this paper, I shall present a corpus-based analysis of Subject-Prodrop in Yiddish, a language in which Subject-Prodrop has not yet been analyzed, and I shall show that, at least in this corpus of this language, the term 'Subject-Prodrop' is a rubric covering phenomena that have diverse syntactic and discourse constraints. In what follows, I shall first describe the corpus and present the syntactic constraints found. Then I shall present the results of a discourse analysis of Subject-Prodrop in Yiddish with respect to Centering Theory. Finally, I shall discuss the implications of these findings. 1. Yiddish Subject-Prodrop: the facts.

this paper. We thank Michal Biran, Mali Gil, Aviah Gvion, and Dafna Wenkert-Olenik for their help in discussions and testing, and the participants for their patient participation. Address correspondence to Naama Friedmann, School of Education, Tel Aviv University, Tel Aviv 69978, Israel. E-mail: naamafr@post.tau.ac.il, http://www.tau.ac.il/~naamafr

Verb raising and subject inversion have long been topics of theoretical linguistic interest in Romance and Germanic languages, amongst others. Bantu languages also exhibit verb raising and subject inversion, though there has been no comprehensive investigation of these phenomena, nor an explanation of crosslinguistic differences. This paper provides a unified account of verb raising and subject inversion in Bantu languages. It shows that subject inversion in Bantu matrix clauses resembles that found in Romance languages. In contrast, however, verb raising to C (similar to V2 in German matrix clauses) occurs only in embedded relative clauses, and only in some Bantu languages. A natural explanation for these phenomena comes from the fact that verb raising interacts with the prosodic status of the relative complementizer, and that Bantu matrix clauses are IPs not CPs. The paper points to the importance of competing interactions between different aspects of the grammar (e.g. prosodic words, syntax) and provides support for the notion of extended projections (Grimshaw

In the Minimalist Program (Chomsky 1995), Merge is a set operation that imposes no intrinsic ordering among its members. However, syntactic structures are linearized into strings of words at PF. This paper proposes that in order for a Merger set to be linearized, its members must be either hierarchically displaced or morphologically fused into a single terminal node.

Root Infinitives (RIs) are default verb forms which young children use in root clauses, where they are generally not possible in the target language. They have been found in the early speech of two-year old children learning a number of different languages, and their basic properties have been extensively documented elsewhere (Weverink 1989; Wexler 1994; Rizzi 1994). For my purposes here their most important property is that they are not distributed randomly throughout a child’s speech—they are almost invariably unmoved verbs (Pierce 1992; Poeppel & Wexler 1993). There is a rapidly growing variety of accounts of RIs, but there is one key property that the majority of them share, and that is that two-year olds produce RIs because they have some deficit of syntactic representation. The details of what children can omit from their representations vary quite dramatically from theory to theory, and range from general deficits in all functional projections (Guilfoyle & Noonan 1988; Radford 1990), to deficits in specific heads (Tense: Guilfoyle 1984; Wexler 1994; COMP: Clahsen et al. 1994), plus intermediate views in which the deficit is in knowledge about what is obligatorily projected