English Object Extraposition: A Constraint-Based Approach

According to the Projection Principle (Chomsky 1981), expletives have no semantic content and thus cannot occur in theta-marked positions. However, there are many examples where expletive it appears as a direct object, in violation of the Projection Principle. The various attempts that have been made to account for such cases (e.g. the case-based analysis of Authier (1991), the predication analysis of Rothstein (1995), and the Specifier analysis of Stroik (1991, 1996)) all posit movement of the expletive from a nontheta marked position to direct object position. However, these analyses have so far been unsuccessful in capturing several important contrasts, e.g. variable optionality of the expletive it. This paper argues that such contrasts (and the complex behavior of expletive it more generally) follow straightforwardly from a lexicalist, constraint-based analysis in which lexical information and independently motivated constraints interact in subtle ways. 1 Extraposition: the Issue English allows a pattern where a finite or infinitival clause appears in sentence-final (or ‘extraposed’) position (cf. Quirk et al. 1985): (1) a. I made it my objective [to settle the matter]. b. I owe it to you [that the jury acquitted me]. This pattern involves the introduction of expletive (or ‘dummy’) it which, though morphologically identical to the third person singular pronoun, is not referential, and hence is unable to be assigned any semantic role. Expletives also exhibit distinctive syntactic properties, as noted by Postal and Pullum (1988): (2) a. For him to smoke is itself illegal. b. *It is itself illegal for him to smoke. (3) a. my observation/description of it falling b. *my observation/description of it raining (4) a. The animal was now quite large, and it was tough to prevent from escaping. b. *It was tough to prevent from becoming obvious that things were out of control. An earlier version of this paper was presented at the 41st Chicago Linguistic Society at the the University of Chicago on April 8, 2005 and the HPSG 2005 at the University of Lisbon on Aug 24th, 2005. We thank the audiences of the two conferences. In particular, we thank Frank Van Eynde, Stefan Müller, Carl Pollard, and Peter Sells for suggestions and clarifications. These contrasts illustrate the differences between anaphoric and expletive it. Unlike the anaphoric pronoun, the expletive in (2b) does not support an emphatic reflexive itself. In (3), we see that only referential it can appear in the nominalizations that are permitted in of-phrases. Finally, expletives cannot occur as the subject of a tough-predicate, as shown in (4b). According to the Projection Principle (which was proposed essentially without argument by Chomsky (1981) and has been widely assumed within mainstream generative grammar), the expletive pronoun, which has no semantic content, cannot occur in any theta-position. This entails that expletives cannot appear in strictly subcategorized positions. However, it is well known that there are overt cases where the expletive it does occur in a strictly subcategorized object position, as in (5) [Postal & Pullum 1988]: (5) a. Sometimes I find it difficult to read my own writing. b. She’s put it in their mind that it’s going to be really tough. c. I take it for granted that there will be an appeal. A number of attempts have been made to account for such cases, mainly from a transformational perspective. However, to our knowledge, none has provided a satisfactory account of the contrast that we find in examples like the following (cf. Authier 1991, Iwakura 1991, 1994): (6) Group I: I blame *(it) on you [that we can’t go]. Group II: Nobody expected (it) of you [that you could be so cruel]. Group III: John thought (?it) to himself [that we had betrayed him]. With respect to the occurrence of the expletive it in object position, there exists a clear contrast here: the expletive is obligatory in Group I, optional in Group II, and of questionable status in Group III. In this paper, we show that these contrasts, in addition to the distributional possibilities of it in object position, follow naturally from the interaction of diverse constraints in our lexicalist, constraint-based analysis. 2 Movement-Based Approaches Small Clauses. Before presenting our analyses, we briefly review the most promising of the previous approaches that have been taken regarding object extraposition. As already mentioned, Postal and Pullum (1988) provided extensive evidence supporting the claim that expletive it can appear in subcategorized positions. The only way of saving the Projection Principle then seems to be to regard the expletive it in the object position as being the subject of a small clause. The small clause analysis seems to fit cases like the following: (7) a. I believe [it to be obvious that he has lost]. b. We kept [it a secret that Jerome was insane]. However, as Postal and Pullum point out, the small clause account appears to be inconsistent with the existence of examples like (8): (8) a. They never mentioned [it [to the candidate] that the job was poorly paid]. b. We can take [it [for granted] that there will be an appeal]. The matrix PP would have to somehow descend into the embedded clause. There are additional cases where the expletive it functions as a subcategorized element of the main verb. For example, it is hard to deny that the particle out in (9a) is in construction with the main verb in Postal and Pullum’s examples like (9b): (9) a. I figured [it out in about five minutes to be impossible to solve the problem]. b. *I figured in about five minutes it out to be impossible to solve the problem. Despite this fact, as the brackettings in (9a) indicate, the small clause forces us to separate the particle from the verb. Postal and Pullum’s observations thus raise a fundamental challenge to the Projection Principle, one that has been responded to in an interesting paper by Rothstein (1995). In the next section, we review her conclusions briefly, but critically.1 Rothstein 1995. Rothstein (1995) claims that the expletive it is licensed only as subject based on the following two assumptions: (10) Predication Condition: Every syntactic predicate must be syntactically saturated. (Rothstein (1995: (15)) (11) Pleonastics are licensed only as subjects of syntactic predicates that do not assign an external theta role. (Rothstein (1995: (26)) In her analysis, a syntactic predicate is defined to be an open maximal projection that needs to be saturated by being linked to a syntactic argument, its subject. This approach thus implies that there is no pleonastic it in the object position; the pronoun it in the object position is either a subject or a referential pronoun. In examples like (12), for example, the expression following the expletive is to function as the extraposed clause’s predicate: There are two other movement-based approaches to the data in question that we are familiar with: the case-based approach of Authier (1991) and the ‘SPEC of CP’ analysis of Stroik (1991, 1996). These are also flawed in various respects, as noted in Kim and Sag forthcoming. (12) a. I consider *(it) obvious that you should have done that. b. I found *(it) stupid that Mary didn’t say anything. But it is unclear how to reconcile the predication analysis with examples where we can find no possible predicate at all or those where the object it is optional (examples from Rothstein 1995): (13) a. He regretted (it) that he was late. b. You just believed (it) that he would help. c. He never mentioned (it) to the candidate that the job was poorly paid. Rothstein assumes that the pronoun it in (13a) is an event variable bound by the CP. This in turn means that the CP here is predicated of the event object of the matrix verb. And the pronouns it in (13b) and (13c) are linked to the right-dislocated CP.2 As many have pointed out (e.g. Collins (1994) and Huddleston and Pullum (2001)), it is quite difficult to differentiate extraposition (EX) from right dislocation (RD), though some differences are apparent. The prototypical RD construction has an NP shifted outside as in (14), and to the right of the governing clause, whereas the prototypical EX has a nominal clause shifted to the right of the predicate: (14) It causes him a lot of embarrassment, his receding hairline. In addition, the pronoun it in RD has a referential function, whereas the one in EX has no referential power: (15) a. RD: It annoyed us both, having to do the calculations by hand. b. EX: It annoyed us both that we had to do the calculations by hand. Prosody can also serve to differentiate the constructions in general. A RD sentence is normally spoken with two intonational phrases – the first with a primary accent on fun and the second with deaccenting of parasailing. This contrasts with the EX rendition, where there is only one such unit containing an accent on parasailing: Within her analysis, the it + CP sequences have at least two different types of analysis: one as event quantification; the other as right dislocation. Verbs like regret, confirm, resent, and announce receive the former analysis; verbs like suspect, assume, suppose, expect, believe, and mention get the latter. The claimed differences are that only the event-quantification verbs can take gerund complements or occur with an event quantifier: (i) a. Alexander regretted that he had destroyed the city/the prize/destroying the city/the destruction of the city. b. Alexander regretted it every time I had dinner with John. (ii) a. *They suspected/assumed/expected John’s stealing the diamonds. b. *They suspected/assumed/expected/supposed it every time he told a lie. (16) a. RD: It was fun, parasailing. b. EX: It was fun parasailing. In other words, the right peripheral element in RD is intoned as an afterthought. And none of Rothstein’s examples require the RD prosodic pattern. They all allow the primary accent to be realized within the that-clause, i.e. they allow the prosodic pattern that is characteristic of EX, not RD.3 In addition, as Huddleston and Pullum (2001) point out, right-dislocated material is required to be ‘discourse old’, whereas the extraposed constituent may be ‘discourse new’: (17) a. RD:#It’s really interesting, a book I’m reading. b. EX: It now seems that there will be another price increase soon. But there are certainly examples like Rothstein’s in (13b,c) that allow indefinites introducing discourse-new referents, e.g. the following: (18) a. If you could just suppose it that there’s a REAL FIRE downstairs. b. I want you to mention it to the class that there’s a NEW KID there. Rothstein must thus analyze as RD, examples that exhibit neither the prosodic properties nor the discourse properties of RD – a highly undesirable consequence. Similarly, Rothstein’s analysis implies that if the pronoun is obligatory then there must be a predication relation. However, there are quite a few examples where the pronoun it is obligatory without there being any predication relation: (19) a. I depend upon *(it) that their paper will expose crooked politicians. b. I figured *(it) out to be more than 300 miles from here to Tuscon. Her analysis takes prepositional extraposition examples like (19a) as ‘adjunct predicate constructions’ analogous to examples like (20): (20) You can’t count on/depend on him drunk. (Rothstein 1998: (91)) However, this neglects the fact that the CP in (19a) is not an optional element, unlike drunk. Extraposed CPs like those in (19) don’t seem to share any properties with adjuncts. In addition, we can easily find examples where the presence of the object it is obligatory, although nothing is plausibly analyzed as a predicative expression: (21) a. Optimistic leaks had it that the negotiators were making good progress on a statement of “principles”. Note that RD also allows the pronoun that, which EX does not. b. I love it that you’ve asked me to go away.4 Such examples cast further doubt on Rothstein’s proposal. All things considered, an analysis that can treat all of these examples as instances of expletive it is to be preferred. 3 A Lexicalist Analysis Lexical Classes. As we have already seen, it is sometimes thought that the verbs allowing object it-extraposition form a restricted class. For example, it is clear (v. Authier 1991) that verbs that allow a choice between a clausal complement and an NP object will license object extraposition: (22) a. They didn’t even mention his latest promotion/that he was promoted recently. b. They demanded justice/that he should leave. c. He said many things/that I was not the person he was looking for. (23) a. They never mentioned it to the candidate that the job was poorly paid. b. They demand it of our employees that they wear a tie. c. He wouldn’t dare say it that I am not the right man for the job. Unlike these, it seems, at least at first blush, that propositional object verbs like hint and think, which select a single CP complement, cannot undergo extraposition: (24) a. I think *(of) you all the time. b. He hinted *many things/that I was not the person he was looking for. (25) a. I think (??it) that John had an accident. b. He wouldn’t dare hint (?it) that I am not the right man for the job. However, more careful investigation reveals many naturally occurring examples of object extraposition with such verbs, as can be seen from the following examples found on the internet: (26) a. ...because he really obviously thought it that it was somehow going to work out to his benefit.5 From the BNC www.bazima.com/archives/before/2004/12/not-only-is-she.htm [April 15, 2005] b. The Auditor would not be able to pick it up unless somebody hinted it that the account existed.6 We speculate that the true generalization is that all verbs (modulo certain qualms about verbs taking interrogative complements) that allow CP (or sentential) objects also allow object it-extraposition. To reflect such lexical patterns, we will assume, following much work in HPSG, that parts of speech come in families and can profitably be analyzed in terms of typed feature structures. The part-of-speech types we will assume form the hierarchy illustrated in (27):7