James E. Lavine
On a new (affixal) AUX in Polish


     This paper discusses the status of the predicate-final morphology in Polish -no/-to constructions, such as those given in (1):

	(1)a.	Znaleziono	niemowle	w	koszu.
		found:-NO 	baby:ACC	in	basket
		'They found a baby in a basket.'
	   b.	Stefana 	wzieto 	do	w—jska.
		Stefan:ACC	taken:-TO	to	army
		'They drafted Stefan into the army.'

We begin with the observation dating back (at least) to Oesterreicher (1926) that the ACC direct object in structures such as (1) came to be used productively only after the auxiliary's disappearance. Other well-known properties of this construction include (i) the presence of a syntactically-active pro-arb external argument and the corresponding lack of a passive przez-phrase, (ii) the cooccurrence of /-no/-to/ with unaccusative predicates and REFL sie, (iii) a robust set of "implicit subject properties" (binding and control), and (iv) an obligatory [+past] interpretation. The question is how these seemingly unrelated properties are to be unified in an explanatory way. That is, why, e.g., does the presence of a syntactically-active external pro-arb argument pattern with the obligatory absence of tense-marking auxiliaries?  
     My proposal is that Polish /-no/-to/ in synchronic terms is no longer passive-participial, but instead is interpreted as the head of T(ense). I refer to this proposal as the AUX HYPOTHESIS. The first piece of evidence for this proposal is the complementary distribution of /-no/-to/ and anything else heading T. A second piece of evidence is the obligatory [+past] interpretation. Finally, the AUX HYPOTHESIS predicts a host of other properties of this construction. Note, in particular, that if /-no/-to/ is non-passive-participial, the lack of dethematization of an external argument and the presence of implicit subject properties follows naturally; so does the patterning of Polish -no/-to with unaccusatives and the lack of a passive by-phrase. The relevant data are well-known from Wolinska 1978, Rozwadowska 1992, Dziwirek 1994, among others.
     To be sure, however, Polish /-no/-to/ does not overtly occupy the head of Tense. After establishing the AUX HYPOTHESIS (the main empirical claim of the paper), the second part of the paper is devoted to the problem of adjoining the material heading T (/-no/-to/) with the verb-stem (the main theoretical issue of the paper). Linear adjacency in such cases (i.e., getting the verb and affix together) is an old problem for which a variety of mechanisms have been introduced, ranging from "affix hopping" to covert LF feature-movement. Thus, the theoretical choice would seem to be overt lowering vs. covert raising. The main problem is that /-no/-to/, as an affixal head, cannot be stranded in T. The possible solutions for this problem considered thus far are schematized in (2):

(2)	[TP [T' -no/-to [vP pro-arb [v' v [VP [V' V-no/-to  NP:ACC]]]]]] 
														a.	LF feature-movement (features of lower /-no/-to/ move to T)

	b.	overt lowering (affix-hopping) (/-no/-to/ moves from T to V)

The options in (2a-b) are both unattractive theory-internally: covert feature movement relies on the stipulation of feature-strength, a theoretical tool that is seemingly manipulated to suit any set of data; overt lowering is unconstrained by the Proper Binding Condition, which requires traces of movement to be bound. 
     I conclude, instead, that the problem of linear adjacency between the verb-stem and the material heading T (/-no/-to/) is not strictly syntactic. Following recent work in Distributed Morphology (Halle & Marantz 1993), and building on the fact that stranded affixes essentially constitute a PF-violation, I show that /-no/-to/ and V- "get together" postsyntactically, in a morphological component. Specifically, the proposal entails that the morphosyntactic features of /-no/-to/ merge in the head of T (with no phonetic content), while the mapping to PF merges /-no/-to/ on V- (Late Insertion). This is a particular instance of Morphological Merger (Marantz 1988), whereby a syntactic relation between two heads is replaced in the PF component by an affixation relation. This is schematized in (3):

(3)		[X ... [Y ...]] -> [Y + X]

Distributed Morphology is designed precisely to account for such mismatches between syntactic interpretation and PF-realization (Spell-Out) (cf. clitics). Once properly understood, Polish -no/-to emerges as an enormously interesting testing ground for cases in which non-syntactic movement derives the common interpretation/pronunciation split.

Dziwirek, K. 1994. Polish subjects. New York: Garland.
Halle, M. & A. Marantz. 1993. Distributed Morphology and the pieces of      
 inflection. The view from Building 20, ed. K. Hale & S. J. Keyser, 111-176. 
 Cambridge, Mass: MIT Press.
Marantz, A. 1988. Clitics, morphological merger, and the mapping to phonological
 structure. Theoretical Morphology, ed., M. Hammond & M. Noonan, 253-270. San
 Diego: Academic Press.
Oesterreicher, H. 1926. Imiesl—w bierny w jezyku polskim. Krak—w: Polska
 akademja umiejetnosci.
Rozwadowska, B. 1992. Thematic constraints on selected constructions in English
 and Polish. Wroclaw: Wydawnictwo uniwersytetu wroclawskiego.
Wolinska, O. 1978. Konstrukcje bezmianownikowe we wsp—lczesnej polszczyznie.
 Katowice: Wydawnictwo uniwersytetu slaskiego.