• Mary seems[ t to be here]

  • Mary tried [PRO to be here]

Why can't Mary be generated in spec VP of 'seems' but can in 'tried'?

Instead it looks like it works more like a passive verb:

  • Mary is believed [ t to be here]

This is in context of The Minimalist Program by Chomsky.

  • Seem (and appear, which is close to synonymous) are both 1-place A-Raising verbs. As such, they occur with subject clauses that are ungrammatical without syntactic intervention. If the subject clause is an infinitive, A-Raising is required; if it's tensed, Extraposition is required. I.e, *For him to be tired seems and *That he is tired seems are both bad, but He seems to be tired and It seems that he is tired are both fine. – jlawler Jul 3 '17 at 21:30

The reason is that try is a two-place predicate that assigns an Agent theta role to Mary and a Theme (or similar) role to its infinitival complement, whereas seems is monadic (in your example), must assign just one theta role, to its complement, and cannot have a second argument in its specifier to which it can assign Agent (or any other) role. [Of course, seems may also be two-place, e.g., in It seems to me that Mary is an excellent teacher, but in that case it must have two 'internal' arguments, and assign a Theme role to the subordinate clause and an Experiencer role to the PP complement to me; under no circumstances can it assign an Agent role to the PP to me, because Agent is never assigned to an internal argument in complement position; Agent can only be assigned in Spec v/V position, and, of course, never by a non-agentive verb like seems].

As a consequence of that difference in argument structure between try and seem, in a sentence like Mary tried [PRO to be there] a PRO is needed in the subordinate clause to discharge the Theme role projected by the predicate be there and some other argument (Mary, in this case) is also needed to discharge the Agent role assigned by try in the upper clause to its own Spec V. If you had only one argument, i.e., if you base-generated Mary as the specifier of be (there) in the subordinate clause, and then raised it into the Spec of the higher verb try, the chain formed by Mary in Spec try and its trace in Spec be (or by the two copies of Mary, under the 'Copy Theory of Movement') would receive two theta roles (Agent and Theme, respectively), the Theta Criterion would be violated, and the derivation would be ungrammatical. (Under later versions of Chomskian grammar, of course, that Theta Criterion violation need no longer be invoked to block the derivation: since Mary already has a valued role attribute in Spec be, it could never be raised into a higher thematic position, e.g., Spec try, because there would be no feature to check in Spec try and Economy would automatically forbid that movement. Mary could still raise, of course, but not in search of a second theta role; it could raise directly into Spec Tense, for example, to check its Nominative Case feature there, but it would have to do so skipping Spec try, and therefore would never get assigned a second theta role as Agent of try anyway.

On the contrary, in a sentence like Mary seems [t to be there] (where, recall, seems is monadic and assigns a theta role only to its complement, not to its specifier), replacing 't' with a PRO in the subordinate clause would make the sentence ungrammatical, because, in that case, it would be PRO that would receive the only undischarged theta role available in the sentence (the Theme role assigned by be there). Mary, on the contrary, would not receive any theta role in Spec seems, and, again, the Theta Criterion would be violated (by defect, in this case) and the sentence would be ungrammatical. If Mary is base-generated as a Spec of be there, though, it gets the Theme role assigned by be there, and, as the higher verb seems does not assign any theta role to its own specifier, Mary never receives a second theta role, the Theta Criterion is not violated, and the sentence is well formed.


seem is a raising verb, try is an equi/control verb ('equi verb' and 'contol verb' are usually used synonymously).

Equi verbs have a semantic subject: It is part of the meaning of to try that there is someone who is trying, which is Mary in this case. The verb try selects Mary as an argument, more specifically, as the subject argument. This subject starts of as the subject of the embedded clause and is raised up to the subject of the matrix clause, if you believe in a transformational syntax.
The same does not hold for raising verbs: There isnt's any actual seemer, someone who fills a thematic role in the seeming event. If anything, a grammatical subject in a sentene with seem as the main verb would be realized by it (as in It seems that Mary is here), but then it is clear that this it is just a filler for something that would otherwise be empty, because there is no semantic subject; just like in It is raining, when it is not an actual subject that is raining; rain is a zero-place verb. The it in both senetences is merely a phonetic filler that needs to be there because English doesn't want the subject position to be emtpy, but there are cases where this pseudo-subject isn't an actual subject. The verb seem doesn't have any more theta role to assign, because there isn't a "seemer" - contrary to the verb try, which does assign the theta role of agent (namely the one who is trying). And since Mary is not selected as an argument of seem, it can not be generated as a subject in SpecVP.
It's simply the semantics of the two verbs that make them behave syntactically different.

Also note the difference between subject-oriented and object-oriented raising and equi verbs. seem and try are both subject-oriented: The filler element takes the position of the subject in the matrix clause.
The example you gave with believe is an object-oriented raising verb. The reason why you chose the passive sentence is that Mary is actually the object of the outer sentence: Having started as the subject of the embedded clause (Mary is here), it is raised to the object position of the matrix clause (John believes [Mary] to be here). But Mary is not semantically selected by the verb believe - the verb doesn't assign any theta role that Mary now fills; Mary isn't "the believed" or something it's just there due to the syntactic operation of raisin. With this behavirous, believe is an object-oriented raising verb.
An example an object-oriented equi verb is persuade, as in Peter persuaded Mary to be here: Mary starts as the subject of the embedded clause and is raised to the object of the matrix clause. However, it is also selected by the verb: Mary is the one who is being persuaded; she fills an argument position that is semantically required by the verb - there is someone who is being persuaded, and this role is satisfied by the now raised Mary. This makes persuade an object-oriented equi verb.

All these behave synactically differently, but the main distinction is simply whether or not the NP in question fills a theta role that is semantically required by the verb.

  • The surface subject of 'try to be there' does NOT 'start off as the subject of the embedded clause and is raised up to the subject of the matrix clause', contrary to your statement at the end of your first paragraph. That never happens to 'equi'/'control' verbs like 'try', for reasons I explained carefully in my own answer. It is RAISING verbs like 'seem' that force the subject of their infinitival complement clause to raise, precisely because that subject would violate the Case Filter in situ and 'seem' etc. have no subject of their own that can raise to Spec T and satisfy the EPP. – user6814 Jul 4 '17 at 7:38

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.