Content area
Abstract
Though there have been many analyses of null pronouns in the past decade, few have been constrained by more than the facts of a single language. As a result, there are nearly as many theories of null pronouns, or 'pro-drop', as there are languages under investigation.
A major goal of this work is to bring cross-linguistic types to bear on the analysis of null subjects. It is possible to simply test the success of language-specific hypotheses against a cross-linguistic sample, e.g., in this way it is easily demonstrated that Pro-drop Parameter (which correlates null subjects, subject inversion, and free extraction of subordinate subjects), is misguided. But more to the point, I presume that a preliminary cross-linguistic sampling of null pronoun phenomena most adequately constrains the analysis.
A cross-linguistic examination of Taraldsen's generalization, which claims an intuitively satisfying one-to-one relationship between agreement morphology and null pronouns, reveals a significant number of counterexamples. Following the distinction in Rizzi (1986), where it is argued that all empty categories must satisfy two requirements, it is presumed that some of these counterexamples are failures of identification and the others are failures of licensing.
The relationship between agreement morphology and its associated null pronoun is taken as the core of the theory of identification; by parallel with this relationship, it is assumed that all null pronouns must be governed by their identifier. This constraint is then applied to the analysis of other null pronouns, e.g., null subjects in agreementless languages like Mandarin, PRO (the null subject of nonfinite clauses), the arbitrary null objects mentioned by Rizzi (1986), and a wide range of arbitrary and indefinite null pronouns.
The three remaining counterexamples--languages with agreement but no null pronouns, e.g., Icelandic, the lexicalization of nonthematic subjects, e.g., English there, and the lack of a lexical alternate to PRO--are taken as the core phenomena constraining the analysis of licensing. I motivate two principles (Case causes lexicalization (Bouchard 1984) and an expanded definition of argument position) and two parameters of Case-assignment, and demonstrate that their interaction alone adequately addresses the posed problems. (Copies available exclusively from Micrographics Department, Doheny Library, USC, Los Angeles, CA 90089-0182.)
