Verbal clitic movement in Polish--expanded AGR phrase in the minimalist program perspective


Material Information

Verbal clitic movement in Polish--expanded AGR phrase in the minimalist program perspective
Physical Description:
v, 114 leaves : ; 29 cm.
Miller, Miroslawa M., 1954-
Publication Date:


Subjects / Keywords:
Linguistics thesis, Ph. D   ( lcsh )
Dissertations, Academic -- Linguistics -- UF   ( lcsh )
bibliography   ( marcgt )
non-fiction   ( marcgt )


Thesis (Ph. D.)--University of Florida, 1995.
Includes bibliographical references (leaves 109-113).
Statement of Responsibility:
by Miroslawa M. Miller.
General Note:
General Note:

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 021970811
oclc - 33825065
System ID:

This item is only available as the following downloads:

Full Text








ABSTRACT ..................................................iv


1. INTRODUCTION ......................................... 1

Verbal Clitic Attachment ........................ 3

Gender Feature Proposal ......................... 7

AGR Approach .................................... 8


The Parametric Approach ........................ 11

Approaches to Scope ............................ 13

The DP Analysis Proposal ....................... 18

Case and Theta Chains .......................... 20

ECP and Government ............................. 22

PPA's and SPA's ................................ 24

Quantifier NP's ................................ 31

Inherent vs. Imposed Features .................. 33

AGR Features ................................... 35


The Polish Copula *Byd' ........................ 37

Reduction of Old Polish Preterite Form ......... 38

Modern Polish Status of 'by6e .................. 42

The Conditional 'by' ........................... 44

Clitic of Insistence 'le' ...................... 49

Coordinate and Subordinate Verbal

Constructions .................................. 51

4. AGR PHRASE PROPOSAL ................................ 56

'Knowledge of Language' Case Treatment ......... 60

Inherent Case and Specificity .................. 65

Some Issues of Government and Binding .......... 69

The Minimalist Program ......................... 75

The AGR Features Approach ...................... 87

5. PHONOLOGICAL CONSIDERATIONS ........................ 95

Univerbation ................................... 96

Voicing and Devoicing .......................... 99

Palatalization ................................ 102

The 'jer' Vowel ............................... 103

Concluding Remarks ............................ 105

BIBLIOGRAPHY ............................................ 109

BIOGRAPHICAL SKETCH ..................................... 114

Abstract of Dissertation Presented to the Graduate School of
the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy



Miroslawa M. Miller

May 1995

Chairman: D. Gary Miller
Major Department: Linguistics

This dissertation deals with movement of verbal clitics

in Polish in terms of expanding the AGR phrase. The analysis

follows Chomsky's Minimalist Program. Different contemporary

research approaches to dealing with verbal clitics are

discussed. These all follow the argument that clitics can be

subsumed under the AGR phrase. Though descriptive, these

analyses are not sufficient to explain the conditions for

verbal clitic movement in Polish. Following the historical

progression of the Polish verb paradigm from pre-16th

century to the present, I propose viewing the clitic

movement in terms of gender, number and person features on

the verb. With the Minimalist Program model for syntactic

data, I propose an expansion of the AGR phrase with a

separate gender node. Polish data support my argument of

positing a phonologically recoverable gender feature

condition on the verb, which must be satisfied in order for

verbal clitics to move. Absence of the condition prevents

movement from occurring.



The problem of classifying clitics in Polish rests on

the fact that the typical clitic treatment approaches have

dealt with clitics as pronominals (Borer et al. 1986,

Armstrong 1989). The overt case markings on the pronominal

system make it a good object of study. Unlike cliticizing

formations in languages like French, Spanish and Italian

(Jaeggli 1986, Mallen 1990, Mallen 1992, Manzini 1986, Rizzi

1986, Toman 1986, Wehrli 1986), Polish is much freer in

placement of the pronominal clitics, though Wackernagel's

Law, as interpreted in Anderson (1993), appears to hold in

it as well. Other approaches to classifying clitics have

taken the taxonomic route like Klavans (1982) and Zwicky


In Klavans' differentiation of verbal pronominall) and

2P clitics (enclitic on leftmost node of the surface phrase

marker), verbal clitics are attracted to a specific category

(verb), while 2P clitics are contingent on the linear notion

of 'position' and the notion 'constituent'. In the Polish

pronominal clitic paradigm, there appears to be some

divergence from the Klavans approach (if Aux clitics are in

situ). Unlike Romance clitics, there is no [+tense] for


proclisis and [-tense] for enclisis stipulation as in the

following from Klavans:

(1)a. Di=me=lo

b.*me lo di

'Tell me it'

(2)a. Magdalena esta cantandolo.

b.*Magdalena esta lo cantando.

'Magdalena is singing it.'

Polish follows the pattern of Archaic Spanish :

(3) Que me tu dizies

That to-me you say-FUT

'that you will tell me'

In these instances the category on the host was not as

strict as in Modern Romance, and before 'que em' became


Thus (4) is equivalent to (5):

(4) Ze mi ty powiesz.

That to-me you-NOM tell-FUT

'That you will tell me'

(5) Ze ty mi powiesz.

That you-NOM to-me tell-FUT

'That you will tell me'

The only requirement for ordering pronominal clitics is

in the enclitic position where Dative would precede the

Accusative, as in (6) and (7).

(6) a. ...... V clDAT clACC

b. *.... V clACC clDAT

(7) a. ..... przyniesiesz mi go

bring-FUT me-DAT him-ACC

b. *.... przyniesiesz go mi

bring-FUT he-ACC me-DAT

Polish has corresponding stressed pronominal versions of

pronominal clitics as in the following paradigm:

DAT-mi:mnie, ci:tobie, memu:mojemu, mu:jemu:niemu,

ACC-mil:mnie, ciq:ciebie, go:jego:niego,


The pronominal clitics will therefore appear in

unstressed positions, with Klavans utilizing the same

phenomenon as an explanation for preventing /IM/ from

appearing in isolation as a reply to:

(8) Who is it? Him

Verbal Clitic Attachment

The phenomenon of Polish verbal clitics freely attaching

within the CP boundaries has been problematic for clitic

theories so far. Topicalization and scrambling have been

approached as one answer to the "Free Word Order Phenomena"

(Gorecka 1988) but that approach and the other theories

would not be able to account for the grammaticality of

clitic attachment to scrambled object NP in (10), (12),

(14), (16) below and the barring of verbal clitic movement

in (18) and (20).

(9) (Ja) da=r=em/da=2=am Janowi ksiqkeS.

(I) give-PASTlmSg/fSg John-DAT book-ACCf

'I gave John a book'

(10) ?Ksiqikq=m dar/dala Janowi.

book-ACCf-cllSg give-PASTSgm/f John-DAT

(11) (My) Janowi ksiqtkq.

(We) give-PASTlmPl/lfPl John-DAT book-ACCf

'We gave John a book'

(12) Ksiqtkq.gmy dali/daly Janowi.

book-ACCf=cllPl give-PASTPlm/f John-DAT

(13) (Ty) da-=.ei/da=E--a Janowi ksiftkq.

(You)-Sg give-PAST2mSg/2fSg John-DAT book-ACCf

'You (Sg) gave John a book'

(14) ?Ksiqtke=s da2/daka Janowi.

book-ACCf-2Sg give-PASTSgm/f book-ACCf

(15) (Wy) da-l=i=-cie/da=k=y-icie Janowi ksiatkq.

(You) P1 give-PASTm2Pl/f2Pl John-DAT book-ACCf

'You (Pl) gave John a book'

(16) ?Ksioikq~scie dali/daky Janowi.

book-ACCf=2Pl give-PASTPlm/f John-DAT

(17) (On/ona) da=-/da-ta Janowi ksiqikq.

(he/she) give-PAST3Sgm/f John-DAT book-ACCf

'He/she gave John a book'

(18) *KsiAikq=-/ka da/da Janowi.

book-ACCf=clPAST3Sgm/f give John-DAT

(19) (Oni/one) da=li/da=ky Janowi ksizikq.

(They-mf) give-PAST3Plm/f John-DAT book-ACCf

'They gave John a book'

(20) *Ksiqikq=li/ry da/da Janowi.

book-ACC=cl3PASTPlm/f give John-DAT

Clitic attachment to scrambled object NP can be

illustrated in the following paradigm:





Looking at data illustrating clitic attachment to Comp of

embedded CP), the present tense inflection does not have a

phonologically realized gender feature, and movement is


(21) Maria mysli ie (ja) daj=q

Maria-NOM think-PRES3Sg that (I) give-PRESlSg

Janowi ksizqkq.

John-DATm book-ACCf

'Mary thinks that I give John the book'

(22) *Maria mygli ie=q daj Janowi ksiqike.

thatCllSg give-PRES John-DAT book-ACCf

(23) Maria mysli ze (ty) daj=esz Janowi

that (you) give-PRES2Sg John-DATm



'Mary thinks that you give John the book'

(24) *Maria mygli ze=esz daj Janowi ksiaike.

thatCl2Sg give-PRES John-DATm book-ACCf

(25) Maria mysli ze (on/ona) daj=e Janowi

that (he/she) give-PRES3Sg John-DATm



'Mary thinks that he/she gives John the book'

(26) *Maria myili ie=e daj Janowi ksiqzke.

thatCl3Sg give-PRES John-DATm bookACCf

(27) Maria mysli ie (my) daj=emy Janowi

that (we) give-PRESIPI John-DATm



'Mary thinks that we give John the book'

(28) *Maria mygli ie=emy daj Janowi

thatCllPl give-PRES John-DATm



'Mary thinks that we give John the book'

(29) Maria mysli ze (wy) daj=ecie Janowi

that (you) give=PRES2Pl John-DATm



'Mary thinks that you give John the book'

(30) *Maria mysli ie=ecie daj Janowi ksiqikq.

thatCl2Pl give-PRES John-DATm book-ACCf

(31) Maria mysli ze (oni/one) daj=a Janowi

that (they-m/f) givePRES3Pl John-DATm



'Mary thinks that they give John the book.

(32) *Maria myili ze=a daj Janowi ksiqzke.

thatCl3Pl give-PRES John-DATm book-ACCf

Gender Feature Proposal

Features need to be phonologically realized, since 0

marking for gender will generate structures, i.e. (22),

(24), (26), (28), (30), (32). In (18) and (20) the gender

and person features are not extractable separately from

tense marking, therefore movement is barred. If AGR would be

postulated with available slots for person, number, and

gender, then movement would be barred if gender is not

available as a separate entity for movement from the AGR

gender node, which would be 0.

AGR Approach

In Chapter 2, different contemporary approaches will be

analyzed in terms of their treatment of Polish verbal

clitics. Chapter 3 will present Decaux's (1955) historical

analysis of the evolution of the Polish verbal system and

the ensuing absorption of gender, person and number features

onto the verb. Andersen (1987), Halle (1990) and Jakobson

(1984) will further show the need for approaching the Polish

verbal system with a separate treatment of gender features.

In Chapter 4, after introducing Chomsky's Minimalist Program

and its implications for Polish, I will argue for the
subdivision of the AGR phrase beyond AGRg and AGRo

designation. Chapter 5 will deal with phonological

considerations of clitic attachment, the special status of

the verb 'bye' (to be) in Polish, and show that it is not a

counterexample to my expanded AGR phrase analysis even if

constructions allowing attachment of that verb's

inflectional clitics are possible with a [-tense]



(33) Maria mygli ie (ja) jest=em w domu.

Mary-NOM think-PRES that I) be-PRESlSg at home-LOCm

'Mary thinks that I am at home'

(34) Maria myhli ie=m jest w domu.

that=CllSg be-PRES at home-LOCm

The role of 'bye' as the 3rd person default function and

the special status of the verb 'to be' will also be analyzed

in Chapter 3, where I will show that its special treatment

also has historical basis and does not constitute a

counterexample to my gender feature recoverability condition

on the verb in order for clitic movement to occur.



There have been many different approaches to

classifying clitics. Klavans (1982) and Zwicky (1977), in

taking the taxonomic route, have specified parameters as the

classifying tool. Others, like Decaux and Andersen, have

chosen to reconstruct the history of clitic formation.

Others, like Borer et al., have fitted clitics into a

specific linguistic theory. Keyser and Roper specify a

category-neutral abstract clitic position for English while

stating that the clitic position is "in fact universal"

(1992:89). Booij and Rubach derive clitics in the lexicon,

and Szpyra proposes a looping model of phonology-morphology

interface which is supposed to allow clitic formation.

Aguado and Dogil posit a model in which phonological rules

must have access to representations, whether or not the

representations were derived. Ciszewska argues for a

separate AGR and Tense projection, and Kipka specifies a +/-

Past condition for verbal clitic movement in his treatment

of Polish impersonals. Halpern bases his very extensive

analysis of clitics in the realm of morphology. Clitics are

generated in AGR, according to Armstrong, but without

expanding the AGR phrase. That approach, like the others,

cannot account for the data in Chapter 1.

The Parametric Approach

Klavans sets out to represent clitics within the

theory of morphology and word structure. Cliticization is

one of the processes that shows a mismatch between a

morphological word and a phonological word. It also involves

syntactic processes which affect the internal structure of

the word. Some of the other processes affecting word

internal structure are case-marking, agreement marking,

Noun Incorporation, and phrase morphological marking. The

problem Klavans deals with is that the subject matter is

comprised of sets of words, grouped at the level of syntax,

but at the same time undergoing lexical rules and

phonological rules, the latter selectively. This makes them

behave as if they were single lexical items. Klavans

originally proposed a system which is capable of

constraining clitic positioning, based on the following

assumptions. (1) Cliticization is a unitary phenomenon and

(2) cliticization possibilities can be captured by five

binary parameters.

In her 1982 IULC edition, Klavans compresses her

original five parameters into three. The first two become

part of the lexical entry of a given lexical item, i.e.

specification of the entry itself and specification of the

lexical category.

Enclitics would look like:

(35) -[ [ ]- = enclitic ]
Proclitics would be represented as:

(36) -[ proclitic = -[ ] ]

The three parameters proposed by Klavans would pattern

in the following way:

(37) Parameter One: Initial/Final
Parameter Two: Before/After
Parameter Three: Proclitic/Enclitic

Configurational information is encoded in the first two

parameters, while the third carries phonological information

about the direction of phonological attachment (liaison).

DOMINANCE is encoded in the first parameter while linear

PRECEDENCE is encoded in the second parameter.

Klavans bases her parameters on observations of

placement and attachment possibilities of clitics. From

these observations she draws the descriptive statements from

which she forms the three parameters. In order for this

analysis to fit into a theory of morphology she proposes

that the parameters, in their role as descriptive

statements, are a derivative of the more fundamental

properties of lexical items, i.e. independently motivated

subcategorization frame which is required on lexical items,

and general constraints on applying phonological rules.

Klavans claims that cliticization is in fact phrasal

affixation. This would imply that a reflection of this be

present in the lexical representation of clitic. Clitics

attaching to phrasal nodes comprise the majority of clitics

Klavans researched. The French and Spanish verbal clisis,

which has V as the relevant domain, appears to be an

exception. In order to get around this, Klavans proposes a

constraint on the lexical representation of clitics. A

necessary part of lexical representation of clitics was a

phrasal requirement on the domain of cliticization. She

backs away from this in her later version by stating that

this requirement might be too strong, since it would result

in elimination the Spanish and French type of Verbal

clitics. Klavans analyzes them now as true verbal features,

changing the subcategorizing bracket from V' to V.

Approaches to Scope

The question of scope of cliticization in languages

such as French, Italian and Czech was analyzed by Toman

(1986). In it he states that in these languages the scope of

clitics covers only the VP. This would prevent them from

occupying a position in which they would have maximal domain

of c-command, extending over the CP. When clitics follow

COMP, as in Czech (and also in Polish) they would occupy

that position. Toman (1981:299) expounds on wh-movement to

COMP in Czech and Polish, settling on scrambling. This would

be formally represented as adjunction to S.

Scrambling is also used by Dyla to propose an S'

deletion for Polish. In his analysis, Dyla cannot account

for some exceptions to S' scope of CM (clitic movement) and

scrambling unless the S' boundary can be deleted. Though at

first it would seem that his data show that to be true, a

closer analysis would explain the exceptions as a question

of government and theta-role assignment of certain verbs,

the analysis put forth in Toman (1986) and Mallen (1990,

1992). In (38) (Dyla:328), the clitic Dative pronoun 'mu' is

scrambled all the way into COMP (38-41), generating an

ungrammatical structure when CP (Dyla's S') is crossed (42-


(38) Marek zauwazyk, ze wczoraj ktos

NOM noticed that yesterday somebody/NOM

zabraZ mu ksiqikq.

took him/DAT book/ACC

'Marek has noticed that somebody took a book

from him yesterday'

(39) Marek zauwazyZ, ze wczoraj ktos mu zabraX


(40) Marek zauwazyk, ze wczoraj mu ktos zabrar


(41) Marek zauwazyr, ze mu wczoraj ktos zabra&


(42) *Marek zauwazyk mu, ze wczoraj ktos zabrak


(43) *Marek mu zauwaiyZ, ze wczoraj ktos zabrat


The following are presented by Dyla as apparent

counterexamples to S'-boundedness for CM and NP scrambling


(44) Marek obiecal/przyrzekI ogoli6 siq

NOM promised to-have-shaved REFL

'Marek promised to shave himself'

(45) Marek obiecaZ/przyrzekk siq ogolid.

(46) Marek siq obiecaZ/przyrzekr ogolid.

(47) Marek obiecaa/przyrzekk przynies6 tq ksiaike.

NOM promised to-have-brought book/ACC

(48) Marek obiecat/przyrzekk tq ksiqkq przynies6.

(49) Marek tq ksiqkkq obiecat/przyrzekk przyniesd.

In CM examples (44-46) and NP scrambling in (47-49), the

fact that there is no object subcategorization by the verbs,
'Marek' is (A) for both verbs (following Mallen:1992

analysis) and 'ogoli6' and 'przynie6d' are [-tense], (with

for case and theta chains) can explain the phenomena without

resorting to a stipulatory S' deletion condition. Even Dyla

himself addresses the implication of his S' deletion

analysis in terms of PRO, which with his analysis would not

always be ungoverned, leaving the onus of proof on Case

assignment (336). Thus the structure of (50) would pattern

as (51a) at D-structure and (51b) after Verb Raising,

following Dyla's S'-deletion analysis (333,335):
(50) Mareki kazak Tomkowij umy6 swv6i/j

NOM ordered DAT to-have-washed REFL's


'Mareki ordered Tomekj to wash hisi/j car'

(51a) [S' [S[NP Marek] [Vyp[V kazal] [Np Tomkowi]

[S' [S[NP PRO] [vp[V umyc] [NP' [A sw6j]

[NP samoch6d]]]]]]]]

(51b) [S' [S[NP Marek] [vp[V kazal] [NP Tomkowi] [S[Np PRO]

EVP([ umyc] [NP' EA swoj] [NP samochod]]]]]]]

In Toman's (1986) model, clitics following COMP would

fail to exhibit subject/object asymmetries (144). Toman

analyzes cliticization from NP's in Czech as contrasted with

cliticization from NP's in French and Italian by positing

extraction from NP for Czech and PP for French and Italian.

For Toman, the problematic data is when cliticization occurs

from a genitive complement position, since specifier-head

structure is needed for government. This concerns Toman's

'secondary measure nouns,, which are quantifiers, i.e.

'tolik' (so many), 'tak' (so), 'cela/cely' (all f/m). He

proposes a rule of reanalysis at LF which will change a

syntactic head-complement structure into a specifier-head

structure (142). This rule is presented as (52).

(52) Conversion into the Normal Form:

Xa Y ---> X Y

head modifier specifier head

(where a denotes the appropriate class of nominal


Toman hopes with this rule to have the position from

which cliticization proceeds to be properly governed at the

level of Logical Form, but at the same time preserving

syntactic properties of these NPs, i.e. agreement, at the

syntactic level of representation. Unfortunately he runs

into evident counterexamples with cliticization from the

genitive position. If these cases are head-complement

structures, and the head is the operator in semantic terms

such phenomena will be explained. But stacking reanalysis

rules to explain a structure, only to have to posit another

to account for the discrepancy seems stipulatory. One

possible solution would be to analyze this phenomenon in

terms of Mallen (1990 and 1992).

The DP Analysis Proposal

Based on his analysis of lexical arguments of Spanish noun

phrases being realized in postnominal position (as Ide'-

phrases), Mallen (1990) proposes a hierarchical distribution

of positions. The configuration of the NP would parallel

that of the VP, with the parallelism also extending to the

nominal functional categories. Thus, a nominal functional

head D (Determiner) corresponds to C (Complementizer), the

sentential functional head, and a nominal functional head

NI(NFL) will correspond to the sentential functional head

I(NFL). Thematic roles would be assigned by NI and the

lexical head N rightward. All nominal heads D, NI and N

would also assign the Genitive Case rightward in Spanish.

Nominal arguments would be base-generated in postnominal

position and receive their T-role at d-structure and Case at

s-structure. Any occurrence of an argument in a prenominal

position would be attributed to a movement operation. Mallen

maintains this movement to be the process of cliticization

(head-to-head movement) to D through the intermediate F-

category NI, paralleling the mechanism of cliticization to I

at the sentential level.

In (53a), D selecting NI and NI selecting N parallels

(53b) where C selects I and I selects V. Lexically, C would

be instantiated by complementizers, I by models, D by

determiners and NI by quantifiers. The quantifiers so

problematic for Toman's analysis would be under NI in this

approach.The configurations pattern as in (52) and (54)


(53) (a) DP

Sp D'


Sp"^ NI'


Sp "N'

N i


(b) CP
Sp ^ C'


Sp I$



V Compl

Using Mallen's analysis, the ungrammaticality of

Toman's data would be explained by the Minimality (M-)

barrier NI blocking antecedent government (Mallen 1990:8-9).

(54) X is an M-barrier for Y iff

(a) X includes an Xmax (not necessarily distinct

from X) containing Y and

(b) the head of X c-commands a head Z

c-commanding Y

In analyzing the patterning of Romance clitics and

Toman's Czech examples, L-marking needs to be taken into

consideration. If Romance clitics can attach to V or I

(Mallen 1990:24-25), I is strong enough to L-mark VP in

Italian, but not French. Polish would pattern same as

Italian here. In French, since I is not a possible L-marker,

a clitic can only attach to V. Unless the V node can raise

to I itself, this restriction will apply. This process will

L-mark the VP and allow clitic attachment to I as in (55).

(55) Jean a promise de bien les faire

Jean has promised well to-them do

Otherwise clitics have the scope of V.

Case and Theta Chains

In Chapter 4, the importance of coindexation will be

demonstrated for clitic movement. In Mallen (1992), A'-

binding of an NP-internal element from outside becomes

possible if the element inside the NP is coindexed with a

clitic. The nominal elements in a case chain have to be a

part of the same thematic chain. This accounts for the

patterning in (56) (Mallen 1992:36).
(56) (a) *Me pareci6 eficaz [DP[NIP[NI su2] [NP N'

to-me seemed efficient its(T)

elaboracion pro2] del comite311

elaboration by-the committee(A)

(b) Me pareci6 eficaz [DP[NIP[NI su2] NP pro2

to-me seemed efficient his(A)

[N' elaboraci6n de la propuesta3] ] ] ]
elaboration of the proposal (T)

(56b) is acceptable since the case chain and thematic

chain condition is satisfied. Thus, 'su' is in a case chain

with NI and 'la propuesta' is in a case chain with N. Since

both belong to two different case chains they will also have

two different thematic chains. Agent theta-role and theme

theta-role hold for each chain. Polish patterns the same

way. Since its also inflected for case, its evident that

case chains and theta-role chains follow this pattern. The

interpretation for the following (57) Spanish structure

(Mallen 1992:42) would be the same for Polish (58).

(57) De que libro leiste mi reseda?
of what book(T) you-read my(A) review

(58) Ktorej ksiszki przeczytatef ma, recenzjq?
of what book(T) (you) read my(A) review

ECP and Government

This approach to analyzing clitics is echoed in

Culicover's discussion of analyzing ECP only under a

conjunctive definition of the ECP and with a DP analysis of

the noun phrase. He explains that ne-cliticization in

Italian and en-cliticization in French do not falsify the

claim of adjuncts being unextractable from NP. The clitic is

analyzed by Culicover as either an argument of N or the head

of NP. Here, ungrammaticality would result as a failure of

head government, not antecedent government since DP would be

L-marked by V. There would be no failure of antecedent

government since the trace of the adjunct will always be

antecedent governed. If ECP is to be used as a solution,

then that solution must come under the failure of head

government as defined by Rizzi (1990). Rizzi (6) defines the

two types of government as fully parallel, with the

difference being in the characterization of the classes of

governors. Head governors are the lexical heads as well as

some functional heads (i.e. those containing agreement and

tense specification), with the assumption that Agr and T can

be heads of independent projections and can be associated as

features with other heads. Antecedent governors are

comprised of coindexed categories. XP's not directly

selected by [+V] elements are "inviolable barriers for

government" (6).

(59) Head Government: X head-governs Y iff

(i) XG {A, N, P, V, Agr, T)

(ii) X m-commands Y

(iii) no barrier intervenes

(iv) Relativized Minimality is respected

(60) Antecedent Government: X antecedent governs Y iff

(i) X and Y are coindexed

(ii) X c-commands Y

(iii) no barrier intervenes

(iv) Relativized Minimality is respected

Relativized Minimality is defined by Rizzi (1990:7) as

the variable notion a-government," which ranges over head

government and antecedent government.

(61) Relativized Minimality: X a-governs Y only if

there is no Z such that

(i) Z is a typical potential a-governor for Y,

(ii) Z c-commands Y and does not c-command X

Relativized Minimality would allow antecedent government

in (62) (Rizzi 1990:9)

(62) How do you think [t' that [Bill solved it t"]]

The subject (A specifier as well as V, I, and C

intervene between t' and t" and 'how' and t". They do not

interfere with antecedent government. Instead of analyzing

the data as nullifying Vo intervention by not projecting a

V' level or positing the I system as intrinsically defective

(i.e. its projections never count as inherent or minimality

barriers) or 'that' deletion at LF, Rizzi proposes that

different kinds of government do not interfere with each

other. Therefore, in (62) the A'-chain is the relevant

relation, and under relativized minimality the intervening

heads and A specifiers do not have any blocking power. This

would also hold for the antecedent-government relation

between 'how' and t', since intervening heads and A

specifiers will never interfere with A'-chains in terms of

antecedent government. Chapter 4 will continue the analysis

of coindexation and chain relations.

PPA's and SPA's

Kipka's Polish inflection analysis to explain

impersonals is ingeniously elegant also for explaining Aux.

clitic attachment, given his morphemic division for Tense,

Gender and Number affix in [+past] and Person, Number affix

for [-past].

The phenomena I call Agr. clitics have been called

verbal desinences by Decaux and Andersen, among others.

Kipka calls them affixes in his acronyms, but the dynamics

are the same: they attach to V's, N's, Adj's, Adv's, Conj.

(anything that appears in a CP, through base generation or

wh-movement) and misbehave phonologically, although this

behavior is contingent on whether there is movement

(raising) or not.

Kipka assumes the following tenses (142):

(63) czyta6 'to read'

Past Imperfective: czytalismy 'we were reading'

Past Perfective: przeczytalismy 'we read'

Present: czytamy 'we are reading'

Future Perfective: przeczytamy 'we will read'

Future Imperfective: bqdziemy czytali 'we will read'

bedziemy czyta6 'we will read'

(64) czyt a

root V

1 i

Past number


I my

Past person



(past person affix)

czyt a

root V (thematic)





mple person affix)

In (64) above Kipka's morphemic division into PPA and

SPA explains how movement proceeds through the tree

structure, and why raising is the only option available for

AGR clitics. Kipka posits a Past tense specification as a

prerequisite to movement as in (65).

(65) (a) Kupilimny lustro. mirror-ACC

(b) Lustro kupilimay.

(c) Lustrosmy kupili.

(d) *Kupili lustrosmy. (no lowering possible)

(e) My kupiligmy lustro.


(f) Mygmy kupili lustro.

(66) (a) Wie ze kupilismy lustro. that

'He knows that we bought a mirror'

(b) Wie ze lustro kupili6my.

(c) Wie te lustroSmy kupili.

(d) Wie iedmy lustro kupili.

I will ignore Kipka's choice of labeling this desinence

as 'affix' (it does not attach only to morphemes, it has

freedom of movement, which a technical affix does not).

Analyzing (65) and (66) this left only floatation is

posited by Kipka as available to Past tense only, therefore

PPA's. He generates subjects in the [Spec,VP], they then

move to [Spec,AGRP] and then to [Spec,IP] where they acquire

Nominative case. Kipka's examples are mostly pro, and he

explains that they will be detectable from agreement

phenomena. Theta marking will occur within the maximal

projection of a relevant head if subjects are generated VP-

internal. Person agreement takes place only between AGR and

its Spec, number and gender by Spec-Head agreement. There is

only one AGR node per clause so that no clause contains two

PPA's, two SPA's or one of each. Analyzing (67) Kipka's
process could also work in Pollock's (1989) AGRs and AGRo

model. For Polish, where there's only subject agreement,
AGRs would be the recipient of the movement, the process

making the AGRO 'inactive', or leaving [Spec,TP] as a

possible landing site.

Thus the [-past] 'czytamy ksiakkq' "we are reading a

book" would have the following structure (144).

(67) IP



[-Past] NP AGR'

e -my VP

I 1_^\
pro V NP
czyta ksiqzkiq

The verb stem and its two arguments start out VP-

internal. The verb stem moves to AGR' (an SPA) and then into

I. The subject 'pro' (ipl) moves to [Spec,AGRP], triggers

person agreement and then moves to [Spec,IP] to get case.

For [+past] PPA of the type 'Zostaliimy pobici' "We were

beaten", Kipka assumes the PPA -smy to be in AGR and the

auxiliary and the main verbs to be heading VP's. Then, the

argument (pro) moves from object position to [Spec,VP] of

each VP, then [Spec,AGRP] then[Spec,IP]. The auxiliary then

moves up to the PPA under AGR, and the affixed verb moves to

[+past] I. Minimality will keep the passive participle in

situ. Tense checking in Kipka's model has PPA's, verbal '1'-

forms (i.e. 'czytali') and passive endings marked [+Past]

Anything moving into [+Past] must have all constituents

[+Past]. Anything moving into [-Past] (present, future) has

to have all constituents marked [-Past]. Thus PPA's will

only accept 'l'-form stems, while SPA's accept only non-'l'-

form stems (bare roots). Minimality will prevent passive

participles from adjoining to PPA's or SPA's. Since there is

movement, feature copying or similar mechanism is ruled out.

This would contradict Klavans's copying approach to clitics.

The reflective 'siq' also obeys attachment rules, not

only for pro-clisis or en-clisis, but for ordering of

attachment. It will follow verbal clitics and the

conditional 'by', and only attach to [Spec,CP] if there is

an overt subject. Thus we have the following:

(68) (a) Basia bawi siq na dworze.

Basia play-PRES3Sg REFL on-outside-LOC

'Basia is playing outside'

(b) Basia siq bawi na dworze.

Basia REFL play-PRES3Sg on-outside-LOC

(c) Basia sie na dworze bawi.

Basia REFL on-outside-LOC play-PRES3Sg

(d) *Siq Basia bawi na dworze.

REFL Basia play-PRES3Sg on-outside-LOC

(d) *Basia bawi na dworze sie.

Basia play-PRES3Sg on-outside-LOC REFL

In (69) and (70), any movement would honor the hierarchy

of attachment.

(69) (a) My bawili-=my siq na dworze.

we-NOM play-PASTmlPl REFL on-outside-LOC

'We were playing outside'

(b) My=-my bawili siq na dworze.

weCllPl play-PASTmP1 REFL on-outside-LOC

(c) My-dmy siq bawili na dworze.

weCllPl REFL play-PASTmPl on-outside-LOC

(70) (a) My bawili=by=nmy siq na dworze.

we-NOM play-PASTClCondmlP1 REFL on-outside

'we would have played outside'

(b) My=9my bawili=by siq na dworze.

weCl1Pl play-PASTmPlClCond REFL on-outside

(c) My=by=-my bawili siq na dworze.

weClCondCl1Pl play-PASTmPI REFL on-outside

(d) My=by-&my sie bawili na dworze.

weClCondCllPl REFL play-PASTmP1 on-outside

(e) *My=4my siq bawili=by na dworze.

weCllPl REFL play-PASTmPlClCond on-outside

In Kipka's analysis of the Polish Impersonals, the

present tense Reflexive Impersonals have third person

singular morphology. In [-present] (past and future) the

Reflexive Impersonals display third singular neuter

morphology (135-137).

(71) (a) Je siq cebule.

eat-PRES3Sg REFL onions-ACC

'One eats onions'

(b) Jadlo sie cebule.

eat-PAST3Sgn REFL onions-ACC

'One/we ate onions'

(c) BSdzie siq jadko cebule.

be-FUT3Sg REFL eat-3Sgn onions-ACC

'One/we will eat onions'

What Kipka does note about gender inflection in his

analyses is that third person singular neuter morphology

appears to be a default option (139). Thus impersonal forms

of verbs such as 'ii6' "to go" will pattern in the following


(72) (a) 0 co (ci) chodzilo?

about what (you-DAT) go-PAST3Sgn

'What did you mean?'

(b) Jak (ci) idzie?

how (you-DAT) go-PRES3Sg

'How are things going for you?'

Quantifier NP's

Default inflection is also induced by certain quantified


(73) (a) Nad tym pracowara jedna studentka.

on that work-PAST3Sgf one-NOMf student-NOMSgf

'One student worked on that'

(b) Nad tym pracowary dwie studentki.

work-PAST3Plf two student-NOMPlf

(c) Nad tym pracowaky trzy studentki.


(d) Nad tym pracowaty cztery studentki.


From number five (inclusive) on, the case of the NP is

genitive and the verb defaults to neuter gender inflection.

(74) Nad tym pracowako pi96 studentek.

work-PAST3Sgn five student-GENPlf

Kipka assumes the default agreement occurs along the

lines of QP's not bearing person, number or gender features.

Toman (1986:144) posits the genitive case as invariable for

an adnominal complement (head-complement) structure. The

head is an operator in semantic terms and makes

cliticization from genitive position possible in Czech (75)

with the same holding for Polish (76).

(75) (a) V Pa-fzi majif cela muzea takovych

in Paris they-have whole museums of-such


(b) Majf jichi v Paf'zi cela muzea [ei]

they-have of them in Paris whole museums
(c) [Takovyzh soch]i maji v Pa~fi cela muzea [ei]

of-such-sculptures they-have in Paris whole


(76) (a) W Paryiu maja care muzea takich rzeib.

in Paris have-PRES3Pl whole museums of-such

(b) Majq ichi w Paryiu care muzea [ei]

have-PRES3Pl of-them in Paris whole museums

(c) [Takich rzeb] i majq w Paryiu care muzea [ei]

of-such sculptures have-PRES3Pl in Paris whole


The NP 'muzea' (a container noun) has to be changed by

'cely/caly' for Czech and Polish respectively into a measure

noun. Only then cliticization will be possible. This is

where Toman proposed his head-modifier into specifier-head

conversion. He proposes treating the cliticization by re-

analysis, but treating this mismatch between syntactic

structure and the syntax of logical form in such terms is

not the only possible solution.

Inherent vs. Imposed Features

Zwicky (1986) in the format of GPSG formulates a

structure which would differentiate between imposed and

inherent feature specifications INH and IMP, respectively.

Imposed would deal with rules imposed by either government

or agreement, inherent specifications would not. Zwicky

argues for a multiple feature marking, and the need for

specifications imposed by agreement on a constituent to be

distinguished according to the types of triggering

mechanisms that impose them. This multiple feature marking

would imply that the AGR have INH and IMP features. The

approach again is taxonomic, with Zwicky citing PR (i.e.

possessor) and PD (possessed) structures to support the

claims. He analyzes the German direct object NP 'seine

Mutter' "his mother" in such a way (86). The base 'sein'

would pattern in the following way.





In order to predict inflectional material and to

distinguish inherent specifications of a PR (possessor) noun

from the specifications it bears by agreement with PD

(possessed) the following full form needs to be represented.


In formalizing his proposal, Zwicky turns to the

approach that categories themselves, i.e. sets, can be

viewed as potential members of categories. This would allow

the categories to be layered. Deep layering would then

distinguish different arguments of a verb according to the

layers they would occupy. AGR would be viewed as a subset

instead of a category-valued feature (95). It is interesting

to note that in some of his representations, Zwicky

acknowledges the split of gender, number, and case in AGR,

but does not formally represent it in any way. The resulting

system of rules might appear to work (though very

stipulatorily and taxonomically) for some CNW (cardinal

numeral words) but the list of features necessary to specify

all the needed conditions and contingencies becomes

unwieldingly long and complicated, with problems branching

out into an if/then type of solutions. Likewise in his

treatment of clitics Zwicky (1977) did not attempt a unified

characterization of cliticization.

AGR Features

Many contemporary approaches to forming a theory of

clitics have looked in the direction of the AGR Phrase.

Whether the approach is morphologically based, as

exemplified by Halpern, or heavily dependent on phonological

rules, as in Aguado and Dogil (i.e. phonological rules must

have access to representations whether they are derived or

not), or morphologically and phonologically parametrized as

in Klavans, AGRP appears to be the direction in which the

analyses are heading. Klavans's differentiation of lexical

clitics and phrasal clitics, the latter being defined

syntactically, suggests treating clitics in that category as

agreement markers, which appears to be the right direction

in attempting to deal with Polish verbal inflection clitics.

Booij and Rubach's division into two clitics, lexical and

syntactic, appears stipulatory. By positing separate gender,

number, person nodes in AGRP, the dichotomy of stipulating


lexical clitics with a surface filter and syntactic clitics

would be avoided. What needs to be addressed is the role of

the gender, number and person features, especially gender,

since the presence of this feature impacts on the

possibility of clitic movement in Polish, while its absence

rules out structures which many contemporary approaches have

not been able to uniformly explain. The positing of separate

nodes is not an ad hoc stipulation. Chapter 3 will show

that, historically, person, number, and gender markings had

started out as separate entities, realized in various ways

as verbs, auxiliaries and desinences.



According to Decaux (1955), the appearance of an overt

person-specified pronominal and then subsequent attachment

of the corresponding aux desinence to the V, started around

the 15-16th century, and resulted in the evolution of the

present verb paradigms in Polish. Historically, the

desinences (Infl. clitics, Agr. critics) follow the

development of BE, the Polish copula and verb of existence.

The Polish Copula 'Bvc'

Andersen (1987) follows Decaux's analysis of the gradual

development of 'bye' enclitic forms in Polish (fixed

desinence) from an autonomous word. In Common Slavic the

verb "bye' functioned as an auxiliary (Anderson:23-25,

Decaux:114-118, 127). Old Polish had two forms of 'bye', an

orthotonic form and a phonologically reduced, enclitic form

in unmarked position.

(79) Old Polish orthotonic enclitic Modern Polish

lSg jedm -(e)sm/-(e)m -(e)m

2Sg jei -(e)6 -(e)9

3Sg jest -0 -0


IPI jesm(y) -(e)smy -(e)Wmy

2P1 jeSde -(e)Wde -(e)dcie

3P1 sq -0 -0

iDual jeswa -(e)swa

2Dual jesta -(e)sta

3Dual jesta -(e)sta/-0

The copula 3rd person zero form becomes evident around

the 1300's and continues from there on. The full orthotonic

forms of 'jest', 'jesta', and 'sa' take on the function of

emphatic predicators, continuing this process through the


Reduction of Old Polish Preterite Form

The role of the copula in the Old Polish tense system

was as a part of compound past tense form. After recasting,

the original present tense form of 'bye' "to be" did not

have a corresponding past tense form with person marking.

The preterite tense forms underwent a reduction with the

compound tense forms subsuming the imperfect and aorist

function, in which 'bye' figured as auxiliary. This resulted

in Old Polish having only two compound past tenses, the

perfect (i.e. preterite) and the pluperfect. The latter,

though still present in some dialects, has declined from the

1700's, and is now more or less defunct (Andersen:17). This

also means that the present tense of 'bye' had no opposing

past tense form with person marking. Since 'bye' functioned

as an auxiliary in Common Slavic, with a resultative

participle in the -1 form of verbs, subject agreement only

in gender and number was shown by the participle

(Andersen:25). Thus, the original present tense form of

'bye' becomes reinterpreted as person and redundantlyy)

number marker, making the 1-participle a finite non-present

form, the preterite. The person and number marker,

therefore, was separate from the gender (and number) markers

shown by the desinences of the 1-forms. The presence of 1-

forms in the period between 1400-1600's resulted in regular

omission of the 3rd person markers, the singular 'jest'

being the first to go, followed by the plural 'sa' and the

dual 'jesta'. The pattern of omission occurred earlier for

following 1-form verbs: 'byZ' "was", 'miat' "had", 'm6gr'

"could", 'chciak' "wanted", and 'kazaa' "said"

(Andersen:28). In the post 16th century paradigm there are

no 3rd person markers in either singular or plural in the

preterite or the pluperfect (Decaux:127).

Common Slavic

Sing 1P

Sing 2P

Sing 3P

Dual 1P

Dual 2P

Dual 3P

Pl. 1P



jestr, jesti





Pl. 2P jeste

Pl. 3P sqtf, sqt


pre-16th c post 16th c present







For 1st and 2nd person mar

forms start disappearing around

jesm -(e)m -(e)m

eg (e) -(e)s

st(jesc, je) 0 0

bswa -(e)swa dial.-(e)zwa

sta -(e)sta dial. (e) ta

>sta -(e)sta/0 0

.smy -(e)smy -(e)Wmy

esm) -(e)m dial.-(e)m

iscie (e)cie -(e) ciE

sq 0 0

kers, the orthotonic full

.d the

16th century. Emphatic

predication proceeds initially by the presence of the 3rd

person markers 'jest' and 'sq', whereas in unmarked

predication the markers would be absent. The process of

reduction, therefore, followed from the initial full forms

to a combination of the emphatic predicator (i.e. person

marker) and enclitic person and number marker, to finally

enclitic markers by themselves. Morphologically, the aux

forms were reduced by dropping the initial 'je-' after

vowels, with 'je' becoming 'e-' after consonants.

Furthermore, a nominative personal pronoun justified


omitting person and number markers (81c), as did the

presence of two or more conjoined verbs (Bid).

(81) Old Polish Preterite

a. Wiem ie stworzyciela wszego luda

(I) know that creator of-all mankind

porodzira jeg

bore-Sgf 2Sg

'I know you bore the creator of all mankind'

b. To-m jest oglqdata

that-lSg emph"be" saw-Sgf

'That i did see'

c. Jest ja ciebie zepchn" albo uczynik-em tobie

emph I thee repulsed or did-lSgm thee

co zlego?

any harm

'Did I repulse thee or did thee any harm?'

d. Ani-s mie zepchnar, ani rzucir, ani-6

not-2Sg me repulsed nor deserted nor-2Sg

niekt6re zkofci uczynir

any harm did-Sgm

'Thou neither repulsedst, nor desertedst me, nor

didst me any harm'

e. Bo-cie-m siq caka darowala

for-thee-lSg REFL entire-f gave-Sgf

'For I gave myself wholly to thee'

Modern Polish Status of 'byc'

Though Old Polish raised the '-em' form to N or a

pronominal, in Modern Polish such a construction with the

plural form would be marked (Decaux:24).

(82) a. Ja jestem > jam jest

I be-lSg I-lSg be3Sg

b. my jesteimy > mysmy jest

we be-lPl we-1Pl be3Sg

The markedness would arise out of the orthotonic

function of 'byc', which would take the default 3rd person,

but take the 'jest' or 'sa' depending on the number marking.

Thus, even towards the end of the 15th century (83) would

not be prevalent, even though 'jestem' was widely attested

since 1439 (Decaux:24).

(83) poslanem jest

sent-PASS-lSg be-3Sg

The original 3rd person forms took on the role of

unspecified-person predicators, and continued as verb of

existence and copula in modern Polish. For 1st and 2nd

person, Old Polish used the forms with enclitic person and

number markers. The orthotonic copula (84a) remained current

until the early 16th century. In (84b) the form 'jest' is a

predicator and in (84c) it functions as an emphatic copula

with a person and number marking. The pronominal assistance

for person and number marker alone is shown in (84d), while

(84e) is an example of a person and number marker appearing

alone (Andersen:37).

(84) a. Ty jeh uliczka rajska

'thou art road to paradise

b. Ty, jen- ie-6 jest przez poczatku...

thou who -2Sg emph since beginning

'Thou, who hast been since beginning...'

c. Bo ja jest-em Maria

for I emph-lSg Mary

'For I am Mary'

d. Bo-m ja pan B6g wasz

for-lSg I Lord God your

'For I am the Lord, your God'

e. Ale my, co-imy?

but we what-lPl

'But we, what are we?'

This transition can also be seen in (85) based on Decaux


(85) spa& jesm > ja jesm spaa > jam spat > spatem

The pronominal/clitic construction with the -1

form is still evident in some regional dialects, with the

southern regions (i.e. Cracow) accepting it as dialectal

variation, as opposed to the northern (i.e. Warsaw) regions,

which view this structure as archaic (Andersen:30).

The Conditional 'by'

The conditional clitic 'by' in Polish is generated

phrase final, and can raise to second position a la

Wackernagel into CP, providing there is a conjunction

present. Historical data show this progression. In Polish

law books of 1450, the following constructions appear


(86) gdy taki listy albo zapis6w

if such-Sgm letter-ACCPlm or record-GENPlm

twardoici miat=by

hardship-GENPlf have3SgCondCl

'If such a person would have letters or records of


The 'by' takes on the scope over CP by attaching to


(87) Gdy=by taki......

ifCondCl suchSgm...

Following Decaux's (145) and Andersen's (35) analyses,

the conditional developed first as a stem 'by-' plus

desinences in Old Polish, reanalyzed in modern Polish as

enclitic '-by' + enclitic number/person markers (my AGR

clitics). This historical development accounts for the

ordering of clitic clusters with the conditional preceding

the person/number enclitics.

(88) Common Slavic Old Polish Modern Polish

1st Sg. byxzi bych by-m

2nd Sg. by by by-s

3rd Sg. by by by

1st Pl. byxomu bychom, bychmy by-smy

2nd Pl. byste byicie by-scie

3rd Pl. by9q bych% by

1st Du. byxove bychowa, bychwa dial. byiwa

2nd Du. bysta bysta dial. bysta

3rd Du. byste bysta by

The dialectal dual form is equal to syntactically

pluralizing a semantically singular person while keeping the

style informal. Some dialects preserve this, otherwise this

function has been subsumed by 2nd person plural, used as a

polite form for a singular, i.e. -escie.

The preterite forms and the conditional forms had

corresponding diverse forms at the crucial transition stage

during the 16th century. The paradigm of development of verb

and conditional endings is from Decaux (154).

The residue of the '-ch' forms is evident in some

dialects, i.e. Dobrzyf, Warmia and Podhale, but not Mazowsze

(the northern parts), following the pattern of more archaic

tendencies of the southern dialects.

(89) Preterite

1st Sg. -(e)m


0 (+ ja)

1st Du. -(e)swa, -(e)iwa

(e) chwa

-(e) gma

(0 + wa)

1st Pl. -(e)smy, -(e)gmy


-(e) m

0 (+ my)




by (+ ja)

byswa, byiwa



(by + wa)

bysmy, bymny



by (+ my)

2nd Sg. -(e)9 byi

2nd Du. -(e)sta, -(e)gta bysta, bysta

2nd Pl. -(e)gcie byscie

3rd all 0 by

In the 16th century the 'bym' form replaced the 'bych in

both singular and plural:

(90) staralichmy sig, abym listy swe

try-PAST1Plm REFL to-CondCl letter-ACCPlm ownCl

przeskad mogli

send can-PAST3Pl

'We tried, so that we could send our letters'

As the pronominal became introduced, it carried the

person marking, transferring the marking then to the verb,

with the 'be' verb inflection becoming an aux-clitic. Thus,

the preterite and the conditional forms of Polish

auxiliaries were reduced to a single form in the 16th

century, with the conditional having the same form as the

preterite, but augmented by 'by'. Decaux (126) calls this

the conditional particle. The preterite in Polish patterns

after the enclitic form of the verb 'bye' "to be" as in


The conditional attaches to the aux-clitic, but

according to Decaux, there is no stress shift with the


(91) Ja 'robiI

I do-PAST1Sgm

Ja by 'robit

I CondCl do-PAST3Sg


do-PAST3SgmCondCl Sg

Ja ro'bikem

I do-PAST3Sg

Ja bym 'robir

I CondCllSg do-PAST3Sgm

This is in contrast with AGR clitics changing

penultimate stress on an -1 stem verb only, and

person/number clitics (my AGR clitics) affecting voicing of

stem final obstruents, discussed below in the section on

phonological considerations.

In embedded conditionals, 'by' attaches to Comp. If Comp

is null, then 'by' will follow the first constituent


With [-tense]:

(92) teby nie byd gokoskownym, dam konkretny

soCondCl neg be 'empty-worded' give-FUTlSg specific



'So as not to speak in vain, I will give a specific


In passive participle constructions:

(93) A c6oes chciak?

and whatInsisClAGRC12Sg want-PAST3Sgm

'And what (insistence) did you want?'

ieby ich uczono filozofii

soCondCl they-GENPl teach-PASSn philosophy

'that they be taught philosophy'

i odciqgano od konfesjonak6w?

and pull-away-PASSn from confessionals-GENPlm

'and be pulled away from confessionals?'

In example (93), the clitic is in second position,

attaching to [+wh] in Comp.

With predicate adjectives, 'by' clitics also attach as

second to a leftmost constituent, i.e. Adj, or conj.

(94) Rldbys potem, aby sip

happyCondClAGRCl2Sg after soCondCl REFL

tancem zabawiala.

dance-INSTRSg entertain-PAST3Sgf

'Happy you would be later, if she would entertain

herself with dancing'

The clitic position constraints have not changed, even

though this example is from 1614. The reflexive's position,

though, in modern Polish would require adjacency to the verb

either pro- or en-clisis. Other positioning would make

grammaticality judgements dependent on dialectal parameters

(cf. (44), (45), (68) and Aguado and Dogil:108).

Clitic of Insistence 'le'

Decaux posits an enclitic 'ie' which he calls a mark of

insistence, while Szober calls it epideictic and

augmentative for demonstrative pronouns (134). In respect to

AGR clitics, the clitic 'ie' is proclitic. In example (89)

above, Decaux posits 'ie' followed by '-ei' and analyzes it

as 'ie=s', with the insistence clitic/preterite clitic

cluster comprising of 'le-' + preterite desinence with the

optional (-e) not present.

Looking at the 'ie' examples below, from the 16th

century to the present, the paradigm is quite stable, with

phonological variations fitting Decaux's specifications


(95) From 1564:

Jesli=ie kto z ksiqhtwa kt6rego o to sie na miq

gniewa& bedzie...

'If, from which kingdom someone will be angry at me

for this'

From 1594:

A jedli=- to u inszych najdujemy...

'And if then we find that with others'

From 1597:

[...] Zapraw=ie was Pan B6g w wielkq mikos6 ku

braciey waszey

'Indeed, the Lord God in great love towards your


From 1890:

jak=ie siq miewa i jako=- mu siq powodzi?

'How then is one, and how then are things'

(96) From 1994:

a. Szybko=ze=smy wr6cizy.

fastInsisClAgrCllPl return-PAST3Plf

'How fast we returned'

b. wr6city.

fastAGRCllPl return-PAST3Plf

'We returned fast'

In respect to elements in Comp, the clitic 'ie' is an

enclitic (95), with a simultaneous proclitic function if the

preterite AGR clitics also attach to the cluster (96a). In

Pachowicz, 'ie' as proclitic to 'by' is attested as late as

1803 (192).

Coordinate and Subordinate Verbal Constructions

The analysis of coordinate verbal constructions in

Polish dating from 1398 to the beginning of the 20th century

presents the phenomena of inflectional clitics, the

reflexive clitic and the conditional clitic being present on

the first verb in the series, and being omitted on the

subsequent ones.


tako mi B6g pomoiy i swieci, ei

so me-DAT god help-PRES3Sg and bless-PRES3Sg thai

by2-esm przy tem i widzia=-0

be-PASTmCllSg with-it and see-PASTSgm

'so help me God and bless me, that I was there and

I saw (it)'


(98) Doznali=9my, ach, doznali=0

experience-PAST3PlmCllPl oh experience-PAST3Plm



'We experienced this, oh, we did (experience)'

(99) zgrzeszyka=h i niezgrzeszyra=0

(you) sin-PASTSgfC12Sg and not-sin-PASTSgf

'you sinned, and you didn't sin'

The omission of person and number AGR clitics on verbs

in coordinate structures, dated 1398 for (97) and 1652 for

(98) is still evident in 1880 for (100) (Decaux:30-31). In

this example, there are two verbs, in a series of three,

which have 0 AGR clitics, with the first verb carrying the

person and number clitic.

(100) Odtad nie czuli=6my, nie widzieli=0

since then NEG feel-PASTPlm=1Pl NEG see-PASTPIm

i nie myhleli=0

and NEG think-PASTPlm

'Since then, we did not feel, see, or think'

In coordinate constructions 'sig' originally patterned

overtly on the first verb and was lacking on subsequent


(101) Kompania rozrzuca sil w

company-NOMSgf scatter-PRES3Sg REFL into

plutony, rozwija w druiyny...

platoon-ACCPlm expand-PRES3Sg into squad-ACCPlf

'the company scatters itself into platoons,

expands itself into squads...'

The implications arising from this analysis suggest a

different theta role assignment of verbs in Old Polish. The

verb 'rozwijac sil' requires a patient theta role to be

assigned. Without the reflexive, Modern Polish specifies a

patient theta role in addition to the agentive function.

Omission of the reflexive in Modern Polish results in *

constructions. Since the data from Decaux was acceptable in

Old Polish constructions, any further analysis in this

direction will have to deal with that question.

The 'by' conditional clitic also follows this

coordinate attachment pattern:

(102) wolal=by ja wolak=0

prefer-PASTSgm=CondCl I prefer-PASTSgm

"I would prefer, I would'

When the data has an orthotonic verb of existence, there

is no omission (examples from ca. 1400, based on Decaux:32).

(103) Ja jesm spar i naspak

I be-PRESlSg sleep-PASTSgm and sleep-PAST-PerfSgm

jesm siq


'I slept and am rested'

(104) i wiesielili jesmy siq i

and rejoice-PASTPlm be-PRESIPI REFL and

kochali jesmy sie we wszech dnioch

love-PASTPlm be-PRESlPL REFL in all day-Pl



'and we rejoiced and loved each other in all our


The subordinate verb constructions pattern as in (102).

(105) wiqcej czytali-9my utwory nii

more read-PASTSgmCllPl work-ACCPlm than

studiowali=0 tekst podrqcznika

study-PASTPlm text-ACCm school-book-GENSgm

'we read the works more than we studied the text

of the school-book'

The orthonic form of 'bye' together with the -1 forms of

the verbs would have the gender and person features realized

separately. Only when the AGR clitics are able to attach to

the -1 stems, after the verb+desinence+aux >

verb+pronominal+aux [aux > 0] progression are they able to


move, being recoverable in their representations. Since

historically the desinences started out separately, the

proposal for their syntactically separate treatment follows




The different approaches to clitic movement analyzed in

Chapter 2 have shown that the answer to the problem of

positing a verbal clitic movement theory lies in a closer

scrutiny of the AGR Phrase structure. Polish data in

Chapter 3 present verb paradigms in Old Polish in which

there is a hierarchy of clitic attachment. If this hierarchy

exists, and data cited by Decaux, Anderson, Pachowicz,

Szober, Zwicky, Booij and Rubach, Aguado and Dogil show that

it does, then there has to be a preset, specified structure

that reflects this hierarchy. The Dative pronominal-REFL

construction has been attested in Decaux, Andersen, Aguado

and Dogil, and Szober. Aguado and Dogil specify an ordering

of case clitics and 'floating clitics' (my AGR clitics). The

AGR clitics follow case endings:

(106) Jankow=i=-


Szober posits a structure in which pronominal clitics

inflected in the Dative precede the REFL.

(107) Jak=-e ci siq spato?

how-InsisCl you-ClDAT REFL sleep-3Sgn(impersonal)

This ordering in a syntactic tree structure would

account for the assignment of functions, i.e. case and theta


Borsley, in analyzing wh-movement and Polish equatives,

notices that 'jak' "how" can move if it is inflected for

case (279-280).

(108) a. Z jakim przystojnym mqiczyzna rozmawiaka Anna?

with how handsome man talked Ann

b. Z jakim Anna rozmawiaka przystojnym mgiczyzna?

(109) a. Z jak przystojnym mqtczyznq rozmawiaZa Anna?

b. *Z jak Anna rozmawiaka przystojnym mqiczyzna?

In (108a) 'jak' carries Instrumental case inflection,

contrasted with (109a) where 'jak' is not inflected,

resulting in a construction.

Cichocki, in his Comp analysis, posits the presence of a

wh+complementizer as a construction, with wh+relative as

an acceptable one (61-62). In (62), Cichocki assigns 'ie' as

a relativizer, even though it is an emphatic marker. The

ungrammaticality could also be analyzed as a failure of co-

indexation, using a theta and case chain approach.
(110) a. *Widziat=em chkopcai [ ie kt6r.emui

(I) see-PASTlSgm boy-ACCSg that who-DATm

[ kupi=-eh ksiaqke]]

(you) buy-PAST2Sgm book-ACCSgf

b. Widziatem chlopca [ kt6remu ie=s

(I) see-PASTiSgm boy-ACCSg who-DATm that=2Sg

[kupit ksisike]]

buy-Sgm book-ACCSgf

Zwicky (1986) tries to fit the structure containing AGR

features into a General Phrase Structure Grammar model, but

that results in a cumbersome, taxonomic, ad hoc model, which

cannot account for verbal clitic movement in Polish data

(1986:86-96). Booij and Rubach's analyses run into problems

with case assignment and floating clitics, creating a

paradox in their theory. A possible solution of a post-

lexical checking mechanism, first suggested by Aguado and

Dogil, is an ad hoc assumption. The only justification for a

filter of this sort is the presence of a structure which

cannot otherwise be accounted for, and which will not stop

*structures from being generated. It would be much more

elegant to posit a system of generating only grammatical

structures, and defining the environment for their

generation, instead of overgenerating, producing

ungrammatical structures, and then positing stipulatory

mechanisms to rule them out. Booij and Rubach's attempt to

keep cliticization within the theory of Lexical Phonology

cannot be maintained, in view of Polish pronominal clitics

and the impossibility of their attachment within a major

constituent (Cichocki:56-57, Aguado and Dogil:108).

Nevertheless, Aguado and Dogil's solution to this problem,

i.e. applying phonological rules to either lexically or

syntactically derived representations, cannot account for

all clitics. They still have to stipulate the 'by' clitic

attachment to the inflectional clitics in the lexicon as a

set of inflected (double) clitic forms.

(111) 1 Sing -by=m

1 P1 -by-=my

2 Sing -by=h

2 PL -byascie

3 Sing -by

3 P1 -by

Aguado and Dogil's attempt at a solution to the problem

of 'by' results in postulating morphological word-formation

rules requiring strong syntactic conditions, such as

selecting the position of an affix in the embedded clause by

the main verb. This analysis is abandoned by the authors

since 'such "rules" would add unlimited power to the

morphological component' (107), and still filters would be

needed to eliminate improperly assigned clitics. Aguado and

Dogil leave the question of 'by' unanswered. They also fail

to take into consideration the historical progression of the

'by' paradigm, which inflected parallel to the orthonic

'bye' forms, but which after enclisis of 'bye' in post-16th

century Polish formed a paradigm of 'by' attaching to AGR

clitics only if the latter raised into Comp.

With the shortcomings of all these approaches, one

possible answer appears to lie in Chomsky's Minimalist

Program approach. The case and theta chain assignment work

very well for Polish data, something that was not as

successfully accomplished in the earlier Case assignment

approach in 'Knowledge of Language' (1986b). The interface

model in MPLT also impacts on the notion of Government and

Binding and barriers, following Cinque. With Chomsky's

interface approach, though, other ideas become possible,

such as positing a separate AGR node for gender, number, and

person. Based on Doron's proposal for an AGR Phrase

structure entailing gender, person and number feature

specifications, these can be adapted into the Minimalist

Program perspective to account for the Polish verbal clitic

movement, something previous descriptions have not

successfully accomplished.

'Knowledge of Lanauage' Case Treatment

The first point of divergence in comparing approaches to

Case in 'Knowledge of Language' (Chomsky:1986b) with 'A

Minimalist Program for Linguistic Theory' (Chomsky:1992) is

in the levels of representations. In Chomsky:1986b, there is

D-structure, S-structure, PF and LF, phrase structure and

transformational rules being the rules which generate D- and

S-structure representations, and the rules of PF and LF

components. There is also an assumption of asymmetry present

in the relation of subject and object to the transitive

verb. Though empirical in nature, the assumption is

supported by V-NP movement as a unit, or meaning being

assigned to V-NP compositionally (i.e. throw a fit, throw a

party, etc.). This contrasts with Chomsky:1992 approach of
AGRg and AGRo being on 'equivalent footing' as far as

features are concerned. Fundamentally, of course, there is

the basic divergence of the 'Knowledge of Language' (KL)

approach treating Case as a D- and S-structure phenomena, D-

structure for inherent and S-structure for structural Case.

Raising is obligatory in terms of Case theory that every

lexically realized NP must be assigned (abstract) Case. In

some languages (i.e. Polish) Case is morphologically

realized. Even in languages that do not realize Case

morphologically, the assumption is that Case is assigned in

a uniform way. Thus, objective Case is assigned to the

object of a verb, nominative to the subject of a finite

clause, oblique to objects of preposition. Subject position

of a finite clause saves a Caseless NP when it is moved into

the position from an intransitive verb object position (i.e.

verbs like 'seem'). If a language permits another Case to be

assigned to the complement NP, then there will be no

movement, such as in Spanish or Italian, both null subject

languages (as is Polish) which permit an empty subject to

remain in place, associated with the object of the

passivized verb to which it transmits nominative Case.

(112) e was see-en Bill (by John)

Chomsky in KL does suggest parametric explanation to

determine the choice of a in Move-a, as in wh movement in

Japanese at LF and in English overtly. The differences would

lie in S-structure, with Japanese-Chinese having S-structure

same as D-structure and English S-structure being same as


There is also an adjacency requirement on Case in KL.

Using X-bar theory, Chomsky puts forth a principle that

where Case is not morphologically realized, a Case marked

element must be adjacent to its assigner (with some

variations). Thus *'put [on the table] [the book]' has 'on

the table' interfering between a Case assigner and an

assignee. With morphological Case realization the equivalent

construction in Polish would be well-formed.

A Visibility condition is also imposed. Relating to

theta theory, it states that a noun phrase can receive a

theta role only if it is in a position to which Case is

assigned, or is linked to such a position. Lexical

arguments, therefore, must have Case, or they will not

receive a theta-role and will not be licensed. Chomsky

considers the Case filter to be eliminated as an independent

principle, as it follows from the Visibility condition.

A chain at S-structure becomes a 'history of movement'

of an element as it is moved from the A-position it occupied

at D-structure to its S-structure position. Theta roles and

Case are assigned to chains, with the chain headed by a Case

marked position and terminating in a theta position. The

application of this can be seen in data examples (108a-b),

(109a-b) and (ll0a-b) above. Case is then transferred from

the head to the terminal position of the chain, making the

terminal position visible to receive a theta role, which

then transfers to the argument that heads the chain. Case

transfer will occur in 1. Chains and 2. Pairings of argument

and expletive (i.e. 'there is a man in the room').

In KL as in MPLT, Chomsky put forth the notion of full

interpretation (FI). In KL, FI requires that every element

PF and LF, taken to be the interface of syntax with the

systems of language use, has to receive an appropriate

interpretation, i.e. it must be licensed in the sense that

is indicated (Chomsky:1986b:98).

If language assigns to each expression a structure

(D,S,P,L--D-structure, S-structure, PF and LF) D will

satisfy one formal and one general condition. Formal in that

it conforms to the principles of X-bar theory, and semantic

in that it is a pure representation of the theta structure.

P and L must satisfy FI, i.e. each element must be

licensed in an appropriate way.

Regarding S-structure, Chomsky says that many of its

properties 'may be reducible to the independent condition of

FI holding of PF and LF representation and the conditions on

D-structure, given an appropriate account of the ways in

which the elements....may be related' (Chomsky:1986b:101).

This echoes the FI and PF, LF interface in XPLT.

Case is uniformly assigned under government. A category

governs a maximal projection X'' if a and X'' c-command

each other. If a governs in this way, it will then govern

the specifier and the head X of X" A head a, then,

governs its complements (the core case of government). In
the following construction [yp V NP], where NP= [NP DET [N'

N...]], V will govern NP, DET, and N. Subject and predicate

govern one another, but only lexical categories and their

projections can be governors: i.e. N, V, A, P, NP, VP but

not the complementizer or infinitival INFL. AGR, assumed to

be nominal, in that it has feature of person, number, and

gender, can govern the subject (Chomsky:1986b:162).

The KL approach distinguishes structural Case from

inherent Case in terms of S-structure position (the former)

and D-structure position (the latter). Chomsky also assumes

that the direction of Case marking is uniform, in marked

cases corresponding to the head parameter of X-bar theory.

In English this works fine, except for the genitive Case.

Therefore, KL distinguishes Case-assignment(at D-structure)

from Case realization (at S-structure) for genitive. Both

would fall under government: with N governing and theta-

marking its complement and assigning Case at D-structure,

and N governing both the complement and the subject at S-

structure, so that Case assignment is possible in either


Inherent Case and Specificity

There are some problems with specifying a Case as

inherent or structural. Chomsky does not explain the

assignment of inherent Case by verbs, like 'helfen' in

German assigning the dative instead of accusative, or

'persuade' assigning genitive as an inherent Case to its

second object, as (113).

(113) I persuaded John [of the importance of going to


Lasnik (1992) expounds on Chomsky's KL approach to Case

assignment by raising V-to-I.
(114) [[a car] [Tense bel] [not tI here]]

Such structures would have operation limited to [+aux]

verbs, following SPC. With expletives, Lasnik postulates

that 'unaccusatives' and 'be' is a Case assigner (389), with

no Case transmission but direct Case assignment, with the

difference in auxiliary verb raising. We would have the

following construction.

(115) ?There arrived a bus.

(116) There is usually a man here.

Adjacency effect is apparent in the following:

(117) *There arrives usually a bus (at this time).

With (116) allowing raising, and therefore leaving a t,

there would be no such effects.

There have been suggestions of multiple Case-assigning

possibilities for individual verbs, or an affix having the

ability to assign a particular Case. Turkish and Hebrew data

show that verbs can assign both accusative and partitive

Case. The crucial condition is specificity, with specific

objects being marked accusative, and non-specific having no

overt Case marking (Lasnik:398).

(118) Ali bir piyano-yu kiralamak istiyor.

Ali one piano-ACC to rent wants

'Ali wants to rent a (particular) piano'

(119) Ali bir piyano kiralamak istiyor.

Ali one piano to-rent wants

'Ali wants to rent a (non-specific) piano'

This suggests that accusative Case can be associated

with a semantic property (specificity) with the implication

that semantic properties are not limited to inherent Cases.

Other languages support this:

Hebrew has a morpheme 'et' which shows up on specific

objects and is analyzed as accusative, while a non-specific

object will have no overt Case morphology (Lasnik:398).

In Polish, verbs that assign accusative must assign

genitive when they are negated, making it a structural not

inherent phenomenon. There is a specificity condition in

Russian, with 'definiteness' communicated by retention of

accusative, with genetive being assigned in its absence. The

structural aspect of this assignment is also argued on the .

grounds that non-theta-marked expressions of duration may

also get genitivee of negation, as long as they are not

governed by the negated verb. Also, there are some verbs,

i.e. in Russian 'pomagat' ('help'), and in Polish 'pomaga6',

which assign oblique Cases to their objects, marking their

object dative and in Russian 'upravljat' ('manage'), marking

its object with the instrumental Case. Neither of these will

alternate with a genitive when they are negated, meaning

that they are inherent.

(120) (a) (My) pomagaky=smy mu.

we-NOM help-PASTPlflPl him-DAT

'we were helping him'

(b) (My) nie pomagaly=imy mu.

we-NOM MNEG help-PASTPlflPl him-DAT

(c) My=imy pomagaky mu.

we=CllPl help-PASTPlf him-DAT

(d) My=smy mu pomagaky.

(e) My-=my mu nie pomagaZy.

(f) ?My mu=-my pomagaiy.

(g) ?My nie pomagary.

In examples (120f) and (120g) the pronoun is already

inflected for Case (dative) when the verbal clitic cluster

attaches to it, which would be supported by Chomsky's

suggestion that parametric properties are located

exclusively in the functional portion of the lexicon, the

substantive portion being invariant (Lasnik:401). Therefore,

when negation of a verb changes the Case assigned by it,

that has to be specified in the lexicon. In 'Minimalist

Program for Linguistic Theory' this is echoed in the

features of lexical items being present in the lexicon, with

morphological requirements being the driving force.

Some Issues of Government and Binding

Toman (1986) posited two distinct syntactic structures,

the specifier-head structure for cardinal number phrases,

and head-complement structure in the case of pseudo-

quantified phrases as in those with 'cely', 'cela' (142-

144). The former would be analyzed in syntax, the latter

would be reanalyzed at the level of Logical Form. Cinque

(1990) posits the notion of D-linking, which sheds some

light on Q-phrase movement.

Cinque posits the notion of 'referentiality' (taken to

mean the ability to refer to specific members of a set in

the mind of the speaker or in a pre-established discourse)

to subsume the notion of D-linking. This linking impacts on

conclusions about wh-phrase behavior at LF, and is

consistent with Q-phrase movement at S-structure. Thus, only

D-linked (referential) phrases can enter into a binding

relation, at LF or at S-structure, and non-D-linked (non-

referential) phrases can only enter chains of antecedent

government relation, at both LF and S-structure. D-linking

implies co-referentiality, as is evidenced in the following

examples taken from Cinque (17).

(121) (a) Which boyi started a fight with which girl

wasn't clear even to themi+j.

Since 'which boy' is D-linked, there is co-


(b)*Whoi started a fight with whom wasn't clear

even to themi+j

'Who' is non-D-linked, therefore there is no co-


Extraction of interrogative phrases in indirect

questions depends on the character of the extracted wh-

phrase. Only D-linked (referential) wh-phrases can enter a

binding relation with their trace, thus their insensitivity

to wh-islands (i.e. weak islands). Non-D-linked (non-

referential) wh-phrases, on the other hand, can only move

successive-cyclically, thus their sensitivity to weak

islands. This account can explain the asymmetry between

questioning (* or ?) and relativization (? or ) since the

relative wh-phrase can be more easily referential than the

interrogative wh-phrase, the latter behaving more like a

non-referential operator.

In order for a phrase to enter a binding relation with

its trace, it has to occupy an A-position, receive a

referential theta role, and have intrinsic referential

properties. Thus NP's in NP-Movement (like Passive, Raising,

etc.) and Clitic Movement are in A-position and receive a

theta role at S-structure. They have to enter a government-

type rather than binding-type relation because of the

independent requirement of theta-role and Case transmission

being a property of chains, and chains can be defined in

terms of antecedent government. In A'-relations there is no

theta-role and Case transmission. Therefore, antecedent

government is not forced on them. Traces of NP-Movement and

Clitic Movement, in contrast to variables, are not

referentially autonomous--they are parts of a discontinuous

referential element, i.e. the A-chain.

According to the 'Barriers' framework, sensitivity to

strong (and weak) islands of the successive cyclic Wh-

Movement results from the requirement that each link of the

successive cyclic chain satisfies antecedent government. The

sensitivity of long Wh-Movement to strong islands follows

from the separate principle of Subadjacency.

Antecedent government then in the 'Barriers' model would


(122) a governs 0 iff a m-commands 0 and there is

no y, y a barrier for 0, such that y excludes a.

Subjacency would be:

(123) 0 is subjacent to a iff there are fewer than two

barriers for 0 that exclude a.


This attempt at unification is not complete, however.

There is an asymmetry if one barrier suffices to derive

strong islands for government, but two are needed to derive

the islands for Subadjacency. Also, two notions of barrier

are postulated, with the 'minimality' barrier for government

and a weak barrier for Subadjacency (i.e. embedded IP or

CP). Furthermore, 'the notion of barrier holding of

government and Subjacency can be given a unified definition

only at a certain cost' (Cinque:22).

Cinque shows barriers by inheritance to be redundant,

since there exist cases where only one (inherent) barrier

triggers Subjacency as in the following.
(124) a. *To whom were they [Ap too angry [cp PRO to

talk t]]
b. *To whom were they Ap angry enough [CP PRO not

to talk j]]

These are degree clauses, i.e. strong islands.

Since there are at least three strong islands where one

inherent barrier is sufficient to trigger a bounding

violation, and since all other strong islands contain an

inherent barrier on top of a barrier by inheritance (IP,

HP), then barrier by inheritance will become redundant for

bounding. Unfortunately for the unification attempts, there

is a residue, weak islands, which constitute a barrier for

bounding (Cinque:29).

In the 'Barriers' approach, following the unified notion

of inherent and inherited barriers, this residue is

expressed in the difference in the number of barriers

relevant for government (one) and for bounding (two). There

is a cost to this, not just in the redundancies, but in the

fact that expressing the difference between the two theories

in the number of barriers crossed will fail in some cases,

where even for bounding one barrier will be sufficient.

Therefore, instead of taking the notion barrier to be the

same with only the number of barriers differing in the two

theories, Cinque proposes that 'one can envisage the

possibility that the notion of barrier itself is (partially)

different for the two theories, while the number of barriers

involved remains constant (one) (29). He expresses the

difference between government and binding in that strong

islands are barriers to both government and binding, since

they are neither L-marked nor theta-marked maximal

projections. Weak islands are barriers to government but not

to binding, since they are not L-marked, but they are theta-

marked maximal projections. Dyla posited an S-deletion in

Polish in order to deal with Case assignment data which

seemed to be violating bounding nodes, while Cinque's

approach addresses the process of government and binding and

shows that Dyla's data can be more elegantly dealt with in

terms of strong islands and weak islands, and theta marking.

Cinque also takes on adjunction to VP and IP, by finding

a 'principled reason' why they behave like L-marked maximal

projections, marked by a [+V] category. XP's that are

directly theta-marked by some head are c-selected,

ultimately being s-selected, by that head, as are indirectly

theta-marked XP's. VP and IP are c-selected by I and C, even

if they are not s-selected by them. I can only take VP as a

complement, and C nothing other than IP. Even if IP is split

into Tense and AGR, C will c-select AGR P, AGR TP, T VP.

This opens up the possibility of using the notion of c-

selection in defining barrier for government--direct

selection. For lexical categories this will result in direct

s-selection, for non-lexical direct c-selection. Though I

and C are not intrinsically [+V], they are compatible with

[+V] elements. In some languages they can host verbs. They

can therefore be treated as nondistinct from [+V] categories


Cinque thus defines government and binding barriers in

the following way:

(125) Definition of barrier for government (final):

Every maximal projection that fails to be directly

selected by a category nondistinct from [+V] is a

barrier for government.

The general notion of selection can also be formulated

here, subsuming direct and indirect theta-marking.

(126) Definition of barrier for binding (final):

Every maximal projection that fails to be

(directly or indirectly selected in the canonical

direction by a category nondistinct from [+V] is a

barrier for binding.

The ineliminable difference between the notions of

barrier is then constituted by:

1. 'Direct selection' for government versus 'direct or

indirect' selection for binding.

2. 'Canonical direction' requirement, for binding alone.

The Minimalist Proaram

In "A Minimalist Program for Linguistic Theory", Noam

Chomsky pares down his EST approach to structural

descriptions as sequences, and representations at D-

structure, S-structure, PF and LF, to two interface levels.

Embedding language in performance systems (therefore

allowing expressions to be used in articulating,

interpreting, referring, etc.), SD's became a set of

instructions to these systems, providing information needed

in their functions.

There are two general types of performance systems: 1.

Articulatory-perceptual (PF) and 2. Conceptual-intentional

(LF). These two systems have their instructions provided for

them by two linguistic levels, the interface levels PF and

LF (A-P and C-I respectively).

Chomsky's other standard assumption here is that there

are two components in a language: a lexicon and a

computational system. Derivation entails choosing an item

from the lexicon and then choosing a computation that

constructs the interface representation pair from the A-P

(PF) and C-I (LF) levels. Variation is limited to the

lexicon only nonsubstantivee parts of it). Thus only one

computational system and one lexicon are necessary.

The principles-and-parameters approach is assumed by

Chomsky with UG providing 'a fixed system of principles and

a finite array of finitely-valued parameters'(1992:5). The

choice of values for the parameters is reduced by language-

specific rules. This eliminates the notion of grammatical

construction (i.e. verb phrase, relative clause, passive).

These constructions are now explained through the

interaction of UG principles with the language-specific

parameter values.

The minimalist program determines what a possible

derivation and a possible derived object is by excluding all

possibilities except those that satisfy the interface

conditions in an optimal way. Then the only possibilities

are PF and LF, and not D- and S-structure representation of

the EST framework, the minimalist approach doing away with

D- and S-structure. Chomsky further restricts the status of

PF and LF by narrowing their legitimacy. They cannot both be

legitimate unless they can be paired. A derivation converges

only if it converges at PF and LF, (convergence being taken

as yielding a legitimate SD). Convergence is determined by

independent inspection of the interface levels. Chomsky also

gets rid of the "well-formed" classification, treating the

concept of grammaticality as having no significance.

X-bar Theory, taken as being fundamental, is the form

items from the lexicon are presented in, making them

accessible to the computational system. Projections of heads

selected from the lexicon compose the X-bar structure. Basic

relations involve the head as one term and are typically

local. Chomsky treats the head-complement relation as 'more

local' (being associated with 0 relations) and the SPEC-head

relation as 'elsewhere'. Other relations are the head-head

(verb to head of NP complement (selection)) and the 'chain

link'. The minimalist program wants to dispense with the

notion of head-government, therefore it needs to account for

the relations which require head government in some other

ways. This impacts on Case Theory.

In standard treatment SPEC-head relation determined

structural Case for the subject position, with the object

position receiving Case under government (by V) including

ECM constructions. Contrary to previous treatments (like

LGB) where Case was unified under government and understood

as m-command that includes SPEC-head relations, in the

minimalist approach m-command plays no role. All these

instances of structural Case assignment are redone in

unified X-bar theoretic terms, under SPEC-head relation.

Pollock's (1989) inflection theory fulfills the need if his

clause structure is adopted:

(127) CP


C AGRs*'


Ad_ TP




SPEC-head relation (NP, AGR) then manifest both

agreement and structural Case. Case properties will be

dependent on T and V of VP characteristics. T will raise to
AGRs, which will yield [AGR T AGR] and V will raise to AGRo

yielding CAGR V AGR]. These will form the complex which

includes the 0 features of AGR (gender, number, person) and

the Case feature provided by T and V. This will result in a

symmetry between the subject and the object inflectional

systems. Chomsky here overlooks the possibility of an NP

raising to [SPEC,T] for Case assignment and then raising to
AGRs for agreement features, a possibility I also ignore.

Looking again at (127), AGR in the minimalist program

would be abstracted as a common property of adjectival

agreement and the subject-object inflectional systems. The

Case Filter, in whatever version, will call for two

occurrences of AGR if two NP's in the VP will require

structural Case. The arrangement will follow the structure

of CP in (127). If there is only one NP, then one of the two

AGR elements will be 'active', with the other inert or maybe

missing. There will be two options, depending on whether the

language in question is Nominative-Accusative or Absolutive-

Ergative. For the subject of a transitive clause
(Nominative-Accusative languages) AGRg will be the choice.

For the Absolutive-Ergative languages AGRo will be selected

with properties of the object of a transitive clause for the

NP. According to Chomsky, the distinction between the two

types of languages is a question of morphology.

The element, be it AGRs in Nom/Acc languages or AGRo in

Erg/Abs languages, will typically assign a less-marked Case

to its SPEC, SPEC being higher in the extractability

hierarchy. The less-marked Case will be compensated by

richer overt agreement, i.e. richer with Nominative and

Absolutive than with Accusative and Ergative. This is

treated by Chomsky as a tendency.

Licensing of pro will take place in the SPEC-head
relation to a strong AGRg or when it is governed by certain

verbs V*. Thus Dro is licensed only in SPEC-head relation to

[AGR aAGR], where a is [+tense] or V, AGR strong or V=V*.

This will place Pro under Case Theory, also extending to


The minimalist approach also makes the concept of

government dispensable, with head-government being

'naturally' expressed in terms of the more fundamental

notion of X-bar Theory, and antecedent-government being

taken as a property of chains, expressed in terms of c-

command and barriers. Local X-bar theoretic relations to the

head of a projection would take care of head-government and

chain-link relation would take care of antecedent-


ECM by a V is interpreted as raising of NP to the SPEC

position of the AGR P dominating V. Since Chomsky assumes

the VP-internal subject hypothesis (also followed by Ouhalla

and Ciszewska for Arabic and Polish respectively), this
raising of object to [SPEC, AGRO] and subject to [SPEC,

AGRg] will result in crossing instead of preserving nested

paths. This crossing is the only answer possible to keep the

object NP from being 'frozen in place', unable to get Case
if it cannot be raised to [SPEC, AGRo] where a trace of the

moved VP internal subject NP is. The subject NP has to cross

to [SPEC, AGRs], leaving the [SPEC, AGRo] for the object NP

to move into to get Case by checking. V would move into
AGRo. This is made possible by the notion 'shortest

movement' stated as (128) from Chomsky (1992:24).

(128) If a, 0 are in the same minimal domain, they are

equidistant from T.

Therefore, two targets of movement are equidistant if

they are in the same minimal domain. Minimal domain would

include only categories locally related to the heads.

Complement domain would be the subset of the domain

reflexively dominated by the complement of the construction,

and domain of a head (a) is the set of nodes contained in

MAX (a) that are distinct from and do not contain a.
(129) XP1


ZPI-- X'


H X2

Minimal domain of X is {UP, ZP, WP, YP, H}, the

complement domain is YP and the residue is {UP, ZP, WP, H}.

The complement domain of X is its internal domain and the

residue domain is its checking domain. This terminology

indicates that the internal domain contains elements that

are internal arguments of a, while the checking domain is

involved in checking inflectional features. In order to keep

the checking domains from being 'redefined' every time one

of its elements (if it is a nontrivial chain) is moved, the

notion is defined derivationally not representationally. In

this way in (129), head H will have no minimal, internal, or

checking domain because it has raised from another position

to form the following chain CE= (H,.... ) with subdomains

being assigned for the position that now occupies. If in

analyzing (129) we take X to be a verb, then YP is an

internal argument of X, and is the only element in the

internal domain of X. If we take X to be AGR and H is a verb

raised to AGR thus forming a chain CH=(H,t), then the

specifier (ZP) and UP and WP, being adjoined, of the

checking domain of X and the chain CH will get agreement

feature by local relation to X and will get Case features by

their local relation to CH. What Chomsky proposes is that
generally V will raise to AGRO, thus forming a chain

CHv-(V,t). The complex [V AGRo] ultimately raises to adjoin

to AGRg. The subject in [SPEC, AGRg] is in the checking

domain of AGRg and therefore will agree indirectly with V.

This again underscores Chomsky's notion in the

minimalist program that 'the basic elements of a

representation are chains', either one-member or nontrivial

(130) CH= (al,..... an) with n > 1.

Move a then can be restated as 'Form Chain'. Successive

cyclic wh-movement of arguments won't treat intermediate

steps as adjunct movement, which it would if it were a

sequence of applications of Move a.

All conditions are interface conditions, with a

linguistic expression being the optimal realization of such

interface conditions. A representation i at PF has to

satisfy the condition of Full Interpretation (FI). If there

is a failure, it's because n fails to give appropriate

instructions to the performance systems. FI is then a

convergence condition. A derivational D converges at PF if

it satisfies FI. If it does not (i.e. a [+high, +low] vowel,

the D will crash. PF is a representation of universal

phonetics, having no indication of syntactic elements and

their relations, i.e. binding, government, etc. A PF

representation t must be construed of legitimate PF objects.

Likewise at LF, each legitimate object is assumed to be a
chain CH=(al,... an), Ch being a head, an argument, a

modifier, or an operator-variable construction. A

representation X satisfies FI at LF if it consists only of

legitimate objects, i.e. a derivation X converges at LF if

it satisfies LF. If not, it crashes. Chomsky states quite

emphatically that 'expressions have the interpretations

assigned to them by the performance systems in which

language is embedded: period' (1992:38).

Nouns are drawn from the lexicon with all their

morphological features, including Case and 0 features, and

these are checked in appropriate positions, here at

specifier of AGR, which may include T or V. Then DP is

paired properly with V. Morphological features appear to be

the decisive element, i.e. variation 'follows only from

morphological specifications of lexical items'


The morphological features of Tense and AGR have two

functions: 1. They check properties of the verb that raises

to them; and 2. They check the properties of the Noun Phrase

(DP) that raises to their specifier position.

Case as an S-structure condition was postulated on the

grounds that Case features appear at PF but must be visible

at LF. Therefore Case must be present by the time the

derivation reaches S-structure. Under checking theory that

argument will collapse: the assumption is that the Case

Filter is an interface condition, a condition that all

morphological features must be checked for convergence,

somewhere. Similarly, S-structure conditions on

raising/lowering are gone. Instead, we have morphological

properties of lexical items.

Parametric variation of wh- in situ will not exist:

language differences will be reduced to morphology, i.e. the

internal morphology of wh-phrases. Chomsky assumes that

operations are 'driven by morphological necessity: certain

features must be checked in the checking domain of a head,

or the derivation will crash' (1992:45). Operator raising to

specifier of CP has to be, therefore, driven by such a

requirement. The assumption is that C may have an operator

feature (either Q or wh-) and this feature is a

morphological property of operators such as wh-. For an

appropriate C, operators are raised for feature checking to

the checking domain of C: [SPEC, CP] or adjunction to a

specifier (absorption). This will satisfy their scopal

properties. If the operator-feature of C is strong, it will

result in overt movement.

Looking at economy of representations, the principle is

FI: every symbol must receive an 'external' interpretation

by language-independent rules. Therefore, there is no need

for the Projection Principle or Theta-criterion at LF, since

they play no role here.

Derivations are legitimate only if they are necessary

for convergence. NP raising, then, is driven by the Case

Filter, assumed to apply only at LF: if the Case feature of

NP has already been checked, NP may not raise (1992:46). To

see this, consider (131).

(131) (a) There is [a a strange man] in the garden

(b) There seems to [a a strange man] [that it is

raining outside]

In (131a), a is not properly positioned for case

checking. It raises at LF, adjoining to LF-affix there and

leaving the trace t. The phrase a is now in the checking

domain of the matrix inflection, the matrix subject at LF

being [ a-there]. It is an LF word with all features

checked, but they are interpretable only in the trace t

position of the chain (a,t), the head being 'invisible'.

In (131b), on the other hand, a satisfies Case

properties internally to PP, therefore it is not allowed to

raise, and there remains free-standing. Semantically there

has no interpretation, and even though it is in a theta

position it is unable to receive a theta role. The

derivation will converge but will emerge as semi-

intelligible. There can be no 'rewriting' to make a

derivation intelligible. Derivations are driven 'by the

narrow mechanical requirement of feature-checking only, not

by a 'search for intelligibility or the like' (1992:47).

Move a will apply to an element a only if the morphological

properties of a itself are not satisfied. Movement to

benefit other elements is not allowed.

The minimalist approach has moved toward a theory of

generalized transformation (GT). Substitution has the

property that complements can be introduced cyclically,

before wh extraction, whereas adjuncts are introduced

noncyclically, adjoined to the wh-phrase after raising to

[SPEC, CP]. Again, wh movement is driven by morphological

The AGR Features Approach

Chomsky assumes AGR to be nominal, in that it has

features of person, number and gender (1986b:162). In

addition it can govern the subject. At the same time, Edit

Doron (1986) proposed the phonological realization of

feature bundles in dealing with clitics in Hebrew.

(132) dani hu ha-more

Dani he the teacher

'Dani is the teacher'

The above structure is analyzed as a nominal sentence.

In it hu 'he', pronoun 'Pron', is a clitic which is the

phonological realization of 'unattached' agreement features

that have absorbed Case' (Doron:313).

The sentence does not contain a copula since the present

tense conjugation for 'be' h.y.y. is not available in Modern

Hebrew. In past or future 'be' is obligatory.

(133) dani yihe more ba-universita

Dani will-be teacher at the university

'Dani will be a teacher at the university'

Pron cannot be analyzed as a suppletive form of 'h.y.y.'

since the following presents evidence against it.

(134) (a) *dani hu yosev ba-kafeteria leitim krovot

Dani he sitting in-the cafeteria often

(134) (b) dani haya yosev ba-kafeteria leitim krovot

Dani was sitting in-the cafeteria often

This parallels Polish in terms of Doron's unattached AGR

Features, the 'Pron" as a clitic being a phonological

realization of a feature bundle:

{[person] [number][gender][Case]}, but not an independent NP


Polish { [person] [number] [gender) ({[Case]}) features

will appear in the following paradigms (Szober:222-225):


Case iSg 2Sg 1PI 2P1

Nominative ja ty my wy

Genitive mnie, ciebie nas was

Dative anie, mi tobie, ci nam wam

Accusative mnie, mie ciebie, cie nas was

Instrumental mnan toba, nami wami

Locative (o) mnie (o) tobie (o) nas (o) was

Third person singular paradigm:

(136) masculine feminine neuter

Nominative on ona ono

Genitive jego, go, niego jej, niej jego, go, niego

Dative jemu, mu, niemu jej, niej jemu, mu, niemu

Accusative jego, go, niego jq, nia je, nie

Instrumental nim nia nim

Locative (o) nim (o) niej (o) nim

Third person plural paradigm:

(137) masculine-animate feminine-neuter-non-animate

Nominative oni one

Genitive ich, nich ich, nich

Dative im, nim im, nim

Accusative ich, nich je, nie

Instrumental nimi nimi

Locative (o) nich (o) nich

The pronouns are all marked for Case.

Polish appears to have a non-recoverable gender feature

in non-past formation for 3rd Person Sg and Pl. Hebrew

diverges in the future tense, since there is a copula

available for it. In Polish, on the other hand, future is

formed from a compound construction of 'byc' in the future +

infinitive. The 'byc' construction does not realize gender

features, the Hebrew future construction does. Compare (133)

above with the following:

(138) Dani bqdzie nauczycielem na uniwersytecie.

Dani be-FUT3Sg teacher-Inst at university

'Dani will be a teacher at the university'

(139) Dani bqdzie czyta6 w bibliotece.

Dani be-FUT3Sg read in library-Loc

'Dani will read in the library'

Looking back to data examples (9-32) in Chapter 1, no

co-indexing is possible for the clitic chain with AGR in

INFL as part of the chain: since gender is not

phonologically realized, movement will be barred. Non-

recoverability will those sentences which won't match AGR

gender features with the moved clitic, since those features

are phonologically 0. AGR=A' position, scrambled object NP

is in A' position and in Comp of CP also A'. Postulating an

A' position would bar an intervening A position head from

invoking minimality.

A possible solution to clitic movement:
(140) [[INFL EAGR [person] [number] [gender]]i

where the feature bundle is coindexed with either 'zei'

in Comp for structures (21-32), or the scrambled NP
'ksiazkei' for data (9-20). The ungrammatical examples would

be ruled out by the non-realization of the gender feature in

the feature bundle. This analysis would also support Lasnik

and Saito's observation about the asymmetry in movement from

indicative and subjunctive clauses in Polish, with an

element in Comp blocking a chain formation for co-indexing.

Cichocki's example in (110) above as well as Dyla's (42-

43) also show the scope of cliticization to be CP in Polish.

Granted, there is historical data in which AGR clitics

appeared only in the main clause, with the -1 stems in the

subordinate clause, suggesting the scope of cliticization to

encompass the main and subordinate clause (Decaux:32).

(141) Na rzekach babilonskich tam-o siedzieli

on rivers-LOC of-Babylon there=ClEmph sit-PASTPlm

jesmy i prakali, gdy wspominali Syon

be-orthoniclPL and cry-Plm when remember-Plm Zion

'We sat by the rivers of Babylon and cried when we

remembered Zion'

This example dates from 1400. On closer analysis, rather

than AGR clitic deletion, we have a transitional form of the

verb paradigm, in which the orthonic 'be' is being lost.

This coincides with the appearance of the pronominal which

then transferred the AGR clitics to the main verb. Decaux

notes many variations when the paradigm was in flux, as it

is here, since the pronominal has not appeared yet. As

recently as 1949, Decaux notices regional variations in

coordinate constructions.

(142) (a) gdzie jedlismy i pisali

where eat-PASTIP1 and write-Plm

'where we ate and wrote'

(b) gdzie-smy jedli i pisali

which is acceptable in the Warsaw dialect for the fully

realized inflectional form 'jedlismy i pisalismy'.

Lasnik and Saito posit a structure of head of CP

coindexing with the CP and blocking a moved element from co-


(143) *Co Maria mysli, ze Janek kupik?

what Maria thinks that Janek bought

'What does Maria think that Janek bought?'

This would yield the following structure:
(144) [CP ie lip Janek [yp kupir [NP till]]

Failure of theta chains would also account for the

following structures:

(145) Komu=Scie Jacek powiedziak (*:e) dali

who-DAT=C12P1 NOM say-PAST3Sgm that give-PASTPlm



'To whom Jack said you gave the book'

Clitics moving through Comp cannot leave a trace as part

of chain, and the derivation crashes.

Polish exhibits the following crossover effects (Lasnik

and Saito:158):

(146) Co Maria chce ie-by-= kupir t?

what Maria wants thatClCondCl2Sg buy-Sgm

*What does Maria want that (you)pro buy?'

(147) (a) Co Maria chce, ieby Janek kupir (t)?

what Maria wants that Janek bought

'What does Maria want that Janek bought?'

(b) *Kto Maria chce, zeby (t) kupit chleb?

who Maria wants that bought bread'

Who does Maria want that bought bread?'

Ordering of clitic elements has been applied in the

Abstract Clitic Hypothesis for English to account for the

re- 1.0. incompatibility in some constructions (Keyser and


(148) (a) We gave him money.

(b)*We regave him money.

Abstract Clitic Hypothesis (ACE) states:


a. All verbs in English have an invisible Clitic

position that may be occupied by markers such as the

one we have called dative.

b. Re-, like dative, is one such marker.

Analyzing (148) above would be done in the following

(150) we[yp[V[v give] Clitic] money]

The presence of the clitic blocks formation with re-,

since re- is a marker.

In Polish, postulating co-indexation of AGR bundle

features with elements they attach to appears to follow

Gender Feature Phonological Realization (GFPR). This

condition means that if gender is not phonologically

realized, and therefore not co-indexed, the resulting post

clitic movement construction will be *. Since Polish, unlike

English, has morphologically overt AGR features that can be

co-indexed, their presence or absence can account for the


phenomena observed in movement of Polish verbal clitics.

This must combine with the occurrence of 'jest' as a stem of

'bye'. Hence the verb clitic movement in Polish is driven by




Polish clitics exhibit a structured ordering in a given

clause. There is leeway as far as pro-clitic or en-clitic

placement with the reflexive 'sie', the Dative-Accusative

sequence of pronominal clitics and the reflexive 'siq' with

Dative clitics. Decaux (26) specifies the 'mi-si,' cluster

as invariable, whether the cluster is pre- or post-verbal as


(151) podoba mi siq ta piosenka

appeal-PRESSg me-DAT REFL this-f song-ACCSgf

'This song appeals to me'

(152) to mi sie podoba

it-NOM me-DAT REFL appeal-PRESSg

'It appeals to me'

The conditional clitic 'by' must attach to Comp if there

is an element present in it. If it is an AGR clitic cluster

it has to follow the tense/gender desinences and precede the

person/number desinences.

The matrix clause has the 'by' conditional clitic

following '-la', tense-gender-number and preceding the '-m',

person number. In the subordinate clause 'by' attaches to

the 'gdy' in CP.

Full Text
xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd

xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID EUVGOQWZU_4499XS INGEST_TIME 2013-03-12T14:05:28Z PACKAGE AA00013575_00001