The structure of null subject DPs and agreement in Polish impersonal constructions 137
d. Pracowało się jako dziennikarki.
work.past 3 sg.n się as journalists.F
‘[One/People/They] worked as a journalist/as journalists.’
As is noted by Kibort (2008: 272), the default [+human] interpretation of a null subject
DP can be overridden by providing a different referent for the unspecified agent some-
where in the context, for example:
(19) Gdy się jest bocianem, gniazdo buduje się wysoko.
when się is stork.inst.m nests.nom.m build.pres 3 sg się high
‘When one is a stork, one builds the nest high up.’
(Kibort 2008: 272)
The analysis of the phi-feature specifications allows for the following generalizations.
Firstly, the example above illustrates that the feature [+human] is not consistently
present in impersonal constructions, contrary to Egerland (2003) and Siewierska
(2008) among others. The [+human] specification can be overridden by providing a
different [–human] referent for an unspecified subject somewhere in the context. This
observation is in accordance with Moltmann’s (2006) hypothesis according to which
impersonal subject DPs can have a more general reference; that is, they are restricted
not so much to humans, but rather to conscious beings. The fact that null subject DPs
in impersonal constructions cannot have non-conscious referents is illustrated by the
acceptability of (19) above as opposed to (20).
(20) ??Jeśli jest się szafą, stoi się w kącie.
if is się wardrobe.inst stands.3sg się in corner.
‘If one is a wardrobe, one stands in the corner.’
Although the covert subject in the example (20) has a referent in the preceding phrase
that is non-conscious, the only meaning that this sentence can convey is that if a con-
scious being acts as a wardrobe (e.g. in a play), then it usually stands in the corner. This
sentence does not and cannot express a generalization (i.e. standing in the corner) that
could be considered common to all or some wardrobes. In other words, the reference
of impersonal subjects cannot be stretched to include non-conscious referents. It is,
nevertheless, clear that the more restricted [+human] interpretation of the null subject
DPs remains the default one unless a different [–human] but still [+conscious] refer-
ent for the subject is provided by the context (see D’Alessandro and Alexiadou (2003)
for a similar observation on the feature specification of impersonal subjects in the
Italian SI construction). Secondly, the null subject DP is not inherently specified for
phi-features, as the predicative complement may carry any number and gender mark-
ers corresponding to an agent/undergoer provided by the context.
I do not, however, assume that it is necessary to introduce a separate feature [+/–
conscious] into the featural make-up of impersonal subject DPs. The [+conscious]/