The structure of null subject DPs and agreement in Polish impersonal constructions 139
such as for example ‘chair’, ‘lamp’ and ‘walk’, which are selected from the Lexicon to the
Numeration. The values of their feature make-up are established before they become
elements of narrow syntactic operations and as such they are interpretable without any
syntactic feature matching (Sigurðsson 2004).^8 In contrast, pronouns, in this approach,
are bundles of interpretable but unbound features that get bound in the course of the
derivation. They are not atomic elements as they are subject to Agree and matching
and their final form is construed in the Narrow Syntax.^9 That is, an argument comes
with a Φ-variable that is bound by clausal Φ/Λ-features. To be a bit more specific, if
the referent of the θ-role is identical to the referent of ΛA, we get 1st person, if it is
identical to the referent of ΛP, we get 2nd person, otherwise we get 3rd person (if the
referent is neither ΛA nor ΛP). What remains to be established is (a) how the reference
is established in the case of 3rd person pronouns (see the following paragraph), and
(b) a difference between 3rd person and ‘no person’, as it is clear that -ΛA and -ΛP are
not the same as 0-person.
Under the second hypothesis adopted here, the interpretation of a referential pro
(3rd person) in the subject position is dependent on a matching relation with a specific
type of Topic: the so-called ‘Aboutness-shift’ Topic (henceforth A-Topic) (Frascarelli
2007 : 693). The A-topic has the function of introducing a new topic (or proposing a
topic-shift) in the discourse (Frascarelli & Hinterhölzl 2007). It is the constituent that,
Frascarelli (2007) argues, identifies (i.e. provides a referential value for) an argument
pro. In other words, null subjects of tensed clauses have the same function as clitic pro-
nouns in Topic constructions: they serve as resumptive pronouns and as such they are
pronominal variables. Frascarelli (2007) maintains that the A-Topics and subjects share
basic properties since they are both connected with given information and provide a
starting point for the event described by the predication. Within discourse, ‘predica-
tion’ can imply a multi-clausal domain, in which chains of clauses are combined and
refer to the same A-Topic. The presence of silent topics thus follows. Holmberg (2010a)
refines the analysis of Frascarelli (2007) and postulates the existence of the D feature
in T which is valued by the A-Topic. The presence of this D feature guarantees that all
null elements will be interpreted as definite.
It is proposed here that the combination of very similar types of mechanism is at
work in the case of null underspecified subjects in Polish impersonal constructions,
both when it comes to establishing agreement with predicative complements and to
- Words, on the other hand, are, according to Halle & Marantz (1993) formed in syntax.
Namely, the noun walk is derived by merging the root walk and n, whereas the verb walk is
derived by the merge of the root and v. - These features, being interpreted at both interfaces, must be present in Narrow Syntax.