For the prelim class last week, I happened to read in quick succession Chapter 2 of Richards’ ‘Uttering Trees’ and Gallego 2006 on T-to-C movement in relative clauses. Both of them present an account of the amazing ban on DP relative operators in Romance languages (1) and in English infintival, but not finite, relatives (2). (To see that DP relative operators are good in English finite clauses, check the well-formedness of the gloss of (1b) as an English NP).
1. a. l’homme [avec qui [j’ai dansé t ]]
the’man [with who [i’ve danced t]]
b. *l’homme [qui [je connais t ]]
the’man [who [I know t ]]
c. l’homme [ que je connais e ]
the’man [ that I know e ]
2. a. a person [ with whom [ to dance t ]]
b. *a person [ who [ to know t ]]
c. a person [ to know e ]]
Richards proposes that this distribution represents an example of his Distinctness effect, such that the configuration required for the DP relative operator cases in (1b) and (2b) would bring the DP relative operator into the same phasal domain as the head D of the whole NP, and hence introduce a linearization difficulty, due to two identical labels occurring in the same spell-out domain. The DP relative operators contained inside PPs do not induce such problem because a) PPs are distinct from DPs and b) the DP relative operator that is the complement of the PP is in a separate spell-out domain, i.e. PPs are phases. The difference between English finite and nonfinte clauses has to do with there being a phasal boundary between the target of operator movement in the finite clauses but not in the nonfinite clauses, i.e. English finite clauses have more and finer-grained structure in the left periphery than the nonfinite clauses, or Romance relatives in general, do, very much in the spirit of Bianchi’s (1999) analysis.
Gallego 2006 instead proposes a kind of economy-of-checking account based on Pesetsky and Torrego’s that-t effect machinery. He proposes that a relative C head, which has both [uT] and [uRel] features that require checking and deletion. This attracts the internal DP to spec-CP, [who person], in a normal English finite DP relative, to give [The [ [who person] [ t loves Mary]]], and then [person] subextracts from that DP to a higher spec-cP to yield the final word order [The [person [ [who t] [ t loves Mary]]].
In that-subject-relatives, like ‘The person that loves Mary’ things are a little more complicated. The core idea is that the null relative operator can’t pied-pipe ‘person’ along to spec-CP. So C gets its [uT] feature checked by that (which, following P&T, is a T head), explaining why it’s required to appear in subject relatives. C gets its [uRel] feature checked by Agree with the null operator, which remains in situ. Then finally the head N person subextracts from the relative DP in spec-TP and moves to spec-cP to check c’s [uPhi] feature.
So there’s no optional that with a null operator in finite subject relative clauses in English. With object relatives, that IS optional, which means that either the subject can check C’s uT feature or T (that) can do so, but then he’s left with a problem: His account predicts that The car which that John sold should be ok, with T checking C’s uT feature. However it’s not, and he doesn’t know why not, so that’s an overgeneration problem. But when the subject checks C’s uT feature, you get The car which John sold, which is good, so at least he does generate all the good versions of English finite relatives. And, bonus, he predicts that there will be no null-operator version of PP relatives, since null-operators can’t pied-pipe, so the contrast between 3a and 3b is accounted for:
3. a. The school in which I studied.
b. *The school in I studied.
Ok, now back to the problem of no DP relatives in Romance relatives and English nonfinite relatives. Gallego proposes that Romance subjects in general, and English PRO subjects of infinitives, can’t check C’s uT feature. For him this is because they’ve checked their own uT feature AND had it deleted, not just valued it but valued and deleted it. This is because, he proposes, there’s a phasal boundary at TP in Romance (and English infinitives). TP is a phase, not vP, in Romance.
Gallego’s account, based on the idea that Romance subjects (and English PRO subjects) are unable to check C’s uT feature, predicts the impossibility of null C complement clauses (a la Pesetsky and Torrego) AND the impossibility of null C wh-Subj-V exclamatives, which are fine in English finite clauses but bad in Romance:
4. a. I know [CP John [TP t called her]]
b. [CP How intelligent [CP Mary [ TP t is t ]]]!
5. a. *Sé Juan la llamó. (Sé que Juan la llamó is ok, I assume).
b. *¡Qué intelligente Maria es!.
If (5b) in particular represents the same problem as (1b), Richards’ account cannot be correct, since (5b) doesn’t involve any issues with two DPs in a single phase.
On the other hand, Gallego’s account, if I understand everything correctly, predicts that (2c) should be ungrammatical, since PRO cannot check C’s uT feature. Instead, Gallego predicts that English should have to use “*A person that to know,” which is flatly out.