158 Małgorzata Krzek
marking on T appears to be caused by the fact that Φ-specification of null imper-
sonal pronouns is ‘richer’ than the ability of the Polish morphological system to
reflect the values of those Φ-features on T. As a result, in the process of agreement
the values of some or all of the Φ-features of some nominal elements do not get
copied on T and are not represented in the verbs’ morphological marking. In Polish,
just like in Finnish, for overt DPs only those with nominative case can but do not
have to, as illustrated by the sentence in (37) above, trigger agreement ( Holmberg
2010b: 209). Therefore, since it has been assumed that agreement between null
impersonal pronouns and T is established by default, it is not entirely obvious that
the null impersonal pronouns have nominative case. It becomes even less so in the
light of the mechanics of case assignment adopted here. Namely, it is assumed here
that morphological case is not tied to specific syntactic positions, and therefore
it should not be, even indirectly, linked to DP licensing. Rather, as noted, among
others, by McFadden (2004, 2006 ), Marantz (1991), and Sigurðsson (2003), cases
are assigned according to a dependency relationship, with accusative case assign-
ment being dependent on there being another higher argument within the same
phase to which structural case has been assigned. Traditionally, this other struc-
tural case could only be nominative. More recently, however, it has been postulated
that dative and other non-canonical cases also can be regarded in certain environ-
ments as structural (see Dotlačil & Šimik 2012), not lexical/semantic. This means,
according to the hypothesis advocated here, that when they appear on a higher nom-
inal element, the lower one is assigned accusative.
Thus, these assumptions appear to complicate the matter even more, as they make
the standard diagnostics (e.g. ability to trigger subject agreement marking, presence
of accusative case on a direct object) used to detect case on a null element difficult
to rely on. In this respect, the situation with Polish impersonal pronouns is different
from the one with a Finnish generic pronoun (henceforth G-pronoun), as the latter
triggers subject agreement marking on T. This, for Holmberg (2010b), is taken to be as
proof that in such environments the G-pronoun is nominative. Holmberg (2010b) also
shows that the Finnish G-pronoun occurs in environments where cases other than
nominative are assigned. It can, for example, appear with necessive predicates that
take subjects with genitive case. In other words, the Finnish G-pronoun appears to be
able to occur in any case-environment where overt subjects occur. The situation is not,
however, exactly the same for Polish impersonal pronouns. That is, although they do
appear with predicates whose overt subjects are assigned nominative (as illustrated by
the examples in (39)), they do not seem to occur in environments where non-canoni-
cal (e.g. genitive, dative) cases are assigned to subjects.
(39) a. Marysia pracowała dużo.
Marysia.nom work.past 3 sg.f a-lot
‘Marysia worked a lot.’