This paper investigates cross-linguistic variation in extractability out of subjects. Our new acceptability judgment experiment, which controlled for a ‘long-before-short’ preference in Japanese, shows that Japanese subjects are as transparent to extraction as objects. Based on this finding, we classify languages into three subtypes with respect to subject opacity: (a) languages in which subjects are never transparent (English), (b) languages in which only in situ subjects are transparent (Russian, German), and (c) languages in which subjects are always transparent (Japanese). To account for these data, we offer a preliminary formal analysis that links this typology to the existence of subject Agreement and the EPP: (a) languages where subjects undergo Agreement and EPP movement, (b) languages where subjects only undergo Agreement, and (c) languages where subjects do not undergo either operation.

We present and analyze novel data on the northeast Caucasian language Archi illustrating a typologically unusual phenomenon of apparent agreement between 1^st person pronouns and absolutive-marked arguments. Apart from their typological significance, these facts challenge current approaches to agreement, which hold that Agree relations can be established only between heads and phrases. The apparent agreement between a 1^st person pronoun and an absolutive DP can be reduced to a more conventional agreement, namely, agreement between the absolutive DP and a series of v heads. We show that Archi has a contrast between strong and weak pronouns; the latter lack noun-class feature specification and must therefore copy a class feature from the closest v. In addition, Archi has complex pronouns (1^st person inclusive) which are composed of  1^st person exclusive pronouns and the focus marker -ejt’u. This focus marker is a D head which requires a noun-class feature and copies that feature from the closest v head. 

This article examines the knowledge of topic and subject particles in heritage speakers and L2 learners of Japanese and Korean. We assume that topic marking is mediated at the syntax-information structure interface, while subject marking pertains to narrow syntax. In comparing phenomena mediated at different levels of linguistic organization, we provide evidence for the hypothesis that information structure-level phenomena present greater challenges for bilingual speakers than those mediated within syntax. While these results may be interpreted as evidence of generalized interface-related deficits, we show that such a global explanation is not supported. Instead, a more nuanced account is developed, based on the recognition of different types of topic (anaphoric, generic, and contrastive) and different types of subject (descriptive and exhaustive). Under the proposed account, the non-native speakers’ deficits follow from three unrelated effects: the status of topic as an interface category, structural complexity, and the memory demands necessary for its interpretation in context.

This paper investigates the geometry of phi-features with a special

emphasis on number and gender in Spanish. We address (i) whether number

and gender evidence single- or multi-valued systems for their respective

features, and (ii) whether number and gender are bundled together or split.

Given the lack of consensus on these issues based on primary data, we approach

these questions experimentally, using the phenomenon of agreement

attraction: a situation in which ungrammatical sequences are perceived as

grammatical when one of the NPs is erroneously identified as determining

agreement. Our results indicate that number but not gender in Spanish is

multi-valued and offer novel support in favor of an agreement model in which

number and gender are in separte projections and are valued independently.

We also offer some considerations concerning parametric variation in such

valuation.

A description of Tsez syntax, created as part of the volume “A Grammar of Tsez” prepared by Bernard Comrie and myself. Because this work is still a draft, chapters are not numbered, and each chapter has its own example numbering. I am posting it now in hopes of getting comments, criticisms, and suggestions before the description is sent to the publishers.