+ All Categories
Home > Documents > A Minimalist Account of Auxiliary/Main Verb Asymlnetries in English

A Minimalist Account of Auxiliary/Main Verb Asymlnetries in English

Date post: 12-Feb-2022
Category:
Upload: others
View: 7 times
Download: 0 times
Share this document with a friend
36
A Minimalist Account of Auxiliary/Main Verb Asymlnetries in English Justin Kelly ([email protected]) 1. INTRODUCTION The relationship between auxiliaries and main verbs in English has been a carefully studied topic throughout the history of generative syntax, beginning with the transformational framework of Chomsky's early work, all the way through late Government and Binding (GB) and early Minimalist models. Consider the data in (1) (3). (1) (2) (3) a. John eats b. John doesn't eat c. *John eatsn't a. John has eaten b. John hasn't eaten c. *John doesn't have eaten d. John has a book e. *John hasn't a book f. John doesn't have a book a. John is a man b. John isn't a man c. *John doesn't be a man d. John is eating e. John isn't eating f. *John doesn't be eating The basic facts in (1) - (3) can be described as follows. Main verbs in English cannot move to TO as is evident by the interaction between eat and negation in (I); in (1 c), we see that the verb cannot move through negation to TO. Instead, do-support is a necessary [51 ]
Transcript

- - .. - A Minimalist Account of

Auxiliary/Main Verb Asymlnetries in English

- Justin Kelly ([email protected]) .. 1. INTRODUCTION

- The relationship between auxiliaries and main verbs in English has been a carefully

studied topic throughout the history of generative syntax, beginning with the

- transformational framework of Chomsky's early work, all the way through late

Government and Binding (GB) and early Minimalist models. Consider the data in (1)

- (3).

- (1)

.. (2)

.. - (3) --

a. John eats b. John doesn't eat c. * John eatsn't

a. John has eaten b. John hasn't eaten c. *John doesn't have eaten d. John has a book e. *John hasn't a book f. John doesn't have a book

a. John is a man b. John isn't a man c. *John doesn't be a man d. John is eating e. John isn't eating f. *John doesn't be eating

.. The basic facts in (1) - (3) can be described as follows. Main verbs in English cannot

move to TO as is evident by the interaction between eat and negation in (I); in (1 c), we - see that the verb cannot move through negation to TO. Instead, do-support is a necessary

- [51 ]

..

• J. Kelly

•last resort operation to support the clitic form of negation, as in (1b). In (3), we see that

be always raises to 'fl, both when be functions as an auxiliary (3e) and as a copula (3b); .. do-support is not an available option, as is shown in (3c) and (3f). With have, the pattern

is a bit different. In (2b), we see that when have functions as an auxiliary it may raise to .. TO, signaled by the cliticization of n 't onto have. On the other hand, when have functions

as a main verb, it may not move to TO, as indicated by the ungrammaticality of (2e). .. Instead, do-support is necessary.

Many researchers have devised systems to account for the difference between .. auxiliary verbs and main verbs in English, contrasting with French, where auxiliary and

main verbs act identically in finite clauses (cf. Pollock, 1989). Attempts have also been .. made to account for the difference between the bifurcating auxiliary/main verb status of

have when interacting with negation, and how this contrasts with the uniform functioning

of be, both as an auxiliary and as a main verb. I refer the reader to the primary sources for • further information regarding previous analyses I.

In this paper, I propose an analysis that provides a unified account of the data in -(1) - (3). The basic course I will take is as follows. First, I discuss a theory proposed by

Kayne (1993) whereby both be and have are underlying1y an abstract verbal head BE, -and the spell-out of BE is conditioned by head movement. I reject many of Kayne's ..assumptions, but I maintain a unified theory of be and have under BE. Next, I introduce

the notion of predicationa1 features, which are similar to the verbal features discussed by ..Chomsky (1995) in the early Minimalist work. These predicationa1 features are present

on verbal and predicationa1 heads and can be interpretable or uninterpretab1e, depending ..on the head on which they are present. I propose that these features are checked in the

syntax through Agree operations, but all head movement is post-syntactic, occurring in a

morphological component of the grammar that intervenes between the syntax and the .. phonology. Although head movement actually occurs in the morpho10g/, it is -I For an early transformational analysis, see Chomsky (1957; 1965). Canonical GB-era formulations can be found in Pollock (1989) and Lasnik and Saito (1984; 1992) and are summarized well in Lasnik (1999; ..2000). An early Minimalist account is provided in Chomsky (1995) but much of the mechanisms of this framework are discarded in Chomsky's later works. 2 Throughout this paper, I refer to head movement as a "PF process." I believe that head movement actually occurs in some morphological component akin to Distributed Morphology (DM) (Halle & ..

[52] • ..

- .. Auxiliary/Main Verb Asymmetries in English

.. constrained by the syntax; head movement is necessarily preceded by Agree relations in

the syntax, and strong phases serve to constrain head movement, even after Agree

.. relationships hold. I propose that head movement may not occur across strong phase

boundaries, while it may occur across weak ones. Utilizing this formulation of head

movement, I show how constructions in English that involve verb movement to TO all

involve a vP that is a weak phase, while in constructions where no vO-to-TO movement is.. possible, vP is a strong phase. In this system, phase strength reduces to the projection of a

specifier, as discussed by Chomsky (2000). The major contribution of this paper is that it.. provides a uniform analysis of verb raising paradigms in English, not only involving be

and have, but also for main verbs. Secondarily, this paper provides support for theories of- post-syntactic head movement and for a morphological component of the grammar.

The structure of the paper is as follows. In Section 2, some theoretical background .. is provided. I briefly describe the Minimalist tenets that are important to my discussion,

as well as some other issues that are central to my topic, such as the status of

predicational phrases (PredPs) and the theory of auxiliary selection devised by Kayne -(1993). In Section 3, I provide a unified analysis of have and be in English. In Section 4, ... I discuss the auxiliary/main-verb asymmetry of have in the scope of negation, using the

analysis from Section 3 to provide a novel explanation to this historically recalcitrant

problem. I also use the proposal from Section 3 to motivate an analysis of -auxiliary/copular uses of be in its interactions with negation. Section 5 provides an .. analysis of Control structures involving have and be, following from the theory in

Sections 3 and 4. In Section 6, I discuss small clauses in light of the theory developed in .. Sections 3 and 4, and in Section 7 I analyze ECM constructions involving embedded

have and be. Section 8 provides a brief conclusion, as well as some avenues for further .. research .

...

.. Marantz, 1993) or Lexeme-Morpheme Base Morphology (Beard, 1995), although I do not take a stand regarding a specific choice of post-syntactic morphological theory to which I subscribe. Therefore, I simply refer to head movement as a PF process in following the literature on this topic (cf. Boeckx &

II1II Stiepanovic, 2001) to signal that it is a post-syntactic operation.

.. [53]

..

.. J. Kelly .. 2. THE FRAMEWORK

In this section, I discuss some preliminary assumptions that are important for the .. discussion throughout this paper. First, I point out some aspects of the current Minimalist

system that are pertinent to my arguments. Then, I briefly discuss the notion of .. predication, which will be refined as we proceed through the paper. Finally, I present a

summary of Kayne's (1993) system for auxiliaries be and have. Although I reject the •majority of Kayne's analysis, I utilize his central concept as a basis for my analysis of be

and have in English. .. 2.1. STANDARD MINIMALIST ASSUMPTIONS .. In this paper, I will be using a minimalist framework based on the one described by Chomsky (2000; 2001). Within this formulation of syntactic theory, much of the .. theoretical baggage of the GB framework is discarded in favor of a more elegant system

which relies on key operations and general constraints on the syntax. Unless otherwise

noted, I assume the Copy theory ofmovement and the Phase-based system3 as outlined in • the references cited.

One relevant point of discussion that Chomsky (2000) takes up is the notion • 'Multiple Spell-Out.' For Chomsky, Phases are critical points in the derivation; at the end ..of the Phase, the information is sent to the phonological component (i.e. Spell-Out). .. 3 Chomsky (2000) introduces is the notion of the Phase as a theoretical construct, and this is central to the literature on Phase-based theories. Universally, vP and CP constitute Phases in the syntax. These Phases are relevant to movement operations because they serve as natural 'clause boundaries' to which Agree operations are subject. The Phase Impenetrability Condition, as stated in (i), is the constraint on the syntax .. that makes Phases relevant.

(i) The Phase Impenetrability Condition (PIC) (Chomsky, 2000) In phase a with head H, the domain of 11 is not accessible to operations outside a, only 11 .. and its edge are accessible to such operations.

The PIC is crucial to Chomsky's system because these phases serve to delimit notions such as 'islands' and .. 'subjacency', terms familiar from the GB literature. Strictly cyclic movement is accommodated easily within the Phase system by reference to the "edge" of a phase in the PIC. The edge of a maximal projection H is defined by Chomsky as the specifier (or specifiers) and head ofH. Thus, strictly cyclical movement through [Spec, CP] (i.e. Wh-movement) falls out as licit movement because the Wh-element moves .. through the edge of a phase. The definition of a phase becomes a bit more murky when extended CPs are taken into account, such as those found in constructions with a significant left-periphery. I do not discuss this issue here, but I refer readers to the literature regarding the left-periphery, such as Rizzi (1997; 2002) to verif'y that the definition ofphases is problematic in these cases. ..

.. [54] ..

- - Auxiliary/Main Verb Asymmetries in English

.. Thus, Spell-Out occurs at various points in the derivation. It is also important to note that

the head of a Phase is no longer available to initiate Agree relationships after Spell-Out of- the Phase. "The head of a Phase is "inert" after the phase is completed, triggering no

further operations." (Chomsky, 2000; p. 107) An additional distinction that Chomsky

makes is the notion of strong and weak Phases, which will become crucial to my theory.

"Some Phases are strong and others weak- with or without the EPP option, respectively, - hence relevant or not for Spell-Out." (Chomsky, 2001, 14)

..

.. 2.2. A NOTE ON PREDICATION

With the development of theories of verbal shells as initiated by Larson (1988) and

extended by Chomsky (1995) and Kratzer (1996), the verbal system within generative

syntax is no longer comprised of a simple VP. Following Chomsky, I take the verbal - system of transitive verbs to consist of a light verb head (vo) which projects a specifier

position and which takes a VP complement. In this manner, the vP is the locus of the ., external argument in the verbal field, and specifier and the complement positions in the

VP are positions reserved for complements of the verb; the verb is base generated in yo and moves to VO (overtly in English; I assume that head movement is a PF phenomenon, -as conjectured by Chomsky (2000) and further supported by Boeckx and Stjepanovic .. (2001)). This has generally become the null-hypothesis in Minimalist syntax.4 The case

with unaccusatives, raising verbs and adjectives, and copular constructions, where

agenthood is not relevant, is a bit different. Most of the literature regarding these verbal -constructions (cf. Chomsky, 1995; Adger, 2003; Haegeman, 1991) analyze these

., constructions as involving a bare VP, without the light verb projection. I will not discuss

raising/unaccusative constructions, but I do wish to discuss how Baker (2003) analyzes

copular constructions. -Within a VP shell or similar system, copular constructions are generally analyzed

as in (4). The copula, a yO, takes a small clause complement, and the "subject" position -of the small clause is taken to raise to specifier position of the copula before moving to

[Spec, TP] for Case assignment/checking. - - ..

[55]

-

., J. Kelly

• (4) /\,

/\p •  •

V A • DP XP

•Small clauses have been generally problematic for the theory, as structures like those in

(4) do not reflect standard X' and Bare Phrase Structure assumptions. To alleviate this

issue, Baker (2003) (citing Bowers, 1993) develops a system where small clauses are • defined as consisting of functional projection PredP, for "predication." This PredP ... projection may serve as the locus for theta role assignment for small clauses, and be is

necessary simply to bear finite tense and agreement morphology. This type of structure is ... given in (5). .. (5) A

r

 • A

vI'

...A v A ...DP A

Pred XP ... Baker gives evidence that be is not actually an instantiation of the Predo head, based on ...small clause contexts where be is not present, as in (6). The motivation for such an

analysis is based on Baker's formulation oflexical categories. Briefly, PredP is necessary

• 4 I abstract away from the possibility of additional verbal shells, which Larson generally assumes to be projected for adverb placement (Larson, 1988,384, fn. 49). •

[56) • •

- - Auxiliary/Main Verb Asymmetries in English

because only verbs license specifiers. When the second argument in a small clause is not

verbal, a predicational element is necessary to conjoin the elements of the small clause. -- (6) I consider John smart/a genius

Throughout this paper, I will assume small clauses to have the structure depicted in (5). - This notion of the PredP is central to my discussion of the data throughout this analysis,

and I will continue to refine the description ofthe PredP as the paper progresses. -.. 2.3. AUXILIARY SELECTION

Another influential paper dealing specifically with auxiliary selection was written by

Kayne (1993) in the beginning of the Minimalist era; many researchers take the theory - presented there as a basis for their discussion of auxiliaries, but there are several gaps in

the system which have not been explored or questioned (to my knowledge). Kayne

basically proposes that have and be are not distinct, but rather are derived from an

abstract verbal head BE, and have is a Spell-Out instantiation of an abstract prepositional

determiner that has incorporated to BE. 5 Let me briefly summarize Kayne's theory. For

--

Kayne, an abstract prepositional determiner (DIP) is the head of the complement clause

of BE. Consider (7). -(7) John has a book-Kayne gives the structure in (8) for an intermediate Merge of (7).-(8) [BE [O/PP D/po

[AgrP John [AgrO a book]]]

To derive (7), Kayne assumes that the DP John must move successive cyclicly out of the -small clause through [Spec, DIPP] to the matrix [Spec, TP] for Case/EPP licensing

reasons. The crucial assumption is that [Spec, D/PP] is an A' -position, and movement -from an A-position (the base-generated position in the small clause) through an A'-

position, and back to an A-position is an "improper movement" (Kayne, 1993; p. 7). In --..

[57]

-

• J. Kelly .. order to "rescue" the derivation, Kayne states (citing Baker (1988» that the DIP can

incorporate to BE, thus converting [Spec, D/PP] to an A-position and legitimizing the .. movement of John through [Spec, DIPP]. This incorporation causes BE to surface as

have. .. In this construction, the DIP selects a small clause complement, headed by Agr.

Kayne does not specify the structure of BE, nor does Kayne explicitly describe the nature .. of Agr, but we can extend Baker's system to these cases, assuming that Agr is

predicational (Le., a PredP). Thus, we can re-write (8) as (9). .. (9) [BE [D/PP D/po

[PredP John [Predoa book]]]

•Kayne uses the system described above to account for participial constructions in

English. Consider (10), with the intermediate Merge structure of (11). .. (10) John has eaten

(11) [BE [D/PP D/po [vp John [eaten]]] .. Instead of a small clause complement of D/PP, Kayne generates a full verbal phrase as .. the complement, and in later portions of the paper, he extends this to a sentential

complement structure, complete with tense and subject and object agreement nodes. I will • not discuss the arguments Kayne presents for this structure, but I will simply accept it at

face value. Assuming this, we can rewrite (11) as (12), assuming a Minimalist analysis, .. with agreement being reflexes ofT and v. ..(12) [BE [D/PP D/po [TP TO [vp John [vo [vp eaten]]]]]

We can analyze such sentences in the same fashion as in the possessive constructions .. discussed above. Thematic relations are satisfied within the vP, but the non-finite tense

cannot serve to assign Case to John. Thus, John must raise successive cyclically through .. [Spec, D/PP] to the matrix [Spec, TP]. In order to make this movement licit, DIPP must .. 5 Kayne extends the theory proposed by Freeze (1992), with crucial evidence from Hungarian, as described by Szabolcsi (1981; 1983). I have not directly consulted these sources, so I simply present Kayne's system. ..

.. [58] ..

- - Auxiliary/Main Verb Asymmetries in English

incorporate into BE, making [Spec, D/PP] an A-position, and again causing BE to surface

- as have.

--..

One major question is evident if we assume this system for the derivation of have. How

do we derive normal copular sentences and progressive forms, such as (13) and (14)

respectively?

(13) John is a man

(14) John is eating

.. -

For these types of constructions, we must assume that D/po does not incorporate into BE,

for this operation would force BE to surface as have, but some mechanism is necessary

whereby [Spec, D/PP] becomes an A-position so that John can move through [Spec,

D/PP] to the matrix [Spec, TP]. It is not immediately evident how such sentences could

- be generated within such a system, but Kayne provides us with some insight, using

evidence from a Central Italian dialect. In this dialect, participial constructions are

--..

sensitive to the person of the subject; be is used everywhere except with third person

plural subjects. For this reason, Kayne proposes a subject agreement node (which is

generally reduced to TO in Minimalist frameworks), and subjects occupy the specifier

position of this agreement projection (the embedded [Spec, TP] for our purposes), these

subjects may move through [Spec, D/PP] provided that DIP incorporates to BE,

transforming [Spec, D/PP] to an A-position. This is schematized in (15) .

-- (15)

 [ (spec) BE [DIPP (spec) D/po [TP subject-[1 st or 2nd person] TO ...

t I incorporation

--

For third person plural subjects, when they occupy the embedded [Spec, TP], certain

agreement features of TO are 'activated' which allow TO to incorporate to DIP, making

[Spec, D/PP] an A-position. Further incorporation to BE is unnecessary, and BE surfaces

- as be (inflected in the Italian dialect, of course).

.. [59] ..

• •

.. J. Kelly ..I.•(16) [ (spec) BE [DIPP (spec) D/po [TP subject.[3rd person] TO ... ..t I .. Although Kayne does not discuss how (15) and (16) would be derived in his paper, we

could assume that some functional head lower than D/po must incorporate into D/po in

sentences like (13) and (14) so that John can move through [Spec, DIPP] on its way to the • matrix [Spec, TP] position, and since D/po has not incorporated into BE, BE surfaces as .. be. This is schematized in (17).

1 • (17) [ (spec) BE [DIPP (spec) D/po [TP sUbject... FO ... ..t I ..Although I do not adopt Kayne's system directly, I do employ his notion of BE, as well

as a the idea that incorporation effects the spell-out of BE. My analysis of English ..auxiliaries utilizing these tools is presented in the next section.

..3. A UNIFIED ACCOUNT OF HAVE AND BE

In this section, I provide a unified theory of be and have in English, accounting for both

their auxiliary and their copular/main-verb forms in non-negated contexts. The driving .. force behind by proposal focuses on the nature of predication, namely the nature of Pred°

and vO. In my proposal, these functional heads have an associated predicational feature, .. and it is the checking of this feature through the syntactic operation Agree that conditions

head movement. Before I begin to analyze the relevant data, I would like to discuss some -issues surrounding vO. .. 3.1 ON v'l AND OTHER VERBAL HEADS

Since its proposal by Chomsky (1995) (and by Kratzer (1996) as Voiceo), vO has been .. readily adopted by many researchers as the locus of the external argument of the verb. I

assume that VO plays a dual role in the syntax. When VO merges with a VP, it plays .. [60] •

..

- - Auxiliary/Main Verb Asymmetries in English

precisely the role described by Chomsky; in these situations, VO supplies the external - 9-role of the verb, so that the subject is merged in [Spec, vP], and VO checks accusative

case on the object of the verb. Thus, we derive Burzio's generalization, whereby when - the verb does not assign a 9-role to its subject, it does not assign Case to its object

(because it lacks va). I will show that VO can also have other properties. I assume that VO

- can merge with vP or PredP, and in these cases, VO is instantiated as the auxiliary BE as

described above, with some modification. So in this system, the form that VO takes

- depends on the structure with which it merges.

Since va is inherently verbal in nature, I assume that there is an uninterpretable

... predicational feature associated with va, call it [uPred]. Furthermore, assume that verbal

heads yO also have a predicational feature, but for verbs the feature is interpretable (i.e.,

- [iPred]). These features must be checked in the syntax through the operation Agree; an

uninterpretable predicational feature must be checked against an interpretable

- predicational feature in its c-command domain. Head movement is parasitic upon these

feature checking mechanisms; a head u may not move to another head unless the

- features of are checked by u. Thus, for standard cases, where va merges with YP, we

can motivate head movement of yO to va, even if we maintain that head movement is a

- PF -phenomenon. This hypothesis seems fairly intuitive, and these [Pred] features seem

to be similar to the Y-features proposed by Chomsky (1995). This process is sketched in

- (18). .. Probe/Agree

- (18) [vp John VO [vp loves Mary]]

I t head movement

- 3.2 ENGLISH AUXILIARIES

Now, how can the relation described above account for the auxiliary system in English? I .. believe that Kayne's (1993) account was essentially correct, although I suggest a

modification to the way it was implemented. A more elegant system can account for the - data by coupling Kayne's notion of BE with the system of [Pred] features. Let's see how

this can be derived. - First, consider sentences like (19), repeated from (14) above . .. [61]

-

.. J. Kelly .. (19) John is eating .. Assuming that BE is the instantiation of vO (that merges with vP), an intermediate

structure of (19) is given in (20). .. (20) [vp BE [vp John vO [vp eat]]] .. In (20), the lowest VO has [uPred] and probes eat, which has [iPred], checking and valuing

the feature on va. Now, since BE is also inherently a verbal element, it also has the .. [uPred] feature, so it probes the vP that it dominates, checking [uPred] of BE against the

newly valued feature [iPred] of the embedded va. This Agree relationship causes the .. embedded vO to incorporate into BE; after TO probes BE-vo to check [uPred] on TO, BE-vo

moves to ro and spells out as an inflected form of be. .. A similar set of relationships accounts for English sentences with the auxiliary

have, as in (21) (repeated from (10) above), with the structure given in (22). .. (21) John has eaten .. (22) [vp BE [vp John vO [vp eat]]]

In order to derive (21), the relationships are almost identical, with one minor difference. • In (22), the lower VO with [uPred] probes eat with [iPred], checking and valuing the

features on the light-verb head. Then, eat incorporates into vO, and BE checks its [uPred] .. against the vo_yo complex with [iPred]. Because eat has incorporated into va, this

complex does not further incorporate with BE, so BE interacts with TO and spells out as .. an inflected form of have.

At first glance, this system seems rather stipulative, but it has significant benefits. .. We can reduce the differences between have and be in English to a minor issue of head

movement/incorporation, utilizing the basic ideas of Kayne (1993) without the need for • added functional projections. One question that you may ask is regarding the differences

in verb forms between (9) and (11): why do participial YOs incorporate into vO while .. progressive forms do not? I believe that my system accounts for this issue in a nice way

as well. In (9) and (11), eat takes its progressive and participial forms, respectfully. Now .. ..

[62] ..

..

.. Auxiliary/Main Verb Asymmetries in English

.. if these forms are not base-generated as such, but rather are generated simply as non-

finite eat, we can assume that the incorporation process affects how the individual forms

.. are spelled out. Thus, in (9), VO does not incorporate into VO but VO incorporates into BE;

the lack of yO_vo incorporation causes eat to spell out as eating, and the lack Va_va

incorporation allows for vO-BE incorporation, which gives us be and progressive aspect.

Conversely, in (11), yO_vo incorporation causes eat to spell out as eaten, and due to Va_va.. incorporation, VO-BE incorporation is not available, resulting in BE spelling out as have. 6

I do not have an answer to why the two different possibilities for incorporation .. .. result in different auxiliary forms, but it makes sense intuitively that if be and have are

the same (aspectual) head, a slight difference in the featural makeup of the syntax could

trigger differences in the morpho-phonological spelling of such heads. I leave this

technical problem unanswered, anticipating that future research will shed some light on.. the issue.7

Before I proceed to the next sub-section I would like to briefly discuss the

thematic- and Case- relations that are relevant for (19) and (21). In (20), the only DP that -we need to consider is John, which receives its 8-role from va. I also assume that the .. matrix TO probes John, checking the un interpretable <p-features ofr and the Case feature

of John, and John moves to [Spec, TP]. In this case, John does not move through the .. specifier associated with BE because John has a 8-role in [Spec, vP]. An identical set of

thematic/Case relations holds for structures like (22) as welL .. 3.3 COPULAR BE VS. POSSESSIVE HA VE

til In the previous sub-section, I developed the groundwork for an account of some data

regarding the auxiliary system of English. In order for this account to be explanatorily .. adequate, we must see how it handles other instances of have and be; a system that

.. 6 Throughout this paper, I systematically ignore passive instantiations of be. I believe that these instances of be could be accounted for within this system by modifying the featural makeup of the vO that immediately dominates the lexical verb. I leave the derivation of these constructions to future research.

..

.. 7 In proposing that have and be are distinct spell-outs of a single light-verb head BE, the fact that have and be indicate different forms of aspect is obscured. It is possible that aspectual heads generated in the inflectional field higher than the verbal projections (cf. Cinque, 1999), coupled with head movement to BE, trigger this variable spell-out. I leave this possibility to future refinement of the analysis proposed in this paper .

.. [63]

..

• J. Kelly

•involves homophonous elements that are derived from completely different structures

should be avoided, if at all possible. However, it is evident that the account of sentences •like (19) and (21) will differ somewhat from what we develop for sentences like (23) and

(24), repeated from (13) and (7) respectively. • (23) John is a man

(24) John has a book • Let me begin by deriving (23). Of course, the structure of (23), given in (25), is • necessarily different than its counterpart containing auxiliary be in (20).

•(25) [vp BE [PredP John Predo a man]]]

In Section 2.2, I discussed some issues associated with predication and the Predo element. • Before I discuss (25), it is necessary to expand the discussion of Predo. First, note that my

structure for copular constructions is consistent with that laid out by Baker (2003), with a • light-verb head VO merged with a PredP. Now, we must consider what roles Predo and vO

play in (25). • In Section 3.1, I showed that inherently verbal elements have a [Pred] feature that

is similar to the V-features discussed by Chomsky (1995). I want to extend this feature to •Predo as well. The reason for attributing [Pred] to Pred° follows from our assumptions

about Baker's predicational head; verbal elements serve to set up a relation between other •elements in a sentence, and Predo is simply a covert analogue of a verbal element. For

this reason, Predo licenses a specifier (see Baker, 2003, 2006 for details regarding his

theory of lexical categories). I also assume a conventional notion of predication as • instantiated in Predo, namely that Pre dO is associated with a a-role and a "predicate." In .. other words, the specifier of Predo has a a-role and the complement of Predo gets

something akin to "predicational force." This notion of "predicate" associated with the •complement of PredO is necessary because the elements that typically merge as the

complement of Predo (i.e., DPs, APs, and PPs) are not inherently predicational, so this •trait must be inherited from Predo. Thus, for the complement of Predo to function as a

predicate, Predo must impart this trait upon its complement. I hypothesize that [iPred] is

what allows Predo to function in this manner. This set of relations is schernatized in (26). • [64] •

- -

PredP -Auxiliary/Main Verb Asymmetries in English

(26) .. 

G-role PradO - 'predlcate"

Now that I have made explicit the precise features of Predo in this system, we can easily .. account for (23) via the structure in (25), repeated here. .. (25) [vp BE [PredP John Predo a man]]]

.. In (25), Predo has [iPred], so it transmits a a-role to John and gives predicational force to - a man. When BE is merged with PredP, BE has [uPred], which can check against the

[iPred] of Predo. This feature checking causes Predo to incorporate into BE, and after the

BE-Predo complex interacts with the matrix TO, the complex spells out as be. In (25),

John receives a a-role in [Spec, PredP] and moves to [Spec, TP] after Agree holds .. between John and TO for checking. I assume that when a man receives the

- predicational force from Predo, this serves as a kind of Case-checking operation.

I propose that the underlying structure for clauses involving possessive have is identical

to the structure found in (25), with one crucial exception. For these possessive- constructions, I assume that Predo lacks the [iPred] feature. With this feature absent on

Predo, the relationships necessary for convergence are a bit different. We saw that [iPred] .. on Predo allows for the relations schematized in (26), so when [iPred] is absent, we can

think of Predo as defective, similar to the defective TO proposed by Chomsky (2000, .. 2001). When we have a defective Predo, the effects mirror Burzio's Generalization; Predo

lacking [iPred] fails to assign a a-role to its specifier, and it also fails to give.. predicational force to its complement. The structure corresponding to such situations are

given in (27). -(27) [vp BE [PredP Predo a book]]] ..

.. [65]

-

III

• J. Kelly

• Before I can discuss how sentences with possessive have are derived, I must first discuss

the nature of defective elements, like defective TO. Consider a sentence like (28a), with •the structure in (28b).

(28) a. John seemed to win

b. [TP TO [vp seem [TP TO [vp John vO [vp win]]]]]

" Chomsky (2001) states that there are two possibilities for movement of John out of the

embedded clause. Under one alternative, the DP must move through the intermediate • [Spec, TP], although the ro head of the embedded clause is defective (Tdef); Tdef contains

a partial set of <I>-features (perhaps only [person], following Chomsky (2000, p. 124» • which must be checked against the DP which has a full set of <I>-features {[person],

[ number]}. Only total agreement between a head and an XP licenses <1>-feature deletion of • the XP (and subsequent Case checking), so the DP remains with <I>-features, making it

visible for further Agree operations (i.e., the DP may continue its movement to higher • positions in the structure). Under the second alternative (Epstein and Seely, 1999), Tdef

does not project a specifier and therefore John moves directly from the embedded [Spec, -vP] to the matrix [Spec, TP]. The partial of <I>-features of Tdef are checked through Agree

with John. • In my analysis of (27), I assume the Epstein-Seely conceptualization of

defectiveness. So in (27), repeated here, Predo is defective, thus it does not assign a 9-role -to its specifier and the specifier is not projected.

" (27) [vp BE [PredP Predo a book]]

So if (27) is a valid intermediate-Merge structure for (24), two things must occur for the .. derivation to converge: John must be introduced into the derivation, and the [uPred]

feature of BE must be checked. Since Predo lacks [iPred], BE cannot enter into an Agree • relationship with Predo, and it seems like we are stuck. In this situation, I propose that

John can Merge as the specifier of BE, checking [uPred] of BE; this operation of " checking the interpretable feature on BE by Merging as the specifier of BE allows John

to pick up a 9-role from Predo. In these cases where a nominal element is merged in the • [66] •

- - Auxiliary/Main Verb Asymmetries in English

BE [u Pred]

f-tt+t 

.. specifier of BE and no verbal head incorporates into BE, BE spells out as have. This

process is schematized in (29) .

Ag ree checks q,-features of BE and case of a book .. (29)

/'OhO ,A,,,-.. Merge John fttPf'eetftt +I- 

.. Pred o This checks [ U Pred] /

[-Pred] l' q>J on BE and gives John Pred 0

a O-role [-Predj I' '1'1 no predicational force to a book BE spells out

as haveno specifier (and no extemal !l-role)

Another effect of the defectiveness of Predo in (27) (i.e., the lack of [iPredD is that a book .. receives no predicational force, and since predicational force is a Case-like feature (for

nominals), BE must enter into an Agree relationship with a book for cp-feature/Case .. checking. 8

So in concluding this sub-section, we can see how the optionality of the [iPred] .. feature on Predo drives the difference between copular be and possessive have in English.

In the next sub-section, I discuss how the same system accounts for predicate adjectives .. in similar constructions . .. 3.4 PREDICATE ADJECTIVES

In the previous three sub-sections, we have seen how a predicational feature on inherently

verbal heads, coupled with head-incorporation, can account for the differences between -auxiliary constructions, as well as the differences between copular be and possessive .. have, in English. The analysis in Section 3.3 lends itself particularly well to an analysis of

predicate adjectives in English, which is the goal of this sub-section_ I will be particularly

concerned with licit sentences like (30) and illicit sentences like (31) in this sub-section. --

8 I assume that the Agree relationship between BE and Predo in (25) that checks [uPred] of BE also checks the un interpretable <\I-features of BE, so BE does not enter into an agree relationship with a man, the complement of PredO. -

- [67]

-

., J. Kelly

• (30) John is happy .,(31) * John has happy

To begin, I would like to propose an analysis of (30) that follows naturally from the • analysis of (23) above. Consider the structure for (30), given in (32). .. (32) [vp BE [PredP John Predo happy]]]

•As with (23), Predo has [iPred], thus it assigns a 8-role to John and predicational force to

the adjective happy. Then, BE (with the feature [uPred]) enters into an Agree relation •with Predo, checking the uninterpretable feature on BE. Finally, the matrix r enters into

an Agree relation with John, checking the relevant <jl-features and Case, and John moves •to [Spec, TP]. Recall that John does not move through the specifier position of BE

because John already has a 8-role. ., Now, for the derivation of (31), we begin with a structure similar to that for (24),

namely one where Predo lacks [iPred] and thus lacks the ability to project a specifier and .. provide its complement with predicational force. The structure for (31) is given in (33).

(33) [vp BE [PredP Predo happy]] • .,

In Section 3.3, I discussed the idea that predicational force is not an inherent trait of

adjectives. In order for (predicate) adjectives like that in (31) to be licensed, they must be .,licensed through a relationship with Predo that contains [iPred].9 This contrasts with the

typically accepted licensing condition for DPs, namely that they must have Case. Of

course, yO generally has [iPred], but if VO selects a complement, a 8-role is associated • with that complement position; that 8-role would go unassigned if a predicate adjective

were merged as the complement of a yO, leading to the ungrammaticality of sentences .. like (34). ., (34) *John bought happy

9 I abstract away from the possibility of a licensing condition for attributive adjectives in this paper. •

[68] • .,

- Auxiliary/Main Verb Asymmetries in English

,. The reason why (31) cannot converge lies in the fact that the adjective happy is not

assigned predicational force by the defective Predo. All other operations should function -properly: BE does not enter into an Agree relationship with happy because the adjective .. does not have $-features. BE cannot enter into an Agree relationship with Predo because

Predo lacks [iPred]. John would Merge as the specifier of BE, allowing John to pick up a

O-role and checking [uPred] on BE, causing BE to spell out as have, and the matrix TO -would then enter into an Agree relationship with John, checking the relevant $-features ..

-and Case.

To conclude this sub-section, I would like to discuss instances when predicate

adjectives can appear in constructions where they directly follow the verb, as in (35), or

they directly follow the complement of a verb, as in (36). ,. (35) John ate asleep (meaning that John was effectively asleep while eating)

(36) John ate dinner asleep -.. I believe that there are two possible ways to account for these structures. I will

concentrate on (35), leaving a more detailed analysis of these constructions for further

research. The first possibility is one where Predo lacks [iPred], with the structure given in .. (37). .. (37) [TP John TO [vp Jelm vO-ate [vp [vp ate] [PredP Predo asleep]]]]

.. In (37), PredP is a secondary prediacte adjoined to VP, and since Predo lacks [iPred],

-Predo does not project a specifier and asleep does not receive predicational force internal

to PredP. We could hypothesize that asleep receives this force from the higher verbal

.. head vO, and John can merge directly in [Spec, vP] to receive its O-role before movement

to [Spec, TP] .

- The second possible account that I can imagine for sentences like (35) would

employ sideward movement (Nunes, 2004). Crucially, for a system like this to work, we

- must also assume that movement into O-position is a licit operation, as has been argued

for quite convincingly by Hornstein (1999; 2001). Let me briefly sketch how sideward

movement could account for (35) . .. [69]

-

III!! J. Kelly

• Assuming that PredP is adjoined to vP, the first step in our sideward-movement

derivation is to Merge the PredP. In this case, and differing from (37), Predo would have • [iPred], so asleep could receive predicational force from Predo and John receives a a-role

in [Spec, PredP]. In parallel to the construction of the PredP, the vP would also be •constructed, up until the point where vOmerges with VP. At this point in the derivation,

two distinct subtrees are present; John is copied and merged from its position in [Spec, • PredP] to the position in [Spec, vP], where it receives a second a-role. Then, PredP is

adjoined to the vP. The derivation fmishes up with movement of John to [Spec, TP] after •Agree holds between TO and John.

As we can see, it is possible to account for sentences where adjectives follow the .. verb, as in (35), without affecting the claims I made about the licensing predicate adjectives in constructions with copular be and possessive have. My analysis of (35) is ... not fully developed, but is a topic left for further investigation.

•3.5 SUMMARY

In Section 3 of this paper, I have utilized previous implementations of syntactic theory,

along with some novel theory of my own, to provide a unified treatment of be and have in • English. Namely, I have integrated Kayne's (1993) influential notion of BE, a null

auxiliary, and his system of head-incorporation into BE that results in the variable spell -out of the auxiliary, into a system that makes explicit some ideas about the features of .. verbal heads in the spirit of early Minimalist work. I have shown that the presence or

absence of this verbal feature [Pred] can have a far-reaching impact on the syntax, and ..the theory I have developed shows a fairly wide range of empirical coverage. In the next

section, I will show how my proposal can account for the long-standing issue of the

asymmetry between auxiliary and main-verb uses of have, as contrasted with uses of be, • when interacting with negation. .. 4. THE AUXILlARYIMAIN-VERB ASYMMETRY

III!!In this section, my goal is to describe a system that can account for the behavior of have

and be when interacting with negation. The discussion relies crucially on the theory ..developed in Section 3. To summarize, have and be are simply different instantiations of

[70] •

"

- Auxiliary/Main Verb Asymmetries in English ..

.. the abstract element BE. When a verbal head incorporates into BE, BE spells out as be.

When no incorporation occurs, BE spells out as have. 10

.. The data that I will be centrally concerned with in this section are given in (38)-(41).11,12

(38)-.. (39)

.. (40)

(41) -..

a. John is eating b. John isn't eating

a. John is a man b. John isn't a man

a. John has eaten b. John hasn't eaten

a. John has a book b. *John hasn't a book c. John doesn't have a book

Cashing in on generally accepted descriptions of (38) - (41), the relevant generalization .. is as follows: be always raises through negation to TO, as an auxiliary or a main verb, as

in (38) and (39), while have raises through negation to TO when it functions as an

auxiliary, as in (40), but fails to raise to raise through negation to TO when it functions as -.. 10 Note that this system is quite different than that ofKayne (1993), who predicts the opposite. For Kayne,

when a head incorporates into BE, BE spells out as have, but when no element incorporates into BE, BE spells out as be. It is worth reiterating that BE simply stands for VO that merges with vP or PredP as opposed to VP, so the fact that BE relies on incorporation to be realized as be in my system does not conflict with .. the innate nature of BE in any way . II Throughout this paper, I ignore many of the inconsistencies of be when it interacts with negation. For instance, the fact that n'l cannot cIiticize onto am (i.e., *1 anm't) is particularly puzzling. 1 assume that this is not a syntactic phenomenon, but rather occurs in the morphological component of the grammar and is

..

.. conditioned by lexico-syntactic features. For some discussion of the issue, see Kayne (1989). For a discussion of post-syntactic morphological frameworks, I refer the reader to work on Distributed Morphology (Halle and Marantz, 1993; Marantz, 1997) and Lexeme-Morpheme Base Morphology (Beard, 1995) . 12 In this paper, I do not discuss issues of English negation not. I assume that not is a specifier of a functional phrase NegP, while n 'I is the head ofNegP. Works on the syntax of negation, such as Zanuttini (1997) allow for this possibility. It would follow that ifnot is a specifier, it would not interact morpho-phonologically with raised verbal heads because these heads would move through the empty NegO head to - TO; this would allow for these raised verbal elements to interact with the subject (cf. You're not happy, You've done it again). This contrasts with the cases where n't, a NegO, is present; in these instances, a ..

... [71 ]

-

• J. Kelly

• a main verb. In the case of (41), do-support is necessary, as is the case with all other

lexical verbs in English. •Before I explain how my theory can account for the data in (38) - (41), I would

like to discuss an issue of phases that is central to my theory; this is the notion ofphase •strength. In his discussion of the nature of phases, Chomsky (2001, 14) alludes to this

notion: "Some phases are strong and others weak- with or without the EPP option, •respectively, hence relevant or not for Spell-Out." I believe that this claim can provide us

with some insight as to the asymmetry illustrated by the data in (38) - (41). I will assume •that Chomsky's reference to the "EPP option" reduces to the projection of a specifier, as

this will be maximally relevant to my discussion.

Another recent development in syntactic theory that will be relevant in this -section is the notion that head movement may, in some instances, be a PF phenomenon.

This was alluded to by Chomsky (2000) and discussed by Boeckx and Stjepanovic • (2001). I attempt to show that although this may be the case, PF head movement should .. still be constrained by the syntax in some manner.

Now, let me begin to derive the asymmetry found in (38) (41). • •4.1 AUXILIARY BE AND HAVE AND NEGATION

In this sub-section, I will discuss how have and be interact with the negative clitic n't.

First, let's consider the structures that I derived for (38a) and (39a), repeated here from • Section 3.

• (42) [vp BE [vp John VO [vp eating]]]

(43) [vp BE [PredP John Predo a man]]] • For both of these structures, I predicted that thematic relations were satisfied inside the .. lower vP and PredP, respectfully. Thus, when John moves out of the clause, it does not .. raised verbal head must move through NegO on its way to 'fl, thus providing the familiar instances of negation cliticized onto have and be discussed in this paper. • ..

[72] ..

.. - Auxiliary/Main Verb Asymmetries in English

move through the specifier of BE in either (42) or (43). In both instances, BE is spelled .. out as be as described in Section 3 .

The same structures are present in the negated forms of the sentences (38b) and .. (39b). These structures are given in (44) and (45) .

- (44) [TP John r [NegP n'tO [vp BE [vp JeI:m VO [vp eating]]]]]

(45) [TP John TO [NegP n'tO [vp BE [PredP JeI:m Predo a man]]]]]

- Considering the highly parallel structures in (44) and (45), where John moves directly

from its base-generated position to [Spec, TP] and BE spells out as be (due to the - incorporation of VO and Predo), it is evident that n'l can c1iticize onto the inflected form of

BE. Note that the vP associated with BE in (44) and (45) does not project a specifier, and

,.. thus is a weak phase. In the spirit of Chomsky's discussion of weak phases as not

"relevant" to Spell-Out, I assume that weak phases do not act as barriers to movement, - and thus the movement of John from its base generated position to [Spec, TP] is not a

violation ofthe Phase Impenetrability Condition (PIC) (Chomsky, 2000). - Now, let's look at the structures associated with (40a) and (4Ia), repeated here

from Section 3 as (46) and (47). - (46) [vp BE [vp John VO [vp eaten]]]

.. (47) [vp John BE [PredP Predo a book]]

The structure in (46) is almost identical to that found in (42), the only difference being ... the method of verbal-head incorporation. In (46), no verbal element incorporates into BE,

so BE spells out as have. John receives a a-role internal to the PredP, so John can move - directly to [Spec, TP). In (47), the case is a bit different. As we said in Section 3, Predo in

(47) is defective in lacking the [iPred] feature. This defectiveness means that Predo does - not project a specifier and cannot check [uPred] of BE. An alternative method of

checking [uPred] of BE is available, namely merging John directly as the specifier of BE.

- This operation satisfies the thematic requirement of John and checks [uPred] on BE, and

John can move from this position to [Spec, TP] . .. ..

[73]

-

• J. Kelly - When structures like (46) and (47) are embedded under negation, we get

structures like those in (48) and (49). • (48) [TP John TO [NegP n'tO[vp BE [vp Jelm VO [vp eaten]]]]] ..(49) [TP John TO [NegP n'tO[vp JelmBE [PredP Predo a book]]]]

The crucial difference between (48) and (49) can be found by examining the status of the • phrase headed by BE in each case. In (48), John does not move through the specifier of

BE, and when BE spells out as have, have may raise to TO with the corresponding • cliticization of n 't. In (49), on the other hand, John must raise through the specifier of BE

in order to receive a a-role and check [uPred] on BE. Since BE projects a specifier in • (49), it functions as a strong phase, while in (48), BE projects no specifier and thus is a

weak phase. Now, if we step back and examine (48) and (49) together with (44) and (45), • a pattern seems to emerge. In (44), (45) and (48), the vP headed by BE is a weak phase,

and BE (plus any incorporated heads) may raise out of the vP through negation (with • cliticization of n 't) to TO, while in (49), the vP headed by BE is a strong phase, and verb

raising is impossible. Instead, do-support is necessary for convergence (cf. the contrast • between (41 b) and ( 41 c)). So there seems to be some correlation between phase strength

and the ability of a verb to raise out of the phase. .. Now, let me sketch my hypothesis regarding this observation. If head movement

is indeed a PF phenomenon, it is possible that PF movement could still be constrained by • phase boundaries. Chomsky claims that "Some phases are strong and others weak ...

hence relevant or not for Spell-Out" Now we know from the definition of the phase that .. the head of a (strong) phase becomes "inert" in that it cannot trigger further operations,

but the head of a phase is still accessible as the goal of operations initiated by heads in a •higher phase. Since a strong phase is "relevant" for Spell-Out, it is possible that although

the head of a phase may be syntactically "active" for further computation, it is "frozen" .. by Spell-Out and thus may not undergo PF operations. This is the heart of my proposal to

account for (44) - (45) and (48) (49). In (44) (45) and (48), BE is the head ofa weak .. phase (which is not relevant for Spell-Out) so BE can undergo PF movement through

negation to TO, whereas in (49), BE is the head of a strong phase and thus is frozen, not .. able to undergo PF movement out of its vP.

[74] • ..

- - Auxiliary/Main Verb Asymmetries in English

.. -

Let me elaborate a bit more on this discussion, since it is the fundamental issue

that I am addressing in this paper. The intuition behind my theory is fairly simple: Agree

is a syntactic operation that drives head movement, but head movement is a PF

phenomenon that is constrained by syntactic factors. 13 The syntactic constraint on head

movement can be phrased as in (50) .

...

-(50) PF Head Movement Constraint

Head movement at PF cannot occur across strong phase boundaries.

-...

With this constraint in place, I would like to discuss one other syntactic constraint related to phases that is relevant to my discussion. Phases are a naturally occurring locality

constraint on the syntactic component of the grammar. I believe that there is also a

locality constraint on phases themselves; adjacent phase boundaries are incompatible. So

the structure in (51), where a and are heads of phases, is syntactically illicit. The

'* relevant constraint is given in (52).

... (51)

(52)

[aP a [I3P 

Phase Anti-locality Constraint

... Where a. and are heads ofphases: "''[aP a [I3P 

- In a situation like (51), where two phase boundaries are adjacent, a patch on the grammar

..

.. -

is available. When a merges with if a probes and an Agree relation holds, loses

its ability to head a phase; in (51), loses its status as a phase, and aP is the only phase

relevant to the computational system. With these constraints in place, we can derive the

head movement facts in a variety of English cases. Let's run through the derivations for

the data central to my argument. First, consider (53).

-...

13 The reader should note that my theory of head movement is driven by syntactic feature checking, contrary to the claim by Boeckx and Stjepanovic that head movement may be prosodically driven. If we assume a post-syntactic morphological component that recognizes (the features of) syntactic heads, post-syntactic head movement would naturally follow. However, prosodic ally driven head movement does not appear to be a viable option theoretically, because prosodic operations typically occur post-morphologically, when syntactic heads are no longer available to the computational system .

- [75]

WII'

.. J. Kelly

• (53) John is eating

[TP John TO [vp BE [vp JeI:Ift VO [vp eat]]]] ..I I I I [uPred] [uPred] [uPred] [iPred]

•In (53), VO probes eat to check its [uPred] feature, and the [uPred] feature ofvo is valued,

becoming [iPred]. Then, BE probes yO, with the newly valued [iPred] of VO valuing •[uPred] of BE; this operation ameliorates the structure prohibited by (52), and the vP

headed by vO loses its status as a phase. Now, head movement (and incorporation) ofvo •into BE obtains, and this incorporation forces BE to spell-out as be. Finally, John moves

to [Spec, TP] without moving through the specifier of the phrase headed by BE. Since BE .. fails to project a specifier, the phrase headed by BE is a weak phase, so when TO probes

BE+vo to check [uPred] of TO, BE+vo raises to T>. The incorporation of vO into BE causes .. the morphological spell out of eat as eating, thus deriving John is eating.

A similar derivation proceeds for sentences like (54), where negation is introduced. • (54) John isn't eating

[TP John TO [NegP n'tO[vp BE [vp JeI:Ift vO [vp eat]]]]] .. I I I I

[uPred] [uPred] [uPred] [iPred] .. In such sentences, when TO probes BE to check [uPred] of TO, BE may move (through PF

head movement) through negation to TO since the maximal projection dominating BE is a -weak phase.

&' Sentences involving have are derived in a similar manner. Consider (55).

(55) John has eaten .. [TP John TO [vp BE [vp JeI:Ift VO [vp eat]]]]

I I I I [uPred] [uPred] [uPred] [iPred] ..

OIn (55), the derivation is quite similar to that of (53), with slight differences. To begin v .. probes eat, checking [uPred] on yO, and eat incorporates (through PF head movement) to

yO. BE then probes vO+eat (with its newly valued [iPred] feature), checking and valuing

[76] • .. III

- - Auxiliary/Main Verb Asymmetries in English

[uPred] on BE; this operation ameliorates the structure prohibited by (52), and the vP .. headed by vO loses its status as a phase. Then, John moves to [Spec, TP] without

movement through the maximal projection associated with BE, and probes BE to

- check [uPred] on TO; BE moves to TO (since the phase headed by BE is weak), and the

movement of BE (without the incorporation of vO) to forces the spell out of eat as

- eaten. In a structure like (56), where negation is present, a similar derivation holds; when

TO probes BE to check [uPred] on BE moves through negation to picking up n't

- through head to head PF movement. Since John doesn't move through the maximal

projection associated with BE, said projection is a weak phase so the head movement of

- BE is licit. In sentences like (55), an identical operation holds; the predicational feature of

BE has been valued due to the checking with vO, so TO checks its [uPred] feature against

- the [iPred] feature of BE, and since BE is the head of a weak phase, the movement of BE

out of its maximal projection through negation to TO is licit. In both (55) and (56), since

- no verbal head has incorporated into BE before its movement to TO, BE spells out as

have.

- (56) John hasn't eaten

[TP John TO [NegP n'tO [vp BE [vp JeIm VO [vp eat]]]]] .. I [uPred]

I [uPred]

I I [uPred] [iPred]

.. Now we tum to situations where BE merges with a predicational structure. First, consider

the example in (57). - (57) John is a man

,. [TP John TO I

[vpBE [PredP JeIm Predo a man]]] I I

[uPred] [uPred] [iPred] - In (57), Predo has [iPred] so it can assign a E)-role to John and assign predicational force

- to a man. Then, BE probes Predo to check [uPred] on BE, and Pre dO moves to BE through

PF head movement. Since John receives its E)-role from Predo, John moves to [Spec, TP]

- without moving through the specifier of the maximal projection associated with BE.

Finally, TO probes BE+Predo to check [uPred] of TO against the newly valued [iPred] of

- [77]

,..

• J. Kelly .. BE+PredO, and BE moves to TO via PF head movement. When negation is introdueed into

the derivation, as in (58), the same relationships hold; TO probes BE (to check [uPred] of .. TO against the newly valued [iPred] of BE), and since BE is the head of a weak phase, BE

moves through negation to TO, with the final spell out John isn't a man. .. (58) John isn't a man .,

[TP John  [NegP n'tO[vp BE [PredP :JeI:ffi Predo a man]]]] I I I

[uPred] [uPred] [iPred] .. On the other hand, with sentences like (59), Predo lacks a predicational feature, so Predo

does not license a specifier and a book cannot receive predicational force. In such .. situations, John merges with the projection dominating BE, and this serves to give John a

a-role and to check [uPred] on BE, and BE checks Case on a book. Now that the • predicational feature of BE is valued, TO may enter into an Agree relationship with BE to ,.check [uPred] on TO, but BE may not move to TO because the maximal projection of BE is

a strong phase (i.e., it has a specifier). BE spells out as have since no verbal head

incorporates with BE, and no stranded-affix filter effects arise from a configuration like .. (59) because TO and BE are adjacent.

(59) John has a book -[TP John  [vp:JeI:ffi BE [PredP Predo a book]]]

I I -[uPred] [uPred] .. This contrasts with the structure in (60), where negation intervenes between BE and 

Since BE is the head of a strong phase, it may not move through negation to ro, and since •TO and BE are not adjacent (due to the intervening n 't), do-support is necessary to rescue

the derivation. .. (60) John doesn't have a book ..[TP John TO [NegP n'tO[vp:JeI:ffi BE [PredP Predo a book]]]]

I I [uPred] [uPred] ..

.. [78]

.,

.. - Auxiliary/Main Verb Asymmetries in English

The analysis presented for (59) and (60) carries to typical English verbs as well. Consider .. (61).

.. (61) John eats

[TP John TO [vp :J.e.ltft VO [vp eat]]] I I I

[uPred] [uPred] [iPred] - In a situation like (61), VO enters into an Agree relationship with eat to check [uPred] of

vO, eat moves to vO, and [uPred] on VO becomes valued as [iPred]. Then, TO enters into an - Agree relationship with vO to check [uPred] on TO, but VO is the head of a strong phase, so

vO may not move to TO. As with (59), no stranded-affix filter type violation results - because TO and VO are adjacent heads.

Conversely, in structures like (62) when negation is present, VO cannot move to TO.. because vO is the head of a strong phase, and vO and TO are not adjacent due to the

intervening n't, so do-support is a necessary last-resort operation for convergence. ,. (62) John doesn't eat

_ [TP John TO [NegP n'tO [vP:J.e.ltft vO [vp eat]]]]

.. [uPred] [uPred] [iPred]

.. To conclude this section, I would like to call attention to the contrast in the spell-out

forms of the thematic verbs in (55)/(56) and (61)/(62). In both pairs of sentences, the

thematic verb moves to vO, but the eat is spelled out as the participle in the first pair of- sentences, while in the second pair, eat is spelled out as an inflected form. This contrast

reduces to the presence of BE in the first pair of sentences. In (55)/(56), [iPred] of vO ,. checks the predicational feature of BE, while in (60)/(61), [iPred] of VO checks the

relevant feature of TO; it is precisely this difference in checking relationships that causes - the different morphological forms ofeat to be spelled out. .. .. - [79]

...

J. Kelly

..

.. 5. INFINITIVAL CLAUSES

Now that I provided a theory to account for various instantiations of be and have in •English, I want to illustrate how my system can account for Control constructions

involving have (and marginally, be). First consider (63). • (63) John has to eat

[TP John T'l [vp Jeha BE [PredP Jeha Predo b[TP PRO TO-to [vp PRG VO [vp eat]]]]]]] -I I I I I [uPred] [uPred] [uPred] [uPred][iPred] ..

In deriving sentences like (63), I assume that a PredP (with [PRO to eat] as its

complement) is generated, and the head of the PredP lacks the predicational feature. ,. Within the sentential complement of PredO, V

O probes eat to check [uPred] on vO and TO, in

tum, probes vO for the same reason. Since Predo lacks the predicational feature, it does not .. project a specifier and no predicational force is assigned to the CP complement of Predo

(I assume that the CP is inherently predicational (or propositional) and thus does not need • to receive predicational force from Predo, in contrast with adjectives). BE cannot check

its [uPred] feature against Predo, nor can BE check its feature with a head lower in the .. structure due to locality/minimality constraints. However, John can merge as the specifier

of the projection headed by BE, thus checking the [uPred] feature of BE and receiving a

8-role. This movement causes the vP headed by BE to be a strong phase. The TO can -check its [uPred] against the newly valued feature of BE, but BE does not move to  due .. to (50). The merger of John as the specifier of BE forces BE to spell out as have. In

sentences like (64) where negation is present, the status of the vP headed by BE as a .. strong phase comes into play; BE cannot raise through n't to TO and BE and TO aren't

adjacent, so do-support is a necessary last-resort operation (cf. *John hasn't to eat). .. (64) John doesn't have to eat

[TP John r [NegP n't [vp Jeha BE [PredP Predo

I I [uPred] [uPred]

[cP[TP PRO T'l-to [vp PRG VO [vp eat]]]]]]]] .. I I I

[uPred] [uPred][iPred] .. .. ..

[80] ..

- - Auxiliary/Main Verb Asymmetries in English

Now, consider cases where be can be used in a Control structure, as in (65). Such .. constructions aren't prevalent in my dialect of English (they seem archaic or extremely

formal), but they nevertheless exist.

--

(65) John is to eat (at 6 o'clock) [TP John TO [vp BE [PredP Jeflft Predo bb PRO r-to [vp PRG vO [vp eat]]]]]]]

I I I I I I [uPred] [uPredJ [iPred] [uPred] [uPred][iPred]

- The crucial difference between sentences like (63) and (65) is the presence of an

- interpretable predicational feature on Predo. Within the CP complement of Predo, the

same checking relationships hold, and since Predo has [iPred], John receives a a-role

- internal to the PredP. BE checks its uninterpretable predicational feature against the

interpretable predicational feature of Predo with Predo incorporating into BE, forcing the

- spell out of BE as be. Finally, TO checks its [uPred] against the newly valued [iPred] of

BE+Predo, and John moves to [Spec, TP]; since John did not move through the specifier

of the vP associated with BE, said vP is a weak phase, so BE+Predo moves overtly to rD. .. ..

When negation is present in these constructions, as in (66), the status of the vP associated

with BE as a weak phase is relevant; in such constructions, BE can move out of the weak

phase through negation to TO, so do-support is not necessary .

--

(65) John isn't to eat (after 10 o'clock) [rp John r [Negp n't [vp BE [PredP J.ehH Predo (CP[TP PRO TO-to [vp PRG VO [vp eat]]]]]]]]

I I I I I I [uPred] [uPred] [iPred] [uPred] [uPred][iPred]

- 6. MORE ON SMALL CLAUSES

If small clauses are properly treated as PredPs, an analysis of these constructions should

- also follow from the theory that I have developed in this paper. First, let's look at some

standard examples of small clause constructions. Consider (66).

- (66) John considers Mark a friend

- [TP John [vp Je.l:m: vO [vp consider [PredP Mark Predo a friend]]]] I I I

[uPred] [iPred] [iPred]

- [81 )

-

J. Kelly

..

.. In (66), I need to assume that Predo has the predicational feature [iPred] , and I will

generalize this assumption to all small clause constructions. Within the PredP, Mark • receives a 9-role and a friend receives predicational force, all stemming from the

interpretable predicational feature on Predo. Since the relevant feature on Predo is .. interpretable, this feature does not need to be checked, and the checking procedure for the

predicational features associated with vO and consider is straightforward. Note that .. selectional restrictions of consider are satisfied because the PredP is propositional, which

is required of the complement of consider. A similar analysis can be assumed for .. sentences like (67). ..(67) John considers Mark smart

[TP John [vI' Jelm VO [vp consider [?redP Mark Predo smart]]]] I I I ..

[uPred] [iPred] [iPred] ..Other constructions involving verbal small clause constructions can also be accounted for

within the system that I have developed. Consider (68).

• (68) John saw Mark fight Ken

[TP John r [vp Jelm vO [vp see [vp Mark VO [vp fight Ken]]]]] •I I I I I [uPred] [uPred] [iPred] [uPred] [iPred] ..

In (68), both see andfight have interpretable predicational features that serve to check the

un interpretable features of dominating heads. In the embedded vP, Mark receives its .. 9-role from VO and Ken receives its 9-role from fight, with Case checking on Ken via the

lower vO and Case checking on Mark via the upper vO. .. 7. ECM CONSTRUCTIONS INVOLVING HA VE AND BE .. In Section 6, I presented an analysis of some small clause constructions in English. In this

section, a similar analysis carries over to certain ECM constructions. First, consider (69). .. (69) John considers Mark to be a friend

hp John TO [vp Jehft VO [vp consider [TP Mark ro-to [vp BE [Pred Mafk Predo a friend]]]]]] til!

.. [82]

- Auxiliary/Main Verb Asymmetries in English -

In sentences like (69), Predo has the interpretable predicational feature, giving Mark a- O-role and assigning predicational force to a friend; the feature [iPred] on Predo also

checks [uPred] on BE, with incorporation of Predo into BE, so that BE spells out as be.- Since the embedded TO is non-finite, no head movement obtains between BE and TO,

Thus, when negation is present, not is obligatory because the clitic n '[ cannot be- supported by BE (which fails to raise) or do (which cannot be inserted in non-finite TO),14

This sort of analysis can also account for sentences like (70). -(71) John considers Mark to be smart

_ [TP John Tl [vp JeI:ta VO [vp consider [TP Mark Tl-to [vp BE [pred Mafk: Predo a friend]]]]]]

Likewise, with ECM constructions involving have, the analysis is comparable, with the - only difference being that Predo lacks the predicational feature. Consider (72). - (72) John considers Mark to have a problem

[n John 1" [vp JeI:ta VO [vp consider [n Mark Tl-to (vp Mafk: BE [Pred Predo a problem] .. In cases like (72), Predo lacks [iPred], so no specifier is projected and a problem fails to .. receive predicational force. In addition, the lack of the feature on Predo forces Mark to

merge as the specifier of the projection headed by BE so that Mark can be assigned a

O-role and so BE can have its [uPred] feature checked. This merger forces BE to spell out - as have. Case on Mark is checked via the upper vO and Case on a problem is checked .. through the lower vO.

The theory that I have developed can also handle ECM constructions involving .. progressive and participial embedded clauses. First, let's look at a progressive

construction, as in (73).

- (73) John considers Mark to be eating

[TP John 1" [vp JeI:ta VO [vp consider [TP Mark r-to [vp BE (vpMark VO [vp eat]]]]]]]-14 This is evident in sentences like (i), although neg-raising often occurs.

(i) John considers Mark to not be a friend -- [83]

-

III

.. J. Kelly .. In cases like (73) in the embedded clause, eat with [iPred] checks [uPred] on vO, and VO ..incorporates into BE, causing the spell out of BE as be and eat as eating. This

incorporation process checks [uPred] on BE since VO has a recently valued predicational

feature. In turn, the valued feature of BE serves to check [uPred] on TO. A similar form of

checking relationship holds between consider, VO and TO in the matrix clause of (73). A ..parallel derivation holds for sentences like (74), with the only difference being that eat

incorporates into the embedded vo, forcing eat to spell out as eating. This incorporation .,process precludes VO from incorporating to BE, so BE spells out as have.

(74) John considers Mark to have eaten .. b John ro [vp Jelm: VO [vp consider [TP Mark TO-to [vp BE [vp Mark VO [vp eat]]]]]]] ..

8. CONCLUSION

In conclusion, let me summarize how the theory presented in this paper contributes to the .. field of syntactic research. I began by showing the familiar asymmetry of head movement

found in the various instances of be and have in English. I proposed a unified treatment .. of be and have by assuming a functional head BE that spells out differently given

different configurations involving head movement. I made the mechanisms syntactically .. explicit by utilizing a predicational feature, which is akin to the V-feature that was

present in Chomsky's (1995) early minimalist work. Finally, I showed how head .. movement, although conditioned and constrained syntactically, can be treated as a PF

phenomenon, and that these syntactically constrained notions of PF head movement can .,account for the long standing issue ofthe asymmetry between be and have. This theory is

also relevant for thematic verbs in English, as a PF head movement account also can

drive the need for do-support when negation is present with thematic verbs. It

Ultimately, research regarding other languages where head movement is a visible liltphenomenon will falsify or reinforce my claims. Some research regarding the

parameterization of phases, such as that of Angel Gallego (p.c.), claims that vP is not a ..phase in some Romance languages, but the relevant phase is TP. If this is the case, my

theory would be further supported, as the lack of a vP phase would naturally allow for the ..v-to-T movement that we see in Romance. Of course, much of the research regarding

[84] .. III

- - Auxiliary/Main Verb Asymmetries in English

topics like PF head movement and the nature of phases is relatively young. Hopefully, .. further investigation will shed further light on these phenomena, allowing us to attain a

better understanding of how the various linguistic modules are structured and how they - interact.

.. - - .. - .. --.. -.. - -- [85]

-

III!

J. Kelly

III! 9. REFERENCES

Adger, D. 2003. Core syntax. Cambridge: Cambridge University Press. .. Baker, M. 2003. Lexical categories. Cambridge: Cambridge University Press.

------. To Appear. The syntax ofagreement and concord. Ms. Rutgers University. III!

Boeckx, C. and S. Stjepanovic. 2001. Head-ing toward PF. Linguistic Inquiry 32: 345-III!355.

Chomsky, N. 1957. Syntactic structures. The Hague: Mouton. .. ------.1965. Aspects ofthe theory ofsyntax. Cambridge, Mass.: MIT Press.

------. 1995a. The Minimalist Program. Cambridge, Mass.: MIT Press. III!

------. 1995b. Bare Phrase Structure. In Government and Binding and the Minimalist Program, edited by G. Webelhuth, 383-429. Oxford: Blackwell. ..

------. 2000. Minimalist inquiries. In Step by step: Essays on minimalist syntax in honor ofHoward Lasnik, edited by R. Martin, D. Michaels and 1. Uriagereka, 89-156. III!

Cambridge, Mass.: MIT Press.

------. 2001. Derivation by phase. In Ken Hale: A life in language, edited by M. III!

Kenstowicz. Cambridge, Mass.: MIT Press.

Cinque, G. 1999. Adverbs andfunctional heads. Oxford: OUP. III!

Epstein, S. D. and D. Seely. 1999. SPEC-ifying the GB 'subject:' Eliminating A-chains and the EPP within a derivational model. Ms., University of Michigan and .. Eastern Michigan University.

Haegeman, L. 1991. Introduction to Government and Binding Theory. Oxford: III! Blackwell.

Halle, M. and A. Marantz. 1993. Distributed Morphology and the pieces of inflection. In .. The View from Building 20, edited by K. Hale and S. 1. Keyser. Cambridge, Mass: MIT Press. ..

Hornstein, N. To Appear. Deriving c-command. University of Maryland.

Kayne, R. 1989. Notes on English Agreement. In Parameters and Universals. Oxford: OUP.

------. 1993. Toward a modular theory of auxiliary selection. Studia Linguistica 47: 3-31. .. ..

[86]

III!

III


Recommended