PROSODIC DIAGNOSTICS FOR PHRASAL MOVEMENT IN POLISH Piotr Banski The paper addresses the question of how to distinguish empirically between remnant movement of a phrase from which all non-head material has been extracted and movement of the head of that phrase alone. 1. The problem The case in question are constructions involving predicate adjectives in Polish, such as zmeczeni 'tired-VIRILE,PL', glodni 'hungry-VIRILE,PL', etc., exemplified under (1). (1) a. zmeczeni=smy bardzo tired=1PL very "We're very tired" b. bardzo=smy zmeczeni very=1PL tired The two options of movement possible for (1a) are head-movement of the adjective (2a) or remnant movement of the entire AP (2b), as sketched below, where the projection the clitic is in is marked as X for simplicity. (2) a. [X0 zmeczeni [X0 smy]] ... [AP bardzo t ] b. [XP [AP t zmeczeni ] [X0 smy] ] ... bardzo t I argue (together with e.g. Borsley & Rivero 1994) that it is option (2b) - phrasal movement - that can only take place in this context, and that this is unambiguously shown by the way in which the lexical phonological rules of Polish (especially the rule of Stress Assignment) treat the resulting host+clitic complex. 2. Stress patterns in clitic constructions as diagnostics for the phrasal status of the host As I argue elsewhere, auxiliary clitics in Polish are sensitive to whether they attach to heads or entire phrases (this sensitivity boils down to the difference in the resulting prosodic structure of such complexes). Descriptively, there is variation in whether the plural clitic integrates with the Prosodic Word projected by the host or not, depending on whether the host is a head element. This is illustrated under (3), where the stressed vowel is capitalized and bold-faced. (3) a. czytAli=smy vs. czytalI=smy read-PRT=1PL "We read" b. zEby=smy vs. zebY=smy SUBJ=1PL c. chcial, zebysmy znikneli want-PRT SUBJ=1PL vanish-PRT "He wanted us to vanish" Example (3a) involves the so-called 'l-participle' of the verb czytac 'read', and example (3b) the subjunctive complementizer zeby, as used in e.g. (3c). Thus, if the plural clitic attaches to a syntactic head, two stress patterns (penultimate and antepenultimate) are possible. Whenever it attaches to an element which is undoubtedly a phrase, however, the antepenultimate pattern is the only option, as shown below. (4) a. kOgo=smy vs. * kogO=smy (widzieli?) who,ACC=1PL see-PRT "Who did we see?" b. JAcka=smy vs. * JackA=smy (widzieli) Jacek,ACC=1PL see-PRT "We saw Jacek" c. czEsto=smy vs. * czestO=smy (tam chodzili) often=1PL there go-PRT "We used to go there often" Examples in (4) involve a wh-element, a fronted object and an adverb - all of them having uncontroversially phrasal status. Note that it is crucial that clitics are sensitive to the phrasal status of the host (understood in prosodic terms) rather than its category, hence the dividing line is between phrases and heads rather than non-verbs and verbs (recall the data in 3b in this context). The stress pattern of the combination of a predicative adjective and a plural clitic (1a) can only be antepenultimate, which I treat as clear indication that phrasal movement of the entire AP must take place in such cases. 3. Other issues Apart from the main points summarized above, the paper also briefly discusses the reason for the stress pattern variation found in clitic constructions (or lack thereof) and sketches a theoretical account for the descriptive facts presented above, which crucially depends on the analysis of the Polish copula jest 'be' which I develop elsewhere. Briefly, I claim that head movement of the adjective in (2a) would be non-local - barred by a copula projection intervening between the location of the clitic and the AP. This impossibility of head-movement forces (sometimes remnant) phrasal movement in cases where the adjective is positioned before the clitic. References: Borsley, Robert and Marķa Luisa Rivero (1994). "Clitic auxiliaries and incorporation in Polish". Natural Language and Linguistic Theory 12:373-422.