Non-Verbal Predication and Head Movement

Dissertations Glot International, Volume 3, Issue 8, October 1998 11

(6) NON-VERBAL PREDICATION AND a.2 b. HEAD MOVEMENT COP(NP1, NP2) NP (NP) 1 1

In equatives then, it is the abstract predicate COP, not the nominal predicate, that undergoes by Andrew Carnie head movement. The COP morpheme is realized phonologically with the subject agreement fea- tures of the INFL head, in the form of a pronomi- reviewed by Peter Svenonius nal element (é/í/iad). Both nominals appear in argument positions. This is seen in (7):

Summary (2) (7) by the author CP CP

1. Introduction C IP One of the goals of the recent Chomskyan C IP Is minimalist movement is the simplification (often INFL COPP on conceptual grounds) of the mechanisms found in syntactic theory. In this thesis, I further one INFL VP such reduction and attempt to justify it with subj COP’ empirical evidence. Primarily using evidence from copular constructions in Modern Irish, I argue for subj COP attribute an underdetermined theory of phrase structure V obj where a p-marker’s behavior in the syntax deter- In (7), the COP predicate bears inflectional fea- mines its X-bar status rather than the X-bar tures which it checks by head moving through the One of the most remarkable features of Modern stipulations driving the p-marker’s behavior. functional heads to the highest position. The Irish is the fact that this predicate-subject order, Since the advent of generative grammar, the arguments move to their case positions, in a an order involving head movement in its deriva- notions of phrase and head have been viewed as manner parallel to normal VSO order. This, then, tion, is also found with non-verbal predicates. primitives by many syntacticians (see for exam- derives the two basic word orders of Irish copular Consider the following copular construction found ple, Chomsky 1957). More recently some authors clauses. A summary of clause types is given in (8). have claimed that they can be derived from other with nominal, individual level predicates (see Doherty 1996 for a discussion of the distribution structural relations (such as terminality (i.e. a (8) head) or being dominated by an element that is of this and other Irish copular constructions): Comp Infl Spec,VP VP,comp not a projection of the head (i.e. a phrase) (see for (3) ()PARTICLE ()PREDICATE ()SUBJECT ()OBJECT/COMP example Speas 1990, Chametzky 1996, Chomsky Is dochtúir (é) Seán 1994, 1995). Under both of these conceptions of C doctor agr John Ní fhaca Seán an dochtúir Verb phrasality, however, the standard assumption is ‘John is a doctor’ NEG saw John the doctor that whether a phrase marker (henceforth p- Ní dochtúir Seán Indef N marker) is a head or a phrase determines its In this construction, which I call the predicative NEG doctor John behavior with respect to the rest of the syntax. By copular construction, the non-verbal predicate contrast, I claim the “phrasality” or “headness” of dochtúir appears between the complementizer is Ní hé Seán an dochtúir Def NP a phrase marker is determined solely by the and the agreement morpheme é. In chapters 4 and NEG COP+ AGR John the doctor function and behavior of that p-marker. “Phrases” 5 of this thesis (and in later work like Carnie and “heads” in this conception are thus simply forthcoming), following a suggestion in Collberg In chapter 7, I consider alternative analyses to artifacts of the behavior of the p-markers in- (1990) and related work on Breton by Hendrick word order alternations in copular constructions, volved. What limits the behavior of p-markers are (1994), I analyze this as the head movement of the including the unified be analysis of Heggie (1988) other properties of the human language computa- non-verbal predicate to the same position as and others, DeGraff’s (1992) theory based on tional system (such as the interface with morphol- verbal predicates in VSO sentences: resumptive pronouns, and Doherty’s (1996) analy- ogy/phonology and the interface with the semantic sis and show that they are inadequate to the task component), instead of a structural definition or (4) CP of dealing with all the facts of the equative/predi- stipulation of the p-markers’ status as a phrase or cative alternation of Irish. head. The head-movement of non-verbal predicates is IP forms the basis for my claim that phrase structure 2. The facts and assumptions underlying is underdetermined with respect to phrasality. the argument INFL SC (= Small Clause) Modern Irish is a VSO language as seen in (1): 3.1. The mysterious phrasal predicates subj NP The analysis sketched above runs into prob- (1) lems when it comes to complex nominal predicates Leanann an t-ainmní an briathar i nGaeilge like that in (9). The whole predicate appears in follow.PRES the subject the verb in Irish attribute ‘The subject follows the verb in Irish’ the position associated with the head-moved element. Since head-movement is, by definition, Following McCloskey (1983) among many others, This construction can be contrasted with the one I the movement of heads, not of phrasal categories, I assume that this order is derived from an under- call an equative copular construction seen in (5): it seems unusual to claim such movement is lying SVO order. In chapters 2 and 3 of this the- possible for what appear to be phrases. sis, I argue for a particular analysis of VSO order (5) involving the raising of the verb to the highest Is é Seán an dochtúir (9) inflectional head around the subject, following C agr John the doctor a. Is [dochtúir capall] é ‘John is the doctor’ COMP [doctor horses.GEN him Sproat (1985) among others. This analysis in- ‘He is a doctor of horses’ volves a split VP, a flipped TP/AgrS structure and In the equative construction, where two definite b. Is [amhrán aL bhuailfidh an píobaire] “Yellow COMP [song COMP play.FUT the bagpiper a VP-embedded AgrO/Asp functional structure. It or referring NPs are equated, neither NP appears accounts for a wide variety of facts about infini- Submarine” in the privileged head-moved position between the ‘‘Yellow Submarine’ is a song which the bagpiper is going tives, aspectual clauses and EPP effects, and is complementizer and the agreement morpheme. to play’ outlined more thoroughly in Carnie and Harley Instead, both NPs appear to the right of the (1997) and in forthcoming work by Carnie and agreement morpheme. I claim that the difference At first glance these sentences would appear to Harley. What is important to the argument here, between the equative and predicative construc- argue against a head movement approach to non- however, is simply that the verb moves to a posi- tions reduces to the controversial difference in verbal predicates in Irish. Under standard as- tion lower than the highest complementizer argument structure. Equative constructions sumptions, the phrasal predicate must, if it is in a particle and higher than agreement morphology involve an abstract equative predicate (COP) derived position, appear in a specifier (see Doher- and the subject, and for the purposes of this brief which takes two arguments (6a). Predicative ty 1997 for one such analysis). There is evidence, summary I abbreviate this derivation as in (2). constructions by contrast involve a single argu- however, that these elements are not in specifiers, ment, with the other non-verbal predicate func- nor are they in actuality XPs. Instead I claim that tioning predicatively (6b). these are p-markers whose phrasality is underde- Dissertations Glot International, Volume 3, Issue 8, October 1998 12 termined, behaving outwardly phrasal with (13) pose that the ability to bear tense and agreement respect to the morphology, but X°-like with re- C + Infl Spec,VP Comp, VP R-adj features is a property only associated with ele- spect to several syntactic tests. ments that undergo head to head-movement (this Ní fhaca Seán an teangeolaí inné notion will be articulated more precisely below), NEG 3.2. Evidence from wh-extraction saw John the doctor today whereas the ability to bear case features is a One piece of evidence in favor of the X°-like property associated with element that undergo XP status of complex indefinite nominal predicates Given that I have claimed predicates in copular movement (A or A-bar). Notice that the relevant comes from wh-extraction. The argument is as clauses are in Infl, then when elision occurs, then criterion for what is an XP and what is an X° here follows. If predicates have undergone head move- the predicate should remain. At least for the is how they behave, both with respect to bearing ment like X°s, then subcomponents of these adjectival and prepositional predicates which features and with respect to movement. predicates should not be able to extract via wh- appear in this construction, this is true (see I suggest, partially following Chomsky, that movement. Before proceeding to the actual test, it (14), (15)). the following are some possible criteria for the is worth noting that an account of such violations XP-ness or X°-ness of a p-marker. Recall that a in terms of island effects and subjacency is unten- (14) p-marker can be both an XP and an X° at the able in Irish, however, as Irish does consistently Q: An le Seán an Subaru? A: Is leis ‘Yes’ same time, so it is not the case that any one of the Q with John the Subaru COMP INFL allow subjacency/ECP type violations (McCloskey ‘Does John own the Subaru?’ COMP with.him following properties are necessarily the definition 1979). If the speaker leaves a resumptive pronoun of an XP or an X°. Rather, a p-marker can have at the extraction site and changes the highest (15) any number of properties of both X°s and XPs and complementizer from aL to aN, then a sentence Q: An ceart mo chuimhne A: Is ceart ‘Yes’ thus behave accordingly. We now have a straight- with a subjacency violation is rendered grammati- Q right my memory COMP INFL forward account of why a “phrase-like” element in ‘Is my memory is right?’ COMP right (from Doherty 1996) cal (see McCloskey 1979 for more details). Given Irish appears in a position associated with X°: its that such extraction is licit, we can use wh-extrac- phrasal status is underdetermined. Syntactically In sentences with referential NP attributes, tion as a test for the X°-like status of a nominal, this element behaves like an X°, but phonological- similar behavior also occurs. Recall that in the in contrast to the situation found in English. If ly, morphologically, and p-marker internally it analysis sketched above, referential NP attributes wh-extraction is licit, then the sequence of mor- behaves phrasal. are not X°s in an functional projection, rather, phemes is behaving like a fully phrasal p-marker; A more important question remains, however. they are the argument of an abstract COP predi- if wh-extraction is illicit, then the sequence is Why such behavior is typologically rare and a cate. Thus in sentences with definite or referen- behaving like an X°. This distribution is exactly restricted phenomenon? Why is it the case that, in tial NPs, we expect only the pronominal what we find with nominal predicates. An indefi- the vast majority of cases, such as French, or even agreement realization of the abstract COP predi- nite, nonreferential NP predicate like that in (10) Irish verbal predicates, this mismatch between cate to remain after ellipsis. This prediction is does not allow extraction, despite the fact that phrasality and X°ness is not found? We must now also true (16). Irish normally allows extraction out of nominal account for the fact that most of the time, there is islands. This is consistent with the idea that these (16) a strict alignment of phrasality, where a p-marker are really functioning as X°s. Q: An é an feirmeoir Liam? A: Is é functions consistently as either X° or an XP, but Q.COMP INFL the farmer Liam COMP INFL usually not both. One possible answer is that such (10) ‘Is Liam the farmer?’ ‘Yes’ mismatches only apply when the syntax is forced a. Is [ amhrán [ aL bhuailfidh an píobaire t ]](é) NP i CP i by morphological reasons to head-move elements COMP [ song COMP play.FUT the piper The situation is more complex with indefinite that aren’t single morphological units. The reason “Yellow Submarine” nonreferential nominal predicates (17) which I AGR that complex predicates are not allowed to adjoin ‘‘Yellow Submarine’ is a song which the bagpiper is going argue appear in Infl. In these cases the predicate to heads in languages like English follows from to play’ does not surface, but is replaced by the dummy the fact that they are not allowed to bear tense L b. * Cén píobairej arb [NP amhráni [CPa bhuailfeadh pronominal ea: which piper WH-COMP song COMP play.COND and agreement features. Adjoining a complex sé t ]](é) “Yellow Sub” predicate p-marker to an inflectional head would j i (17) he AGR a. An dochtúir Seosamh? b. *Is dochtúir cause the derivation to crash, since none of the * ‘Which bagpiper is ‘Yellow Submarine’ a song which he/t i Q doctor Joseph ✓Is ea appropriate features could be checked. The is going to play’ ‘Is Seosamh a doctor?’ phrase/head status of the element is thus deter- mined by its behavior with respect to the rest of This can be strikingly contrasted with the definite This may well be similar to ‘do support’. This the computational system. Irish, on the other NP attributes, which are not predicates and do dummy pronominal shows up when there is a hand, is special, since it allows complex predicates not undergo X° movement. In these sentences wh- nonreferential indefinite predicate. What is to bear tense and agreement features, which are extraction from the definite NP is licit. crucial here is that the element appearing in the then forced to check their features in a head–head Infl head is retained (via the pro-form ea) in (11) relation. The narrowness of the phenomenon is responsives, supporting the analysis that these a. Is é “Yellow Submarine”[NP an t-amhráni thus derived from the fact that such mismatches COMP AGR the song complex nominal predicates are part of Infl. will only show up with morphological irregulari- [ aL bhuailfidh an píobaire t ]] CP i ties, such as atypical feature association. In the COMP play.FUT the piper 4. An underdetermined theory of phrase ‘‘Yellow Submarine’ is the song which the bagpiper is thesis, other similar cases from Tagalog, Persian going to play’ structure and other languages are seen to exhibit similar b. Cén píobairej arb é ‘Yellow Submarine’ [NP an I propose simply that when an element is behavior under similar circumstances. Which piper WH-COMP AGR the behaving like an X°, it is treated by at least some t-amhrán [ aL bhuailfeadh sé t ]] i CP j i parts of the grammar as an X°. With respect to 5. Summary song [COMP play.COND he ‘Which bagpiper is ‘Yellow Submarine’ the song which other components of the grammar that same p- The theory I suggested in this thesis is that he/ti is going to play’ marker may appear phrasal. Complex nominal phrasal status not be stipulated or structurally predicates in Irish are treated by the grammar derived, but rather is merely epiphenomenal, with To summarize, wh-extraction is generally allowed like X°s and are allowed to undergo head-move- other output constraints (such as constraints on from phrases of all types throughout the grammar ment to adjoin to functional categories just like the configuration of feature checking) resulting in of Irish. However, extraction from NPs that ap- verbs: the surface appearance of such artifacts. pear in initial (predicate) position is disallowed. (18) References 3.3. Evidence from the responsive system the = Xo Carnie, Andrew (forthcoming). Copular Construc- There is some further evidence that these tions in Modern Irish. Cambridge MA: Beech complex predicates are behaving like X°s. This the man Stave Press. evidence comes from the responsive system. Irish Carnie, Andrew and Heidi Harley (1997). The has no words for yes or no; instead, the verb is EPP and the licensing of PRO in a verb Initial man who repeated in either the positive or negative form, language. Paper presented at LSA, 1997, as seen in (12) (where the negative form is indi- Chicago cated by an adjoined negative complementizer): who Saw Carnie, Andrew and Heidi Harley (in prep). Argu- ment licensing in a Verb Initial Languages. (12) (simplified tree) t Saw Ms. Harvard University. a. An bhfaca tú an teangeolaí? b. Ní fhaca OR c. Chonaic Chametzky, Robert A. (1996). A Theory of phrase Q saw you the linguist NEG saw saw ‘Did you see the linguist?’ ‘No’ ‘Yes’ saw him markers and the extended base. Albany: State University of New York Press. This can be analyzed as the elision of everything What mechanisms determine whether a p-marker Chomsky, Noam (1957). Syntactic Structures. to the right of the V+Infl complex in a manner is an X° or an XP? I propose that the notions X° Janua Linguarum 4. Mouton, The Hague. familiar from VP ellipsis (see McCloskey 1991 for and XP are simply artifacts of the behavior of the Chomsky, Noam (1994). Bare Phrase Structure. more discussion). For example, the shaded parts p-marker with respect to other components of the MIT Occasional Paper in Linguistics 5. of the sentence schematized in (13). computational system. For example, let us pro- Chomsky, Noam (1995). Bare Phrase Structure. In Gert Webelhuth, ed., The Principles and Dissertations Glot International, Volume 3, Issue 8, October 1998 13

Parameters Approach to Syntactic Theory: A Irish example sentences and attentively discusses tion that he has relegated the head-phrase dis- Synopsis, Basil Blackwell, 385–439. [Revised previous work on Irish. He notes problems for tinction, as I will discuss below. version of Chomsky 1994.] Doherty’s (1996) analysis of Irish copular con- What, then, is the head-phrase distinction, if Collberg, Sheila Dooley (1990). An Expanded structions, which locates the subject in an under- it is not what is shown in Table 1? In theories INFL syntax for Modern Irish. Lund Univer- motivated rightward specifier. Carnie provides an without head movement, like those of Pollard & sity Working Papers 36, 1–17. alternative which does not make use of a right Sag (1994) or Brody (1997), it is possible to argue DeGraff, Michel (1992). The Syntax of Predication specifier, involving movement of the predicate to that it is a function of the lexicon and the morpho- in Haitian. Proceedings of NELS 22, 103–117. the left. logical component: what is listed in the lexicon or Distributed by GLSA. For several aspects of Irish syntax, Carnie produced by the morphological component is an Doherty, Cathal (1997). Predicate initial construc- argues persuasively for a specific analysis within X0, and what is constructed in the syntax is an XP tions in Irish. Proceedings of West Coast the framework assumed, approximately that of (cf. Chomsky 1994 for a weaker version of this). Conference on Formal Linguistics 15, 81–95. Chomsky (1993). For example, his arguments that This is essentially expressed in the fifth and last Stanford: CSLI. the copular element is is located in C rather than distinction in Carnie’s table. However, this posi- Doherty, Cathal. (1996). Clausal structure and the in Infl seem cogent. But when he deviates from tion is not a widely held one in the theoretical Modern Irish copula. Natural Language and that framework, proposing theoretical modifica- space in which Carnie moves. There, it is general- Linguistic Theory. 14, 1–46. tions, he is less careful to motivate the innova- ly assumed that head movement constructs X0 Hendrick, Randall (1994). The Brythonic Celtic tions or explore their consequences. My focus in complexes that correspond to a single morphologi- Copula and Head Raising. In: D. Lightfoot this review will be on the broader implications of cal item (e.g. Halle & Marantz 1993, which Carnie and N. Hornstein (eds.), Verb Movement. Carnie’s main theoretical proposal, a revision of adopts, p. 205). Cambridge University Press, 163–188. the distinction between head and phrase. It is an I suspect that the most common understand- Heggie, Lorie (1988). The Syntax of Copular interesting and potentially important proposal, ing of the head-phrase dichotomy is as a primitive Structures. Doctoral Dissertation, University and although ultimately I suggest that it cannot distinction of X-bar theory: a head projects up to a of Southern California. be maintained on the basis of the evidence provid- phrase (cf. Muysken 1982; Bloomfield’s use of the McCloskey, James (1979). Transformational ed, it raises serious and useful questions. word head was broader in that each Xn would be Syntax and Model Theoretic Semantics: A the head of the Xm containing it). The node that Case Study in Modern Irish. Reidel Publica- 1. Properties of heads and of phrases does not project any further is the XP (XMAX). The tions, Dordrecht. The most radical theoretical contribution of node that is not projected is the X0. Intermediate McCloskey, James (1983). A VP in a VSO lan- the work at hand is the claim that a phrasal nodes (X’) are generally taken to have some kind guage. In: G. Gazdar, G. Pullum, and I. Sag element can move to a head position. Carnie of defective status (cf. Chomsky 1994). Whether a eds. Order Concord and Constituency. Foris, suggests (pp. 184–191) that this fits naturally more general theory of Structure Preservation Dordrecht, 9–55. with the theory of phrase structure outlined in (Emonds 1976) or of Uniformity of Chains (Chom- McCloskey, James (1991). Clause Structure, Chomsky (1994). Carnie points out that the sky 1994) is invoked or not, it is furthermore Ellipsis and Proper Government in Irish. The phrase-head distinction has to be stipulated in generally assumed that only phrases adjoin to syntax of Verb Initial Languages. Lingua that framework: Chomsky suggests that a head is phrases and only heads adjoin to heads (this is Special Edition 85, 259–302. a terminal element, but as Carnie notes, complex- most explicitly derived from more basic assump- Speas, Margaret (1990). Phrase Structure in es formed by head-to-head adjunction would then tions in Kayne’s 1994 theory); this is enough in Natural Language. Kluwer, Dordrecht. not count as heads. Carnie argues instead that some frameworks to ensure that only phrases Sproat, Richard (1985). Welsh Syntax and VSO the head-phrase distinction is not primitive, but is appear in Specifier positions (for example in Structure. Natural Language and Linguistic determined by the “behavior” of a p-marker (e.g. Hoekstra 1991 or Kayne 1994). These various Theory 3, 173–216. p. 185). It is not always entirely clear what Carnie commonly held beliefs are rarely made explicit means by this. On p. 202 he lists five features and even more rarely justified (cf. Kornai & which are supposed to be characteristic of XPs Pullum 1990), and Carnie is therefore right to Review and X0s, in Table 1, attributing them to Chomsky. question them. by Peter Svenonius Carnie suggests that “a p-marker can have any number of properties of both X0s and XPs and 2. The Irish copular construction In a typology of works in linguistics, one impor- thus behave accordingly” (ibid.), apparently Carnie follows Chung & McCloskey (1987) tant division would be between the sort that take meaning that a p-marker can have some proper- and McCloskey (1996), inter alia, in taking VSO a theoretical device (often motivated by a broad ties from the left of Table 1, and some from the order in Irish to be derived by movement of the range of empirical evidence) as the starting point right, in any combination. The case of the Irish finite verb to Infl (more specifically, Carnie as- and then develop it against a range of linguistic copular construction would be a case in which the sumes that V moves to T1, the highest node in the data, and the sort that take a construction type or non-verbal predicate bears Tense and/or agree- Infl system, above AgrS; pp. 110–112). He extends set of construction types as the starting point ment features, according to Carnie, thereby that analysis to copular constructions like that in (assuming some basic theoretical framework, of undergoing head movement, while being the (1a), where the non-verbal small clause comple- course) and then derive theoretical claims based output of syntactic processes like Merge, rather ment to Infl is labeled SC, and the trace of the on that evidence. This dissertation belongs firmly than of the morphological component. nominal predicate is labeled tNP; the copular in the latter category. In fact, the abstraction of However, many of the assumptions apparent particle is is glossed ‘COP’. the title belies the earthly nature of the content: in Table 1 are subject to challenge. For example, (1) the dissertation is really primarily about Irish the link between Tense and Agreement features and heads is unclear. DPs are widely assumed to a. Is [IP [I’ dochtúir [SC Seán tNP ]]] clausal syntax. Forays into other languages are COP doctor John but brief, and although claims are made about bear Agreement features, and assuming some sort ‘John is a doctor’ of head feature percolation, TPs at least must syntactic theory, the focus is consistently on how b. [IP [I’ Is [NP dochtúir]] Seán] COP doctor John it relates to the Irish data. bear Tense features. As for Case features, it is far That is not to say that the dissertation fails to from clear that only phrases bear them. First of Carnie proposes that is is a complementizer significantly treat non-verbal predication or head all, the D or N inflected for Case would ordinarily element, outside IP, and that the nominal predi- movement: it provides a head-movement analysis be assumed to bear Case features. Secondly, on cate dochtúir ‘doctor’ moves to the left, across the for the non-verbal predicate in copular construc- the assumption (common since Chomsky 1993) subject of the small clause. He provides argu- tions in Irish. Nor do I intend to intimate that that feature checking occurs when features match 0 ments (pp. 258–260) against the earlier analysis there are no interesting theoretical proposals with within a checking domain, Agr must bear Case of Doherty (1996), sketched in (1b), in which is relevance for other languages, in fact I will dis- features in order to check Case features on a DP occupies Infl, the nominal predicate remains in cuss one at length below. But the crux of the in SpecAgrP. At the heart of Carnie’s formulation situ, and the subject is base-generated in a right dissertation is its careful and detailed treatment of this second pair of properties is an opposition specifier of Infl. of Irish syntax, especially for copular construc- between V-features and N-features, apparently Carnie suggests specifically that the predi- tions. Carnie lays out a wealth of (often original) adapted from Chomsky (1993). It is to that opposi- cate moves to Infl by head movement; while acknowledging that complex XPs appear in the Table 1. same position, he suggests that XP movement to X0 positions must be allowed in principle. He Properties of X0s Properties of XPs gives three reasons internal to the Irish copular construction, and then provides some supporting theta markers theta marked evidence from other languages. I examine the bear Tense and Agreement features bear Case features (undergo XP movement) arguments based on Irish first, and briefly discuss (undergo head movement) the material from other languages in §5 below. The three reasons are: [i] (p. 143) taking the select for complements are selected for order in (1) to be derived by head movement don’t have reference may have a real world reference unifies that construction with the VSO order standard in other clauses; [ii] (pp. 192–194) wh- input to/output from the morphology not input/output of the morphology movement is impossible from the predicate Carnie Dissertations Glot International, Volume 3, Issue 8, October 1998 14 takes to be in Infl; and [iii] (pp. 195–197) respons- tactic feature specifications are shared between a Advertisement es to Yes–No questions in Irish generally require head and its projections (cf. Gazdar, Pullum, & that the element in Infl be repeated; when copular Sag 1982). Thus non-percolating head features sentences like that in (1) are made into questions, are an important part of Carnie’s theory; for Atomism and Binding they are answered (affirmatively) with the copula example, whatever feature triggers head move- plus a dummy element ea, which Carnie takes to ment of the verb in a V2 construction must not be located in Infl. percolate to VP. A better move, given Carnie’s The second and third arguments can be other assumptions, would be to assume that edited by rapidly dispatched. Doherty (1997):86 falsifies [ii] Economy favors head movement where possible, Hans Bennis, Pierre Pica & Johan Rooryck with examples like the relative clause construc- so that if a head can move (rather than a phrase) tions in (2). (2a) contains a resumptive pronoun it must (cf. Chomsky 1995 in which movement of A’-bound by the relative operator, and (2b) is the more material than the feature needing checking same example with a gap, only possible in literary is likened to pied piping). varieties. Nevertheless, Carnie will still be forced to postulate non-percolating phrasal features, as the (2) trigger for XP movement, stipulating that they a. an fhoireannj ar captaen dój Parnell are attached at the maximal node, rather than to the team CCOP captain to.it Parnell ‘The team the captain of which is Parnell’ some lower node. The assumption seems problem- b. an fhoireannj ar captaen _j Parnell atic. Wh-movement, for example, involves XPs, so the team CCOP captain Parnell the wh-feature must be a phrasal feature (despite being morphologically realized on certain heads). As for argument [iii], the idea is that if Infl were Or perhaps wh-movement is movement of an empty in the copular construction, then the af- operator, and there are semantic reasons that firmative reply would presumably consist only of only XPs can be operators. But it seems unlikely the copula (the empty Infl node being unpro- that all phrasal movement could be due to seman- nounced). However, as Doherty (1997: 88 fn. 9) tic factors. Take, for example, DP movement to a points out, on Carnie’s account it is mysterious Case position: Carnie must claim that the fea- that it should be necessary to replace the predi- tures checked in SpecAgrP do not appear on the cate with a pro-form, just when it is a non-verbal head D, even if they are morphologically realized predicate, and that the pro-form cannot replace there; otherwise the head N0 or D0 might raise to such predicates elsewhere. Thus, while the exact the specifier position. Or take the idea that the status of ea remains unexplained, Carnie’s con- EPP involves a categorial feature, N or D (Chom- tention that it is a pro-form of the predicate raises sky 1995). For Carnie, this would have to be a The sixteen papers in this volume provide a representa- more questions than it answers. feature that is attached only to the phrasal level. tive overview of the broad range of issues relevant to Argument [i] is more substantial, promoting the study of binding phenomena in the generative It would be very interesting to see such implica- framework. as it does uniformity in the derivation of clauses. tions explored, and I feel it is a shortcoming of the However, there are at least two ways to maintain dissertation that they are not. Since the inception of the theoretical interest in co- parallelism in Irish copular and non-copular referential relations in generative grammar, there has To summarize, Carnie’s proposal requires a been a debate with respect to the question as to clauses without allowing an XP to move to a head basic distinction between heads and phrases in whether and how (co)-reference should be repre- position. I discuss them in §4, but first, in §3, I two places: both in the feature checking system, sented in the grammar. Notions such as “(co)- probe the deeper implications of Carnie’s propos- and in the association of features with p-markers. indexation”, “disjoint reference”, “free” and “bound” al. Although V-features can only be checked in a play an important role in this discussion. Their rele- head position, and N-features can only be checked vance is explicitly analyzed in many articles of this 3. Distinguishing heads from phrases in in a specifier position, there is no constraint on volume. the morphology what sort of syntactic object moves to those posi- The role of thematic information in Binding Theory Carnie claims (p. 185) that “[w]hat limits the tions to check them, beyond what is imposed by constitutes another important line of inquiry. Various authors argue that thematic information largely influ- behavior of p-markers are other properties of the the phrasal or non-phrasal nature of the morpho- human language computational system (such as ences binding phenomena, although the specific pro- logical features being checked (semantic con- posals expressing this relation differ quite substantially. the interface with morphology/phonology and the straints are also an option, given the passage interface with the semantic component), not the The discussion whether Binding Theory is a component quoted at the beginning of this section). Nor, since of sentence grammar is also pursued in this book. p-marker’s status as a phrase or head.” It requires the head-phrase distinction has no syntactic Many papers address this issue quite explicitly and a little bit of effort to determine just what is being status, can there be anything like Rizzi’s (1990) delineate the respective roles of sentence and dis- proposed here. As it turns out, Carnie’s vision Relativized Minimality, or Chomsky’s (1994) course grammar in novel and intriguing ways. mainly involves shifting burdens from X-bar Uniformity Condition on chains; their effects Further topics investigated in this volume include the theory to the morphological component. must be derived in some other way, though Carnie decomposition of morphosyntactically complex ana- Carnie assumes (p. 27) a basic distinction does not suggest how. phors, the relation between switch-reference and between N-features and V-features, apparently binding, and the role of binding with respect to con- derived from Chomsky (1993). The position 4. Alternatives trol. Carnie takes is that N-features must be checked Consider a possibility available if we reject Although the particular issues discussed differ consid- in a specifier position, while V-features must be Carnie’s opposition of V-features to N-features, erably, their unity lies in the empirical domain covered checked in a head position. Thus the labels are but retain the assumption that only heads can and the theoretical framework adopted. This collection misleading: if head-movement of N0 to D0 occurs, of papers offers a unique picture of the state of the art move to head positions in the syntax. Then in generative research on binding. for example, it must be triggered by a V-feature. Carnie’s suggestion that tense and agreement The volume contains contributions by: For Irish, Carnie assumes that there are features in Irish can be associated with a nominal Stephen Berman & Arild Hestvik strong V-features in T that must be checked (pp. predicate takes on a different cast. If a predicate George Aaron Broadwell 102–103, 113–114); this is what forces V move- Hamida Demirdache NP bears features that can check the strong Robert Fiengo & Robert May ment. Carnie furthermore suggests (pp. 202–204) features on T, then on the feature checking theory Zygmunt Frajzyngier that Irish exceptionally allows tense and agree- of Chomsky (1993), NP will have to move to the Robert Freidin ment features to be attached to a nominal predi- Jeff Gruber checking domain of T. The specifier of T, SpecTP, James Higginbotham cate, which we can assume for argument’s sake to is the nearest XP position in the checking domain Hajime Hoji be an NP; these are the features that can check of T, and NP will therefore be forced to move there Jan Koster the strong features in Agr. Because they are V- Howard Lasnik (equivalently, assuming something like Hoekstra Seth Minkoff 0 features, they can only be checked in T . Thus, 1991 or Kayne 1994, NP adjoins to TP). This is Gertjan Postma Carnie claims, NP movement to T0 is forced. But essentially the proposal in Doherty (1997), modulo Eric Reuland & Sigridur Sigurjónsdottir this is only assuming that the features attached Ken Safir the node labels (it is also that of Massam & Small- Christopher Tancredi. to the predicate NP cannot appear on its head; if wood 1997 for Niuean). It represents a minimal they did, then we would get N0 movement to T0, In their introduction, the editors attempt to pull to- adjustment of Carnie’s claims, and yet eliminates gether some of the main theoretical threads which run without the rest of the NP being moved. Carnie the need to move an XP to a head position. It is through the papers. suggests (pp. 205–207) that the reason that these also similar to other proposals in which the same features cannot be realized on N0 is morphologi- strong feature can be checked either by XP move- FORIS PUBLICATIONS cal; there is no lexical item in Irish corresponding ment or by X0 movement, for example Alexiadou ISBN 90-6765-535-x | xx+412 pp. | january 1998 | NLG 65 (ex. VAT, ex. P&P) to a tensed N0. Thus, the features remain at the & Anagnostopoulou (1995) and Svenonius (1996). phrasal level. Another alternative, which maintains the Holland Academic Graphics [ s c i e n t i f i c [ d o c u m e n t ] p r o c e s s i n g ] He also suggests (p. 202) that when phrases parallelism between copular and non-copular are prevented from moving to head positions, it is clauses but which requires abandoning neither P.O. Box 53292 e-mail: mail@ hag.nl 2505 AG because they fail to bear the right featural specifi- Carnie’s N-feature/V-feature distinction nor the The Hague http: www.hag.nl 31 70 4480177 cations. The usual assumption is that morphosyn- syntactic head-phrase distinction, would be to The Netherlands fax: + Dissertations Glot International, Volume 3, Issue 8, October 1998 15 assume that VSO order in non-copular clauses is ‘verbal’, at least in Chamorro, as in Ancient He- does he show reasons for rejecting the reasonable derived not by head movement, but by VP move- brew, judging from Steiner 1997: 165–166.) alternatives available in the literature. ment. This is made possible by the analysis of At any rate, it seems that the evidence so far Irish clause structure in Bobaljik & Carnie (1996), for allowing phrases to move to head positions is 5. Conclusion by which both the subject and the object evacuate rather feeble. In the next section I briefly discuss Carnie (1995) is a careful and detailed study from the VP to higher specifier positions in tensed Carnie’s cross-linguistic evidence (Carnie provides of Irish clausal syntax. It is to be commended for clauses. The VP, then, contains only V, just as in another Irish-internal argument (pp. 215–218), its clear discussion, ample supply of example Hinterhölzl’s (1997) analysis of West Germanic based on a case of putative movement to D0 by an sentences, sensible organization, attentiveness to Verb Raising as involving VP-remnant movement, NP. However, the comments I made above regard- previous work, and considerate citation etiquette. or in Kayne’s (1998) analysis of certain English ing alternatives to NP to Infl movement generally No expert on Irish syntax, I found the guided tour constructions. Such a proposal is not compatible apply in that case as well). through the literature helpful and illuminating. with Carnie’s assumptions about minimality (pp. In addition, it makes some provocative theo- 104 ff.): he assumes, following the theory of mini- 5. Examples from other languages of P- retical claims, but these seem to me to be less well mality motivated in Bobaljik & Jonas (1996) and markers with mixed phrase-head motivated and developed. I have examined the formalized in Chomsky (1993), that the subject characteristics main one in detail, the bold proposal that the X- and object cannot both leave VP without V having Carnie provides (pp. 211–228) some cross- bar theoretical distinction between heads and moved to some higher position. However, this linguistic evidence for “an ambiguity between phrases be jettisoned. I believe I have demonstrat- need not stop us: the foundations for that theory phrasal and X0 behavior” (p. 210) which is intend- ed that Carnie’s system preserves the head- of minimality are called into serious question by ed to support the proposal discussed above. He phrase distinction, but relocates it to the Holmberg (1997). briefly discusses about a half-dozen examples of morphological component. Having systematically If VSO is derived by VP movement in general, cases which are supposed to show a mix of head examined his empirical arguments for his revision then VOS languages like Tzotzil are more similar and phrase properties. However, he provides no of X-bar theory, I have concluded that they are to SVO languages like Irish than previously indication as to how his theory is to account for wanting, and I have pointed out two alternatives thought; and languages like Chamorro, in which these cases, nor even whether they are consistent to his specific proposal for non-verbal predicate VOS and VSO are alternatives, could be analyzed with his proposal. In most cases there are perfect- constructions. as raising the object out of VP only optionally. Of ly plausible accounts which do not rely on aban- However, there are at least three ways in course, in order to derive VSO order, substantial doning the head-phrase distinction. The section which even this proposal represents headway. leftward movements would have to be postulated suffers from the fact that he has not precisely First, it improves on the most thorough analysis and motivated, in order to evacuate the VP before identified the head-phrase properties at issue, nor of Irish copular constructions to have preceded it VP movement, much as in recent Kaynean analy- explained why they should be associated with (Doherty 1996) in that it derives the predicate- ses of OV order. It must also be explained why heads or phrases. I will not discuss all six exam- subject order by leftward movement of the predi- such fronting does not occur in non-verbal predi- ples here for reasons of space, but will briefly note cate, a move which is adopted in Doherty (1997). cates; Case-marking is not the whole story, as PPs two, to illustrate my point. Second, it makes a first stab at capturing the never front along with the Irish verb. But this For example, he discusses some Yoruba parallelism between sentence-initial verbal heads seems to be an interesting area for investigation constructions with phrasal complexity (citing and sentence-initial non-verbal predicate phrases, (cf. the similar yet intriguingly different proposal Pulleyblank & Akinlabi 1988, which I have not a distinction which looks likely to remain with us for Irish in Duffield 1995). seen), suggesting that they are headlike in being for some time. Third, it calls attention to the The cross-linguistic evidence warrants an islands for extraction and anaphora (properties, it stipulative nature of our fundamental assump- approach based on universal distinctions between will be noted, which do not appear in Table 1). tions about phrase structure, widely assumed but verbal and non-verbal predicates. Carnie treats But the connection between islandhood and rarely discussed. Even if I have rejected the the parallelism of verb-initial clauses and non- head status is far from clear; certainly, phrases specifics of Carnie’s proposal, the fact remains verbal predicate-initial clauses as parochial when can be islands, as demonstrated by Ross (1967). that the X-bar distinction between head and he suggests that Irish non-verbal predicates Worse, if excorporation is possible, as argued in phrase is generally underexamined. exceptionally bear verbal features (p. 202). How- Roberts (1991) or Koopman (1994), then heads are ever, it seems that the pattern is widespread; for not islands for movement. Nor are they anaphoric Note example Carnie himself notes that it holds for islands, if N-incorporation structures are heads, Thanks to Jim McCloskey, Cathal Doherty, Tagalog (p. 212), and Chung (1990):570 notes it as Carnie assumes (p. 207, citing Baker 1988; but and David Adger for discussion of this material. for Chamorro, giving examples like the one in Baker explicitly uses the referential transparency All opinions expressed and metaphors mixed are (3a), while Massam & Smallwood (1997) note it of incorporated nouns to motivate a syntactic my own. for Niuean, as shown in (3b) (from their p. 268). account; cf. Baker’s pp. 78–81). Abbreviations appearing in glosses are: More interestingly, the Yoruba phrases show AGR(1S) first person singular agreement, ART arti- (3) derivational morphology. But this might be sub- cle, CL classifier, CCOP complementizer copula, COP a. Ginin i chi’lu-hu esti na katta. sumable under a theory of phrasal affixation, copula, L linker, P preposition, POSS possessive, from the sibling-AGR(1S) this L letter ‘This letter is from my sister.’ such as that of Miller (1991), Anderson (1992), or PRES present. See the works cited for explanations. b. Ko e kamuta a au. Halpern (1995). P ART carpenter ART I In any case, Carnie says nothing about how References ‘I am a carpenter.’ his theory would handle the Yoruba examples, Alexiadou, Artemis, and Elena Anagnostopoulou being content to note them as a case of mixed (1995). SVO and EPP in Null Subject Lan- Cursory glances at other VSO languages suggest head and phrase properties. Thus it remains guages and Germanic, FAS Papers in Lin- that it holds there as well, as suggested by the unclear whether his theory “predicts” the exist- guistics 4: 1–21. FAS, Berlin. examples in (4a–c). ence of cases like Yoruba, as he claims (p. 226). Anderson, Stephen (1992). A-Morphous Morpholo- Carnie also mentions (p. 223) a case which is gy. Cambridge University Press, Cambridge. (4) a. ‘O le fili o le ‘iole le pusi. in a sense the opposite of the Yoruba case, that of Baker, Mark C. (1988). Incorporation. The Univer- PRES ART enemy POSS ART rat ART cat separable prefix verbs (he mentions only Yiddish sity of Chicago Press, Chicago, Ill. ‘The cat is the enemy of the rat.’ in this context, but German and Dutch are identi- Bloomfield, Leonard (1933). Language. Holt, b. Caw bakich no’ ha-txitam tu’. cal in the relevant respects). Separable-prefix Rinehart, and Winston, New York. very fat CL your-pig that ‘That pig of yours is very fat.’ verbs have some headlike properties, for example Bobaljik, Jonathan, and Andrew Carnie (1996). A c. ’ala¯ lmma¯ ‘idati kita¯bun. the prefixes are typically not phrasally complex, minimalist approach to some problems of on table book but on the other hand they can be separated by V Irish word order, in The Syntax of the Celtic ‘A book is on the table.’ movement. Languages, ed. by Robert Borsley and Ian Once again, Carnie does not propose an Roberts, pp. 223–240. Cambridge University (4a) is Samoan, from Mosel & Hovdhaugen (1992): analysis, but is content to present this as a case of Press, Cambridge. 500, (4b) is Jakaltek, from Craig (1977): 138, (4c) mixed head-phrase properties. There are two Bobaljik, Jonathan, and Dianne Jonas (1996). is Classical Arabic, from Kaye (1987): 684, though different ways out of this problem other than the Subject positions and the roles of TP. Linguis- in Classical Arabic, definite subjects generally abandonment of the head-phrase distinction. One tic Inquiry 27.2: 195–236. precede a non-verbal predicate, so (4c) might be a is to accept excorporation as a possibility (cf. Brody, Michael (1997). Mirror theory. UCL Work- red herring. above). After all, Carnie does not suggest why ing Papers in Linguistics 9, University Col- Quite possibly, then, all VSO languages have excorporation should be impossible. The other lege London. in common that non-verbal predicates, when they approach that does not involve weakening the Chomsky, Noam (1993). A minimalist program for front, front by phrase movement, rather than by head-phrase distinction is to assume that separa- linguistic theory, in The view from Building head movement. If that is the case, then it cannot ble prefix verbs are not heads at all. Zwart (1993) 20: essays in linguistics in honor of Sylvain be due to some accidental morphosyntactic prop- and Taraldsen (1998) both provide analyses in the Bromberger, ed. by Kenneth Hale and Samuel erty. (Care must be taken in formulating this which the ‘prefix’ occupies a phrasal position to Jay Keyser, 1–52. MIT Press, Cambridge, Ma. putative generalization. Chung (1990): 570 shows the left of the verb. Chomsky, Noam (1994). Bare phrase structure, that adjectival predicates can strand complements The other examples Carnie presents are MIT Occasional Papers in Linguistics 5, to the right of the subject, suggesting that they generally similar in that he provides no specifics Cambridge, Ma. undergo head movement; so they are sufficiently as to how his theory would handle such cases, nor Chomsky, Noam (1995). The Minimalist Program, Dissertations Glot International, Volume 3, Issue 8, October 1998 16

MIT Press, Cambridge, Ma. Hinterhölzl, Roland (1997). A VO-based approach Miller, Philip (1991). Clitics and Constituents in Chung, Sandra and James McCloskey (1987). to verb raising, in Proceedings of the North Phrase Structure Grammar. PhD dissertation, Government, barriers, and small clauses in East Linguistic Society 27, ed. by Kiyomi Rijksuniversiteit Utrecht. Modern Irish. Linguistic Inquiry 18.2: 173– Kusumoto, 187–202. GLSA, University of Mosel, Ulrike, and Even Hovdhaugen (1992). 237. Massachusetts, Amherst, Ma. Samoan Reference Grammar. Scandinavian Chung, Sandra (1990). VP’s and verb movement Hoekstra, Eric (1991). Licensing Conditions on University Press, Oslo. in Chamorro, Natural Language and Linguis- Phrase Structure, PhD dissertation, Rijksuni- Muysken, Pieter (1982). Parameterizing the tic Theory 8: 559–619. versiteit Groningen, Grodil, Groningen. notion ‘head’, Journal of Linguistic Research Craig, Colette (1977). The Structure of Jacaltec. Holmberg, Anders (1997). The true nature of 2: 57–75. University of Texas Press, Austin, Tx. Holmberg’s Generalization, in Proceedings of Pollard, Carl and Ivan Sag (1994). Head-Driven Doherty, Cathal (1996). Clausal structure and the the North East Linguistic Society 27, ed. by Phrase Structure Grammar. The University of Modern Irish copula, in Natural Language Kiyomi Kusumoto, 203–217. GLSA, Universi- Chicago Press, Chicago, Ill. and Linguistic Theory 14: 1–46. ty of Massachusetts, Amherst, Ma. Pulleyblank, Douglas and Akinbiyi Akinlabi Doherty, Cathal (1997). Predicate initial construc- Kaye, Alan (1987). Arabic, in The World’s Major (1988). Phrasal morphology in Yoruba, Lin- tions in Irish, in The Proceedings of the Fif- Languages, ed. by Bernard Comrie, 664–685. gua 74: 141–166. teenth West Coast Conference on Formal Oxford University Press, Oxford. Rizzi, Luigi (1990). Relativized Minimality. MIT Linguistics, ed. by Brian Agbayani & Sze- Kayne, Richard (1994). The Antisymmetry of Press, Cambridge, Ma. Wing Tang, 81–95. CSLI, Stanford, Ca. Syntax. MIT Press, Cambridge, Ma. Roberts, Ian (1991). Excorporation and Minimali- Duffield, Nigel (1995). Particles and Projections in Kayne, Richard (1998). Overt vs. covert movement, ty, Linguistic Inquiry 22.1: 209–218. Irish Syntax. Kluwer, Dordrecht. to appear in Syntax 1.2. Ross, John Robert (1967). Constraints on Varia- Emonds, Joseph (1976). A Transformational Koopman, Hilda (1994). Licensing heads, in Verb bles in Syntax. PhD dissertation, MIT. Approach to English Syntax: Root, Structure- Movement, ed. by David Lightfoot and Nor- Steiner, Richard (1997). Ancient Hebrew, in The Preserving, and Local Transformations. bert Hornstein, 261–296. Cambridge Univer- Semitic Languages, ed. by Robert Hetzron, Academic Press, New York. sity Press, Cambridge. 145–173. Routledge, London. Gazdar, Gerald, Geoffrey Pullum, and Ivan Sag Kornai, Andras, and Geoffrey Pullum (1990). The Svenonius, Peter (1996). The optionality of parti- (1982). Auxiliaries and related phenomena in X-bar theory of phrase structure, Language cle shift, Working Papers in Scandinavian a restrictive theory of grammar, Language 58: 66: 24–50. Syntax 57: 47–75. 591–638. Massam, Diane, and Carolyn Smallwood (1997). Taraldsen, Tarald (1998). V-movement and VP- Halle, Morris, and Alec Marantz (1993). Distribut- Essential features of predication in English movement in derivations leading to VO order, ed morphology and the pieces of inflection, in and Niuean, in Proceedings of the North East paper presented at VO and OV workshop in The View from Building 20, ed. by Kenneth Linguistic Society 27, ed. by Kiyomi Kusumo- Tromsø. Hale and Samuel Jay Keyser, 11–76. MIT to, 263–272. GLSA, University of Massachu- Zwart, Jan Wouter (1993). Dutch Syntax: A Mini- Press, Cambridge, Ma. setts, Amherst, Ma. malist Approach. PhD dissertation, Universi- Halpern, Aaron (1995). On the Placement and McCloskey, James (1996). On the scope of verb ty of Groningen. Grodil, Groningen. Morphology of Clitics. CSLI, Stanford, Ca. movement in Irish, Natural Language & (1992 Stanford PhD dissertation). Linguistic Theory 14: 47–104.

(1) THE COPY THEORY OF MOVEMENT [ John [ was [ arrested John ] ] ] (2) a. John was arrested. AND LINEARIZATION OF CHAINS b. * John was arrested John. Another conceptual problem with the compu- IN THE MINIMALIST PROGRAM tational system as proposed in Chomsky (1994, 1995: chap. 4) is that Merge is taken to be an operation in its own right in certain cases, and a suboperation (of Move) in other cases. In an optimal system, we should in principle expect by Jairo Nunes Merge to have the same theoretical status in every computation. Finally, as is emphasized by reviewed by Hans-Martin Gärtner Brody (1995), if chain formation and Move express the same type of relation, a theory which contains both notions is redundant. This dissertation develops a strictly Minimal- Summary “lower” copies must be deleted in the phonological ist version of the copy theory of movement which by the author component is provided, the notion of a trace as a overcomes the conceptual problems raised above primitive is reintroduced. To put it more general- and has a broader empirical coverage than the A considerable amount of research within the ly, the simplest and, therefore, most desirable versions developed in Chomsky (1993, 1994, 1995: Principles and Parameters Theory has been version of the copy theory of movement should chap. 4). It proposes that the fact that a chain devoted to properly characterize the properties of take traces and heads of chains to be subject to cannot have more than only link overtly realized movement, traces, and chains. In the recent the same principles and be accessible to the same (see (2b)) follows from Kayne’s (1994) Linear developments of the Principles and Parameters operations. Any difference between heads of Correspondence Axiom (LCA), according to which Theory which have culminated with the proposal chains and traces, such as phonetic realization, the linear order of a PF sequence is determined by of a Minimalist Program for linguistic theory (see for instance, should follow from independently asymmetric c-command. Under the assumption Chomsky 1995), these issues arise anew in face of motivated properties of the computational system, that the two copies of John in (1) are “nondistinct” the elimination of much of the rich theoretical rather than being idiosyncratic properties of the (they relate to the same element in the initial apparatus previously available. chain links themselves. numeration), no linear order can be established in Chomsky (1993) incorporates into the Mini- Deletion of traces (lower chain links) becomes accordance with the LCA. Given that the verb was malist Program the “copy theory of movement”, even more enigmatic, if we adopt the core Minimal- in (1), for instance, asymmetrically c-commands according to which a trace is a copy of the moved ist assumption that economy considerations play a and is asymmetrically c-commanded by the element which is deleted in the phonological role in determining the set of admissible deriva- “same” element, namely John, the LCA should component, but remains available for interpreta- tions in a given language or universally. Consider require that was precede and be preceded by tion at LF. Under this view, the operation Move is for instance the structure in (1) below, where John John, violating the asymmetry condition on linear a complex operation comprised of (at least) three has moved to the subject position and left a copy order. Put simply, deletion of all but one link is suboperations (see Chomsky 1993: 22; 1994: fn. behind. The derivation of (2a) from (1) requires one forced upon a given chain CH in order for the 13; 1995: 250): (i) a suboperation of copying; (ii) a application of deletion targeting the lower copy of structure containing CH to be linearized in ac- suboperation of merger; and (iii) a suboperation of John, apparently being less economical than the cordance with the LCA. The derivations of (2a) trace deletion. In addition, Move should be fol- derivation of (2b), which involves no application of and (2b) therefore cannot be compared for econo- lowed by an operation of chain formation relating deletion. Thus, if the derivations of (2a) and (2b) my purposes, because only the former yields a PF the relevant copies. were to be compared in terms of economy, we would object. There are several conceptual inadequacies in wrongly predict that the derivation of (2b) should The next question to be addressed then is this picture. First, if no explanation for why rule out the derivation of (2a). why it is the case that only traces are deleted for purposes of linearization, but not heads of chains.