Holmberg (2005) Evaluates the Following Two Hypothesis Regarding the Status of Null Subjects

Two Kinds of Subject pro?

Pilar P. Barbosa

Universidade do Minho, Portugal

Holmberg (2005) evaluates the following two hypothesis regarding the status of null subjects against data from Finnish, a partial pro-drop language, and concludes that hypothesis B is favoured over A: (A) Agr is interpretable in Null-Subject Languages (NSLs), and pro is therefore redundant; (B) Null subjects are specified but unpronounced pronouns which assign values to the uninterpretable features of Agr.

In this talk, I will argue that a version of hypothesis A is superior to hypothesis B for consistent NSLs of the rich agreement type. I will focus on Romance consistent NSLs and I will show that their grammar differs in important ways from closely related non-NSLs such as French or partial pro-drop languages such as Brazilian Portuguese (BP). In the former case, subjects (overt or null) are not subject to displacement and remain in situ (cf. also Alexiadou and Anagnostopoulou l998, Pollock l997, among many others); this yields the well known property of “free inversion” associated with the NSLs with rich agreement morphology, which is lacking in French/English or modern spoken BP. A close inspection of the cluster of properties subject to change in modern spoken BP shows that (i) these are deep properties that pertain to narrow syntax (they can be detected in constructions with overt subjects); (ii) they are closely related to the presence of rich agreement morphology; therefore, it cannot be the case that the role of rich agreement reduces to matters of sentence processing, as suggested by Holmberg (2005).

The assumption that the phi-features in T are interpretable in consistent (rich agreement) NSLs entails that there are no unvalued features in T in these languages (as opposed to the non-NSLs or partial pro-drop languages). This, in turn, renders it possible to unify null pronouns in languages with rich agreement morphology and in languages without agreement (Huang l989): they are minimally specified bare nouns (Holmberg 2005) in both cases and what these languages have in common is that there are no unvalued phi-features in T; hence such a minimally specified null pronoun can be licensed.

This leaves out partial pro-drop in Finnish and Brazilian Portuguese, for which hypothesis B may be correct, given that the phi-features of T are uninterpretable and trigger displacement. In sum, the unification proposed in this paper is not between partial and consistent null subject languages (as proposed by Holmberg 2005) but rather between rich agreement and no agreement (consistent) subject drop.