The overgeneration problem and the case of semipredicatives in Russian 51
6.2.3 Why ordinary adjectives always have an instrumental option
Assuming a PredP, the reason why ordinary adjectives always have an instrumental
option has to be that a structure with Pred such as (57a) is always possible for predicate
APs. The agreeing structure, on the other hand, does not involve Pred, but rather just a
small clause FP. Adopting the MTC then gives us something like (59) for the two vari-
ants in (24), repeated as (58):
(58) Ivan prišel domoj grustnyj/grustnym.
Ivan.nom came home sad.nom/inst
‘Ivan came home sad.’
(59) a. [IvanNOM prišel domoj [FP IvanNOM F [AP grustnyjNOM]]]
b. IvanNOM ... [PredPIvanNOM Pred [FPPROINST F [APgrustnymINST]]]
Agreement is thus impossible whenever movement is,^31 but instrumental is always
available. Of course, the presence of PredP has semantic implications (at least, when
both options are viable) and there are much investigated interpretative differences
between the two structures, with and without a Pred. I put these aside in this paper,
referring the reader instead to studies such as Pereltsvaig (2007), Richardson (2007),
Bailyn (2012), or Bondaruk (2013).
Returning in this light to Polish (51), the case facts suggest that FP (whatever the
identity of F) is absent in this language. Thus Pred cannot select for FP in Polish, which
means its complement must agree when adjectival, as in (60a), but is assigned instru-
mental when nominal, as in (60b):
(60) a. [JanNOM był [PredP JanNOM Pred [AP szczeryNOM]]]
b. JanNOM był [PredPJanNOM Pred [NP szczerym człowiekiemINST]]
The opposition in (60) is a simple consequence of APs agreeing but NPs being assigned
case directly.
6.2.4 Why semipredicatives must agree where ordinary adjectives can
The question of why semipredicatives always agree in contexts where ordinary adjec-
tives can agree (or be instrumental) was the point of departure for Madariaga (2006).
It also subsumes the overgeneration problem. In the system outlined above, agree-
ment for ordinary adjectives is a local relation. It should be the same for the semipred-
icatives, which, following Madariaga, I take to be QPs. Since QP is not an admissible
complement of Pred, the semipredicatives have no choice but to agree in a structure
- In (59b), movement is not possible from [Spec, FP] directly to [Spec, TP], i.e. over [Spec,
PredP].