The Weekly Newsletter of MIT Linguistics

Colloquium 11/20 - Susi Wurmbrand

Speaker: Susi Wurmbrand (University of Connecticut)
Title: Fake indexicals, feature sharing, and the importance of gendered relatives
Date: Friday, November 20th
Time: 3:30-5:00 PM
Place: 32-141

A popular trend in binding theories is to view binding as a dependency between the bindee and a functional verbal head (v, T, C), rather than a direct dependency between the antecedent DP and the bindee (Reuland 2001, 2005, 2011, Chomsky 2008, Kratzer 2009). In this talk, I will show that binding is not sensitive, nor can it be assumed to be driven or mediated by functional heads. Instead data are provided that argue for a return to the traditional view that binding requires a direct dependency between the antecedent and the bindee/variable. I propose that this dependency is formalized as Reverse Agree (Wurmbrand 2011 et seq.), constrained by a locality condition reminiscent of Rule H (Heim 1993, Fox 1998) and the concept of feature sharing proposed in Pesetsky and Torrego (2007).

The empirical focus of the talk will be bound variable interpretations, in particular (fake) indexical pronouns in four Germanic languages—English, Dutch, German, and Icelandic. In all four languages, bound variable interpretations are available for 1st and 2nd person pronouns in focus constructions such as Only I did my best—in these contexts, the 1st person pronoun my is not interpreted as the speaker in the set of alternatives (no one else did their best), but as a variable, hence the term fake indexicals. Fake indexicals are, however, restricted in relative clauses of the form I am the only one who takes care of my son/(*)who did my best: English and Dutch allow a bound variable interpretation of my in such contexts, whereas German and Icelandic prohibit fake indexicals (my can only be interpreted as referring to the speaker).

Kratzer (2009) proposes a morpho-syntactic spell-out approach for English vs. German, in which the feature sets of the relative pronoun, T, v, and the possessive pronoun unify, leading to conflicting 1st/3rd person feature specifications on T and the possessive pronoun, which leads to a fatal spell-out dilemma in German. In English, on the other hand, markedness rules allow ignoring certain features, and the spell-out dilemma can be resolved in favor of person for the possessive pronoun (my) and in favor of gender for verbs (with gender corresponding to 3rd person). This account does not address why only some languages have such markedness rules, in particular not why Dutch patterns with English and Icelandic with German.

Based on a series of word order differences, I show that the nature of the verbal inflection is irrelevant for the licensing of fake indexicals, but that the crucial factors are: c-command by the antecedent DP, and a locality condition favoring feature sharing with the closest possible binder whenever possible. I argue that the relevant difference between the two language groups lies in the morphological make-up of the head DP of the relative clause (and in German also the relative pronoun): the relative DP shows morphological gender distinctions in the singular in German and Icelandic, but not in English and Dutch. I propose that the lack of gender features allows by-passing the closest binder (the relative DP/pronoun) and long-distance binding by the indexical matrix subject directly, whereas morphologically fully specified relative pronouns/DPs restrict binding to the relative DP. Various consequences supporting this approach will be discussed, among them non-indexical bound variable contexts which show the same locality effect when gender agreement is attempted across a relative pronoun/DP with different morphological features.