Control, PRO, and the Projection Principle

CONTROL, PRO, AND THE PROJECTION PRINCIPLE Peter W. CulicoverWendy Wilkins University ofArizonaUniversity of Washington This paper presents a theory of control (predication), in terms of thematic relations, which makes no use of the element pro in the syntax. Important consequences of the theory are that the T-Criterion must be relativized to particular local domains, and the Projection Principle cannot be maintained. A number of syntactic arguments against pro are summarized, and the arguments of Koster & May 1982 in favor ofpro are addressed. It is concluded that, given a thematic relation-based account ofpredication, the Projection Principle in its current form is not a useful postulate in the theory of grammar.* 1. In previous work (Culicover & Wilkins 1984, henceforth LLT), we have assumed that infinitives in general are not derived from S' complements. This means that we question the existence ofthe abstract empty NP, usually referred to as pro, which is assumed in much current work to be the syntactic subject of embedded infinitival complements. The issue of the existence of pro as a syntactic element is of great importance, given its central role in the theory of Government and Binding (Chomsky 1981, and much other work). Pro is necessary to avoid violations of the Projection Principle (hereafter PrP), which states that: 'Representations at each syntactic level (i.e. L[ogical] F[orm] and D- and S-structure) are projected from the lexicon, in that they observe the subcategorization properties of lexical items' (Chomsky 1981:29). In particular, this means that a verb which requires a propositional complement in LF would require a sentential complement at D- and S-structure. Where such a verb apparently occurs in the syntax with a bare infinitive, the PrP requires that the infinitive be analysed as a full S. This untensed S, with no overt subject, would have pro as its subject—at least in English (and similar languages) where the subject is not optional in the expansion of S. The PrP is stated formally as follows (Chomsky 1981:38): (i) If ß is an immediate constituent of ? in [7 ... a ... ß ...] or [7 ... ß ... a ...] at L¡, and y = ?, then a ?-marks ß in y. (ii) If a selects ß in ? as a lexical property, then a selects ß in y at L1. (iii) If a selects ß in 7 at L1-, then a selects ß in ? at L7. In our theory we do not assume the PrP; specifically, we take issue with statements (ii)-(iii). This means that not all thematic information—the ?-marking of (i)—has an overt syntactic representation in terms of distinct categories at each syntactic level. In other words, subcategorization requirements can be satisfied without necessarily presupposing that the logical/semantic requirements of a verb have a one-to-one correspondence with the syntactic categories in syntactic structure. Because of the theory of coindexing which we present here, a given NP may bear a thematic role with respect to more than a single * This work was funded in part by grants from the National Science Foundation and the Sloan Foundation. We gratefully acknowledge their support. We would like to thank Joe Emonds, Ann Farmer, Eloise Jelinek, Chisato Kitagawa, Fritz Newmeyer, Richard Oehrle, and Geoffrey Pullum for their very helpful comments. The authors' names appear in alphabetical order. 120 CONTROL, PRO, AND THE PROJECTION PRINCIPLE121 verbal (or relational) element. Then, because the PrP is not assumed, there is no reason why the mapping from syntactic to semantic structue cannot introduce arguments, or rather representations of arguments, as under conditions of predication. The advantage of our approach over one which includes the PrP is that the non-syntactic nature of pro is immediately explained. That is, the apparent inconsequentiality of pro for many syntactic phenomena ceases to require explanation . Pro is a logical element, not a syntactic one. Koster & May 1982 claim to demonstrate conclusively the theoretical advantages of assuming that infinitival complements contain a syntactic pro subject , and are sentential. Our opposing argument begins, in §2, by presenting our theory of predication, which makes the no-PRO theory interesting. In §3, we discuss some arguments against the syntactic element pro. In §4, we summarize our response to Koster & May's arguments. Our general conclusion (§5) is that little syntactic evidence exists...