CLITIC AUXILIARIES and INCORPORATION in POLISH* the Following Sentences Illustrate an Intriguing Feature of Polish Syntax

ROBERT D. BORSLEY AND MARIA LUISA RIVERO

CLITIC AUXILIARIES AND INCORPORATION IN POLISH*

This paper discusses the two alternatingsyntactic patterns of Polish past and conditional sentences from a Slavic perspective.We argue that what are often referred to in Polish as past tense verbs, for example widziale.s'you saw', are in fact combinations of a past participleand a perfect auxiliary,e.g., widzial 'seen' and s 'you have'. These combinationsare the result of syntacticIncorporation in the sense of Baker(1988). Whennot combinedwith the participle,the auxiliarycan appearalmost anywhereto the left of the participlewithin the same clause. We argue, however, that it alwaysoccupies the same syntacticposition, only to undergoPF-cliticization. The auxiliarycombines with a varietyof elementsbecause phrasalfrontings such as Wh-movementand Scramblingallow a varietyof categoriesto immediatelyprecede the I-node occupiedby the auxiliary.The proposalthat the auxiliaryappears in the I-node alone or incorporatesthe participleexplains why certainitems can host a clitic auxiliarywhile others cannot. A second auxiliarythat incorporatesa participleis the conditionalauxiliary, as in widzial+ bys 'you would see'. However, the conditional auxiliaryis not a clitic and hence, unlike the perfect auxiliary,can appearin initial position. We argue that Polish is uniqueamong West and South Slavic languagesin havingIncorporation. Bulgarian sentences like cel sum 'I have read' and counterparts in Czech, Serbo-Croatian,and Slovak appear similar to Polish examples involving Incorporation.However, they are the productof Long Head Movement, i.e., the movementof the participledirectly to the C-positionacross the auxiliary.We argue that Polish sentencesinvolving Incorporation differ in syntacticproperties from Long Head Movementconstructions in the other languages.

1. THE Two SYNTACTIC PATTERNS OF PASTS AND CONDITIONALS The following sentences illustratean intriguingfeature of Polish syntax:

(1)a. Ty widzial- es' tq ksiq&q. you seen + MASC + SG- 2SG this book

* Researchfor this articlehas been partlysupported by the EuropeanScience Foundation under the Eurotyp Project, and by Grants 410-88-0101and 410-91-0178from the Social Sciences and HumanitiesResearch Council of Canadato the second author. Preliminary versionswere presentedto the joint meetingof the Cliticsand the Complementationgroups of the EurotypProject in Viennain October1991, and to the LinguisticAssociation of Great Britainin Brightonin April 1992. We thankEwa Jaworskaand Jacek Witkosfor help with the Polish data, Riny Huybregtsand Henk van Riemsdijkfor helpful comments,Danijela Kudraand LjiljanaProgovac for discussionof the Serbo-Croatiandata, three anonymous reviewersfor insightfulcriticism, and FrederickNewmeyer for very useful editorialadvice. Usual disclaimersapply.

NaturalLanguage and LinguisticTheory 12: 373-422, 1994. (D 1994 KluwerAcademic Publishers. Printed in the Netherlands. 374 ROBERT D. BORSLEY AND MARfA LUISA RIVERO

(1)b. T-s widzial te; ksi#?kq. you 2SG see + MASC + SG this book You saw/have seen this book. In (la), we have what is often referred to as a past tense verb form: widziates''you saw'. In (lb), the element -s, which one might conclude from (la) is a verbal inflection, is attached to a subjectpronoun. In this paper, we will look in some detail at this phenomenonand argue that the so-called 'past tense' is in fact a syntacticcomplex formed by a participle and a perfect auxiliary.We will also consider how the auxiliarycan be combinedwith other elements of the sentence; on the face of it, when the auxiliary does not form a syntactic complex with the participle, it can appear almost anywhereto the left within the same sentence. Consider, for example, the following: (2)a. Ty jego widzial- es. You him see + MASC + SG- 2SG b. Ty- s jego widzial You- 2SG him see + MASC + SG c. Ty jego- s widzial You him- 2SG see + MASC + SG You saw him. In (2a), the auxiliaryattaches to the preceding participle,in (2b) to the subject, and in (2c) to the object. There are even examples in which the auxiliaryappears to be inside a phrase. Thus we have both (3a) and (3b): (3)a. Ewy ksi#?kqczytal- es. Ewa's book read + MASC + SG- 2SG

b. Ewy- s ksi4zkqczytal. Ewa's-2SG book read + MASC + SG You read Ewa's book. However, there are restrictionson the position of the auxiliary:it cannot appear in a higher clause than the participle, attach to items to the right of the participle,or follow the negative particleor transitiveprepositions. We will argue that the combination of participle and auxiliaryis the consequence of syntacticIncorporation in the sense of Baker (1988: 22): the participle raises to the auxiliaryin the syntax. We will also propose that when the auxiliaryis not attachedto the participle,it alwaysoccupies CLITIC AUXILIARIES AND INCORPORATION IN POLISH 375

the same syntactic position, the lo-node, and subsequentlyundergoes a process of PF cliticization.In Polish, clitic auxiliariesare not in a special second position in the clause and can attach to a variety of elements because independentlymotivated phrasal movements allow these items to immediatelyprecede them. Polish auxiliariesfail to attach to items which cannot prepose independently. We have similar phenomena in conditional sentences. The following resemble the examples in (1):

(4)a. Ty wizial- bys' tq ksi#zkq you seen + MASC + SG- COND + 2SG this book

b. Ty bys widziat t ksi#ikq. you COND + 2SG seen + MASC + SG this book

c. Bys widzial te ksiBzkq. COND + 2SG seen + MASC + SG this book You would see this book.

In (4a) the conditionalauxiliary is attachedto the participle.In (4b-c) it is not attachedto the auxiliaryor anythingelse. Like the perfect auxiliary, the conditionalcan on the face of it appear anywhereto the left of the participleincluding inside a phrase. Thus we have the examplesin (5) and (6), parallelto those in (2) and (3):

(5)a. Ty jego widzial- bys. You him see + MASC + SG- COND + 2SG

b. Ty bys jego widzial. You COND + 2SG him see + MASC

c. Ty jego bys widzial. You him COND + 2SG see + MASC You would see him.

(6)a. Ewy ksi#kq czytal- bys. Ewa's book read + MASC + SG COND + 2SG

b. Ewy bys ksi#?kqczytal. Ewa's COND + 2SG book read + MASC + SG You would read Ewa's book. 376 ROBERT D. BORSLEY AND MARfA LUISA RIVERO

We will argue that the conditionalauxiliary is very much like the perfect auxiliary,except that it is not a PF clitic. An importantfeature of the analysesthat we develop here is that they minimizethe differencebetween Polish, other West Slavic languages,and South Slavic languages. Polish has Incorporation, whereas the other languages have Long Head Movement, that is, the movement of the participleto the complementizeracross the auxiliary.Otherwise they are very similar. In addition, our analysis accounts for the distributionand positional restrictionsof Polish auxiliaries, providing an explanation of why certain items can host a clitic auxiliarywhile others cannot. The paper is organizedas follows. Section 2 discussesPolish Incorpor- ation of the verb to the auxiliary, and its syntactic contrasts with Long Head Movement of the verb to the complementizerin other Slavic languages. Section 3 deals with the position of the auxiliaryin non-incorporated patterns, arguingfor two main conclusions:first, the auxiliaryis in the I?-position,just like a finite verb; second, word order results from the frontingof the phrasepreceding the auxiliary.This latter conclusionshows that Polish is a configurationallanguage with movement properties attested in many other languages.Section 4 considersnegation in the Polish clause and its interaction with Incorporation.Section 5 argues, on the basis of conditionals in subjunctiveand counterfactualclauses, for the hypothesisthat Incorporationin Polish is a syntacticrather than a lexical operation.

2. SENTENCES WITH INCORPORATION FROM A SLAVIC PERSPECTIVE In this section, we develop our analysis for Polish sentences involving Incorporation.Polish past and conditionalsentences are particularlyinter- esting from a general Slavic perspectivebecause they differ in important ways from superficiallysimilar sentences in other Slavic languages. We argue that this is because they involve the incorporationof a participleto the auxiliaryin the syntax, a process whichis absent from the other Slavic languages.Syntactic differences between Polish and languageslike Czech, Slovak, and Bulgarianmotivate this conclusion.

2.1. Past TenseParadigms Let us begin with the full range of past tense forms of a typical Polish verb such as widziec 'see' in (7): CLITIC AUXILIARIES AND INCORPORATION IN POLISH 377

(7) SINGULAR PLURAL 1P: a. widzial-em 'I saw' widzieli-smy 'we saw' seen + MASC + SG-1SG seen + VIRILE + PL-1PL1 a'. widziala-m widzialy-smy seen + FEM + SG-1SG seen + NON-VIRILE+ SG- 1PL 2P: b. widzial-eg 'you saw' widzieli-scie 'you saw' seen + MASC-2SG seen + VIRILE + PL-2PL b'. widzialaA widzialyAcie seen + FEM + SG-2SG seen + NON- VIRILE + PL - 2PL

3P: c. widzial 'he saw' widzieli 'they saw' seen + MASC + SG seen + VIRILE + PL c'. widziala 'she saw' seen + FEM + SG widzialy seen + NON-VIRILE + PL c". widzialo 'it saw' seen + NEUT + SG

This paradigmmakes it clear that the verb stem is followed by forms that show numberand gender and that an additionalmarker indicating person and number completes the word. For instance, in the first person masculine singularform widziatem'I saw', widzi- is the stem, -al- shows number (Sg) and gender (Masc), and the suffix -em indicates number (Sg) and person (1P). Imperfective futures consist of an auxiliary and a participle as free forms, and as illustratedin (8), the participleswhich follow the future auxiliary are identical to the combination formed by the stem and the internal markers in the paradigmin (7). For instance, the bound form widzial- in the masculine singularpersons in (7) is identical to the free participlein (8a), which is a masculinesingular form. Gender is a charac- teristic of participles,but not of other verb forms:

1 'Virile' is the traditional term in Slavic philology for the plural [+male] form, as opposed to the feminine/neuter plural form. 378 ROBERT D. BORSLEY AND MARfA LUISA RIVERO

(8)a. Chiopiecbvdzie widzial ksi#?kq.2 boy will see + MASC + SG book The boy will see the book.

b. Dziewczynkabqdzie widziala ksi?k&. girl will see + FEM + SG book The girl will see the book.

c. Chlopcybqd# widzieli ksi?kq. boys will see + VIRILE + PL book The boys will see the book.

d. Dziewczynkibqd# widzialy ksi?kq. girls will see + NON-VIRILE+ PL book The girls will see the book. Given the formal identity between the initial fragmentsof the 'pasts' and the participles of imperfective futures, we therefore conclude that the forms in (7a) consist of a participlefollowed by an auxiliaryfunctioning as affix, as in (9): (9) [[PARTICIPLE ] AUX ] a. [[widzial- ] em ] 'I saw' V+MASC+SG 1SG. b. [[widzieli- ] smy J 'we saw' V + VIRILE + PL 1PL A similar analysisseems reasonablefor conditionalslike widzial-bys'you would see'. We therefore conclude that 'conditionals'also consist of a participlefollowed by an auxiliary.

2.2. The AlternatingSyntax of Past Sentences As we noted at the outset, the Polish past does not always appear in the forrmshown in (7), but exhibits two alternativesyntactic patterns. It can consist of a participlepreceding an auxiliary, as in (10), exemplified by (la) and repeated now as (12a), an order which is frequently used in

2 The future auxiliarycan also take an infinitivecomplement. (8a) is equivalentto (i) Chlopiez bedzie widzied ksia&ke boy will see + INF book CLITIC AUXILIARIES AND INCORPORATION IN POLISH 379 modern Polish (Andersen 1987). Alternatively,the auxiliarycan precede the participle,as in (11), exemplifiedby (lb) and repeated now as (12b). This order is now less frequent: (10) (X) Participle-AUXY. (11) X-AUX ParticipleY.

(12)a. Tyeso widzial- t ksisbok. you seen + MASC + SG- 25G this book

b. Ty- s- widzial tq ksi#ikq. you 2SG see + MASC + SG this book You saw/have seen this book. From our perspective,the synonymouspatterns (12a) and (12b), or parallel conditionalsentences, are the resultof Head-movementwith the participle incorporatingto the auxiliaryin the first instance, as in (13a), and representauxiliaries with the participleas complementin situ in the second instance, as in (13b):3

3 The participlesin the VP could be the result of Head-Movementof the verb to a c- commandingAgreement head, as in (i). Such an analysisis in the spiritof Kayne'sproposals for Romanceparticiples with Object Agreement (1989 and later work). Since this analysis is immaterialfor our purposes,we will treat the participleas V: (i) AuxP

Aux Agri)

AgrO VP

V0 380 ROBERT D. BORSLEY AND MARfA LUISA RIVERO

(13)a. IP

ty It

10 VP

widzialieg V0 NP

fj tq ksi#zkq

PARTICIPLEINCORPORATION

b. IP

ty I

10 VP

widzial tq ksi#ikq

UNINCORPORATED PERFECT

It has sometimes been suggested that patterns such as (12b), where the auxiliaryprecedes the participle,are the resultof a process movingcertain inflectionsaway from a verb stem to the left (Gussman1980, Sussex 1980, Borsley 1984, among others). There are a numberof reasons for rejecting such a floating inflection approach to (12b). One is that it involves a questionable movement process which separates an affix from a stem. Another unusualfeature of the analysisis that it appearsto move an affix to a non c-commandingposition located inside a phrase. In our analysis, the 'floating'inflection is an auxiliarythat either cliticizes, giving (12b), or incorporatesthe participle,resulting in (12a). CLITIC AUXILIARIES AND INCORPORATION IN POLISH 381

2.3. Incorporation in Past Sentences The process moving the participle to the perfect auxiliary in the past sentence in (13a) has the formal properties of syntactic Incorporation, pace Baker (1988: 22). Namely, it is an instance of Move a and adjoins a syntactichead to anotherhead. That is, it takes an X? from an independent base structureposition to combine it with anotherX? in the syntax, creating a complex category of the X? level and establishinga syntactic link between two positions in the phrase marker. The proposed analysisimplies that both the auxiliaryin (13b) and the participleplus auxiliaryin (13a) occupy the same slot in Io as the verb in the present in (14), assuming that this last pattern results from head- movement of the verb to I':

(14) Kiedy widzisz sig z Janem? when see- Pres + 2SG REFL with Jan When are you seeing John?

As we shall see in Section 3, the predictionthat Polish auxiliaries,incorporatedforms, and ordinaryinflected verbs have the same syntacticdistri- bution appearsto be correct. The parallelism,as we will argue, is due to the fact that inflectedverbs and incorporatedforms have parallelsyntactic derivations. In the case of the inflected verb in the present, it is well accepted that a stem must raise to an affix. We propose a similarprocess for the incorporatedperfect: a participleraises optionallyto an inflected auxiliary.

2.4. Motivating Incorporation Havingsketched the differencesbetween patternswith and without Incor- poration, let us considerarguments in favor of Incorporation.Support for the idea that Incorporationis responsiblefor sentencesin whichthe participle precedes and is attached to the auxiliarycomes from stress. Polish has penultimate stress, but past forms are exceptional in that they can have either penultimateor antepenultimatestress (see Booij and Rubach 1987 for detailed discussion). Thus, phonologicalevidence indicates that the past czytaliYmy'we have read' is derived in a differentway from the superficiallysimilar czytamy 'we are reading'. Syntacticarguments for the proposal that participle+ auxiliary sequences involve Incorporationare based on their distributionand on the positionof pronominalclitics viewed from a general Slavic perspective. 382 ROBERT D. BORSLEY AND MARIA LUISA RIVERO

2.4.1. Incorporation versus Long Head Movement in Slavic Polish, Bulgarian, Czech, Serbo-Croatian,and Slovak share apparently identicalparticiple + auxiliary+ NP object sequences, as (15) illustrates: (15)a. Kupili- smy lustro. (Polish) bought- 1PL mirror We bought a mirror.

b. Pro6elisme knigata. (Bulgarian) read Pres + 1PL book + the We read the book.

c. Koupiljsi knihy. (Czech) bought Pres + 2SG books You bought books d. Citao sam knjigu. (Serbo-Croatian) read Pres + 1SG book I read a book. e. Nap'sal som list. (Slovak) written Pres + 1SG letter I wrote a letter. However, the paralellismis only apparentand hides a crucial difference. On the one hand, Polish (15a) involves.Incorporation, or the movement of the participleto the head of AuxP, as scheihaticallyshown in (16a). On the other the hand, the other Slavicexamples in (15) have Long Head Movement, or the fronting of the participle to Co, bypassing the finite Auxo altogether, as shown in its essential aspects in (16b) (see Rivero (1994) for Rumanianand Bulgarian;Lema and Rivero 1989 and 1992 for Bulgarian, Czech, European Portuguese, Old Spanish, and Rumanian; and Rivero 1991 for Serbo-Croatian,Slovak, and Czech in more detail): CLITIC AUXILIARIES AND INCORPORATION IN POLISH 383

(16)a. CP b. CP

Co AuxP C0 AuxP

Aux0 VP Vi Aux0 VP

Vi Aux0 li NP ti NP

POLISH INCORPORATION SLAVIC LONG HEAD MOVEMENT Polish is unique within Slavic because it has Incorporationbut no Long Head Movement, while the other languageshave Long Head Movement and no Incorporation. Similarcomments apply to the Polish conditionalin (17a), in contrast with Czech and Serbo-Croatian(17b-c). These patternsmay look identical, but they hide a major difference with important syntactic conse- quences. In Polish they result from Incorporationof the participleto the conditional, as in (16a), while in the other languagesthey involve raising of the participleto the tomplementizer, as in (16b):

(17)a. Czytal bym tq ksi4ikq. (Polish) read- Cond + 1SG this book I would read this book.

b. Koupil bych knihy. (Czech) bought Cond + 1SG books I would buy books.

c. 'itao bih knjigu. (Serbo-Croatian) read Cond + 1SG book I would read a book.

In Polish the syntacticraising of the participleresults in a complexdomin- ated by a single X? in the phrase-marker,and this X? containsthe auxiliary; in addition, in surfacestructure this Polish X? is in 10,where ordinary inflected Vs are also located. In the other languages, the raising of the participleresults in two differentXos in the syntacticphrase-marker: the 384 ROBERT D. BORSLEY AND MARfA LUISA RIVERO higherX? is the complementizer(CO) and containsthe participle,and the lower X? containsthe auxiliary. We will now justifythe claim that Polish participle+ auxiliarysequences are the product of Incorporationof the participleto the auxiliary. Such sequences are unrestrictedin distribution,in that they are found in root and non-root environments,including complements and relative clauses, as in (18a) and (18c). Incorporationis optional, so (18a), the patternwith Incorporation,alternates with (18b), the patternwithout Incorporation:

(18)a. Janek powiedzial, ze pojechal- es do Warszawy. Janek said that gone- 2SG to Warsaw Janek said that you went to Warsaw.

b. Janek powiedzial, ze-s pojechal do Warszawy. c. Piotr powiedzialdokladnie, co kazal es Piotr said exactly what ordered-2SG Piotr said exactly what you ordered

By contrast,Slavic LHM is a root phenomenon,so participle+ auxiliary sequences occur in main and independent clauses but not in embedded clauses with non-root characteristics,such as relativeclauses, or the complement clause in (19):

(19) *Znam ce proceli sme knigata. (Bulgarian) I + know that read Pres + 1PL book + the I know that we read the book.

Notice too that participle+ auxiliarysequences have the same distribution in LHM languageswhere clitic auxiliariesmust necessarilybe in the second syntactic position in the clause, such as Czech, and in languages where they need not, such as Bulgarian.In Bulgarian,the clitic auxiliarycan be in the third syntacticposition in the clause with no ill-effects, as in Znam [ce nie ste sme proceli knigata]'I know [that we will have read the book]'. This order is impossible in Czech. The root nature of these sequences follows from the hypothesis that the non-finite verb moves to Co. This movement is parallel to V2 in Germanic, where the finite verb raises to C? in sentences with root-like properties. The hypothesis that Polish participle + auxiliary sequences are the result of Incorporationof the participle to the auxiliary in I and not movement of the participle to C thus accounts for the less restricted CLITIC AUXILIARIES AND INCORPORATION IN POLISH 385

distributionof such sequences in Polish compared with the other Slavic languages.

2.4.2. Clitic Pronouns in Slavic A second argument for the view that Polish participle + auxiliary sequences are the result of Incorporationcomes from the distributionof clitic pronouns. The contrast between Polish and other Slavic languages again provides the basis for our conclusion. Consider first the following Bulgarianexamples:

(20)a. Toj go e vizdal. (Bulgarian) he him Pres + 3SG seen He saw/has seen him.

b. Vizdal go e. He saw/has seen him.

Sentence (20a) showsthe auxiliaryfollowed by the participleas its complement. In (20b), the participle is in initial position, preceding both the auxiliarye 'has', and the clitic pronoun go 'him'. The fact that the clitic pronoun intervenes between participle and auxiliary suggests quite stronglythat they do not form a complex X? in the syntax. In other words, it supports the view that participle + auxiliarysequences in Bulgarian are the result of LHM, i.e., the movement of the participle to C. The same fact holds for Serbo-Croatian,old and conservativeRomance, and Romanian. Polish cliticpronouns appear in a varietyof positions, and their distribution is quite free (see Spencer 1991: 267-269). For example, they may follow any firstmajor constituent such as the first Wh-wordin the multiple Wh-questionin (21) (Rudin 1988). Notice that examplessuch as (21) show the clitic to be an X? which is independentfrom the verb in syntax:

(21) Kto siq komu podoba? who Refl to + whom likes? Who likes who?

By contrast,Bulgarian clitic pronounsmust appearstrictly adjacent to the inflectedverb or auxiliaryand follow all Wh-phrases.However, like clitic pronouns in other Slavic languages, Polish pronouns may not appear in initial position, as (22)-(23) illustrate: 386 ROBERT D. BORSLEY AND MARfA LUISA RIVERO

(22)a. Maria go kocha. (Polish) Maria him loves Maria loves him.

b. Kocha go. She loves him.

c. *Go kocha.

(23)a. Nie go vidiaxme. (Bulgarian) we him saw We saw him.

b. Vidiaxme go. (We) saw him.

c. *Go vidiaxme. Given the variety of positions in which Polish clitic pronounscan appear, we would expect it to be possible for a clitic pronounto appearbetween a participleand the followingauxiliary, as in Bulgarian.This is not possible, however, as the contrastbetween (24) and (25) illustrates: (24)a. Zmusil-em go do zrobieniatego. forced- 1SG him to doing this I forced him to do this.

b. Dal- es mi go. given- 2SG me it You gave it to me.

c. Widzial-es go. seen- 2SG him You saw him.

(25)a. *Zmusil-go-mdo zrobieniatego. b. *Dal-mi-go-s. c. *Widzial-go-s'. The ungrammaticalityof the examples in (25) is expected if the participle and auxiliary form a complex X?. Here, then, we have some further evidence for the view that Polish participle + auxiliary sequences are CLITIC AUXILIARIES AND INCORPORATION IN POLISH 387

the product of Incorporationof the participleto the auxiliary.Although participle + auxiliarycannot be separated by a clitic pronoun, auxiliary + participlesequences can, as (26) illustrates: (26) Kiedy-s go widzial? when-2SGhim seen? When did you see him? This is predictedby our analysissince auxiliaryand participledo not form a complexX0. It is also possibleto have a clitic pronounbefore a participle + auxiliarysequence, as (27) illustrates: (27) Kiedy go widzial-es? when him seen- 2SG When did you see him? Again, this is as we would expect since the participle+ auxiliarysequence forms a complex X? with the same distributionas the inflected verb in (22a). We have now presentedtwo argumentsthat Polishparticiple + auxiliary sequences are the productof Incorporationof the participleto the auxiliary and not of LHM to Co, which is responsiblefor superficiallysimilar sequences in other West and South Slavic languages:these sequences are not restrictedto root contexts, suggestingthat they are not the result of movement to CO;they cannot be separatedby a clitic pronoun, suggesting that they form a complex X?.

2.4.3. Conditionals Polish conditionalpatterns can be analyzed in the same way as those of the perfect. That is, the conditionalauxiliary allows Incorporationof its complementparticiple, as illustratedin (29) for (28): (28) Czyj artykulchial-byg przeczytac? whicharticle liked-Cond+ 2SG read? Which articlewould you like to read?

(29) [AuxP [Au' [AWo [chiali [AXO bys]] [vp ti [ przeczyta6]]]] The argumentthat the conditionalinvolves Incorporationand that participle + conditional sequences are syntacticXos and occupy I is identical to that alreadydeveloped for the perfect. A variety of items can separate auxiliaryand participlewhen the two appear in that relative order: for 388 ROBERT D. BORSLEY AND MARfA LUISA RIVERO instance, in (30a) demonstrativeto splits.the two. However, when the participlebegins the clause and precedes the auxiliary,the two items are not separable,indicating that Incorporationhas appliedto form a syntactic complex dominated by a common X?. Thus (30b) is fine, but (30c) is ungrammatical.As predicted, participle+ conditionalsequences are also possible in subordinateclauses, providingmore supportfor Incorporation as opposed to LHM:

(30)a. Bym to zrobil. Cond + 1SG it done I would do it.

b. Zrobil-bymto.

c. *Zrobilto bym.

There are alternativeanalyses for conditionalsthat might be proposed, but they all have seriousdefects. One possibilityis a versionof the analysis proposed in Rivero (1991) for the Slovak conditionalin (31):

(31) Ja by som napisal list. (Slovak) I Cond Pres + 1SG writtenletter I would write a letter.

Rivero proposesthat such examplesinvolve the structurein (32). Namely, the conditionalmarker by is the Specifierof the AuxP headed by som:

(32) [A' by [Au' Aux [vP V 111

If this analysis were extended to Polish, the conditional auxiliarywould result from the combining of the perfect auxiliary and its specifier by. Something like this analysis is proposed for Polish in Dogil (1987) and Booij and Rubach (1987). However, it has two defects. Firstly, it would not, generate the grammatical(33a), while excluding the ungrammatical (33c). We analyze (33a) as in (33b): the verb must incorporate to the (inflected)auxiliary, as in the case of the perfect chia1-ef'you have liked', which excludes (33c). Secondly, (33d) would involve Incorporationof the auxiliary to its specifier, something which is not allowed on standard assumptions(Baker 1988). In our view, (33d) does not involve Incorpor- ation and shows the inflected auxiliaryfollowed by its VP-complement: CLITIC AUXILIARIES AND INCORPORATION IN POLISH 389

(33) a. Chcial- bys. liked- Cond + 2SG You would like.

b. [Auxo Vi [AUxO Cond]] tj c. *By chcial-es. d. Bys chcial. You would like.

A second possibility, suggested by an anonymousreferee, treats by as the head of a Modal Phrase with the maximalprojection of the perfect auxiliaryas its complement.This is similarto the analysisof the Bulgarian futureperfect Ste sum pro6el 'I will have read' proposedin Rivero (1991). On this analysis, we would have structureslike the following:

(34) [MP [Mo by][AuxP [Auxo 'I [VP [vochial]]]

However, in this approach, as in the previous one, it is not clear how (33c) could be excluded. Moreover, (33a) would involve a ratherdubious non-local Incorporationprocess. A third possibility would be to treat the perfect auxiliary as a head taking a complementheaded by by as auxiliaryor verb. By in turn would take the main verb as complement, as in (35a). This analysis must be rejected because it requires that by always raise to the perfect auxiliary as in (35b), to prevent the ungrammatical(35c) from being generated. Under this approach, (28) would require raising the main verb cheial to by, with the complex subsequentlyraising to the perfect auxiliary:

(35)a. Basic structure:AuxP[Auxo[-s] vpt vo[by-]vp[chcial]]]

b. By-Inc (oblig.): AUxp[Auxo[byi-sI vp[ vo[t,] vp[chcial]]] c. *Czyj artykul-es by chial przeczytac? which article-2SGCond liked read?

Our analysis, in which the conditionalauxiliary is a lexical entry different from the perfect auxiliary,avoids the problemsthat face the alternative approaches.In addition, under this treatmentit follows naturallythat by 390 ROBERT D. BORSLEY AND MARfA LUISA RIVERO should be the item with person markingwhen it precedes the participle, as in (33a), since this conditionalitem is an inflected auxiliary.4

2.5. Incorporation and LHM: a General Perspective We have arguedthat Polish perfect constructionsshare the basic structure of perfect constructionsin other Slaviclanguages: the clitic auxiliaryheads a phrase and takes a VP-complementheaded by a participle.However, Slavic languages differ in the type of Head-movementthat may affect participles:Incorporation for Polish and LHM for the other languages.In root clauses, these two movements may result in apparentlyidentical invertedperfects (or conditionals).However, when Incorporationapplies in Polish, the resultingsequences form a complex dominatedby a unique X? in syntax because the participleadjoins to the auxiliaryand forms a syntactic unit with it. In contrast, when LHM applies in the other languages, the two items are separateconstituents and do not form a syntactic X?. As a consequence,the incorporatedPolish past or conditionalformed by participle and auxiliaryoccupies the same X?-position as a present tense verb. By contrast,the participle+ auxiliarysequence in other Slavic languages does not behave like a present tense verb because it is not a syntacticX? resultingfrom Incorporation. Incorporationis unique to Polish within Slavic but has counterpartsin other language families. In Old Romance, futures and conditionalsare formed by the syntactic movement of an X? to another X? (Lema and Rivero 1991), just like Polish perfects and conditionals.For example, Old Spanish ver-emos '(we) will see' and ver-iamos '(we) would see' result from Incorporationof the infinitivever 'see' to the future and conditional auxiliariesemosliamos '(we) will/would'. Like their Polish counterparts, Old Romance verb + auxiliarysequences resulting from Incorporation occur in root and non-root environments. In the light of the proposed analysis, three types of languagescan be distinguishedwith respect to non-finiteV?-raising in the presence of clitic auxiliaries.The first type is representedby South and West Slavic, which offers two options as to the position of the non4lnite V. In languagessuch as Czech, Bulgarian,Serbo-Croatian, and Slovak, the non-finiteV? either

4 In rejectingthe alternativeapproaches, we are not denyingthat the conditionalauxiliary is plurimorphemic.However, we would suggestthat the perfect auxiliaryis plurimorphemic in just the same way. Perfect and conditionalconstitute separate lexical entries and carry agreementmarkers for person.Wedo not discuss these markers,but it could be that they correspondto an AgrP that takes the maximalprojection headed by the conditionaland perfect auxiliariesas complement:AgrP[Agr? AuxP[Auxo vp[V?]]]. CLITIC AUXILIARIES AND INCORPORATION IN POLISH 391

moves by LHM to C0 or remains in situ and follows the auxiliary,as we will show in section 3. The second type of language is exemplified by medievaland 'conservative'Romance. Here, languagessuch as Old Span- ish and 19th centuryEuropean Portuguese show either LHM of V0 to C0 (parallelto Slavic) or Incorporationof Vo to Auxo (like Polish), with the two patterns in almost perfect complementarydistribution (Lema and Rivero 1991). Polish representsa thirdtype of languagewith either Incor- poration of V0 to Auxo, as discussedin this section in detail, or the non- finite V0 in situ in patterns which do not involve Incorporation.Polish Incorporationis optional, so constructionswith V0 incorporatedto Auxo and those with V0 in situ are in free variation in almost all cases. In addition, Polish differs from Slavic and medieval and conservative Ro- mance in that it lacks Long Head Movement, i.e., the movement of V0 to C0 across Auxo.

3. AUXILIARIES IN SENTENCES WITH NO INCORPORATION Let us now turn to sentences without Incorporation,as in (lb) repeated here as (36):

(36) Ty-s widzial tq ksi#ikq. You've seen this book. Polish sentences where the participleand the auxiliaryare not fused are importantin that they provide motivationfor the hypothesisthat, despite its relative free word order, Polish is configurational(in the sense of Hale 1983). We will argue that the apparentflexibility in the position of the auxiliariesis the effect of phrasal movement rules which apply within a hierarchicalclause structure, in the spirit of recent work on other free word order languagessuch as Japaneseand German(Saito and Hoji 1983, Webelhuth 1985). We also demonstratethat the proposed movements, i.e., Wh-movementand Scrambling,always respect syntacticconstituency and do not operate on the basis of phonologicalstructure. We will show that Polish movements, which are varieties of Move a, obey the same constraintsas do well-documentedsyntactic processes in languageswhere word order is more restricted.

3.1. The Structureof Past Sentenceswithout Incorporation in Slavic Let us begin by showinghow Polish auxiliarystructures without Incorpor- ation resemble those of-other Slavic languages. We propose that Polish 392 ROBERT D. BORSLEY AND MARfA LUISA RIVERO

(36) is structurallyparallel to the Bulgarian,Czech, and Slovak examples in (37):

(37)a. Nie sme proceli knigata. (Bulgarian) we Pres + 1PL read book + the We read the book.

b. Tehdy jsem koupil knihy. (Czech) then Pres + 1SG bought books Then I bought books.

c. Ja som napisallist. (Slovak) I Pres + 1SG writtenletter I wrote a letter.

In our view, the auxiliariesin (36) and (37) head their own Xmaxand take a VP complement headed by the participle, as schematicallyshown in (38):

(38) AuxP

Aux VP

Polish: widzial Bulgarian: sme pro6eli Czech: jsem koupil Slovak: som napisal

In these languages, the auxiliariestraditionally labelled 'clitics' cannot stand in initial position and must be supportedby any precedingconstituent, regardlessof its category. As shown by the deviance of the examples in (39), the Polish perfect auxiliaryis like the perfect auxiliaryof the other Slavic languagesin this respect: CLITIC AUXILIARIES AND INCORPORATION IN POLISH 393

(39) a. *9 widzial tq ksi#ikeq. (Polish) 2SG seen this book

b. *Sme proceli knigata. (Bulgarian) Pres + 1PL read book + the

c. *Jsem koupil knihy. (Czech) Pres + 1SG bought books d. *Som napisallist. (Slovak) Pres + 1SG writtenletter In view of the parallelismbetween Polish and other Slavic languages, we propose that a similarPF-process of cliticizationfuses the auxiliaryenclitic and the precedingconstituent in all these languages, licensing the auxiliary. When no constituentprecedes the auxiliary,as in (39), the structure is ungrammaticalbecause the auxiliaryis not licensed.5 The Polish conditionalin (40a) shows auxiliary+ participleorder similar to exampleswith the perfect auxiliary.However, the conditionalauxiliary is not a PF enclitic. That is, the conditional auxiliarydoes not undergo the PF-processwhich fuses perfect auxiliaryand the precedingconstituent, and as a result it may stand in absolute initial position, as shown in (40b): (40)a. Gdzie bys chcial i'd? where COND + 2SG liked go Where would you like to go?

5 The type of cliticizationwe proposecannot be a word-formationoperation in models such as LexicalMorphology, where word-formation applies in the lexiconbefore items are inserted in the syntactictree. This is because the process attachingthe auxiliaryto the pronounin (36) operates across syntacticboundaries. Thus, in our analysispersonal pronouns such as ty 'you' are not inflectedfor person and tense in the lexicon and insertedin the syntactic tree with such markers;rather, clitic auxiliariesbearing person and tense attachto personal pronounsin PF. See Booij and Rubach(1987) for the lexical approachwe reject and Dogil (1987) and Aguado and Dogil (1989) for criticism.Dogil (1987) treatsthe auxiliaryas a clitic in Agreementwhich is copied on its differenthosts. Our analysisdoes not requireany such rule. Toman (1981b: 306) reports that in colloquialCzech, the 2SG auxiliarymay be op- tionallyreduced, as in (i):

(i) Ty jsi pfisel. -* Tyspfisel. (Czech) you Pres + 2SG come You have come. Withinour approach,the contractionin (i) is the result of the PF processof encliticization that Czech shareswith Polish. 394 ROBERT D. BORSLEY AND MARfA LUISA RIVERO

(40)b. Bys chcial isc. COND + 2SG liked go You would like to go.

A second characteristicthat the Polish perfect auxiliaryshares with the perfect auxiliariesof the other Slavic languagesis that it does not license VP-Preposing.Consider the contrastin grammaticalitybetween (41a) and (41b), noted by Kipka (1989): (41)a. Lustro- smy kupili. Mirror-1PL bought We bought a mirror.

b. *Kupili lustro- smy. Bought mirror-1PL

The ungrammaticalityof (41b) seems unexpected because Polish word order is flexible, and clitic auxiliariesneed not be in second position in the clause. This example has a fronted VP similar to the English VP of Bought a mirror, I have, and shows that the Polish perfect auxiliary disallows VP-Preposing. The same restriction is shared by the perfect auxiliaryin all the Slaviclanguages we have mentioned(Lema and Rivero 1989, Rivero 1991). From this perspective Bulgarian (42a) and Serbo- Croatian(42b) are parallelto Polish (41b):

(42)a. *Procel knigata sum. (Bulgarian) read book + the Pres + 1SG

b. *6itao knijguje. (Serbo-Croatian) read book Pres + 3SG

Example (41a) is unproblematicin a 'free word order' language such as Polish. It has an initialNP-object, and this constituentprovides the support requiredin PF by the clitic auxiliary. Polish VP-preposingis, however, licensed by the future auxiliary, as shown in (43), which should be contrastedwith (41b) (note that the VP may be headed by a participleor an infinitive):

(43)a. Kupowalilustro bTd. bought mirrorwill + 3PL Buy a mirror,they will. CLITIC AUXILIARIES AND INCORPORATION IN POLISH 395

(43)b. Czytac ksi4zkq bd. read + INF book will + 3PL Read a book, they will.

(43) has a counterpartin Czech. In Czech, VP-Preposingis impossible with the perfect auxiliarybut possible with the future auxiliary(Rivero 1991): [Kupovatknihy] budu 'Buy books, I will.' vs. *[Koupilknihy] jsem '*Boughtbooks, I have.'.

3.2. The Position of the Auxiliary In the following discussion, we attempt to determine the position of the Polish auxiliary in auxiliary + participle sequences. We argue that in structureswithout Incorporation,the Polish perfect auxiliaryis a clitic under I and does not float to another position in syntax, except in one well-definedcase discussed at the end of this section. We will show that Polish is not a languagewith Wackernageleffects for clitic auxiliariesand in this specific sense differs from Czech, Slovak, and Serbo-Croatianbut resemblesBulgarian. Identifyingthe position of the auxiliaryin some structuresthat do not involve Incorporationis straightforward.For example, sentence (36), (Ty-s widzial tq ksiqzkq. 'You've seen this book.'), is like an English sentence containinga contractedauxiliary: the auxiliaryis transparently in I?. The analysisis also quite straightforwardwhere we have null subject sentences in which a variety of constituentsprecede the auxiliary:

(44) Kiedy-s widzial krolika? when- 2SG seen rabbit When did you see the rabbit?

(45) Jego-s widzial. him- 2SG seen You saw him.

(46) Jerzego-s lubila. Jerzy- 2SG liked You liked Jerzy. 396 ROBERT D. BORSLEY AND MARfA LUISA RIVERO

(47) Daleko- m poszla. far- 1SG gone I went far.

This order is the result of two processes which have well-knowncounter- parts in many languages:movement to the CP-specifierposition in cases like (44), i.e., Wh-movement,and movement to a position adjoinedto IP by Scramblingin the following examples. In both cases, cliticizationof the auxiliary to the preceding item applies across the empty subject, parallel to English examples such as Who do you think's been here? (Schachter1984). There are other examples that are less straightforward,although there is no reason to thinkthat they involve movementof the auxiliary.Consider for example (2c) repeated as (48), which has an the NP-object between subject and auxiliary:

(48) Ty jego- s widzial. you him 2SG seen You saw him.

This shows that Polish clitic auxiliariesneed not be in the Wackernagel position (second in the clause). In (48), in syntacticterms the auxiliaryis in the thirdposition aftersubject and object constituentsand, in phonological terms, after two different accented words. We suggest that the position of the object NP in (48) is the result of Scrambling.As shown in (49), the object jego has adjoined to IP, and the subject ty has adjoined to IP, or Spec-of-CP:

(4) Tyj lIP jegoi [IP tj [Io S] [VP ti widzial tif]]

If this is correct, there is no reason to assume that the auxiliaryis in a position other than I? in (49). Where the auxiliaryappears to be inside a phrase, a non-movement analysisis not so obvious. (3b), repeated now as (50), and (51)-(55) are typical examples of this situation:

(50) Ewy- s ksi#2kqczytal. Ewa's- 25G book read You read Ewa's book. CLITIC AUXILIARIES AND INCORPORATION IN POLISH 397

(51) gwiadomy-s tych problemowbyl. aware- 2SG theseproblems been You were aware of these problems.

(52) Ktorego- s mqzczyznewidzial? which- 2SG man seen Which man did you see?.

(53) Ladne- s kwiaty kupil. nice- 2SG flowers bought You bought nice flowers.

(54) Bardzo-s ladnie wygl#dalwczoraj. very- 2SG nice looked yesterday You looked very nice yesterday.

(55) Ksizkq- s o Polsce napisal. book- 2SG about Poland written You wrote a book about Poland. In (50) we appear to have a pre-nominalmodifier as specifierseparated from its head by an auxiliary,while in (51) we seem to have an auxiliary separating a head from its complement. If Polish determinerssuch as kt6rego 'which'are heads of DP's in the sense of Abney (1987) and take an NP complement,then (52) is parallelto (51). If determinersare speci- fiers of NP, (52) will be analogousto (50). In (53) and (54), the auxiliary separatesa premodifierfrom the associatedhead, as in (50). In (55), the auxiliaryseparates a head from its complement, as in (51). It may be temptingto think that examples (50)-(55) involve movement of clitic auxiliaries.However, there are parallelcases in which the present tense verb appears in the position of the auxiliary.The following corre- spond to (50)-(55):

(56) Ewy czytasz ksi?zkq. Ewa's read + Pres + 2SG book You are readingEwa's book.

(57) Swiadomyjestes tych problem6w. aware be + Pres + 2SG theseproblems You are aware of these problems. 398 ROBERT D. BORSLEY AND MARf A LUISA RIVERO

(58) Ktorego widzisz mqzczyznq? which see + Pres + 2SG man Which man are you seeing?

(59) Ladne kupujesz kwiaty. nice buy + Pres + 2SG flowers You are buying nice flowers.

(60) Bardzo wygl#dasz ladnie dzisiaj. much look + Pres + 2SG nice today You are looking very nice today.

(61) Ksi4ikq piszesz o Polsce. book write+ Pres + 2SG about Poland You are writinga book about Poland. There are also parallel cases involving Incorporationof the participleto the perfect auxiliary,as in (62)-(67): (62) Ewy czytal-es ksi#zkq. Ewa's read- 2SG book

(63) gwiadomybyl- es tych problemow. aware been- 2SG theseproblems

(64) Ktorego widzial-es mqzczyzne? which seen- 2SG man

(65) Ladne kupil- es kwiaty. nice bought-2SGflowers

(66) Barzdo wygl#dal-es ladnie wczoraj. much looked- 2SG nice yesterday

(67) Ksi#2kqnapisal-es o Polsce. book written-2SGabout Poland The above parallelismsfollow from our assumptionthat a clitic auxiliary is either in I alone, resultingin (50)-(55), or forms an X?-complexwith the participlewhich optionally incorporatesto it, resulting in (62)-(67). In addition, the verb in simple or one-word tenses such as the present moves to I, resultingin (56)-(61). Thus it follows that inflected auxiliar- CLITIC AUXILIARIES AND INCORPORATION IN POLISH 399

ies, inflectedparticiple + auxiliarycomplexes, and inflectedordinary verbs have the same distribution. It is clear, then, that rather than being a peculiarity of clitics, the generalizationuniting (50)-(67) involves the syntacticprocesses allowing certaintypes of constituentsto precede the lo-position,namely Wh-movement and Scrambling.

3.3. Scrambling Before we consider how Scramblingaccounts for the propertiesof sentences with auxiliaries and no Incorporation,we need to flesh out its properties.We proposethat in Polish, Scramblingis adjunctionto an Xmax such as AuxP or VP and subsumes topicalization, which we regard as adjunctionto IP. The process resembles Wh-movementin that it licenses parasiticgaps. Considerhere the following relevant data: (68)a. Jan skrytykowalten film [przed obejrzeniem*(go)] Jan criticized thisfilm before seeing it Jan criticizedthis film before seeing it.

b. Ktory film Jan skrytykowal[przed obejrzeniem] Whichfilm Jan criticized before seeing Which film did Jan criticizebefore seeing?

c. Ten film Jan skrytykowal[przed obejrzeniem] Thisfilm Jan criticized before seeing This film Jan criticizedbefore seeing.

d. Jan ten film skrytykowal[przed obejrzeniem] Jan thisfilm criticized before seeing This film Jan criticizedbefore seeing. Example (68a) shows that the verbal noun obejrzeniem'seeing' requires an overt object undernormal circumstances. (68b-d) show that Wh-movement, Scramblingto IP, and Scramblingto VP respectivelylicense para- sitic gaps. It follows therefore that Scramblingis movement to an A-bar position. . Another piece of evidence that Polish Scramblingmust be movement to an A-bar position is that it neither decreases nor increases anaphora possibilities, unlike movements to A-positions, which do (see the dis- 400 ROBERT D. BORSLEY AND MARIA LUISA RIVERO cussion in D6prez 1989, Mahajan1990). In this connection consider the following data:

(69)a. Jan spotkal swojego brata. Jan met REFL brother Johni met hisi own brother. b. Swojego brata Jan spotkal. REFL brotherJan met

(70)a. *Sw6j brat spotka Jana. REFL brothermet Jan *Hisi own brothermet Johni. b. *Janasw6j brat spotkal. Jan REFL brothermet

(69b) shows that topicalizing,or in our terms Scrambling,the object NP to IP does not decrease binding possibilities for anaphors. (70b) shows that Scramblingdoes not increasebinding possibilities: the object precedes and c-commandsthe subject, but binding of the anaphorremains impossible. Similarly,the following data show that Scramblingto VP does not affect bindingpossibilities:

(71)a. Jan kazal swojemubratu spotkac Jerzego. Jan told REFL brothermeet Jerzy Jani told hisi own brother to meet Jerzy. b. Jan Jerzego kazal swojemubratu spotkac. Jan, told hisi own brother to meet Jerzy.

(72)a. Jan kaza Marii umydsw6j samoch6d. Jan told Mary wash REFL car Jani told Maryjto wash hisi/herjcar. b. Jan sw6j samoch6d kazal Marii umyd. Jani told Maryjto wash hisi/herjcar. In (71b) and (72b), object NPs have been scrambled, but the binding possibilitiesfor the possessive anaphorremain unchanged. We feel that it is safe to conclude then that Scrambling in Polish adjoins a phrase to a CLITIC AUXILIARIES AND INCORPORATION IN POLISH 401

Xmaxin the clause and creates a chain with the characteristicsof Wh-chains (and see Witkog1993 for additionalparallelisms and discussion).

3.4. Scrambling and the Position of Auxiliaries We returnnow to the position that auxiliariesoccupy in sentences without Incorporation.First, as example (50) (repeated below as (73)) shows, a specifiermay be separatedfrom the associatedhead ((56) and (62) illustrate the same phenomenon).6Not obeying the Left Branch Conditionis a common Slavicfeature (for discussionof Polish see Borsley (1981, 1984) and Corver(1992), which appearedafter this paper was written):

(73) Ewy- s ksi#kq czytal. Ewa's- 2SG book read You read Ewa's book. Given the assumptionthat Polish is underlyinglyVO, (73) and similar

6 Exampleswhere the clitic auxiliarystands between the constituentsof a quantificational Wh-expressionmay belong to the type with determinerseparated from the associatedhead. Considerthe followingfrom (Dogil 1987): (i) ?Jako-m -s mu pomogL. how-lSG Indefhim helped I helped him somehow. Here the auxiliarystands between the Wh-wordand its indefinitemarker. According to our consultants,this example is marginal,and the pattern is restrictedto certain grammatical persons. For example, (ii), a pluralcounterpart of (i), is ungrammatical: (ii) *Jako-9my-s mu pomogli. how- 1PL-Indefhim helped Similarcomments apply to patternswith the clitic-auxiliarystanding between a Wh-phrase and its genericmarker. Thus, while both (iii. a) and (iii. b) seem fine, only the (a) example in (iv) is grammatical: (iii)a. Co- kolwiekzrobil-es. what-ever done- 2SG Whateveryou (SG) did b. Co-g-kolwiekzrobit. (iv)a. Co- kolwiekzrobili-scie. what-ever done- 2PL Whateveryou (PL) did b. *Co-scie-kolwiekzrobili. Given such restrictions,it is perhapsthe case that examples(i) and (iii.b) are the product of more idiosyncraticprocesses than the patternsdiscussed in the text. 402 ROBERT D. BORSLEY AND MARIA LUISA RIVERO cases involve adjunctionof a verbal complementto VP and adjunctionof the specifierof this complementto IP, as in (74):

(74)a. IP b. IP

I0 VP 10 VP

9 czytal NP NPi VP

NP N' NP N'

Ewy Ewy czytal ksi#?kq ksi#zkq

a. BASE b. SCRAMBLINGTO VP

c. IP

Ewy IO VP

NPi VP

tk N' czytal t

ksi#zkq

c. SCRAMBLINGTO IP We conclude that the second process is adjunctionto IP, rather than to the CP-Specifierbecause this same word order is found in subordinate clauses (Witkos 1993), as the following illustrates: CLITIC AUXILIARIES AND INCORPORATION IN POLISH 403

(75) Wie, ze Ewy- s ksi4tkv czytal. He + knows that Ewa's-2SG book read He knows that you read Ewa's book.

Examples such as (75), in which in embedded clauses the auxiliaryneed not immediatelyfollow the complementizer,provide further evidence that Polish clitic auxiliariesare not in the Wackernagelposition. Here we find a contrast with Czech, Serbo-Croatian,and Slovak. Serbo-Croatianis a typical Wackernagellanguage in this respect:

(76)a. Jovan tvrdi da je Marijadosla. (Serbo-Croatian) John claims that has Mary come John claims that Mary came.

b. *Jovantvrdi da Marijaje dosla.

The idea that Scramblingis the two-step process illustrated in (74) is supported by the fact that Polish allows Scramblingjust to VP. So correspondingto (50), (53), and (54) we have the following:

(77) Ty- s Ewy ksi#k& czytal. you-2SG Ewa's book read

(78) Ty- s ladne kwiaty kupil. you-2SG nice flowers bought

(79) Ty- s bardzo ladnie wygl4daIwczoraj. you-2SG much nice looked yesterday

Analogously, Scramblingto IP alone is also attested. We find this in sentences in which a present tense verb separates a specifier from the related head, as in (56), and in those in which a participleand auxiliary combinationseparates the two, as in (62). Let us now consider examples in which the clitic auxiliaryseparates a head from its complement, as in (51), Swiadomys tych problemow byl 'You were aware of these problems', and parallel cases. Such sentences are not instancesof Long Head Movement, which, as we have noted, is not possible in Polish. Rather than being an X? that has undergoneLHM to Co, swiadomyis the head of a phrase that has moved as a Xmax. This becomes clear in examples where this head and its premodifierprecede the auxiliary,as shown by (80). Since both the modifierand the auxiliary, that is bardzoswiadomy, have been fronted, we conclude that the process 404 ROBERT D. BORSLEY AND MARfA LUISA RIVERO in (80) is phrasalmovement and not X0-movement,which affects the head of the phrase in isolation: (80) Bardzo gwiadomy-s tych problem6wbyl. much aware- 2SG theseproblems been Examples such as (51) involve the scramblingof a complementout of the verbal complementto a pre-verbalposition and the scramblingof the remainderof the verbal complementto a pre-auxiliaryposition. We have adjunctionto the AuxP headed by byl followed by adjunctionto IP, as in (81).7 Notice that byl is the head of the complement of the perfect auxiliary,explaining how it is able to undergoIncorporation to the perfect auxiliary,as illustratedby byl-es in (63):

7 The analysisin (81) is partlybased on treatmentsproposed for Germanexamples such as (i) (Webelhuth1985, 1992, Koster 1987, den Besten and Webelhuth1990): (i) Gelesen hat Hans das Buch nicht. (German) read had Hans the book not John did not read the book. For the sentence in (i), which den Besten and Webelhuthconsider a case of 'remnant topicalization',the hypothesisis that the VP containinggelesen fronts, while das Buch as complement of V remains behind because it has been scrambledout of the VP. Van Riemsdijk(1989) proposesthat in (i) a non-maximalverbal projection is fronted,which the Barriersframework (Chomsky 1986) does not contemplateas a possibility. 8 The derivationin (81) raises interestingproblems which are beyond the scope of this paper, and severalof them are mentionedin the cited referencesfor the parallelDutch and German Scramblings;however, as far as we can tell, none of these problems would be particularto Polish, given our proposals. R. Huybregtshas pointed out to us that the preposedphrase in cases like (81c) has an unboundtrace, a situationwhich will also arisein ordinaryVP-Preposing in manylanguages, if subjectsare consideredVP-internal, and with preposedPassive VPs, howeversubjects are analyzed (They said he was believedto be mad and believedto be mad he was). For the parallelDutch and Germanscrambled patterns, den Besten and Webelhuth(1990) propose a reconstructiondevice. D. Adger has pointed out to us that in (81c) the phrase scrambledto AuxP intervenes between the phrase scrambledto IP and its trace, in apparentviolation of Relativized Minimality(Rizzi 1989)if the two phrasesare of the A-bartype, as we assume.One solution wouldbe to considerthat some scramblingshave mixedproperties - withinour terminology, that they are A and A-bar at the same time (see Webelhuth(1992) and Ddprez (1989) for differentapproaches to this issue). One way to implementthis idea wouldbe to assumethat the movement in (81b) is to the Spec-of-AuxPinstead and counts as A-movement for RelativizedMinimality but has A-bar propertiesfor BindingTheory. On the basis of some WeakCross Over phenonema, Witkos (1993: Chapter 7) also concludesthat other scrambling patternsshow mixed properties. CLITIC AUXILIARIES AND INCORPORATION IN POLISH 405

(81)a. IP b. IP

Io AuxP 10 AuxP

9 byl AP s NPi

gwiadomytych problemow tychproblem6w AuxP

byl AP

Owiadomyt,

a. BASE b. SCRAMBLINGTO AUXP

C. IP

APk IP

swiadomyti AuxP

\ NPi

tychproblemow AuxP

\ ~~byl AP

tk c. SCRAMBLINGTO IP

The second fronting in (81) is adjunctionto IP, not movement to Spec- CP, because the same word order is found in subordinateclauses, as in (82): 406 ROBERT D. BORSLEY AND MARfA LUISA RIVERO

(82) Wie, ze swiadomy-s tych problemowbyl. He + knows that aware- 2SG theseproblems been He knows that you were aware of these problems.

As one would expect, examples involvingjust the first of these movement processes, Scramblingto AuxP in (81b), are possible, as in (83)- (84), thereby supportingthe idea that the derivationinvolves a two-step procedure:

(83) Ty- s tych problemowbyl gwiadomy. you-2SG theseproblems been aware

(84) Ty- ? o Polsce napisal ksi4kA. you-2SG about Poland written book

Thus, examples in which a head and its complement are separated by verbalforms, includingclitic auxiliaries,can be analyzedin essentiallythe same way: they involve the scramblingof a phraseto a position preceding the inflected auxiliaryor verb which is in 10. If sentences in which a clitic-auxiliaryseparates a head from its complement arise through movement of the complement, such patterns should occur only when it is in fact possible to move the complement. This seems to be the case, providingadditional motivation for our analysis.For example, an auxiliarycannot be attached to a transitivepreposition, as (85) illustrates:

(85) *Do-s' Poznaniapojechal. to- 2SG Poznan' gone You went to Poznan'.

This is an automaticconsequence of the fact that Polish does not allow prepositionstranding, as (86b) illustrates:

(86)a. 0 czym Jan rozmawia? about what Jan talks What is Jan talking about?

b. *CzymJan rozmawiao?

Thus we need no special stipulationto account for the ungrammaticality of examples like (85). This account also predicts correctlythat it should CLITIC AUXILIARIES AND INCORPORATION IN POLISH 407 be possibleto attacha clitic-auxiliaryto a frontedor scrambledintransitive preposition: (87) Dookola-smy przeszukali. around- 1PL searched We searched around. Finally, Polish has a restructuringprocess which combinesa preposition with the first constituent of its complement (see Borsley and Jaworska 1988, 1989). As a result of this and the lack of Left Branch Condition effects, we have examples like the following, where prepositionand Wh- word have been fronted together:

(88) Z ktorym Jan rozmawiamq?czyznq? with which Jan talks man Which man is John talkingto? As Henk van Riemsdijk has pointed out to us, it should therefore be possible for a clitic auxiliaryto apparentlybe inside the complementof a preposition.This is indeed the case, as the following illustrates:9

9 In the light of examples such as (i.b) it has been claimed that the position of clitics in Serbo-Croatianis not defined syntacticallybut phonologically.In this view, which differs from our syntactictreatment of Polish clitic auxiliaries,clitics appearafter the firstaccented word in the clause and may breakup a syntacticphrase (see the discussionin Spencer1991: 355ff.): (i)a. Taj piesnik mu je dao autogram Thispoet him has given autogram This poet has given him an autogram. b. Taj mu je piesnikdao autogram. The sentences in (i) resemble the Polish exampleswhere the clitic auxiliaryappears to be inside a phrase, which we have proposedinvolve the frontingof xm, ratherthan a clitic placement process sensitive to phonology. For Serbo-Croatian,(:avar and Wilder (1992) have independentlycome to a conclusionthat resemblesour proposalfor Polish. They point out that Serbo-Croatianis not sensitiveto the Left-Branchcondition, so (i.b) can be analyzed as involvingthe frontingof taj. In their view, the clitic cluster can interrupta phraseonly when the portionof the phrasewhich precedes can be extractedindependently. For instance, PPs can be fronted,but Ps in isolationcannot, so a clitic cannotinterrupt a PP in cases such as (ii), pointed to us by LjiljanaProgovac: (ii)a. Petarje krenuo[pp prema njemu]. Peter has moved towardshim b. [pp Premanjemu] je Petar krenuo. c. *Premaje njemuPetar krenuo. 408 ROBERT D. BORSLEY AND MARIA LUISA RIVERO

(89) Z tym-? mqzczyzn4rozmawial. with this-2SG man talked You talked with this man. We have argued in the foregoing that there is no need to assume that the clitic-auxiliaryis moved in examples in which it appears to be inside of a phrase. Rather, such sentences can be analyzed as the result of Wh-movementor Scrambling,with the auxiliaryremaining in I.

3.5. A Special Positionfor Auxiliaries There are, however, certainsentence types which do involve a movement restricted to auxiliariesof various types, including non-clitics. Consider first (90a-c), which have the inflected auxiliary after two fronted Wh-phrasesand before the main verb: (90)a. Co komu- s dal? whatto + whom-2SGgiven What did you give to whom?

b. Gdzie kto by poszedl? where who Cond + 3SG gone Who would go where?

In connection with the phonologicalapproach to clitic position, an anonymousreviewer calls attention to examples such as (iii), which our consultantD. Kudrafinds literarybut grammatical: (iii) U novom je hotelu ziveo. In new has hotel lived He has lived in a new hotel. This sentence, which resemblesPolish (89) in that the clitic auxiliaryfollows a preposition and an adjective, could be open to the syntactictreatment we propose for Polish if the prepositionrestructures with an adjectiveas head of X"'a. However, unlike the situation discussedfor the Polish auxiliary,je in (iii) occupies a position in which ordinaryinflected verbs cannot appear.The same reviewermentions that when an adverbmodifies the adjective, the auxiliarymay follow the adjective,as in (iv.a), but not the adverb,as in (iv.b): (iv)a. U vrlo novom je hotelu ziveo. In very new has hotel lived He has lived in a very new hotel b. *U vrlo je novom hotelu ziveo. This situationappears compatible with syntacticapproaches such as the one we advocate, but is problematicfor phonologicalapproaches. CLITIC AUXILIARIES AND INCORPORATION IN POLISH 409

c. Co komu bldziesz rozdawal? what to+ whom will+2SG given + away What will you give away to whom?

In cases like (90), it can be maintained that the auxiliaryis in Io if we accept Rudin's analysisfor multiple Wh-movementin Polish. Under this analysis,the first Wh-phrasefills the CP-specifierposition, and other Wh- phrases are adjoinedto IP and precede the I position. (90a-c) have variants(91a-c), in which the two Wh-elementsare separated by the auxiliary,suggesting that auxiliariescan also be adjoined to IP:

(91)a. Co- s komu dal? what-2SGto + whom given

b. Gdzie by kto poszedl? (Toman1981a) where Cond + 3SG who gone? c. Co bedziesz komu rozdawal? what will+2SG to+whom given+away

Motivationfor this auxiliaryadjunction comes from the fact that when we have three Wh-elements,the auxiliarycan intervenebetween the first and the second, or the second and the third, as in (92):

(92)a Co by komu kiedy Jan dal? what Cond+3SG to+whom when Jan given What would Jan give to whom when?

b. Co komu by kiedy Jan dal? whatto+whom Cond+3SG when Jan given

The sentences in (91) and (92) lead to two independentconclusions, the first of which has been already established on other grounds. First, the orderingof Polish auxiliariesabove is not a second position phenonemon, so the auxiliaries are not in a special Wackernagel slot. Second, the auxiliary raising in (92) cannot be movement to C, since the second Wh-phrasein (92b) is adjoinedto IP. Thus Polish contrastsnot only with Wackernagellanguages but with Germanic V2 languages, in which the auxiliaryfills Co in cases like Whathave you given to Mary? The combinationof participleand auxiliarywhich results from Incorpor- 410 ROBERT D. BORSLEY AND MARfA LUISA RIVERO ation cannot separate a pair of Wh-elements.Thus, in contrastwith (90a) and (91a), the following is impossible, except as an echo-question: (93) *Co dal- es komu? whatgiven-2SG to+ whom Hence, it must be concludedthat there is a process applyingto the various finite auxiliariesin Io to optionally adjoin them to IP but that finite Vs and finite Incorporation-complexescontaining V cannot be raised further than I. This process could be adjunctionof an X? to IP, which has been postulated by Kayne (1991) for Romance infinitives, and is compatible with the tenets of the Barriersframework.10 As far as we can see, however, this does not challenge our central conclusion, namely that the clitic- auxiliaryis in I0 in examples (48) through(55).

3.6. Summary We have argued that auxiliarystructures without Incorporationin Polish share the general characteristicsof the auxiliary+ participlestructures of West and South Slavic. We have also arguedthat Polish is a configurational language, ratherthan one with flat structures.From this point of view, its freedom in word order is due to the effect of syntacticrules such as Wh- movementand Scrambling,which are sensitive to structuraland hierarch- ical relationsin the phrase-marker,rather than PF arrangementsof phonologically defined groupings. We have also argued that Polish auxiliaries are generally in I, the exception being where they move to a higher position in the clause. This position is neitherthe Wackernagelslot associated with Czech, Slovak, or Serbo-Croatian,nor the C0 associated with

10 The contrastin movementproperties between Polish auxiliariesand verbs partiallyparal- lels English. Pollock (1989) arguesthat Englishauxiliaries such as have can raise to T/Agr while verbs cannotbecause a Theta-criterionviolation ensues in the second case, as a result of the propertiesof EnglishAgr. When V raisesto Agr, it formsa chain whichprevents the transmissionof the Theta roles this item assigns to its arguments;since auxiliariesdo not assign Theta-roles,their movement is unproblematic.In contrast, in French, transmission of Theta-rolesis possibleafter V-raising, so both verbs and auxiliariescan raise. If we accept this analysis,it could be suggestedthat Polish shares some propertiesof both French and English. On the one hand, Polish seems parallelto Frenchin that auxiliariesand verbs can appearunder the node we have labelled P; thus, raisingV to I is not detrimentalto Theta- role assignmentin Polish. On the other hand, Polish resemblesEnglish in that movement to a positionhigher than 1 is possibleonly for auxiliariesbut not for verbs. Thus, it can be suggestedthat in Polish this second raisingprocess creates a movementchain which prevents the transmissionof Theta-rolesto the argumentsof V and that it is viableonly with auxiliaries since those have no Theta-rolesto assign. CLITIC AUXILIARIES AND INCORPORATION IN POLISH 411

GermanicV2 phenomena;in this respect, Polish contrastswith these two types of languages. This process of raising from I applies to clitic and non-clitic auxiliaries alike, but does not apply to participle+ auxiliary combinations.

4. NEGATION In this section we examine the distributionof negativeparticles in auxiliary structureswith and without Incorporation.The determinationof the position of negation in Polish is importantfrom two points of view. The first is contrastive. It has been suggested by Rivero (1991) that there are two types of Slavic languages with respect to the position of Neg: (a) those such as Bulgarianand Serbo-Croatianwhere Neg takes T(ense)P(hrase) as complement; (b) those such as Czech and Slovak where Neg is the complementof T. We arguebelow that Polish belongs to the second type. Second, Incorporationis unique to Polish within Slavic and, as we will see, is at the root of what makes the distribution of negation in this languagedifferent from that of other Slavic languages. We can begin by illustratingthat the negativeparticle nie cannotprecede the perfect or conditionalauxiliary. As shown in (94), nie can precede the participle+ auxiliarysequence but not the auxiliaryalone:

(94)a. Nie widzial-em Marii. not seen- 1SG Mary I did not see Mary. b. *Nie-m widzial Marii.

c. Nie widzialbymMarii. I would not see Mary. d. *Nie bym widzial Marii.

Thus, Incorporationof the participleto the auxiliarymay appear obliga- tory in this context, while in other environmentsbasic and incorporated variantsalternate. We will show that the contrastsin (94) do not reflect restrictionsimposed on the rule of Incorporationbut can be attributedto the structuralposition of NegP in the clause, which groups Polish with Czech and Slovak. 412 ROBERT D. BORSLEY AND MARfA LUISA RIVERO

4.1. The Position of Negation As we have noted above, there are two types of Slavic languages as far as the position of Neg is concerned. The first group includes languages such as Bulgarianand Serbo-Croatian,in which negation must precede the perfect auxiliary:

(95) (Az) ne sufmprocel knigata. (Bulgarian) (1) not have read book+the I have not read the book.

For this first group Rivero argues that Neg heads a NegP complementof CP and takes the highest functionalprojection within IP as complement. That is, Neg c-commandsTense and Agreement(s), whether these categories head different projectionsor are conflatedunder Io as in our discussion, and the perfect auxiliary. The second group includes Czech and Slovak, with negation following the perfect auxiliary:

(96)a. Tehdy jsem nekoupil knihy. (Czech) then have not + bought books I did not buy books then.

b. Ja som nenapisal list. (Slovak) I have not+ writtenletter I did not write a letter.

For this type, Rivero argues that Neg heads a NegP and is c-commanded by the perfect auxiliaryinstead of c-commandingit. More precisely,AuxP takes NegP as complement. Adopting this analysis, let us consider negation in Polish, beginning with the perfect auxiliary.Examples such as (97) where Incorporationhas not applied indicate that Polish belongs to the second type of Slavic language:

(97) My-smy nie czytali tej ksi#zki. we- 1PL not read this book We did not read this book.

In Polish, then, Neg is c-commandedby 10 and therefore (97) has the structurein (98): CLITIC AUXILIARIES AND INCORPORATION IN POLISH 413

(98) IP

NP I'

My 10 NegP

smy Neg0 VP

nie czytalitej ksi#?ki

As for the conditionalauxiliary, (99a) shows that it is parallelto the Czech conditionalin (99b) and must be located in the same structuralposition as smy in (98), with Nego lower than I in the basic representation:

(99)a. My bysmy nie czytali te ksi#2k9. we Cond + 1PL not read this book We would not read this book.

b. Pan by nekoupil. (Czech) man Cond + 3SG not + bought A man would not buy. In brief, in Polish the AuxP headed by the perfect or the conditional auxiliarytakes NegP as complement, as in Czech and Slovak.

4.2. Negation and Incorporation We can now return to (94a), Nie widzial-emMarii 'I did not see Mary', and negative sentences in which Incorporationhas applied. Given that Neg in Polish heads the complement of AuxP (Aux - Neg - V), the order in (94b), *Nie-mwidzial Marii, could arise only if nie fronted to a position preceding the auxiliary -m if Incorporationhas not applied. If such a processis somehow disallowed,(94b) will be excluded, as required. As for (94a), it will be derived if Neg raises with the participle when Incorporationapplies. Clearly,then, we must considerwhy it is impossible for Neg to front in isolation in cases like (94b) and how it is possible for Neg to raise together with the participlein cases like (94a). Similarcom- ments apply to the conditional sentences in (94c) and (94d): nie cannot 414 ROBERT D. BORSLEY AND MARfA LUISA RIVERO

precede bym unless Incorporationof the participle to the auxiliaryhas applied. We suggest that these featuresof Polish are a consequenceof a process adjoiningNeg to the following V, similar to that discussed in Section 3, combininga prepositionwith the first item of its complement. Given this process, Neg and V will become a complexunit. As a result, they will raise togetherif Incorporationapplies; Neg cannot raise alone. Why shouldNeg and P behave in the same way? We suggest that both are functional categories forming an extended projection with their complements (van Riemsdijk1990, Grimshaw1990). It seems, then, that certainPolish functional categories undergo a restructuringprocess, which combines them with the first item of their complement. Since lexical categories will not undergo this process, the scramblingdiscussed in the last section will not be affected. An alternativeapproach might be to combine the participlewith Neg through Incorporation,raising the participle to nie. As Riny Huybregts has pointed out to us, on this approach, (94a) and (94c) would involve Incorporationof a complexNeg into the auxiliary,necessitating the raising of Neg but only if it forms a complex with the participle,excluding (94b) and (94d). This last stipulationseems undesirable. In brief, we propose that the negation and the participle incorporate as a complex to the auxiliary,which gives the Neg + participle+ auxiliaryorder of (94a) and (94c).

4.3. The Effect of Negation in LHM Let us now turn to a majorcontrast between Polish on the one hand and Czech and Slovak on the other, which results from the effect of Long Head Movement in the latter two languages. Recall that we assume that the Czech or Slovak participlewhich precedes the auxiliaryin sequences like Nap'sal som 'I wrote' has undergone Long Head Movementto C0 ratherthan Incorporationto the finite auxiliary in I?, as is the case in Polish. As a result of this difference, a Slovak negatedsequence such as Nenapi'salsom 'I did not write'is linearlyidentical morpheme-per-morphemeto Polish Nie widzial-em'I did not see', but the similaritybreaks down at the structurallevel: the Slovak Neg + participle complexnenap'sal occupies Co, while the complete Polishsequence including the auxiliaryis a unique X? in 10. As a consequenceof their different syntacticstructures, Slovak Neg + participle+ auxiliarysequences are restricted to root clauses, while analogous Polish sequences appear in all types of clauses including subordinatestructures, since they fill I and CLITIC AUXILIARIES AND INCORPORATION IN POLISH 415

leave Co unaffected. This explains both the similarityand the contrast between the sentences in (100) and (101). In Slovak (lOOa)and Polish (lOla), the negated main participleis below Io, so they are both grammatical. Slovak (lOOb)is ungrammaticalbecause if the negated participle raises, it cannot incorporateto the auxiliaryand has no landingsite since C0 is filled. Polish (lOlb) is grammaticalbecause the negated participle incorporatesto the auxiliaryin Io ratherthan moving to CO: (100)a. Spytal sa Zisom nenapisal list. (Slovak) asked Refl if Pres + 1SG not + written letter He asked if I did not write the letter. b. *Spytalsa ci nenapisalsom list. (101)a. Janekpowiedzial, ze- s nie pojechal do Warszawy. Janek said that-2SG not gone to Warsaw

b. Janekpowiedzial, ze nie pojechal- es do Warszawy. Janek said that not gone- 2SG to Warsaw

4.4. Negation and the two types of Slavic auxiliaries There is a furthersimilarity between Czech/Slovakand Polish with respect to the position of Neg in a differenttype of auxiliarystructure. In all three languages,the future auxiliaryfollows the negative particle, in contrastto the conditional and perfect auxiliariesin sentences like (96), (97), and (99): (102)a. Nie bqdq czytal ksi#ki. not will + 1SG read book I will not read the book.

b. Nebudu kupovat knihy. (Czech) not + will + 1SG but books I will not buy books. Such differences are presumablydue to the lexical propertiesof the two types of auxiliariesand the differentstructural positions in which they are generated. The Czech and Polish future auxiliaryselects VPs headed by imperfective Vs and thus imposes selectional constraintson its complement. This contrastswith the conditionaland perfect auxiliary,which can take any type of VP as complement. The selectional properties of the 416 ROBERT D. BORSLEY AND MARIA LUISA RIVERO future auxiliarysuggest that it has a more complex lexical structurethan the other auxiliaries.Perhaps the future counts as a lexical auxiliarygen- erated in the V-slot, a position c-commandedby both I? and Nego. If so, (102a) has the structurein (103), which should be contrasted with that proposed for past sentences in (98): (103) IP

10 NeP

Neg0 VP

nie V0 VP

bqde czytal tq ksiaz'kq Neg combines with the future auxiliaryand the resulting complex raises to I0, just like ordinaryV's with argumentstructure, giving the order in (102). As noted above, the perfect and conditionalauxiliaries lack selectional properties,so they are functionalrather than lexical categories. The syntactic consequence of this is that they are generated in a higher position than the future auxiliary, c-commandingNeg. We have not discussed recent proposals where I? is split into several functional heads (Pollock 1989), so we simplysuggest that the perfect and the conditionalauxiliaries occupy a functionalposition in the clause, and we label this position Io, as shown in (98). In other Slaviclanguages such as Serbo-Croatian,where the future auxiliarylacks selectional propertiesand is used with all types of verbs, this auxiliarybehaves syntacticallylike the perfect and the conditional auxiliariesin every respect and in our terms is a functionalcate- gory."

" For distinctionsbetween functionaland lexical auxiliariesin Slavicand Old Romancesee (Lema and Rivero 1989, 1991, 1992). Another contrast,pointed out in section 3, is that lexical auxiliariessuch as the future license VP-Preposingwhile functionalauxiliaries such as the conditionaldo not. CLITIC AUXILIARIES AND INCORPORATION IN POLISH 417

4.5. Summary In Polish, Neg heads a NegP complement of the functional category Io containingthe perfect and conditionalauxiliaries and takes a VP headed by lexical categoriessuch as the future auxiliaryor the verb as its complement. Neg is a bound morpheme which obligatorilycombines with the head of its complement, i.e., with the future auxiliaryor the main verb. In this sense, Polish is similar to Czech and Slovak, in which Neg is a bound morpheme in a position below Io in the syntacticphrase-marker. However, Polish differs from these languagesin that it has Incorporation of participlesto the perfect and conditionalauxiliaries. As a consequence of the basic position of Neg and the effect of Incorporation,two syntactic patternsalternate in negativesentences in Polish. Neg + participle+ auxiliary sequences result from Incorporation,of the Neg + participle-complex to Aux, and auxiliary+ Neg + participlesequences result when Incorpor- ation does not apply. Neg + auxiliary+ participle sequences are impossible.

5. INCORPORATION IS SYNTACTIC This section motivates the claim that Incorporationin Polish is syntactic ratherthan lexical, on the basis of the fact that it is 'bled'by the application of syntacticmovements. The behaviorof the conditionalauxiliary in sub- junctive complementsand counterfactualconditional clauses providesthe primaryevidence for our conclusion.

5.1. Conditional Contexts with and without Incorporation Conditionalsshow both incorporatedand basic forms in three syntactic environments:root clauses (104), relativeclauses (105), and complements of Vs of knowledge, traditionallylabelled 'Indicatives'(106): (104)a. Ja to zrobil-bym. b. Ja to bym zrobit. I would do it.

(105)a. Mqzczyzna,ktorego odwiedzil-bys, jest man who visited- Cond + 2SG is bratem Marii. brother Mary 418 ROBERT D. BORSLEY AND MARiA LUISA RIVERO

b. Mq?czyzna,kt6rego byg, odwiedziljest man who Cond + 2SG visited is bratem Marii. brotherMary The man you would visit is Mary'sbrother.

(106)a. Wie, ze lustro bysmy kupili. he + knows that mirrorCond + 1PL bought b. Wie, ze lustro kupili-bysmy. He knows that we would buy a mirror.

However, in three other syntacticcontexts, Incorporationof the participle fails to apply, and the auxiliaryforms an inseparableunit with the complementizer: Subjunctivecomplements of Vs of volition, as in (107), the protasis of the counterfactualconditional construction,as in (108), and certain purpose clauses, which we do not discussfor lack of space:

(107)a. Chcq, ze- byg widzialkrolika. I + want COMP-COND+ 2SG seen rabbit I want you to see the rabbit.

b. *Chc9, ze widzial-byskrolika.

(108)a. Gdy- bym mial czas, poszedl-bym COMP-COND+ 1SG had time, gone + COND + 1SG do kina. to cinema If I had the time I would go to the cinema.

b. *Gdy mial-bymczas, poszedl-bymdo kina.

(108a) manifestsboth syntacticcontexts. On the one hand, the counterfac- tual clause shows the necessarily unincorporatedparticiple, while the apodosis as root clause shows (optional) Incorporationof the participle to the conditional. In complement clauses such as (107a), the complementizer ze can be omitted, but Incorporationremains impossible. CLITIC AUXILIARIES AND INCORPORATION IN POLISH 419

5.2. Incorporation in syntax Assuminga model insertingalready inflected items in a syntactictree, the above patterns show that Incorporationof the participleto the auxiliary cannotbe a morphologicalor exclusivelylexical operation.This is because the environmentswhere Incorporationcannot operate must be defined syntacticallyrather than morphologically,and they involve the notion of clausaltype. For instance, Incorporationcannot apply in the complement clause selected by a verb of volition. The detailed investigationof these patternsis beyond the scope of this paper, but we suggest that the conditionalmoves to the complementizer in the syntax, in a way which precludesIncorporation of the participleto the auxiliary.12 For instance, it could be that the conditionalincorporates to Co, which functions as prefix, or that the auxiliarycliticizes onto C0 obligatorilyin syntax, so as to be adjacent,and that Incorporationof the participle to the auxiliarydestroys the necessary adjacency. Motivation for the idea that the complementizerand the auxiliarycombine is provided by the fact that in (107a) and (108a), they cannot be separated. NPs and other constituentscannot stand between the two, as illustratedin (109): (109)a. *Chcq, ze ty bys widzial krolika. I + want COMP you COND + 2SG seen rabbit

b. Chc9, ze- bys ty widzial krolika. I + want COMP- COND + 2SG you seen rabbit I want you to see the rabbit. By contrast, in indicativecomplements containing conditional auxiliaries, a complementizerand an auxiliaryare two separate constituents,so NPs and other constituentscan intervene between the two, as demonstrated by (106a). In this sense, Polish is parallel to Czech, where the conditionalis clitic- ized to clause-initiala in subjunctives(Comrie 1991): (110) Ucitel mi rekl, a- bych mluvil hlasite. (Czech) teacher me told, that-Cond + 2SG spoken loudly The teacher told me to speak loudly. If the approachwe have outlinedis correct,the absenceof Incorporation in subjunctiveand counterfactualclauses requiresa common explanation,

12 Dogil (1987) proposes to generate oneCOMS by in and another one in INFL. The first is selected by the main V, the second is independentfrom it. 420 ROBERT D. BORSLEY AND MARfA LUISA RIVERO so we offer one furthersuggestion to unify these two syntacticstructures and to answer the question why the conditional auxiliary raises to the complementizer in the two cases. First, consider the complements in (107a), which are clauses selected by a main V. In this case, raising of the conditional to C could be triggered if the selection requirementsof the verb can be satisfiedby conditionalby and not by the complementizer ze, because it is inert. Now considercounterfactual clauses which contain the complementizergdy, as in (108a). If this complementizeris also inert, clausal operator properties could be activated by raising the conditional auxiliaryto C. This could provide a common accountfor the two syntactic structureswhere the auxiliaryforms a complex with the complementizer. To conclude, we have identifiedsyntactic contexts which disallow Incor- poration in Polish and concludedthat the process is syntacticrather than lexical, given its sensitivityto the type of clause structure.

6. SUMMARY In this paper we have discussedtwo syntacticpatterns in Polish past and conditionalsentences. One type consistsof an inflectedauxiliary preceding a VP complement headed by a participle, with general characteristics found in other Slavic languagesand similarto English examples like rve seen a book. This analysis contrastswith earlier treatments, which view the auxiliaryas a unique-to-Polishmoving affix. Withinour approach,the auxiliaryhas the usual distributionof items which occupy the I? position, includingfinite non-auxiliaryverbs. It can be preceded by constituentsof many types because Polish is a configurationallanguage and has several types of syntacticfronting processes. The second type of past and conditionalsentence containsthe complement participleforming a complex X? with the inflected auxiliary,as the result of syntacticIncorporation. This second pattern is unique to Polish withinSlavic because other languagesin this familylack Incorporation,but it has close counterpartsin medievaland conservativeRomance languages, which form futures and conditionalsby a similarsyntactic process.

REFERENCES

Abney, Stephen:1987, TheEnglish Noun Phrasein its SententialAspects, unpublished Ph.D. dissertation,MIT. Aguado, Manueland GrzegorzDogil: 1989, 'Cliticsin LexicalPhonology: Alleged Countere- vidence', LinguistischeBerichte 120, 99-116. Andersen,Hennig: 1987, 'FromAuxiliary to Desinence', in M. Harrisand P. Ramat(eds.), HistoricalDevelopment of Auxiliaries,Mouton de Gruyter,Berlin, pp. 21-51. CLITIC AUXILIARIES AND INCORPORATION IN POLISH 421

Baker, MarkC.: 1988,Incorporation: A Theoryof GrammaticalFunction Changing, Univer- sity of ChicagoPress, Chicago. Besten, Hans den and Gert Webelhuth:1990, 'Stranding',in G. Grewendorfand W. Sterne- feld (eds.), Scramblingand Barriers,Benjamins, Amsterdam, pp. 77-92. Booij, Geertand Jerzy Rubach: 1987, 'Postcyclic versus Postlexical Rules in LexicalPhonolo- gy', LinguisticInquiry 18, 1-44. Borsley, Robert D.: 1981, 'Wh-Movementand UnboundedDeletion in Polish Equatives', Journalof Linguistics17, 271-288. Borsley, RobertD.: 1984, 'Free Relativesin Polish and English',in J. Fisiak (ed.), Contras- tive Linguistics:Prospects and Problems,Mouton de Gruyter,Berlin, pp. 1-18. Borsley, Robert D. and Ewa Jaworska:1988, 'A Note on Prepositionsand Case Marking in Polish', LinguisticInquiry 19, 685-691. Borsley, Robert D. and Ewa Jaworska:1989, 'On Polish PPs', Linguistics27, 245-256. C(iavar,Damir and ChrisWilder: 1992, 'Long Head Movement?Verb Movementand Cliticiz- ation in Croatian',Arbeitspapier Nr. 7, Institut ftir deutsche Spracheund LiteraturII, JohannWolfgang Goethe-Universitat, Frankfurt am Main. Chomsky,Noam: 1986, Barriers,MIT Press, Cambridge. Chomsky,Noam and HowardLasnik: 1977, 'Filtersand Control',Linguistic Inquiry 8, 425- 504. Comrie, Bernard:1991, 'Complementationin SlavonicLanguages: An Overview',in Com- plementStructures in the Languagesof Europe,N. Vincentand K. Borjars(eds.) Eurotyp WorkingPaper III, 2, EuropeanScience Foundation,pp. 3-16. Corver, Norbert: 1992, 'On DerivingLeft BranchExtraction Asymmetries: A Case Study in ParametricSyntax', NELS 22, 67-84. Dogil, Grzegorz:1987, 'LexicalPhonology and FloatingAffixation in Polish', Phonologica 1984. CambridgeUniversity Press, London, pp. 39-47. Deprez, Viviane: 1989, On the Typologyof SyntacticPositions and the Natureof Chains, unpublishedPh.D. dissertation,MIT. Grimshaw,Jane: 1990, 'ExtendedProjection', unpublished ms. BrandeisUniversity. Gussman,Edmund: 1980, Studiesin AbstractPhonology, MIT Press, Cambridge. Hale, Kenneth:1983, 'Warlpiriand the Grammarof Non-configurationalLanguages', NLLT 1, 5-48. Kayne, RichardS.: 1989,'Facets of Past ParticipleAgreement', in P. Beninca(ed.), Dialectal Variationand the Theoryof Grammar,Foris, Dordrecht,pp. 85-103. Kayne, RichardS.: 1991, 'RomanceClitics, Verb Movement,and PRO', LinguisticInquiry 22, 647-686. Kipka, Peter F.: 1989, 'Impersonalsand Inflection in Polish', MIT WorkingPapers in Linguistics10, 135-150. Koster, Jan: 1987, Domainsand Dynasties,Foris, Dordrecht. Lema, Jos6 and MariaLuisa Rivero: 1989, 'Long Head Movement:ECP vs. HMC', NELS 20, 333-347. Lema, Jose and Maria Luisa Rivero: 1991, 'Types of Verbal Movementin Old Spanish: Modals, Futures,and Perfects',Probus 3.3, 237-278. Lema, Jos6 and MariaLuisa Rivero: 1992, 'InvertedConjugations and V-secondEffects in Romance', in C. Laeufer and T. Morgan(eds.), TheoreticalAnalyses in Contemporary RomanceLinguistics: Selected Papers from theNineteenth Symposium on RomanceLinguis- tics (LSRL XIX), 21-23 April 1989, pp. 311-328, Benjamins,Amsterdam. Mahajan,Anoop: 1990, The A/A-bar Distinctionin MovementTheory, unpublished Ph.D. dissertation,MIT. Pollock, Jean Yves: 1989, 'Verb Movement,Universal Grammar, and the Structureof IP', LinguisticInquiry 20, 365-424. Riemsdijk,Henk van: 1983 'Movementand Regeneration',in P. BenincA(ed.) Dialectal Variationand the Theoryof Grammar,Foris, Dordrecht,pp. 105-136. 422 ROBERT D. BORSLEY AND MARIA LUISA RIVERO

Riemsdijk,Henk van: 1990, 'FunctionalPrepositions', in H. Pinksterand I. Genee, (eds.), Unityin Diversity.Papers Presented to Simon C. Dik on his 50th Birthday,Foris, Dord- recht. Rivero, MarfaLuisa: 1991, 'LongHead Movementand Negation:Serbo-Croatian vs. Slovak and Czech', LinguisticReview 8, 319-351. Rivero, Maria Luisa: 1994, 'Clause Structureand V-movementin the Languagesof the Balkans',NLLT 12, 63-120. Rizzi, Luigi: 1989, RelativizedMinimality, MIT Press, Cambridge. Rudin, Catherine:1988, 'On MultipleQuestions and MultipleWh-fronting', NLLT 6, 445- 501. Saito, Mamuroand Hajime Hoji: 1983, 'Weak Crossoverand Move Alpha in Japanese', NLLT 1, 245-259. Schachter,Paul: 1984, 'AuxiliaryReduction: An Argumentfor GPSG', LinguisticInquiry 15, 514-523. Spencer,Andrew: 1991, MorphologicalTheory: An Introductionto WordStructure in Gen- erativeGrammar, Basil Blackwell,Oxford. Sussex,Roland: 1980, 'On Agreement,Suffixation, and Enclisisin Polish', in C. V. Chavany (ed.), Morphosyntaxin Slavic, Slavica, Columbus,Ohio. Toman, Jindrich:1981a, 'Aspects of Multiple Wh-movementin Polish and Czech', in R. May and J. Koster (eds.), Levels of SyntacticRepresentation, Foris, Dordrecht,pp. 305- 310. Toman, Jindrich:1981b, 'Weak and Strong:Notes on be in Czech', in G. Brettschneider and C. Lehman(eds.), Wegezur universalienForschung. Sprachwissenschaftliche Beitrage zum 60 Geburtstagvon HansjakobSeiler,. Gunter Naar Verlag, Tubingem, pp. 293-302. Webelhuth,Gert: 1985, 'Germanis Configurational',Linguistic Review 4, 203-246. Weblhuth,Gert: 1992, Principlesand Parametersof SyntacticSaturation, Oxford University Press, New York. Witkog,Jacek: 1993, Some Aspects of PhrasalMovement in Polish, Ph.D. dissertation,Adam MickiewiczUniversity, Poznan.

Received 11 March1992 Revised 20 August 1993

Robert D. Borsley, School of Englishand Linguistics,Linguistics Section, UniversityCollege of North Wales, Bangor, GwyneddLL57 2DG. U.K. [email protected]

MariaLuisa Rivero, Departmentof Linguistics, Universityof Ottawa, Ottawa, OntarioKlN 6N5. Canada.