Advances in the Syntax of DPs - Structure, agreement, and case

52 Steven Franks

comparable to (59), i.e. one without PredP. The difference in structures could be that

the semipredicatives are adjoined rather than introduced using FP:^32

(61) [IvanNOM prišel domoj [DP IvanNOM [QP sam]]]

Note that since Polish lacks FP, (61) can also apply to semipredicatives in that lan-

guage. In sum, semipredicatives can agree like APs in addition to being assigned case

like NPs, and they do not involve PredP.

6.2.5 Why agreement under obligatory Object Control is only possible

for semipredicatives

Recall the contrast in (42) versus (43), repeated below:

(62) *Pavel poprosil Ivana [ne idti na prazdnik grustnogo].

Pavel asked Ivan.acc not go.inf to party sad.acc

‘Pavel asked Ivan not to go to the party sad.’

(63) Pavel poprosil Ivana [ne idti na prazdnik odnogo].

Pavel asked Ivan.acc not go.inf to party one.acc

‘Pavel asked Ivan not to go to the party alone.’

While some speakers accept agreement for the semipredicative in (63), none do for

the adjective in (62). The question here is why, under the assumption that odnogo has

the option of agreeing with Ivana in (63), that same option is not similarly available

to grustnogo in (62). This puzzle is resolved by rejecting agreement with Ivana as the

source of accusative case on the semipredicative in (63). If the MTC does not allow

movement of Ivana from the embedded clause to matrix object position, there has to

be an alternative structure in which direct assignment of accusative case takes place

both to the matrix object and to the semipredicative. Crucially, this structure is not

available for ordinary adjectives (and does not exist in Polish at all), since these cannot

be directly assigned case. Whatever assigns accusative case to Ivana also assigns it to

the semipredicative; I leave the details of the account for further research.^33

32. An adjunction structure for semipredicatives is argued for by Despić (2011: ch. 4), al-
    though his structure is somewhat different.


33. Note that this is not unlike Landau’s (2008: 879–80) “PRO-control” version of OC, in
    which the main clause case assigner (for him, T for nominative or v for accusative) assigns
    case to a DP within its own clause, and, over C, to PRO in the embedded infinitival clause
    (with which the semipredicative then somehow agrees). Witkoś (2010) also adopts a “[+mul-
    tiple] probe” approach. By restricting that feature to T in Polish (versus Russian) he is able to
    prevent agreement in the Polish version of (63), although this leaves him with no explanation
    of the contrast between the two Russian examples.