54 Steven Franks
Alternatively, one might reject the multiple probe account in favour of movement of
IvanaACC out of [DP IvanaACC [QP odnogo]] to [Spec,ApP], except that it is not clear to
me how to allow this movement but at the same time block movement out of [FP Iva-
naACC F [AP grustnogoACC]].^34
The problem of variation over an indirect object, as in (14c–d), could be handled
similarly, with two competing structures. The dominant agreement one would be like
simple OC, with the embedded subject moving to the main clause. The SD possibility
would necessarily involve a CP, perhaps with an ApP structure as in (64).
As for variable case transmission to inside of V + N collocations, as in (14e–g), it
seems to me the issue is whether the clause is treated as a complement to the noun, in
which case it is necessarily a CP and the SD ensues, or it is treated as a complement to
a complex verb [V V + N], in which case, like other infinitival complements to V, it is
treated as a TP and agreement through the MTC ensues.
Lastly, consider the observation that čtoby only optionally blocks case transmis-
sion (similarly Polish żeby), as in (11)–(13) and (14b); cf. also fn. 4. Here I suggest that
čtoby is either analysable as a C head, inducing dative, or as a Modal head (so no CP
is projected and no dative can be assigned), leading to OC and agreement. Another
possibility is that the variation relates to the possibility of movement out of čtoby ‘in
order to’-clauses, as in (66), from Bailyn (2012: 101), since čto ‘that’-clauses are islands
in Russian.
(66) Komu ty xočeš’ [čtoby Ivan pozvonil___]?
who.dat you want so_that Ivan called
‘Who do you want Ivan to call?’6.4 Movement, multi-attachment, timing, and feature sharing
In this final section, we step back from the details and consider some larger architec-
tural issues.
First of all, although the literature is silent about how to implement clause-internal
agreement, a reasonable assumption is that some kind of “co-valuation” mechanism is
needed. That is, multiple items share a single set of features, which once valued are
realized on those items for purposes of vocabulary insertion. The advantage to feature
sharing in the context of the present paper is that the case of one item can be made to
depend on the case of another before that case has even be determined in the deriva-
tion. This provides a simple way of dealing with the look-ahead problem.- One thought is that the adjunction structure does not entail domination.