Ever since Phase Theory was first put forth by Chomsky, it has been taken for granted that phases include at least CP and transitive vP (or whatever you think the highest projection in a transitive verb phrase is). More recently, Keine (2017) has presented a very nice argument that vPs, even transitive ones, cannot be phases, at least not in Hindi.
I find Keine’s idea to be very promising, for a couple of reasons. First, there is the issue of the Phase Impenetrability Condition (PIC). As many of you probably know, Chomsky has two versions of the PIC. The first (“PIC1”) is straightforward: once the phasal XP is complete, nothing inside its complement domain (Compl,X) is accessible. This allows things to escape via Spec,XP, a pattern linguists have been finding evidence for since the ’70s. The second (“PIC2”) is a “staggered” version of PIC1: the complement domain of a phasal XP remains accessible until the next phase head Y is merged. Now, if I were a big believer in conceptual arguments, this is where I would tell you that PIC2 makes no sense, and therefore must be wrong. But I’m not, so I won’t. What I do want to point out is that, tracing the history of PIC2, I’m pretty sure the only reason it was put forth (in Chomsky’s 2001 Derivation by Phase; see p. 14) was to account for agreement between T and a nominative object in Icelandic (specifically, in those cases where the object has not undergone object shift). If there is a phase boundary at the verb-phrase level, then PIC1 would rule out such an agreement relation, whereas PIC2 would rule it in. Do you see where this is going? This constitutes an argument that we need PIC2 rather than PIC1 only on the assumption that vP (or something thereabout) is a phase. So the first reason that Keine’s result is appealing is because it allows us to do away with PIC2 (which, whether it makes conceptual sense or not, is clearly more complicated) and return to the simpler PIC1.
The second reason I find Keine’s position appealing is that the vast majority of the evidence we have for the successive-cyclicity of long-distance movement concerns the CP layer. Den Dikken has a project in which he attempts to argue that vP (or thereabouts) is a phase but CP isn’t. One of the striking things about this project is just how hard Den Dikken has to work: the vast majority of evidence for intermediate stopping, by far, is evidence involving the CP layer. In fact, I would venture to say that the only evidence I know of that seems to solidly point to a vP-level phasal category comes from van Urk & Richards’ (2015:127) work on Dinka.
At this juncture, you might be asking yourself something like, “Wait a minute, what about all the other evidence we have for vPs being phases?”
As you might suspect, I don’t think much of that evidence stands up to scrutiny. Here’s why.This is – and not for the first time on this blog – an elaboration of a comment I left on Norbert’s blog once upon a time. Phase theory is supposed to deliver predictions on where things must stop, not where things can stop. Or, if you prefer: it’s the Phase Impenetrability Condition, not the Phase Optional Permeability Condition. Crucially, many of the purported arguments for the phasehood of vP amount to arguments that a Spec,vP stopping-off point is possible, and provide no evidence that such a stopping-off point is obligatory.
Now, one could entertain a theory where the only possible stopping-off points were phase edges. While that is a logically coherent theory, I think we can safely say that it can be discounted on empirical grounds. The following is a digression to demonstrate this point.
· · · · · · · · · · · · · · · · · · · ·
(1) The children seem to her to have liked Mary.
(2) The woman showed the boy to himself in the mirror.
(3) The woman showed the boy to herself in the mirror.
(4) *The children seem to me to have appeared to myself to be clever.
(5) The children seem to me to have appeared to themselves to be clever.
From the disjoint-reference effect in (1) between her and Mary, we can conclude that the to-experiencer of a verb like seem is able to bind into the verb’s infinitival clausal argument. From the success of reflexive-binding in (2) we know that English self-anaphors are not subject oriented. And from the success of reflexive-binding in (3) we know that binding of English reflexives is not subject to minimality (i.e., there is no condition demanding that the antecedent of a reflexive be in the closest position that could possibly bind that reflexive).
Taken together, this means that the failure of reflexive binding in (4) cannot be because (to) me is not capable of binding into the infinitival clause; nor can it be because (to) me is not a subject; nor can it be (merely) because the infinitival contains a closer binder (say, the trace/copy of the children in the subject position of appear). Instead, the reason seems to be that the reflexive myself in (4) is contained in a binding domain (say, the infinitival clause whose main verb is appear) that necessarily excludes (to) me. But if the relevant binding domain necessarily excludes (to) me, then it also necessarily excludes the matrix position of the children (on the benign assumption that binding domains are structurally contiguous).
If all this is so, then how is reflexive binding of themselves, in (5), successful? The only answer, it seems to me, is that the binding domain of myself/themselves in (4)/(5) includes a trace/copy of the children in a position c-commanding the reflexive. And since the relevant position of this trace/copy is not its theta position (which is lower down, in the copular infinitival), the conclusion is that there is a possible stopping off point for movement of the children in the intermediate infinitival clause, likely in Spec,TP of the TP complement of seem.
Now, importantly, the TP complement of seem has no business whatsoever being phasal.Even if you think that some TPs “inherit” certain phasal properties from C(P), the particular TP in question is not in a structural position to do so, being nowhere near any C(P). So there you have it: a possible stopping off point for movement is not an argument for phasehood.
· · · · · · · · · · · · · · · · · · · ·
Taking this into account, it seems like many of purported arguments in favor of the phasehood of vP have the wrong quantificational force, if you will: they are arguments that a certain stopping-off point is possible, not that it is obligatory. Take the famous “scope-trapping” diagnostics used by Fox (2000:164) and Legate (2003:507) to argue for an intermediate landing site in Spec,vP: these diagnostics merely show that a stopping-off point in Spec,vP is possible, not that it is obligatory. Sure, under certain circumstances (where all other possible stopping-off points would yield binding-theoretic ill-formedness), one can force movement to stop off at this position. But, ipso facto, what is then responsible for the necessity (as opposed to the possibility) of stopping off in Spec,vP is binding, not phasehood.
A possible objection, raised by Peter Svenonius, goes as follows. We know that not every intermediate position is a possible stopping off point. If something (e.g. Spec,vP) behaves like it is consistently a possible stopping off point, phasehood provides a very plausible explanation for why this position and not others behaves that way. The problem with this logic is that it applies with just the same force to the Spec,TP scenario, above. And since we know that it delivers the wrong conclusion there (namely, that raising TPs are phases), I conclude that there must be something wrong with the argument itself.
It is illuminating, I think, to compare this state of affairs with some of the evidence we have for CP-level successive-cyclicity. For example, the famous findings by McCloskey (1979:150–156) about the Irish complementizer system show that, in cases of overt wh-movement, every complementizer on the movement path must undergo the go→aL change. Must, not may. The landscape is not quite as clean as one might hope since, as I mentioned above, there is at least one good argument that I am aware of that Spec,vP is an obligatory stopping point (see van Urk & Richards’ 2015:127). If Keine’s position is to be fully vindicated, one must find an alternative account of such facts.
The important thing is, however, that Keine’s position absolutely does not conflict with data like that of Fox or Legate, or anyone else who adduces evidence that Spec,vP is a possible stopping-off point for movement. Phase Theory is a theory of obligatory stopping-off points, not of optional ones. Or, more accurately: Phase Theory is a theory that is supposed to deliver predictions about obligatory stopping-off points, not about optional ones.
|1.↩||This is – and not for the first time on this blog – an elaboration of a comment I left on Norbert’s blog once upon a time.|
|2.↩||Even if you think that some TPs “inherit” certain phasal properties from C(P), the particular TP in question is not in a structural position to do so, being nowhere near any C(P).|