, for Bill to see Monument Valley, etc. are clauses. This conclusion is not consistent with some conditions on sentencehood assumed in TradG: tense-marking, subject verb agreement, and the nominative subject condition. These traditional assumptions are entirely unmotivated. They are easily revised, i.e., discarded. A sentence/clause may be finite or nonfinite. In traditional grammar, strings of words like the world to be round never qualify as sentences or clauses: no tense-marking, no subject-verb agreement. They are never made explicit, and there is never any motivation, explanation, argumentation offered to support them. They are arbitrary restrictions, which constitute a conceptual barrier, as we discussed in Lecture 1 Data and Sample of Problems. Some traditional grammars and grammarians may not even be aware of them. The traditional conditions above jointly define finiteness. Thus, in TradG, sentencehood implies finiteness, rendering the notion nonfinite clause unformulable, hence a conceptual barrier. But note that there is no universal principle at all that dictates this implication. There is no general principle that implies that sentences are of necessity finite. Pseudo-clefting 13

15 5. Finite clauses and infinitives conjoined The argument: If, by hypothesis, only identical categories can be coordinated, and if nonfinites can be coordinated with finite clauses, then nonfinites are clauses. Conjoined finite clauses (cf. a) and conjoined nonfinites of similar morphosyntactic types (cf. b) occur freely: (11) a. [John is writing a novel] and [he expects the world to give it critical acclaim]. b. [To write a novel] and [for the world to give it critical acclaim] is John s dream. Certain NF expressions may also be conjoined to finite clauses: (12) John expected [to write a novel] but [that it would be a critical disaster]. The second conjunct in (12) is clearly a clause. On the assumption that coordinated expressions are of the same category, it follows that the first conjunct (the infinitive) is a clause, too. The picture is not quite as clear as this, but we shall not discuss the complexities now. It still holds that (12) ought to be bad, if infinitives were not clauses. Finite clauses and infinitives conjoined 14

16 6. Subject-oriented adverbs in object control structures The argument: If clauses must have subjects (by the EPP), and if nonfinites must have subjects, then nonfinites are clauses. An adequate account of the distribution of subject-oriented adverbs forces the assumption that infinitives and gerunds always have subjects. Observation: Certain adverbs in English, e.g. intentionally, carefully, etc., are regularly construed as predicated of the subject of the sentence in which they occur: (13) John a married Mary intentionally a. (14) Fred a was willingly a shaved by Sally. Observation that raises a problem: In sentences like (15) the property expressed by a subject-oriented adverb is (equally regularly!) understood as predicated of the surface object: (15) a. John forced Bill b to hit Harry intentionally b. b. I persuaded Bill b to carefully b cut the cake. Subject-oriented adverbs in object control structures 15

17 The Problem If adverbs like intentionally are subject-oriented, how is it that they are objectoriented in sentences like (15)? Conceivable alternative solutions Alt. 1. We might say that the object-oriented interpretation of otherwise subject-oriented adverbs in sentences like (15) is exceptional. This alternative, however, must be rejected, simply because such data cannot be dismissed as exceptional, because (a) such sentences are not rare, and, more importantly, (b) they regularly have an object-oriented interpretation. Alt. 2. We might simply abandon the idea that there are subject-oriented adverbs in English, since there is plenty of evidence to the contrary (cf. the many sentences like (15), in which a subject-oriented adverb (SOA) is object-oriented ). This would not only be no solution to the problem either, but instead would introduce an additional problem, because it would leave even the regular subject-oriented cases unexplained, in addition to failing to say anything about the irregular ones. Briefly, it would create two problems out of one (by re-introducing a non-problem as a problem). Subject-oriented adverbs in object control structures 16

18 Weaknesses of some of the alternatives In addition to the weaknesses just pointed out, these alternatives share another kind of weakness : they are both weak in that the former does not take counter-evidence seriously (by attempting to dismiss it as exceptional ), and the latter does not take the initial observation (and its account) seriously (by succumbing too easily to the weight of apparent counter-evidence). Subject-oriented adverbs in object control structures 17

19 A strong alternative emerges Alt. 3. We may choose to take the bull by the horn and take the initial observation and its account very seriously: Once subject-oriented, always subject-oriented even in sentences like (15), where they are apparently object-oriented. The problems (13) [ Subject John] a married Mary intentionally a (15a) John forced [ Object Bill] b to hit Harry intentionally b? No visible subject in (15a) of which intentionally (an SOA!) is predicated. Intentionally is predicated of an invisible subject. Intentionally is apparently predicated of Bill, the object of force. Intentionally (an SOA!) cannot directly relate syntactically to the object. The SOA must relate to an invisible (phonetically empty) subject, coreferential with the object. The infinitive contains a phonetically empty subject (call it PRO), coreferential with the matrix object. The only visible subject is John, but intentionally is not predicated of that. Subject-oriented adverbs in object control structures 18

20 The apparent irregularity (in fact, an interestingly problematic regularity!) observed in (15) is explained if these examples are assumed to have the following structure: (16) a. John forced Bill 2 [PRO 2 to hit Harry intentionally]. b. I persuaded Bill 2 [PRO 2 to carefully cut the cake]. PRO is controlled by the matrix object (object control). The adverbs, in turn, are predicated of the complement Subject (PRO), and hence of its controller. Summary and conclusions An independent problem, the apparently irregular (or rather: paradoxically regular) behavior of subject-oriented adverbs, has been resolved by showing that the behavior of these adverbs is perfectly regular. By assuming that the apparently subjectless infinitives containing subject-oriented adverbs have a PRO subject, the regular behavior of the adverbs is predicted. (At least some) apparently subjectless infinitives have subjects. They are sentences. Sentences like (15) are biclausal: [Main clause [Infinitival subclause]] Thus, the assumption of a PRO subject in apparently subjectless infinitives is strongly motivated by independent considerations. Subject-oriented adverbs in object control structures 19

21 Terminology: Control: The process of determining the coreferentiality (the antecedent) of PRO. Subject control: PRO is coreferential with (= is controlled by) the matrix subject. Object control: PRO is coreferential with (= is controlled by) the matrix object. Subject-oriented adverbs in object control structures 20

22 7. Bound anaphora The argument: General principles of Chomsky s (1981) Government and Binding theory anaphors and their antecedents must be clause mates. Anaphors in apparently subjectless NF complements must have subject antecedents. (Sentences must have subjects.) NF complements are sentences. Anaphors English reflexives and reciprocals (himself, each other, etc.) are anaphors, because they always occur with a coreferential antecedent. (17) Tom i shaved himself i. (18) * Each other are happy. An anaphor and its antecedent must be clause mates (Binding-theoretic principle) (19) a. Mary thought [that Tom i shaved himself i ]. b. * Mary i thought [that Tom shaved herself i ]. (19b) is ungrammatical because the anaphor herself and its antecedent Mary are not constituents of the same clause, in violation of the principle. Bound anaphora 21

23 A problem (20) a. John i said [it was difficult to shave himself i ]. b. Mary i said [that shaving herself i was a pain in the neck]. c. Helping oneself i would be difficult. If these sentences have the structure as indicated, BT predicts that they are ungrammatical, because they violate the structural restriction on anaphors and their antecedents: In (20a and b), the antecedent is outside the binding domain of the anaphor (the clause that contains the anaphor) it binds. (20c) contains no antecedent at all that binds the anaphor oneself. Assuming BT, oneself in (20c) is bound by an invisible (phonetically empty) antecedent; the anaphors in (20a) and in (20b) are also bound by phonetically empty antecedents, i.e., the sentences cannot have the structure indicated. We must assume a phonetically empty category (the antecedent of the anaphors) in (20a-c). But these sentences are perfectly acceptable. Therefore a grammatical theory must characterize them as grammatical, or else the theory does not meet the requirement of observational adequacy. Why? Because the matrix subject cannot bind the anaphors in the subclause, since it is outside their binding domain, the subclause. Bound anaphora 22

24 Solution Assume that (20a c) have the following structure: (21) a. John 2 said [it was difficult [PRO 2 to shave himself 2 ]]. b. Mary 2 said [that [PRO 2 shaving herself 2 ] was a pain in the neck]. c. [PRO 2 helping oneself 2 ] would be difficult. On these assumptions, the problems noted above all disappear. The structural representations in (21a c) no longer violate BT. In (21a and b) the anaphor and its antecedent (the PRO subject of the infinitive and gerund) are clause mates. The anaphor in (21c) has an antecedent (PRO) in the required position (within the binding domain of the anaphor). Each anaphor is bound by a coreferential antecedent, PRO, which in turn is controlled by a coreferential NP in the highest clause. The only exception is (21c), where PRO is not controlled. This is correct too. This explains how the understood agent of the gerund is an arbitrary person. Such a PRO is known as arbitrary PRO, PRO ARB. Thus, the meaning of the sentences is accounted for just as well as before, or even better. This chain of coreference relations explains how the reflexives eventually (indirectly) co-refer to a noun phrase in the highest clause. Bound anaphora 23

25 Conclusion An adequate account of the data and the need to solve the problems that the data raised on Binding-theoretic assumptions required that we assume a PRO subject in infinitives and gerunds. In other words, we found we MUST assume a phonetically empty PRO subject in apparently subjectless infinitives and gerunds. It seems, then, that all infinitives and gerunds have subjects. If infinitives and gerunds have subjects, and if sentences but not VPs must have subjects by the Extended Projection Principle (EPP), then it follows that infinitives and gerunds are sentences. Final descriptive remarks As it turns out, sentences like (20a and b) are not biclausal, as traditionally assumed, and as is represented in (20a and b), but they contain three clauses, as represented in (21a and b). And (20c) is not monoclausal, but biclausal, as represented in (21c). Bound anaphora 24

26 Terminology: Binding: Binding is a coreference relation between noun phrases. A DP a is said to bind (i.e. be co-referential with) another DP b (and then DP b will be said to be bound by DP a ) iff DP a c-commands DP b and they are co-indexed. C-command (constituent-command): A category A c-commands another category B if neither of them dominates the other and the first maximal projection that dominates A also dominates B. Dominance: The relation between a category and its constituents. A category dominates all of its constituents. Binding domain (minimal governing category): A binding domain (minimal governing category) is the minimal structure (DP or IP) within which the relationships of binding obtain. The issue whether at least some noun phrases are DPs has no bearing on the present discussion or its conclusions. Bound anaphora 25

27 8. Split-antecedent phenomena They and each other The pronominal expressions they and each other share two features: (a) they are pronominal, and therefore may enter into coreference relations with other noun phrases, and (b) they are both notionally plural. An important difference: the pronoun they may have split antecedents but the anaphor each other requires a unitary antecedent: (22) a. JOHN i told MARY j that THEY i+j had to leave. b. [JOHN AND MARY] i like EACH OTHER i. c. * JOHN i talked with MARY j about each other i+j. Terminology Split vs. unitary antecedents: They may have split antecedents in (22a): they is understood as coreferential with the union of two separate NPs John and Mary. John and Mary are split: They are different constituents (cf. John: subject NP, Mary: indirect object NP). [John and Mary] in (22b) is a constituent (two NPs conjoined under one category, a third composite NP), hence a unitary antecedent of each other. Split-antecedent phenomena 26

28 The verb propose The verb propose allows a split-control interpretation: (23) John proposed to Mary to go to the movies. On the most natural reading, the understood subject of the infinitive in (23) is John and Mary, i.e., the reference of this understood subject is jointly determined (technically: controlled) by the NP John and the NP Mary. This is made very clear by the synonymy between (23) and (24) John i proposed to Mary j that they i+j go to the movies. Both the understood subject of the infinitive in (23) and the subject they of the complement sentence in (24) are controlled by the same split antecedents (split control). Split-antecedent phenomena 27

29 Now consider (25) John proposed to Mary to help each other. Two simple observations first: 1. (25) is grammatical. 2. Part of the meaning of (25) is the idea that John and Mary help each other, i.e., each other is coreferential with two distinct NPs. To continue: a problem 3. The anaphor each other must have a unitary antecedent in (25). (This follows (a) from the first observation above that (25) is grammatical and (b) from the requirement which each other imposes on its antecedent, i.e., that it be unitary.) 4. John and Mary cannot be the direct split antecedents for each other (because it requires a unitary antecedent). 5. each other is bound by a phonetically empty (unitary) antecedent. 6. Finally: striking parallelism between the problematic (25) and (26) John i proposed to Mary j [that i+j they 2 help each other 2 ]. It is unproblematic, because the anaphor each other is bound by a unitary clause-mate antecedent, they, as required. Split-antecedent phenomena 28

30 The solution Let us assume that (25) has the following structure (which parallels (26)): (25) John i proposed to Mary j [ i+j PRO 2 to help each other 2 ]. The PRO subject of the infinitive is the unitary antecedent of each other, as required (cf. coindexing by the subscript 2 ). The PRO subject of the infinitive is controlled by two split antecedents in the matrix clause, as is made possible by the verb propose, and thus required for an adequate account of the interpretation of the sentence (cf. coindexing by subscripts i and j ). Conclusion These assumptions are apparently necessary for a satisfactory account of the structure and interpretation of sentences like (25) (and its non-accidental parallelism with (26)). Because one of the necessary assumptions is that the infinitive has a PRO subject, it follows that Infinitives are clauses. Split-antecedent phenomena 29

31 To summarize the main points of the argument: The anaphor each other must have an antecedent in (25). John and Mary cannot be the direct split antecedents for each other. It must be assumed that the antecedent of each other is the PRO subject of the infinitive. Apparently subjectless infinitives must be assumed to have a subject. Infinitives are clauses. Split-antecedent phenomena 30

32 9. The structure at LF The argument: All NF complements will have subjects (arguments) and predicates in any decent semantic representation (and each predicate will have exactly one subject argument). The easier it is for the semantics to match up each predicate with its subject, the better (because it simplifies the theory). It is easiest if each semantic subject is represented in the syntax as a syntactic subject. Apparently subjectless nonfinites must have (phonetically empty) subjects. Nonfinites are sentences. A syntactic theory that reflects the semantic structure of NF complements (and other constructions) more transparently is superior to less transparent theories ( the problem of the syntax semantics interface: the simpler the interface, the better). The structure at LF 31

33 One of the properties of the clausal analysis is that there is a one-to-one correspondence between subject and predicates in non-coordinate structures. That is to say, in the clausal analysis there is no predicate without a corresponding subject, every verb has a subject. Thus each verb in (27) has a corresponding subject: (27) John 2 wants [PRO 2 to try [PRO 2 to date Mary]]. Moreover, when there is no overt subject, there is always a (phonetically null) subject to preserve the pairing of subjects and verbs: (28) [PRO to leave now] is impossible for John. Under the nonfinite-vp (verbal-phrase) analysis the single subject in (27) would be related to three different verbs, and the verb in (28) would not be related to any subject at all. That is, not on the syntactic level of representation. The structure at LF 32

34 References Chomsky, Noam Lectures on Government and Binding. Dordrecht: Foris. É. Kiss Katalin Configurationality in Hungarian. Budapest: Akadémiai Kiadó. Freidin, Robert Foundations of Generative Syntax. Current Studies in Linguistics Series 21. Cambridge, Mass: MIT. Greenbaum, S The Treatment of Clause and Sentence in A Grammar of Contemporary English. In: S. Greenbaum et al. (eds.) (1980) Studies in English Linguistics. London: Longman. Kiefer Ferenc (ed.) Strukturális magyar nyelvtan. Budapest: Akadémiai Kiadó. Koster, Jan and Robert May On the Constituency of Infinitives. Language 58, Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech, and Jan Svartvik A Comprehensive Grammar of the English Language. London: Longman. References 33

35 Huddleston, Rodney, and Geoffrey K. Pullum The Cambridge Grammar of the English Language. Cambridge: CUP. Baltin, Mark R Floating Quantifiers, PRO, and Predication. Linguistic Inquiry 26: Czeglédi Csaba Issues in the Syntax and Semantics of Infinitives and Gerunds in English. Eger: Lyceum Kiadó. References 34