Syntactic structures revised
This section (and the following) will focus on three issues in the analysis of the verbal morphology system from our current perspective:
(1)
Let us take this perspective to include the following assumptions:
(2)
Given (2a), the SS analysis of (1a), where the order of verbal elements is stipulated in the phrase structure rule, must be replaced. Given (2c), the SS analysis of (1b), where this fact is stipulated in the complicated SA of Tq, is not available, so (1b) must be captured in some other way. As noted in the previous section, it is not obvious that current assumptions preclude an account of (1c) that utilizes transformational rules of affix hopping and do-support.
As we know from Emonds (1978) (see also Pollock 1989), the inability of a main V in English to undergo head movement is a parametric distinction (main V can move in other languages, such as French), and therefore this property should fall under a theory of parameters that relates directly to UG. Pollock’s account of this parametric variation, which Lasnik discusses, is based ultimately on a theory of θ-role assignment. His account starts from the natural assumption that main Vs have θ-roles to assign but auxiliaries do not. The inflectional element Agr is morphologically rich in French but poor in English. Pollock assumes that morphologically rich Agr is transparent to θ-role assignment, whereas morphologically poor Agr is opaque. The effect of this is that main V can raise through Agr to T in French but cannot in English. When English main V raises, it cannot assign θ-roles to its complements, thereby leading to a violation of the θ-Criterion.
At this point in Lasnik’s presentation a student asks a penetrating question: how do we know that main V in English moves before assigning its θ-roles? Lasnik responds with a discussion of where interpretation occurs. He then notes that Pollock’s account appears to be based on two assumptions: (1) that θ-role assignment is not a property of D structure, and (2) that either movement does not leave traces or traces cannot be used to assign θ-roles. He notes that these assumptions will cause problems for standard XP movement (e.g., in John was arrested).
Lasnik’s proposal for handling the V-raising parameter, in contrast, has nothing to do with θ-roles. It is based instead on two distinctions: (1) between verbal elements that enter a derivation fully inflected (these he calls “lexicalist”—henceforth V*) and those that do not—what he calls “bare V” (henceforth V+); and (2) concerning the nature of the inflectional element I—either it is “featural” (the quotation marks are Lasnik’s—RF) (henceforth If) or affixal (henceforth Ia). It is a hybrid approach that mixes the purely affixal analysis of SS with the lexicalist/feature analysis of more recent minimalist analyses (e.g., Chomsky 1993, which Lasnik discusses in detail).
Lasnik proposes that all French verbal elements enter the derivation fully inflected, hence are featural. Consider the case where the main verb VM is the only verbal element (3a), a configuration that could be instantiated in four different ways, as given in (3b).
If has strong V-features (in both French and English), so V must raise to I to check those features, as will occur in (3bi). If I in (3a) were affixal, as in (3biii), the derivation would crash because V would not be compelled to move, thereby leaving affixal I stranded—a violation of the Stranded Affix Condition (or possibly Full Interpretation at PF), and even if V moved it could not take on the affixal I because it is already fully inflected. Given that all Vs in French enter the derivation fully inflected, I could never be affixal in French, thereby ruling out (3biii) and (3biv). Lasnik makes the simplifying assumption that I can be either affixal or featural universally; no special stipulations are required. UG excludes the combination of Ia with V*. This applies to English as well, where auxiliary Vs are lexicalist. Therefore, I must be featural when the adjacent verbal element is an auxiliary, the configuration in (3bi). Furthermore, auxiliary Vs must raise to I to check the V-features of I.
Assuming that featural I is universally available creates a problem for the analysis of English. How do we exclude the configuration (3a) in English where I is featural? Notice that if I is not featural, then V will not be able to raise to I, which is exactly what we want. But to account for this, Lasnik has to make a second stipulation: main V in English is bare, hence not featural. From this it follows that I cannot be featural when it occurs with only a main V. Furthermore, given the Stranded Affix Condition, affixal I will have to join with the main V—presumably via affix hopping because there is no motivation for V to raise to I.
To account for negative sentences, Lasnik resurrects the adjacency analysis in SS. Affix hopping is subject to a strict linear adjacency requirement. If affixal I and the bare main V are separated by some syntactic element (e.g., not), then affix hopping is blocked. At this point do-support can apply as a last resort operation to save the derivation. It’s classic SS analysis. In particular, it does not rely on the Head Movement Constraint (HMC) to block the movement of the affix. In fact, it need not rely on the HMC at all. We could assume that in (4) has has moved over not to I (as an instance of V-to-I raising) to check I’s features or to check its own agreement features against the subject John via spec-head agreement.
(4) John has not left yet.
If only the finite form of the auxiliary can check the V-features of I, a natural assumption, then there is no motivation to move any nonfinite verbal element to I. Thus, HMC violations are blocked by Last Resort.
As smoothly as Lasnik’s analysis seems to work, there is a problem that remains to be solved, as he notes. It involves the situation in (5), where I is affixal and V is a main verb and therefore bare.
(5) I not V+
Example (5) will lead to deviant constructions like (6).
(6)
Lasnik cites only the uncontracted version of (6a), where be is the progressive auxiliary (this does not be a problem). His analysis generalizes to copular be (6b) as well as the perfective auxiliary have in (6c). He suggests that the lexicalist option for the auxiliary (is) with V-to-I raising should block constructions like (6a) but says he does not see how to make the analysis precise.
Notice that to get (7), the grammatical counterpart to (6b), requires a further stipulation that copular be can be lexical.
(7) This isn’t a problem.
If we can strengthen “can” to “must,” then the problem of (6b) disappears. Whether this is possible under Lasnik’s theory is not obvious, as I will show shortly.
It is worth noting at this point that the problem of (6b), but not (6a) or (6c), could have also occurred in the SS analysis. Given the SA of Tq, which is the same for Tnot, copular be can be ambiguously analyzed as either a main verb (the first line) or the phonological form be (the fourth line). If it is analyzed as V rather than be, then Tnot will insert n’t between C and V, thereby blocking affix hopping and triggering do-support. This derivation yields the deviant (6b). Only when the copula is analyzed in terms of its phonological shape rather than its syntactic category does the derivation yield the grammatical (7), where n’t has been inserted to the right of be.
Thus, in both the SS analysis and Lasnik’s revision there is no motivation for analyzing the copula as a main V; though how this is to be captured remains to be determined.
In addition to the problem posed by (6), there are two further potential worries about Lasnik’s hybrid analysis. For one, the distinction between affixal and featural I is not completely transparent. The distinction between tense as a feature versus tense as an affix seems quite subtle. Moreover, featural verbs in French can be analyzed as having affixes, in which case the distinction between features and affixes becomes hard to grasp. And in the case of English, we would still have to assume that affixal I must agree with the of the subject, in which case affixal I would, as noted by Matthew Wagers (p.c.), appear to be featural in the relevant sense. Perhaps the distinction can be made in terms of strong versus weak tense, leaving the out of the equation—though this is tantamount to stipulating the facts.
The second worry concerns the notion of bareness for verbs. Lasnik’s analysis conflates the bareness of underlying main V in English with the bare verbal elements, both auxiliaries and main verbs, that occur following a modal or infinitival to. They are, however, different. The main verb that is underlying bare—that is, apparently affixless— may end up at PF with an affix; whereas the bare form following a modal or infinitival to remains bare at PF. Furthermore, it is plausible that the bare forms at PF occur with a phonetically null (zero) affix, in contrast to the underlying bare main V which by hypothesis has no affix at all.
The plausibility of distinguishing the two kinds of bare verbal form increases when we consider the problem of main verbs in English that are not inflected for tense. The discussion in SSR does not cover the hybrid analysis of main V when it occurs with an inflected auxiliary, it is not clear how participial main Vs should be handled. One possibility is that a main V that is conjugated with a fully inflected aspectual auxiliary enters the derivation fully inflected (i.e., with an -ing or -en affix). If so, then the stipulation about main Vs entering the derivation bare will not be a general statement about main Vs but instead will merely describe the complexity of the data.
In Lasnik 1995d, where the hybrid analysis was first explicated, another possible analysis is mentioned but not developed systematically—namely, the general statement about the underlying bareness of main V is maintained. This requires that aspectual auxiliaries can enter the derivation fully inflected along with their associated affix, which would then hop onto the bare main V Alternatively, these auxiliaries can enter the derivation without the associated affix when the following verbal form is another inflected auxiliary. The problem then arises that aspectual have must be entered in the lexicon in two forms:
(8)
(8a) is needed when the progressive auxiliary is present, and (8b) when the perfective auxiliary is immediately followed by a main V. It is not a problem if (8b) is selected to precede the progressive auxiliary because the affix -en would be stranded in violation of the Stranded Affix Condition. However, it would be a problem if (8a) is inserted adjacent to a bare main V. By hypothesis, there would be no features of V+to check because V+ is not featural; so why would bare V be ruled out in this construction? One possible answer is that bare V+is truly affixless, whereas the other bare forms are not, and there is a general prohibition against affixless stems. Of course, the most plausible solution is that the perfective auxiliary selects the past participial form of the following verb (either auxiliary+-en or V+-en). But then it would be more economical to have main V enter the derivation with its participial inflection, thereby avoiding the problem created by the dual lexical entry (8). But then we are back to complex stipulations about how main Vs enter a derivation.
