42
In Journal of Slavic Linguistics Vol. 8, numbers 1-2, 2000, pp.199-237. Impersonal SIĘ in Polish: A Simplex Expression Anaphor María Luisa Rivero. Linguistics, U of Ottawa. Ottawa, Ontario K1N 6N5, Canada. [email protected] Polish resembles Italian, Slovenian, and Spanish, and differs from Bulgarian, Czech, Rumanian, and standard Serbo-Croatian in displaying an arbitrary subject use for reflexive się . Polish shares with other Slavic languages an arbitrary object use for this clitic. Arbitrary się is an indefinite pronoun of the S(implex) E(xpression) anaphor type, as in Reinhart and Reuland (1993). It signals the movement chain of a phonologically null defective NP with a human feature but no phi -features that raises as external or internal argument of the predicate to the “base- generated” się to repair its formal and referential deficiencies, by checking Case and receiving quantificational force. One use of się as SE-anaphor distinguishes Polish both from other Slavic languages and from the Romance languages: as expletive, it can transmit the thematic, binding, and control properties of the external argument to a non-selected Dative. 1. Introduction * In this paper, I propose that Impersonal się in Polish is a Simplex Expression (SE) Anaphor (R(einhart)& R(euland) 1993) (see also Everaert 1991), which provides a new perspective on an item that has attracted much attention. 1 This analysis unifies the well-known arbitrary subject- oriented use in (1) with the less known arbitrary object use in (2), which I dub Null Object Impersonal (NO-się ). It also serves to relate these two pronominal uses to anaphoric (3), which seems central. * Research for this paper has been partially supported by the Social Sciences Humanities and Research Council of Canada (Research Grant 410-97-0242). Preliminary versions were presented at Going Romance in Utrecht in December 1998, the Fundación Ortega y Gasset in Madrid in June 1999, and the Glow Summer School in Mytilene in July 1999. I am grateful to the organizers of Going Romance for inviting me, and to the various audiences for useful comments. I owe many thanks to Anna Boron and Magdalena Goledzinowska for being my main sources of information on Polish. I also thank Bob Borsley, Ewa Jaworska, Adam Przepiórkowski, Ewa Willim, and two anonymous reviewers for much additional discussion, references, and data. For help with Bulgarian, I thank Olga Arnaudova, for Rumanian Constanţa Diaconescu, for Serbo-Croatian Danijela Stojanović. Milena Sheppard and I are researching differences between Polish and Slovenian (Rivero & Sheppard 1999a-b), and I thank her for much useful discussion and information. 1 For Polish, the review in Kubiński (1987), and Dziwirek (1994), Szymanska (1998) for more recent references. I must single out for inspiration Kański’s most interesting proposals (1986) and Dziwirek (1994), although I make a different proposal from theirs. An excellent review of the extensive Romance literature is (Mendikoetxea 1992).

Impersonal SIE in Polish: a Simplex Expression …aix1.uottawa.ca/~romlab/pubs/Rivero.2000.pdfImpersonal SIĘ in Polish: A Simplex Expression Anaphor. María Luisa Rivero. Linguistics,

  • Upload
    dokhanh

  • View
    216

  • Download
    1

Embed Size (px)

Citation preview

In Journal of Slavic Linguistics Vol. 8, numbers 1-2, 2000, pp.199-237.

Impersonal SIĘ in Polish: A Simplex Expression Anaphor

María Luisa Rivero.

Linguistics, U of Ottawa. Ottawa, Ontario K1N 6N5, Canada.

[email protected]

Polish resembles Italian, Slovenian, and Spanish, and differs from Bulgarian, Czech, Rumanian, and standard Serbo-Croatian in displaying an arbitrary subject use for reflexive się. Polish shares with other Slavic languages an arbitrary object use for this clitic. Arbitrary się is an indefinite pronoun of the S(implex) E(xpression) anaphor type, as in Reinhart and Reuland (1993). It signals the movement chain of a phonologically null defective NP with a human feature but no phi-features that raises as external or internal argument of the predicate to the “base-generated” się to repair its formal and referential deficiencies, by checking Case and receiving quantificational force. One use of się as SE-anaphor distinguishes Polish both from other Slavic languages and from the Romance languages: as expletive, it can transmit the thematic, binding, and control properties of the external argument to a non-selected Dative.

1. Introduction*

In this paper, I propose that Impersonal się in Polish is a Simplex Expression (SE) Anaphor

(R(einhart)& R(euland) 1993) (see also Everaert 1991), which provides a new perspective on an

item that has attracted much attention.1 This analysis unifies the well-known arbitrary subject-

oriented use in (1) with the less known arbitrary object use in (2), which I dub Null Object

Impersonal (NO-się). It also serves to relate these two pronominal uses to anaphoric (3), which

seems central.

* Research for this paper has been partially supported by the Social Sciences Humanities and Research Council of Canada (Research Grant 410-97-0242). Preliminary versions were presented at Going Romance in Utrecht in December 1998, the Fundación Ortega y Gasset in Madrid in June 1999, and the Glow Summer School in Mytilene in July 1999. I am grateful to the organizers of Going Romance for inviting me, and to the various audiences for useful comments. I owe many thanks to Anna Boron and Magdalena Goledzinowska for being my main sources of information on Polish. I also thank Bob Borsley, Ewa Jaworska, Adam Przepiórkowski, Ewa Willim, and two anonymous reviewers for much additional discussion, references, and data. For help with Bulgarian, I thank Olga Arnaudova, for Rumanian Constanţa Diaconescu, for Serbo-Croatian Danijela Stojanović. Milena Sheppard and I are researching differences between Polish and Slovenian (Rivero & Sheppard 1999a-b), and I thank her for much useful discussion and information. 1 For Polish, the review in Kubiński (1987), and Dziwirek (1994), Szymanska (1998) for more recent references. I must single out for inspiration Kański’s most interesting proposals (1986) and Dziwirek (1994), although I make a different proposal from theirs. An excellent review of the extensive Romance literature is (Mendikoetxea 1992).

2

(1) Tę książkę czyta się z przyjemnością.

this bookACC read3S się with pleasure

“One reads this book with pleasure.”

(2) Marek się bije.

MarkNOM się fight3S

“Mark fights (others).”

(3) Janek myje się.

JohnNOM wash3S się

“John washes himself.”

If, as I argue, impersonals are SE-anaphors, they resemble reflexives as SELF-anaphors in (a)

being referentially deficient, and (b) displaying partially parallel structures, as in sect. 3. Many

have viewed se/si /się as referentially incomplete. The new idea here lies in the nature of the

deficiency, and ways to resolve it, which makes się resemble either a pronoun or an anaphor of

the traditional Binding Theory (Chomsky 1981). Applying the binding theory of R&R (1993),

the distinction is one between SE and SELF-anaphors, which relates the less general impersonal

subject/object use (the SE-anaphor) to the reflexive use (the SELF-anaphor) as general or

primary.2

Slavic and Romance display in every historical stage with a sa/ se/ si / się clitic the use in

(3), but the impersonal use in (1-2) and the middle/passive use in (4) display variation and may

be absent.

(4) Krowy doją się dobrze.

cowsNOM milk3P się well

“Cows milk well.”

I illustrate in sect. 2 that in Slavic (1) is found in Polish and Slovenian, but not in Bulgarian,

(Czech or Slovak) and standard Serbo-Croatian. Romance exhibits a similar division.

Italian/Spanish are like Polish, but Rumanian lacks this impersonal (most recently Dobrovie-

2 The uses in (1-2) seem amongst the most recent. The received view is that (1) developed after the 16th century. The NO-use in (2) is ascribed to child/colloquial registers in Polish (among others Kański 1986; Kubiński 1987), Bulgarian (Dimitrova-Vulchanova 1996: 143), and Serbo-Croatian (Progovac 1998: 108). The (colloquial) lexicon of all these languages contains sexual activity Vs with się/se in the NO-use, which I do not exemplify but mention as another sign of a young use.

3

Sorin 1998), and I show in sect. 2 that Rumanian resembles Bulgarian and Serbo-Croatian in

ways unnoticed in the past (Rivero 1998). Thus, Italian, Polish, Slovenian and Spanish belong to

one language group, and Bulgarian, Rumanian, and Serbo-Croatian to another.

The subject impersonal use seems (relatively) recent in languages that display it. The

NO-use in (2) also shows signs of youth, with variation along other lines. Slavic languages with

a sa/se/się clitic now display the NO-use under similar conditions (sect. 2). This means that the

Slavic clitic is an arbitrary pronoun (SE-anaphor) in subject and object uses in some languages

(Polish /Slovenian), or in an object use in other languages (Bulgarian, Serbo-Croatian). A similar

use seems absent from Romance, but could be found with Russian sja (Babby 1975), and

Swedish s (J. Rooryck, p.c.).

The passive/middle use also shows variation. Bulgarian, Slovenian, and S(erbo)-

C(roatian) display it, including patterns like (5) with an animate subject and a “passive” meaning

without the property reading of the strict middle (the language of examples not from Polish is

indicated).

(5) Roditelji se poštuju. SC

parentsNOM se obey3P

“Parents {are/should be} obeyed.”

By contrast, the Polish middle in (4) is considered to be restricted (Gołąb 1975: 28; Kardela

1985; Szymańska and Spiewak 1998). A passive use similar to (5) is reported to be

ungrammatical (among others Siewierska 1988). In Romance, this use(s) also shows variation.

On the one hand, Rumanian resembles Bulgarian/Serbo-Croatian with a general passive/middle

use and no impersonal use of type (1). On the other hand, some Italian dialects may lack a

middle/passive use (Cinque 1988), and, reminiscent of Polish, the Spanish equivalent of (5)

without a property reading (parents are “obeyable”) sounds objectionable: Los padres se

obedecen.

The reflexive use thus seems to know no exceptions, which is not the case for the

impersonal or the middle/passive uses. If it is assumed that impersonal się is a SE-anaphor, then

this use can be connected to the central reflexive as a SELF-anaphor, along the lines suggested in

4

Sect 3.3 In sum, important uses of se/si/się fall under principles for SE/SELF-anaphors, as in

(R&R 1993). These various uses can be related to one another under the assumption that the core

or central use of się is as a SELF-anaphor.

The paper is organized as follows. Sect. 2 establishes two language classes based on (1),

examines Slavic (2), and shows that Polish is rich in evidence for się as a defective pronoun

=SE-anaphor. Sect. 3 sketches R&R’s binding theory and provides analyses for (1-2) based on a

syntactic treatment of clitics as heads of a functional projection: the Clitic Phrase. Sect. 4 applies

this analysis to (6), which also proves of contrastive interest.

(6) Tę książke czytało mi się z przyjemnością.

this bookACC readNEU meDAT się with pleasure

Dominant reading: “I read this book with pleasure.”

It is argued that the (human) się in (1) is ‘referential’ in the sense of being an indefinite pronoun.

By contrast, the się in (6) is not a referential but an expletive pronoun that transmits certain

properties to the (unselected) Dative. In this way, się in (6) is reminiscent of there in There is a

man in the garden. This analysis captures contrasts between Polish, the other Slavic languages

with or without the impersonal use, and the Romance languages with the impersonal use.

Constructions resembling (6) are common in Slavic, which hides differences unnoticed in the

past. If Polish (6) relies on (1), as I argue, this is unavailable in languages without this

impersonal. Rather, languages like Bulgarian/Serbo-Croatian rely for the equivalent of (6) on the

middle/passive, which leads to contrasts. Contrasts between Polish and Romance have a different

source. Polish impersonal się may be an expletive pronoun, but its Romance counterpart retains

in similar cases its referential characteristics as an indefinite pronoun. Thus, Polish and Spanish

share the impersonal use in (1), and there are Spanish examples with the morphosyntax of (6),

but they contain a referential and not an expletive se.

3 I leave uncited numerous attempts at unification in the Polish and Romance literature. Within R&R’s system, Moore (1993) relates (Spanish) reflexive and passive/middles, but leaves the impersonal; Dobrovie-Sorin (1998)

5

2. Polish Impersonal się: a comparative perspective

Let us locate Polish among Slavic and Romance languages. Polish się shares with sa /se/si clitics

in the other languages the important number of uses in (7).4

(7) a. Janek (nie) ubiera się.

JohnNOM (Neg) dress3S się

“John {dresses/ does not dress} himself.”

b. Ten samochód powadzi się łatwo.

this carNOM drive3S się easily

“This car drives easily.”

c. Szklanka się rozbiła.

glassNOM się brokeFEM

“The glass broke.”

d. Maria boi się Janka.

MaryNOM fear3S się JohnGEN

“Mary is afraid of John.”

Frequent labels for them are (a) (true) reflexive/reciprocal (as in Janek i Maria kochają się. ”John

and Mary love each other.”) for (7a), (b) middle/(notional) passive for (7b), (c) anticausative/

ergative/ inchoative/ neuter/unaccusative for (7c), and (d) inherent/intrinsic/reflexiva tantum for

(7d). Similar examples could be given in Bulgarian, Czech, Serbo-Croatian, Slovak, and

Slovenian in Slavic, and in at least French, Italian, Portuguese, Rumanian, and Spanish in

Romance.

A rich morphology unifies these uses. (I) The overt NP is Nom(inative) in

affirmative/negative clauses: (7a). (II) The predicate agrees with this NP in Number if Present,

sets aside the (Romance) impersonal and unifies other uses under an (extrinsic) reflexivity marker equivalent to a SELF-anaphor. On these views, unification with impersonals could be via the reflexive as SELF-anaphor. 4 For się as clitic (Kubiński 1982, Kupść 1999). The precise nature of se/si/się as clitic does not seem to impinge on the properties of interest here. Clitics in 2P or the Wackernagel type (Slovenian), those adjacent to Tense (Italian/Spanish), those in still ill-defined positions (Polish), and those that can/cannot be first in the clause all behave alike as to the impersonal, middle/passive, and other uses discussed next.

6

(7a), (7b), and (7d), and in Number/Gender if Past, (7c). The Acc(usative) label is often

connected to the uses of się in (7). In generative grammar, this label reflects the idea that (a) the

(possibly overt) NP checks Nom under usual assumptions, and (b) się is a clitic located in a

functional layer that {“absorbs”/ licenses/ checks/ agrees with} Acc.

Another use shared by sa/se/ si /się in these languages with the exception of French is

illustrated in (8). Dubbed impersonal/indefinite/intransitive/ unaccusative/ unergative, it occurs

with intransitives of all types. The predicate is 3S if Present, and Neuter if Past, and there is no

overt NP.

(8) Tutaj się {pracuje/ śpi} sporo.

here się {work3S/ sleep3S} much

“Here people {work/sleep} a lot.”

Space limitations prevent me from discussing this construction that raises interesting problems

(see most recently (Dobrovie-Sorin 1998)). On empirical grounds, it is useful to separate the

more common (8) from the more restricted (1=9). Dobrovie-Sorin proposes to analyze (8) along

the lines of (7b-c) in languages without (1), which could distinguish (1) from (8) on theoretical

grounds.

Table 1 recapitulates uses shared by Slavic and Romance. The exceptional language is

French, which lacks the intransitive use.

TABLE 1

Uses shared by Slavic and Romance reflexive clitics

refl/recipr. mid./pass. anticaus. inherent intransitive

Yes Yes Yes Yes All but French

2.1. The Subject-oriented impersonal. We just saw that Slavic and Romance share many uses

of sa/se/si/się. However, the impersonal in (1), which may be called indefinite, (pro)-nominative,

nominativeless, subjective, and [- arg] (Cinque 1988), separates the languages that share the uses

in Table 1 into two groups, with Polish and Spanish in one and Bulgarian and Rumanian in the

other. This use is absent in French as well. Consider its morphology in (9), where (9a) repeats

(1).

7

(9) a. Tę książkę czyta się z przyjemnością.

this bookACC read3S się with pleasure

“One reads this book with pleasure.”

b. Tej książki nie czyta się z przyjemnością.

this bookGEN Neg read3S się with pleasure.

“One does not read this book with pleasure.”

c. Tę książkę czytało się z przyjemnością.

this bookACC readNEU się with pleasure

“One read this book with pleasure.”

d. Tej książki nie czytało się z przyjemnością.

this bookGEN Neg readNEU się with pleasure

“One did not read this book with pleasure.”

The overt NP is Acc in affirmative clauses, (9a) and (9c), and Gen(itive) in negative clauses: (9b)

and (9d). The Nom label often associated to this use indicates in generative grammar that (a) się

checks a Nom feature (see sect. 3 for an implementation of this idea), and (b) NP checks

Acc/Gen. V is always 3S if Present, (9a-b-c), and consistently Neu(ter) if Past, (9d), and so does

not agree with the NP. The morphological marking on the verb is often seen as a default option

(Kipka 1989, among others), but the more precise claim here is that T(ense) is void of all phi-

features (Number, Person, Gender) (also Dziwirek 1990).

Sl(ovenian) is similar to Polish, as (10) illustrates with a Plural NP. An apparent syntactic

difference is the 3S Past auxiliary with the NEU V: (10c).

(10) a. Starše se uboga. Sl

parentsACC se obey3S

“One obeys parents.”

b. Staršev se ne uboga.

parentsGEN se Neg obey3S

“One does not obey parents.”

c. Starše se je ubogalo.

parentsACC se be3S obeyedNEU

“One (has) obeyed parents.”

8

However, (11) shows that Polish is similar in tenses with auxiliaries, as in Conditionals. In (11a-

b), Participle incorporates to Aux as in (Borsley and Rivero 1994); (11c-d) are more literary

patterns without Incorporation.

(11) a. Tę książkę czytało-by się z przyjemnością.

this bookACC readNEU-would3S się with pleasure

b. Tej książki nie czytało-by się z przyjemnością.

this bookGEN Neg readNEU-would3S się with pleasure.

c. Tę książkę by się czytało z przyjemnością.

this bookACC would3S się read NEU with pleasure

d. Tej książki by się nie czytało z przyjemnością.

this bookGEN would3S się Neg read NEU with pleasure

“One {a.c. would/b.d. would not} read this book with pleasure.”

Thus, morphology distinguishes the impersonal from most notably the passive/middle, which has

in Polish/Sl a Nom NP, and a predicate agreeing in phi-features (Number/ Person if Present, and

Number/ Gender if Past): (7b) and (12). The Sl Past auxiliary in the middle/passive agrees in

Number/ Person with the NP, as it does in Polish conditionals: (12b) and (13).

(12) a. Starši se (ne) ubogajo. Sl

parentsNOM se (neg) obey3P

“Parents are (not) obeyed.”

b. Starši so se ubogali.

parentsNOM be3P se obeyedMASC

“Parents were obeyed.”

(13) Ten samochód powadził-by się łatwo.

this carNOM driveMASC-would 3S się easily

“This car would drive easily.”

9

Morphology in Polish/Sl clearly identifies the impersonal construction, which is not the only

aspect that characterizes it. Się/se in this use can also bind reflexive anaphors, including the

possessives in (14c) (Siewierska 1988) and (14d).

(14) a. Teraz się myśli tylko o sobie.

now się think3S only of oneselfLOC

b. Sedaj se misli samo na sebe. Sl

now se think3S only of oneselfACC

“Now one thinks only of oneself.” (14a-b)

c. Swoich przyjaciół tak się nie traktuje.

OneselfGEN friendsGEN so się Neg treat3S

d. Svojih prijateljev se tako ne tretira. Sl

OneselfGEN friendsGEN se so Neg treat3S

“One does not treat one’s friends like that.” (14c-d)

If binders are present in the syntax, then morphology, syntax, and semantics conspire to support

the idea that the Gen NP is the object, and się/se signals a syntactically present Nom NP binding

the anaphor. Thus, these are “active” sentences. For binding in Polish see (Reinders-Machowska

1991; R&R 1991; also Kardela 1981, 1985).

Another property of impersonal się is control into adjuncts (gerunds (Dyła 1983);

part(icipials) (Dziwirek 1994)): (15a). For some (Dobrovie-Sorin 1998 and references), this type

of control may be by implicit arguments. In addition to adjuncts, however, się controls into

complements, including passives that to my knowledge have gone unnoticed: (15b). If this type

of control is by an argument present in the syntax (Jaeggli 1986), in (15b) impersonal się signals

such a controller (i.e. the external argument of chce).

(15) a. Tę książkę czytało się siedząc przy kominku

this bookACC readNEUT się sitPART by fireplace

“One read this book sitting by the fire.” (Dziwirek 1994)

b. Chce się być {admirowanym/lubianym}.

Want3S się to.be {admiredINS/likedINS}

10

“One wants to be {admired/liked}.”

Impersonal się is used with many predicate classes5, and thus can associate with various

(subject) Th-roles (Slovenian se seems more restricted). Three classes are of morphological

interest here: (a) copulatives, (16a), (b) passives, (16b), and (c) modals illustrated in (20).

(16) a. Jest się {wysokim/ szczęśliwym}.

be.3S się {tallINS/ happyINS}

“One is tall/ happy.”

b. Bywa się karanym przez przyjaciół.

be3S się punishedINS by friends

“From time to time one is punished by friends.”

Polish copulatives and Romance copulatives/ passives are seen as theoretically significant

(Kański 1986, Cinque 1988 and their references). In my view, the Polish patterns bear additional

interest due to morphology. Adjectives/participles must in these instances be Inst(rumental), not

Nom: *Jest się wysoki for (16a) is ungrammatical. Inst Case is the normal case for predicative

nominals, but not for adjectives. I contend that the obligatory Inst with się indicates here that the

(empty) basic/derived subject is human, arbitrary in reference, but otherwise unspecified: i.e., the

familiar features of impersonal se/ si/ się. In my view, Inst signals a deficient but not totally void

of content subject, and this is supported by the Case connected to arbitrary PRO, which must also

occur with Inst: (17). By contrast, overt subjects and ordinary null subjects (little pro) occur with

Nom adjectives/participles: (18-19).

(17) Być {lubianym/*lubiany} jest miło.

beINF {lovedINST/ *lovedNOM} is niceNEU

“It is nice to be loved.”

(18) (Janek) {jest / był} wysoki.

(JohnNOM) {be 3S/ wasMASC} tall NOM

5 Some uses of psychological Vs exclude the impersonal: (i) from Rozwadowska as cited in (Dziwirek 1994) is deviant when understood in a non-volitional sense, which seems due to lexical structure. (i) *Fascynuje się profesorów swoją wiedzą. fascinate3S się professorsACC Poss knowledge “One fascinates professors with one’s knowledge.”

11

“John/ (he) {is/was} tall.”

(19) (Maria) jest kochana.

(MaryNOM) be 3S lovedFEM

“Mary/she is loved.”

Modals viewed as raising Vs provide support for the same conclusion. Kański

(1986:182) notes that the impersonal is grammatical with Modals that take Nom subjects, powin-

in (20), and impossible with those that do not, trzeba in (21).

(20) a. Powinno się być łysym.

should NEU się be INF baldINST

“One should be bald.”

b. Premier powinien być łysy.

prime.ministerNOM should beINF baldNOM

“The prime minister should be bald.”

(21) a. Trzeba pracować. (* Trzeba się pracować).

must work INF

“One must work.”

b. * Mania trzeba pracować.

Mary NOM must work

In Kański’s view, (20) supports that się is Nom, and I concur. In addition, note that the

Adjective is Inst in (20a), and Nom in (20b). If powin- is a “raising” V (i.e., has an EPP feature),

the argument in (20) is merged in the complement and {raises = is copied} to check the feature

in the matrix. Then, Inst on the Adjective can be interpreted as a signal that the lower subject in

(20a) is incomplete (i.e. it lacks phi-features). Nom on the Adjective is then a signal that the

lower subject contains a complete set of phi-features. On this view, impersonal się is Nom and

while it lacks phi-features, it has a human feature (=arb) (and see sect. 3 for a precise account of

how się can be Nom).

A similar argument can be based on sam “by oneself, alone”, inspired by Kardela (1985:

57), but based on the different constructions in (22).

(22) a. (Marysia) powinna iść tam {sama/*samej}

12

(MaryNOM) should FEM go there {herself NOM/herself DAT}

“{Mary/ she} should go there all by herself.”

b. Powinno się iść tam {samemu/*sam}

should NEU się go there {himself DAT/himselfNOM}

“One should go there all by oneself.”

Overt NPs (Marysia) or ordinary null pronouns (pro) cooccur with Nom sam(a) in the embedded

clause: (22a); with the impersonal in (22b) this item must be Dative. As before, I attribute this

difference in otherwise identical structures to the feature content of the (raised) NP. In (20a), the

raised NP is a (referential) argument of V with a “complete” set of phi-features. In (20b), the

raised category is an “incomplete” or defective argument of V: it has a human feature, but as

before, lacks phi-features.

E. Willim notes that impersonal się can be used with feminine adjectives, as when a

woman speaks (also Dziwirek (1994); Kański (1986) holds the view that women always use a

masculine impersonal). Spanish se behaves like Polish się (the feminine may be used), which

could indicate that certain phi-features are licensed in context, overriding the default setting of

the impersonal, which is masculine singular both in Polish and Spanish.

Many have noted that impersonal se/si/się must denote a human (or personified animals),

similar to arbitrary PRO. The above arguments provide morphological support for the claim that

the human denotation of the impersonal, its usual interpretation, must be encoded formally, and

is not simply pragmatic (contra Kański (1986)). As to interpretation, Siewierska (1988) notes

that Gniazda buduje się wysoko / nests.Acc build.3S się high up/ “One builds nests high up”

refers to people building nests, and not to birds (except when giving advice or speaking to birds,

which personifies them). This is a common observation that applies to intransitives and is valid

for all the languages with the impersonal use. For instance, with the Sp. verb amanecer “to

dawn”, Hoy amanece bien -Lit. Today (it) dawns well- means that the day that begins looks nice

(it is not raining). With impersonal se in Hoy se amanece bien, humans must be involved, as in

“Today people/you/we wake up well (they/you/we are in a good mood)”.

In sum, morphological, syntactic, and semantic reasons support the hypothesis that in Polish the

Nom impersonal is syntactically present, has human denotation but is void of phi-features, so it

usually cooccurs with default morphology. However, this does not make it incompatible with

13

non-default morphology, such as feminine gender. By contrast with this view, Kański (1986,

1992) sees the impersonal as a free variable void of all content, without effect on the

interpretation of a (subjectless) VP. However, this idea cannot be right, and the traditional view

that impersonal się is human must be maintained. Another construction with impersonal się that

supports the view that this item formally encodes some interpretable content that restricts its

denotation is with the Dative, as in (6) Tę książke czytało mi się z przyjemnością. “I read this

book with pleasure.” I argue in sect. 4 that this construction involves a pronoun that functions as

an expletive with an animate feature as part of its interpretable content. The Dative phrase is the

associate of this expletive się and shares this feature because there is an identity requirement

between the two.

It has often been mentioned that się may include speaker or hearer, and be equivalent to

“someone” or to “they”/“everyone” in past, present, or future sentences. There are two semantic

views on how to capture such readings. On the one hand, Kański (1986) considers these

quantificational readings contextual variants, and Cinque (1988: 546 (48)) holds similar views

about Italian. On the other hand, Chierchia (1995) develops an account with the impersonal as an

indefinite pronoun. This indefinite pronoun has intrinsic existential force, but can optionally be

disclosed by adverbs of quantification that provide it with some quantificational variability. In

simple terms, the impersonal is similar to “someone”, but in the presence of an adverb such as

“always” it may be equivalent to “all”, and in the presence of “seldom” it may be equivalent to

“few”, and so on and so forth. This last view captures the properties of Polish as shown in

(Rivero and Sheppard 1999b), so I adopt it here. In sum, subject-oriented impersonal się stands

for a defective argument denoting a human-like participant of the event described by the

predicate, similar to arbitrary PRO. It seems to lack phi-features. However, extending

Chierchia’s views on Italian to Polish, the impersonal is an indefinite pronoun that bears

quantificational force in addition to its human feature.

The impersonal in Romance is rather similar, and linguists working on Polish are aware

of this fact unmentioned in work on Romance. Here I highlight some of the properties discussed

in the literature that illustrate syntactic parallelism. (23a) shows via It(alian) that overt NPs lack

case, but the predicate is consistently 3S (showing an absence of phi- features). By contrast, the

middle/ passive in (23b) displays a V that agrees in phi-features with the NP. Another interesting

14

property of Italian is illustrated by si-impersonal sentences with a 3S auxiliary and a past

participle with plural morphology, as in (23c).

(23) a. Si mangia le mele. It

si eat3S the apples

“One eats the apples.”

b. Si mangiano le mele.

si eat3P the apples

“The apples are eaten.”

c. Si è venuti

si be3S arrivedMASC.P

“People have come.”

Example (23c) illustrates that Italian impersonal si is human, masculine, and plural, so shares the

morphological agreement pattern of arbitrary PRO in this language, which is plural. The

similarity resides in the fact that Polish arbitrary PRO is human, preferably masculine, singular,

and instrumental, and so is impersonal się.

In Sp(anish) a accompanies definite/specific object NPs, which partially compensates for

the absence of morphology on N, as in (24).

(24) En esta escuela se castiga a los alumnos. Sp

in this school se punish3S a the students

“In this school one punishes the students.”

Acc clitics in (25) indicate that that si/se connects to Nom; thus, the impersonal is considered, as

in Polish, “active”. Polish Acc clitics could be used to make the same point.

(25) a. (Le materie letterarie) le si studia in questa università. It

(the humanities) them si study3S in this university

“The humanities, one studies them in this university.” (Belletti 1982)

b. En esta escuela se los castiga. Sp

15

in this school se themACC punish3S

“In this school one punishes them.”

The Romance impersonal occurs with many predicates (copulas, passives, control Vs, and

modals) with a variety of subject Th-roles, as in Sp (26):

(26) a. Nunca se es feliz. Sp

never se be3S happy

“One is never happy.”

b. A menudo se es traicionado por malos amigos.

often se be3S betrayed by bad friends

“One is often betrayed by bad friends.”

c. Siempre se quiere ser admirado.

always se want 3S beINF admired

“One always wants to be admired.”

d. Se {puede/debe} ser admirado.

se {can 3S/must3S } be INF admired

“One can/must be admired.”

Slovenian, Italian and Spanish, then, closely resemble Polish, and belong to the group

with the subject-oriented impersonal. In Bulgarian, Serbo-Croatian, and Rumanian these

impersonal patterns are absent. Subject to some variation, (Standard) SC, which has

morphological Case on N, displays sentences consistent only with the middle/passive, not what I

call impersonal ones: (27a) vs. (27b). The NP in (27) must for many be Nom, and the predicate

must agree in phi-features with this NP, in contrast most notably with Sl (10), a language deemed

close to SC.

(27) a. Roditelji se (ne) poštuju. SC

parentsNOM se (Neg) obey3S

“Parents are (not) obeyed”

16

b. *Roditelje se poštuje.

parentsACC se obey3S

Intended: “One obeys parents”

Bl/Rum(anian) resemble It/Sp in having no case on N. However, NP and predicate must

necessarily agree in these languages: (28-29).

(28) Se {scriu/ *scrie} scrisorile. Rum

se {write3P/ *write3S} letters.the

“The letters are written.”

(29) {Jadat/*Jade} se jabalkite. Bl

{eat3P/*eat 3S } se apples.the

“The apples are eaten.”

As to Acc clitics, constructions parallel to (25) are ungrammatical in the group without

impersonal se: (30a) (Dobrovie-Sorin 1998) and (30b) (SC is similar).

(30) a. *(Stiinţele umane) le se predă in această universitate Rum

(humanities) them se study3S in this university

b. *Literaturni kursove {gi se/se gi} izučava v tozi universitet.

(humanities) them se study 3S in this university Bl

A final difference illustrated with Bl arises with predicate classes. Adjectives (singular or plural)

and passives, for instance, are ungrammatical with se in this type, in contrast with Sp. (26a-b) or

Polish examples (Th-roles for se are also restricted).

(31) a. *Ne se e nikoga dovolni. Bl

Neg se be3S never happyP

b. * Često se e predaden ot prijateli.

17

often se be3S betrayedS by friends

Dobrovie-Sorin (1998) illustrates other contrasts within Romance, including that middle/passive

se cannot be a controller, unlike the impersonal. Speakers of Bl/ SC/ Czech/Slovak will find that

if other similarities obtain, they are with Rum.

To conclude, properties of the Polish subject-oriented impersonal się include:

(a) The same clitic characteristics as non-tonic pronouns.

(b) A rich morphology on overt NP, and predicate without phi-features.

(c) A syntactically present argument with a human feature, arbitrary reference, and

quantificational variability, but also unspecified for phi-features.

(d) Binding of reflexive anaphors, including possessives.

(e) Semantic (Theta) roles otherwise held by an (overt) Nom subject.

(f) A syntactically present controller for complement clauses (and adjuncts).

(g) Use with many classes of predicates (and semantic roles). Passive Vs, copulas

with adjectives, and modals are of particular morphological interest.

There are two classes of languages with respect to a similar impersonal clitic. On the one

hand, Slovenian, Italian, and Spanish share this use with Polish, but offer less rich a picture.

Slovenian resembles Polish in morphology, anaphor binding, and control properties, but

predicate classes are limited with the impersonal, leading to a more restricted syntactic/ semantic

distribution. The main difference is that Slovenian impersonal se is excluded with adjectives, and

in passive sentences (Rivero and Sheppard 1999a–b). Italian and Spanish offer poorer

morphological clues, and a wide distribution based on predicate classes, which enriches the

semantics of the impersonal. Overall, then, Polish offers the most robust morphological,

syntactic, and semantic evidence for the subject-oriented use signaling an arbitrary human as

syntactic argument of the predicate.

On the other hand Bulgarian, Rumanian, and (standard) Serbo-Croatian are

representatives of the second group, which offers a different picture. In these languages, se

displays many uses that characterize Polish się. However, the subject-oriented impersonal use

with transitive Vs displaying binding and control properties is missing, which leads to important

morphological, syntactic, and semantic differences with Polish.

18

Table 2 summarizes the distribution of the subject-oriented impersonal clitic. In

comparison to Table 1, it shows that crosslinguistically the impersonal with transitive Vs and

other symptoms is not as common as other uses. The use with intransitive Vs is much more

common than the use with transitive Vs, which can control and bind.

TABLE 2

The subject impersonal use

Polish

Slovenian

+

+

Slavic

Bulgarian

Czech

Serbo-Croatian

Italian

Spanish

+

+

Romance

French

Rumanian

2.2. The Null Object Impersonal (NO) się.

A prototypical instance of NO-się is (2) repeated now as (32). Kański (1986) calls it right-

oriented middle, and Kubiński (1987) antipassive.

(32) Marek się bije.

19

MarkNOM się fights

“Mark fights (others).”

I consider it the object counterpart of się in 2.1: an arbitrary deficient pronoun with a human

feature and Acc (or, in negative sentences Gen, identical to Acc in tonic/clitic pronouns). This

się is comparable to Rizzi’s (1986) arbitrary little pro in Italian Questo conduce (la gente) alla

seguente conclusione “This leads (people) to the following conclusion.” I repeat, się signals a

structurally realized/ syntactically present null object as internal argument, checks Acc/Gen, has

a human feature ensuring referential character, but is otherwise incomplete (no phi-features).

Subject-oriented impersonals may include the speaker (others=me), which also applies to

NO-się as in (33) (Kański 1986:195). It is significant that (33b) as an imperative depicts an

activity involving a human patient, and does not assign an inherent property to the syntactic

subject as agent (i.e. it does not have a middle-like interpretation).

(33) a. Mania się przezywa.

MaryNOM się name.call3S

“Mary calls others names.”

b. Nie pchaj się, pan!

neg pushIMP.2S się, man

“Stop pushing others, sir!”

In my view, (32-33) display two syntactic arguments (i.e. they are “active”): the Nom NP, Mania

in (33a), and a human indicated by się as Acc. No identity relation exists between the two, so się

is free like a (clitic) pronoun. The two semantic properties of sentences with this pronominal się

are, then, a) a patient with human content indicated by się, and b) an agent that is not assigned an

inherent property, which is the Nom NP. Thus, this Polish construction is not what is sometimes

called a “deponent middle”: an active sentence with a permanent property assigned to the

syntactic subject as Agent, and no syntactic object.

My proposal that the use of się in (32-33) is pronominal and identifies an argument of the

predicate contrasts with Kański’s view (1986). He treats NO-się in a way similar to the subject

20

impersonal: as a free variable void of content in what he calls a right oriented middle (i.e. the

reverse of the usual middle). For Kański, this się does not identify a syntactic argument with

human denotation, which is in contrast to what I propose. Kański’s analysis may serve for

Russian -sja used to describe a prototypical characteristic of a syntactic subject as in Nasa sobaka

(ne) kusaetsja “Our dog (does not) bite”. The contrast between Polish and Russian, if correct,

seems reminiscent of Italian vs. English (Rizzi 1986). Polish NO- się indicates a syntactic

argument similar to Italian object little pro, and Russian -sja could signal its absence as in This

leads (one) to the following conclusion. On this view, Russian may involve a lexical operation

suppressing the internal argument of a predicate with an external argument, which is the reverse

of what Grimshaw (1982) or Babby (1998), among others, propose for the ordinary

middle/passive. Other aspects supporting the pronominal character of the use of się in Polish

(32-33) are (a) that the pronominal use as subject impersonal in 2.1. is robust, while the ordinary

middle is restricted, (b) that the register of this use is colloquial /childish, in contrast with the

(restricted) middle/passive, and (c) that the subject and object uses are the most recent.

Verbs with the use in (32-33) are not numerous, also reminiscent of Italian arbitrary little

pro as object. Precise inventories may vary among speakers. One class in fn. 1 not illustrated

suggests for adult language a colloquial register (i.e. recent). The illustrated class comprises Vs

describing actions appropriate for children consistent with its child language status, another sign

of youth (kopać “kick”, pchać “push”, bić “beat”, calować “kiss”, obejmować/ przytulić

“hug”, drapać “scratch”). I find it significant that Vs describing prototypical animal activities

such as butt (said of cows) are listed with -sja in Russian grammars but do not seem to display

the NO-się use in Polish, another difference suggestive of a pronominal use.

Examples from Bl (Krapova 1998), SC and Sl with “Polish” characteristics are given in

(34-36). These examples may display reflexive, reciprocal, and other readings, but only the NO-

reading is indicated through the use of others.

(34) Ivan iska decata da se bijat. Bl

Ivan want3S children.the da se spank3P

“Ivan wants the children to spank others.”

(35) a. Deca se grle. SC

children se hug3P

21

“The children are hugging others.” (Progovac 1998)

b. On se ujeda.

heNOM se bite3S

“He is biting others (including me).”

(36) Pokazi kako se poljubljaš. Sl

show how se kiss2S

“Show me how you kiss others.”

A “deponent middle” interpretation seems here implausible. A natural reading for (35a), for

instance, is not that children are inveterate huggers, which would correspond to the deponent

middle reading. D. Stojanović reports that (childish) (35b) may be corrected by adults with On

me ujeda “He is biting me.” This suggests a use seen as pronominal in a language that lacks the

subject-oriented impersonal in 2.1 with transitives. I thus propose that Bulgarian, Slovenian, and

Serbo-Croatian are similar to Polish in having a (young) NO-se use that corresponds to an

indefinite pronoun.

Aspect interacts with subject-oriented impersonals (Cinque 1988), which is also true of

NO się/se, as in (37). D. Stojanović notes that in SC (37a), imperfective Aspect/Aktionsart

brings to light the universal-like reading of this use (i.e. others) suppressed with perfective Vs.

(37b) illustrates that Polish is parallel, (and so is Bl, not illustrated).

(37) a. Kad je bio mali, Ivan se strasno {grebao/ ogrebao}. SC

when be3S been little, Ivan se horribly {scratchIMP/scratchPERF}

b. Kiedy był mały, Marek się strasznie {drapał/podrapał}.

when was little, Marek się horribly {scratchIMP/scratchPERF}

IMP: “When little, {Ivan/Marek} would scratch himself/others horribly.”

PERF: “When little, {Ivan/Marek} scratched himself horribly (once).”

A similar effect is seen in (38). With the imperfective V, the (childish) NO-reading of (33a) is

maintained: (38a). With the perfective V in (38b), this reading is absent, and the V interpreted

reflexively acquires a different meaning.

22

(38) Kiedy była mała, Marysia się {a. przezywała/ b. przezwała}

when was little, M. się { call.names a.IMP/b.PERF}

IMP: “When she was little, Mary (always) called others names.”

PERF: “When she was little, Mary gave herself another name.”

Unlike the imperfective version, the perfective version in (38) may sound in need of a secondary

predicate: Kiedy była mała,Marysia się przezwała Ksiezniczka “When she was little, M. call

herself Princess.”

Recall that Chierchia (1995) suggests that the subject impersonal is an indefinite pronoun

that can derive its quantificational force from adverbs of quantification. Applying this idea to the

object use in Slavic, examples (37–38) suggest that imperfective aspect may be similar to an

adverb of quantification, which is not a new idea. That is, we can think of imperfectivity as

semantically equivalent to a generic always that provides (universal-like) quantificational force

to object się as indefinite in the way Chierchia proposes for subject uses. On this view, the

object pronoun use of się may display an interpretation close to a deponent middle with

permanent property for the Agent, and this is due to the universal-like force of the object as

Theme/Patient. In simple terms, in (37b) imperfectivity as an adverb of quantification means that

Marek was scratching nearly everybody, which comes close to saying that Marek had the

characteristic of scratching, which is the middle deponent interpretation.

In sum, NO-się (1) is human, but lacks phi-features or is referentially incomplete. (2) It

may display existential/ universal readings including the speaker. (3) It is sensitive to

aspect/generic time reference, or may derive quantificational force from the equivalent of an

adverb of quantification: imperfectivity. (4) It is Acc/Gen, not Nom. Slavic languages with clitic

se resemble Polish; they display a NO-se with these properties felt to belong to child/ colloquial

registers. Russian -sja may differ from NO-się/se ; the contrast, if correct, seems like the English

implicit object vs. Italian little pro as structural object.

Table 3 summarizes the distribution of the uses for the reflexive clitic in sections 1 and 2. Each use is provided with an English equivalent to facilitate comparison:

23

TABLE 3

Distribution of the uses for the reflexive clitic Uses of sa/ se/si/się Slavic Romance

Reflexive /reciprocal

•John washes

•The children hug each other

+

+

Middle/Passive

•This car drives well

•The house was built today

+

+

Anticausative

•The branch broke

+

+

Inherent

•John is afraid of Mary

+

+

Intransitive

•One works/dies

+

+

(French excepted)

Subject impersonal

•One punishes these children

•One wants to be admired

I

+

(Pol)

II

(Bulg)

I

+

(Span)

II

(Rum)

Object impersonal

•The children scratch others

+

3. Subject/Object Impersonal się as SE-Anaphor.

In this section I argue that the Nom/{Acc/Gen} Impersonal of sect. 2 is a S(implex) E(xpression)

anaphor, which sheds light on how clitics can be incorporated into a binding theory not intended

for them. The idea that arbitrary się is a SE-anaphor is implemented by combining (a) “base

generation” of this clitic and movement to it of a defective (null) NP from inside the VP, with (b)

(the spirit of) R&R’s binding theory. The basic claim is that the chain formed by the null

24

defective NP that raises to the impersonal reflexive clitic in Polish is formally equivalent to the

movement chain formed by the Dutch SE-anaphor zich that raises to an overt subject.

Let us first introduce the general proposal by looking at Tę książke czytało się “One

read this book ” in (9), with a subject-oriented się. I adopt the standard assumption that in this

sentence the verb heads the VP, which contains two NP arguments. The ‘arbitrary’ NP as

external argument or NP1 is equivalent to a pronoun without phonetic content that is defective: it

has a human feature, (structural) nominative Case, but no phi-features (no gender, number, or

person). The internal argument or NP2 is the overt accusative object this book. In addition to the

VP, the structure contains two projections each headed by a functional category. One projection

is the familiar T(ense) P(hrase) headed by T(ense) (in this case Past). It is defective in that it also

lacks a specification for phi-features (no gender, number, or person). TP takes VP as its

complement, and a standard assumption is that V checks features against T. Given that T is

defective, V is Neuter or without phi-features: czytało. The other functional projection dubbed

Cl(itic) P(hrase) is headed by się. A standard assumption is that ClP takes TP as its complement;

hence się is structurally higher than T. However, nothing hinges on this in our analysis, so it

could also be that ClP is the complement of TP. Given these assumptions, the basic skeleton of

(9) minus the adverb is as in (39). No Agreement Projections are postulated.

(39) ...[CLP [CLsię] [TP [T Past] [VP NP1 V NP2 ]]]

Within the structure in (39), NP1 must satisfy formal and referential needs and repair

deficiencies by raising to the ClP with się. That is, NP1 checks features against się, which

makes the null category contrast with the ordinary null subject pronoun little pro, which is often

assumed to check features with a non-defective T in TP. In (39), both NP1 and T lack phi-

features, so adapting ideas on feature checking in (Chomsky 1998), I propose that they cannot

establish an appropriate checking relation with each other. The required checking relation is

established between NP1 and the clitic, when the first moves to the second. These categories are

each equipped with a structural Case feature, which allows them to match for checking. The

raising of NP1 bypassing TP to the ‘based-generated’ clitic się to check Case results in a well-

formed chain interpretable at LF. This chain is comparable to a licit SE-anaphor chain in R&R’s

terms: i.e. a pronoun that repairs its referential deficiency by raising to a higher category. This

25

LF-chain can serve as appropriate input for later levels of semantic interpretation, as in the

semantic analysis of Chierchia (1995), who treats impersonal subjects of this type as indefinite

pronouns with intrinsic existential force. Our syntactic analysis is compatible with Chierchia’s

semantic proposal for a variety of reasons given in informal terms. One, we see next that the

movement chain proposed here counts as a pronoun in terms of the binding principles of R&R.

Two, this chain contains as foot a trace that can count as the variable for the required

quantificational force. Three, the chain also contains as head a combination of a raised NP with a

human feature and a clitic się that can be seen as the operator that provides the existential force.

From this perspective, the null NP that raises to się by LF satisfies a formal need, which is the

checking of a Case feature, and a referential need, which is to receive quantificational force in a

way that allows it to be interpreted appropriately. Syntactically, the chain is equivalent to a SE-

anaphor of a new type, and semantically it is an indefinite pronoun for humans.

The analysis just proposed for the arbitrary subject can easily extend to the arbitrary

object się, in (33) Mania się przezywa. “Mary calls others names.”. In this second case, NP1 in

(39) stands for the overt nominative subject Mania that checks features against T in TP. T is not

defective, and contains a complete set of phi-features. NP2 is the null item with the human

feature, (structural) Accusative Case, but no phi-features. It raises to the ClP with się to check

structural case. Similar to an arbitrary subject, NP2 as object has the pronominal characteristics

of a SE-anaphor, and raises to się to form a chain that is interpretable at LF.

A last question is why się in (39) can attract for checking purposes the defective subject

NP1 in (9), or the defective object NP2 in (33). Inspired by proposals in (Chomsky 1998), I

hypothesize that się as target of the movement for these NPs contains a structural Case feature

with an unspecified value. This feature can thus be used to check any other structural Case

feature, which may be either Nominative, as on the defective external argument NP1 in (9), or

Accusative /Genitive, as on the defective NP2 in (33).

Having outlined the core aspect of the analysis, let us now look in more detail at its two

main ingredients, beginning with the base generation approach to clitics followed by R& R’s

binding theory, and in particular SE-anaphors. As to the clitic, I propose that się is directly

merged in a functional slot in the clause that is outside of the VP, which is traditionally known in

generative grammar as the base generation approach. That is, się is an entry classified in the

lexicon as a functional category. When merged into a phrase marker, it is similar to other clitic

26

pronouns in heading a Functional Projection called here Clitic Phrase (and see Sportiche 1995;

Rivero 1997 for discussion and references). In some analyses, clitics head Agreement Subject or

Agreement Object Phrases of the type in (Chomsky 1991). Mendikoetxea (1992), for instance,

proposes that the subject impersonal in Romance heads AgrSP. Here I do not relate się to AgrP

for two reasons. The main reason is that the core content of an Agr head is phi-features, and clitic

się is in my view phi-less, so it cannot represent this type of head. A less crucial reason is that

Agr has recently come under criticism (Chomsky 1995, 1998), and may not be a distinct

functional head. As agreement-like features play no role in my analysis, and się is not an

agreement head, the diagram in (39) contains no AgrP.

A prominent alternative to base-generation is “clitic movement”, which seems less

desirable for the impersonal as it poses difficulties. The movement option consists in

generating/merging the clitic within the VP, and subsequently raising it to some inflectional layer

such as ClP, with a common assumption being that the clitic is of the functional category

D(eterminer). On this view, arbitrary NP1 in (39) would correspond to DP1, and arbitrary NP2 to

DP2, and się as D would raise. This poses both general and language specific difficulties for the

analysis of the impersonal. A general difficulty is that the impersonal must then be treated as a

DP-expression, while its semantic and syntactic characteristics seem more compatible with the

category NP (and see Chierchia 1995, and Chomsky 1998 on different angles of indefinite NPs).

Another difficulty is that Polish lacks D (Willim 1998), which suggests that the morphosyntax of

clitics, (tonic) pronouns, and (tonic) reflexives/reciprocals cannot be based on this category.

“Base-generation” of CliticP as in (39) thus frees us from calling the impersonal fully referential

in the sense of a DP, or from calling Polish się a morphosyntactic D.

Another assumption I make about się as functional entry of the lexicon is that it bears a

structural Case feature that is unspecified for a value, which can be justified on morphological

grounds. The Structural cases in Polish are Nominative, and Accusative/Genitive. Clitic się has

the same morphological form for all these Cases, it lacks a Dative use, and uses for other cases.

This suggests that się is compatible with uses that involve structural Case. The idea that się has

an unspecified structural Case means that it can match and thus check the structural Case of a

phi-less arbitrary NP that is Nominative, as in the subject use in (9), or the one that is

Accusative, as in the object use in (33). In sum, się is a functional category that can attract for

checking the external or the internal argument of the predicate. Once I introduce R& R’s binding

27

theory in informal terms, and show that the impersonal use is comparable to a SE-anaphor, I will

cast these ideas in terms of (Chomsky 1998).

Now consider binding, R&R propose (1993: sect.1) to distinguish three types of

expressions called respectively Pronouns, SELF-anaphors, and SE-anaphors, using two features

called Refl(exivizing function) and R(eferential Independence). Pronouns are not reflexivizers

and contain a full specification of phi-features, which allows them to be interpreted

independently: [–Refl ; +R]. Him in John hates him is a pronoun and does not make the predicate

hate reflexive. By contrast, SELF-anaphors are referentially defective, and are reflexivizers

[+Refl; –R]. Himself in John hates himself, is a SELF-anaphor, and makes the predicate hate

reflexive. SE-anaphors are like pronouns in not being reflexivizers, and are also similar to SELF-

anaphors, as they do not have a full specification of phi-features, so cannot be interpreted

independently: [–Refl ; –R]. The content necessary for the referential interpretation of a SE-

anaphor is obtained via movement, which makes it similar to a pronoun: [+R; –Refl]. There are

no SE-anaphors in English, but the behavior of this category can be exemplified with Dutch zich.

On the one hand, SE-anaphors pattern with pronouns in that they do not make a predicate

reflexive when they occur on one of its argument positions. This is why the following Dutch

example is deviant: *Max haat zich –Max hates SE-anaphor. Thus the SE-anaphor zich contrasts

with the English SELF-anaphor above, which must reflexive-mark the predicate when it occurs

on one of its argument positions. On the other hand, zich lacks phi-features but is also like a

pronoun, and can obtain the content necessary for a referential interpretation by movement. In

Max legt het boek achter zich “Max puts the book behind him=Max”, zich as SE-anaphor

adjoins in LF to I ( = Agr) to inherit subject features (1993:659), which results in a well formed

A-chain that is +R and Case-Marked. The movement does not make the predicate put reflexive,

but entails the coindexation of zich with Max, as the argument in subject position. In this way,

zich is necessarily “subject-oriented”, since it repairs its referential deficiency by sharing the phi-

features of the subject. In sum, a SE-anaphor in R&R’s view is a defective pronoun that repairs

its deficiency by acquiring phi-features via a movement that coindexes it with the subject.

My proposal is that the impersonal use of the reflexive clitic in Slavic and

Romance is comparable to a SE-anaphor in R&R’ sense, with some interesting differences. To

see this, consider the defective argument in (39). This is a NP for reasons given above, not a DP,

and can be the external or the internal argument. In sect. 2 it was argued that it is phonetically

28

null, present in the syntax so not implicit, saturates a grid position of the predicate, and has a

human feature but no phi-features. This NP is comparable to a SE-anaphor in three ways. One,

lack of phi-features means that it does not project an argument that can be interpreted

independently. Two, movement enables it to obtain the content necessary for its interpretation.

Three, it is similar to a pronoun.

A difference between Dutch zich and the SE-anaphors of this paper is that the movement

operation to repair their referential deficiency is not the same, which has a variety of

consequences. R&R’S SE-anaphors adjoin to Inflection/Agreement to inherit the phi-features of

the subject. This makes them (a) “subject-oriented”, and (b) interpretable as ordinary personal

pronouns with a full set of phi-features. By contrast, the defective NPs as external/ internal

argument of the predicate in (39) repair their deficiency by moving to CliticP with się. A first

contrast then is that our SE-anaphors based on a clitic need not be “subject-oriented” in the sense

Dutch zich must be. In sect. 2 we saw that się can be the equivalent of a subject pronoun in the

arbitrary subject use, thus “subject-oriented” in a rather different sense from zich. We also saw

that się can function as object pronoun or be “object-oriented” but not coindexed with the

subject in the arbitrary object use, again in contrast with zich. Below I propose that these

orientations are a consequence of the Structural Case (Nom or Acc) on the defective NP as SE-

anaphor, and that no phi-features are involved in this aspect. A second contrast that arises in

view of different landings concerns the specific way in which referential deficiencies are

repaired. The SE-anaphors of Dutch obtain a full set of phi-features, so they are interpreted as

ordinary pronouns. NP1 or NP2 as SE-anaphors in (39) move to a target void of phi-features,

which, following Chierchia (1995), contains quantificational force, so they come to be

interpreted not as regular pronouns but as indefinite pronouns.

The movement to repair deficiency in the SE-anaphors behind the impersonal use of the

reflexive clitic is thus different from R&R’s operation. It nevertheless achieves the same

technical result, which is to create a LF-chain that is well-formed (and thus can serve as input for

semantic interpretation). For R& R, an A-chain that is +R and Case-Marked is well formed. In

our case, the movement of the phi-less NP in (39) up to clitic się ensures that this NP as SE-

anaphor can be considered +R. In addition, we see next that the NP that raises checks case

against się, so the chain is Case-Marked. The core idea, then, is that movement (up to the clitic)

of the defective NP on an argument position of the predicate results in a chain that is

29

interpretable at LF. This is the similarity between the SE-anaphors of this paper involving a clitic

as target for movement and the Dutch SE-anaphors, which do not involve a clitic as target.

The third aspect that makes impersonal się like a SE-anaphor is the Binding Theory,

highlighting its pronominal character. SE-anaphors pattern with pronouns in that they occur on

an argument position of a predicate without making this predicate reflexive. On this view too, the

defective NP or SE-anaphor that serves as double for impersonal się patterns like a pronoun. It

fits in unproblematic ways traditional principle B (Chomsky 1981): a pronoun is free in its

governing category. Alternatively, following R&R, it does not make the predicate reflexive when

it occurs on one of its argument positions (external/ internal). R&R propose no principle to

prevent a SE-anaphor from being free, and we saw above that Dutch zich is coindexed with the

subject not because of the binding theory, but due to the movement that makes it inherit phi-

features. The two impersonal uses of clitic się discussed in sect. 2 behave like pronouns rather

clearly, as they are, intuitively speaking “antecedentless” expressions. These two uses, then, are

pronominal as they indicate a SE-anaphor, which does not reflexivize the predicate (–Refl), and

repairs referential deficiency by moving to się (or, in R&R’s terms, changes its value from [–R]

to [+R]).

The SE-anaphors in this paper differ from Dutch SE-anaphors in that they do not repair

their deficiency by inheriting through movement the phi-features of an independent argument of

the predicate (i.e. Max in the Dutch example). Instead, our SE-anaphors raise to a clitic that has

no phi-features, with the consequence that no coindexation independent from the binding theory

clouds their pronominal character. In sum, the chain with the phi-less NP raised to się as head

and its trace as foot may be “subject” or “object” oriented. It functions as an indefinite pronoun

that is free, with the evidence for the impersonal use of się as well behaved SE-anaphor robust.

Now let us return to the formal reason for movement of the NP in (39) to się, adapting

proposals in (Chomsky 1998) to our case. In this paper, Chomsky proposes two operations to

check formal features: MOVE and the new operation AGREE, which is more economic.

Oversimplifying, a category can MOVE to the Spec of a functional head to check features, or the

functional head can check features with a category in its complement via AGREE, without the

need for movement, and the last operation is preferred. One important question in this

framework, then, is why MOVE and not AGREE applies in the impersonal --what forces the

deficient NP as external/internal argument to leave in (39) the VP-layers of the clause, as

30

proposed. Another important question is why MOVE must target functional się. In my view,

structural Case combined with the lack of phi-features can provide a formal answer to these

questions, keeping the insight that MOVE signals the presence of a defective pronoun = SE-

anaphor that repairs its referential deficiency in the way discussed above.

Let us develop the idea that the subject Impersonal in 2.1 involves movement for Case

checking, first recalling crucial assumptions. Się (a) has an undifferentiated structural Case

feature (perhaps equivalent to an E(xtended) P(rojection) P(rinciple) feature), and (b) is located

in a CliticP that c-commands T(ense). As to this (finite) T, (c) it lacks phi-features, since as

illustrated in sect. 2.1 the predicate must be 3S/Neut. As to the NP as external argument of the

predicate in the VP, (d) it also lacks phi-features for all the reasons in 2.1. (e) It has, in addition

to the human feature, an uninterpretable structural Case feature as in (Chomsky 1998). On these

assumptions, then, the structure for the subject impersonal construction is roughly as in (40),

with a phi-less T and a phi-less NP as external argument of the VP-layer:

(40) [ CL [ T [VP NP V ...]]]

Absence of phi-features on T and NP prevents these two categories from establishing the

AGREE relation that could ensure feature checking without movement. That is, (finite) T and NP

are in a local relation and the features of one could be checked against the features of the other

under matching, which is how AGREE operates. Recall that AGREE is more economic than

MOVE. However, using Chomsky’s terms T cannot be the probe for NP, and NP cannot be the

associate of T or its active Goal for AGREE to apply. The two categories fail to share phi-

features, which is the crucial ingredient for AGREE to match features. While T has no phi-

features in need of removal, the NP has an uninterpretable Case feature that must be deleted,

which cannot be done via AGREE with T, as just explained. Thus, NP must MOVE to CL się,

which removes the offending structural Case feature under identity with the similar feature on

CL. It could be that T in (39) has an EPP feature that needs to be removed. If this is so, the NP

can stop in T on its way to CL.T has no phi-features, so this intermediate step will not remove

Case from the NP. If T lacks an EPP feature and the category has no intervention effect, then NP

can bypass it to reach CL (equivalent to long movement). In brief, absence of phi-features, a

prominent characteristic of the impersonal construction, makes the NP raise to CL to satisfy

31

formal needs (feature checking for Case). Once this formal deficiency is repaired, the chain can

be considered interpretable at LF in a way comparable to what R&R envisage for SE-anaphors in

Dutch, but with the differences discussed above.

A similar analysis can be sketched for the object Impersonal in 2.2. Chomsky (1995)

proposes that a transitive verb involves two layers of syntactic structure. The upper layer is a

light verb v that contains the subject in its projection vP. The lower projection is VP with the

lexical V and the object: [vP NP1 v [VP V NP2]]. Thus, the object that raises to się must cross

both v and T on its way to the clitic. The assumption is that absence of phi-features in the NP as

internal argument of the predicate prevents it from establishing the appropriate relation with the

light v in the vP (i.e. NP2 cannot be the active Goal for v). Again, NP2 must move up to się to

remove its Structural Case feature (ACC) (perhaps stoping at vP), and this operation renders the

chain interpretable at LF.

In sum, się with unspecified Structural Case is in CL. NPs without phi-features do not

enter into an AGREE relation with T/v, and check structural Case against się. This results in a

LF-chain interpreted as an indefinite expression with a pronominal character, which is the SE-

anaphor.

In the introduction, I noted that the reflexive use is common while the impersonal use of

the reflexive clitic is more restricted. It thus seems desirable to formally relate the less common

impersonal use to the common reflexive use, so let us see how this can be achieved. R&R

propose rather similar structures for SE and SELF-anaphors. If reflexive się is seen as a SELF-

anaphor—a clitic that doubles a null NP that is -R and +Refl—then the impersonal as SE-

anaphor can be formally related to this reflexive. R&R see anaphors as NPs with a D in their

Spec. Recall, however, that D seems unjustified for referentially deficient expressions, and that

Polish has no D. Thus, I propose that the Spec of this SE-anaphor NP, notated E(xternal)

A(rgument) in (40), can be filled to the same semantic effect by categories such as Adjectives. In

sum, I keep NP and omit D, assuming that what may look like a D+NP structure is a NP with an

AP in the specifier position. On this view, the structure of the impersonal SE-anaphor as NP

should be as in (41a): a SE-anaphor as EA and, similar to pronouns, a one-argument or non-

relational N with a human feature. EA saturates/discharges the argument in the N-grid in the way

R&R discuss. The NP raises to się for the reasons given for (39) (an alternative is just the one-

argument N, as suggested by R&R (1991) for Polish siebie). The NP for reflexive się seen as an

32

extrinsic marker of reflexivity and a local coreference anaphor could then be (41b): SE-anaphor

as EA, and a 2-argument/relational N coupled to an identity relation. EA saturates or discharges

one argument in the N-grid, and the other argument must be found elsewhere in the way

discussed by R&R. This reflexive NP also raises to się, which results in the clitic as local

anaphor, not a long-distance one. This is the well-known locality of reflexive clitics, which

usually encompasses clitic climbing as in Jan się chce umyc “John wants to wash himself.”

Tentatively, the NP for reciprocal się (always local as well) could be (41c): SE-anaphor as EA,

and a 2-argument or relational N coupled to a disjoint (non-identity) relation. Intuitively, here SE

stands for the equivalent of each, and N for other. The difference as to the internal structure of

the impersonal and the reflexive then is that the first is a non-relational N and the second a

relational N.

(41) a. NP b. NP c. NP 2 2 2

EA N EA N EA N

SE e <1> SE e <1=2> SE e <1≠2> Impersonal się Reflexive się Reciprocal się On this view, się does not alter the argument structure of the predicate and stands for an

ambiguous SE-anaphor with properties of both pronouns and anaphors. When it displays the

structure of a pronoun (Subject/Object Impersonal się in sect. 2), it is a non-reflexivizer that

lacks full-fledged referential properties: (41a). When it displays the structure of a SELF-anaphor

(Reflexive się in (41b)), it still lacks full-fledged referential properties but is a reflexivizer.

4. Referential vs. Expletive Impersonal się

This section shows how the above analysis can accommodate the alternation in (42a) vs.(42b

=6), discussed most prominently by Dziwirek (1994) as (relational grammar) Inversion.

(42) a. Tę książkę czytało się z przyjemnością.

this bookACC readNEU się with pleasure

“One read this book with pleasure.”

33

b. Tę księżkę czytało mi się z przyjemnościa.

this bookACC readNEU meDAT się with pleasure

Dominant reading: “I read this book with pleasure.”

I argue that (42a) is the core construction, with się an indefinite pronoun indicative of a chain

with the deficient argument raising to the clitic as in sect. 3. I consider the second pattern

derivative, or requiring an additional operation, which for reasons stated below is language

specific. The two sentences have the same impersonal morphology. However, (42b) displays a

Dative (clitic or phrase) that is (a) unselected or not part of the argument structure of the

predicate, (b) human/animate, and (c) preferentially interpreted as bearing the Th-role of the

External Argument (Agent with this particular V). I propose that when an unselected Dative

clitic/phrase is {added /merged} as in (42b), the (human/animate) się functions as an expletive

pronoun that transmits properties to the Dative (rather like Passive transmission to by-phrases).

Się as pronoun contains a human/animate feature. When it functions as an expletive, it imposes

an agreement requirement on the unselected Dative, which must thus be human/ animate.

For Dziwirek, the Dative phrase (a) bears the Th-role otherwise held by an ordinary

subject, (b) it can be the binder of an anaphor, and (c) it can control into participial adjunct

clauses. Dyla (1983) reports variation as to the last aspect: for many speakers, the Dative cannot

be a controller of adjuncts.

My analysis differs from Dziwirek’s, who considers (42b) the core construction for (all)

subject-oriented impersonals. Oversimplifying, Dziwirek proposes an overt /covert Dative

responsible for human flavor in both (42a-b), while I propose an operation on (42a) to obtain

(42b) in the case of Polish, when the Dative is added. For Dziwirek the Polish Dative

construction is the unmarked case for impersonals. For me, it is marked, and involves a restricted

language specific procedure that finds no exact counterpart in the other languages that share the

impersonal.

One advantage of my proposal is that it serves to capture previously unnoticed

differences with other Slavic languages, and with Romance languages where impersonal se does

not function as an expletive pronoun. Beginning with the Slavic languages without impersonal as

in sect. 2.1, compare Bl (43), the parallel construction, to (42b):

34

(43) Četjaxa mi se knigi. Bl

readPAST.3P meDAT se books

“I felt like reading books.”

A morphological difference is that Bl se is what in 2.1 I called middle-passive, with predicate

and overt NP agreeing in phi-features. Nouns have no morphological Case in Bl, but knigi is

arguably Nominative, so checks Case against T. A semantic contrast tied to this is that se is not a

SE-anaphor that as pronominal expression can transmit properties to the unselected Dative, in

contrast with Polish się in (42b). Two important consequences of this difference left for future

research are (b) the semantic role of the Dative, and (b) the “factivity” of the event. First, mi in

(43) is not interpreted as Agent but Experiencer, which I take to indicate that this se is not a

transmitter of a Theta role. This is in contrast with Polish impersonal się above which transfers

the appropriate role from the predicate (in an imperfect way with an “accidental” reading

(Golab 1975)). Second, the event described by the predicate need not have occurred in (43),

which also contrasts with (42b). That is, the Polish sentence is contradictory if followed with a

coda such as... but I could not read it. By contrast, Bulgarian implies no contradiction with codas

such as... but I could not read any (the definite/indefinite character of the NP seems irrelevant).

This difference extends to intransitives, left undiscussed for lack of space. Bl Prispa mi se

/slept.3S me.DAT se/ “I came to feel like sleeping.” (Dimitrova-Vulchanova 1986) is a statement

about my predisposition, while Polish Spało mi się dobrze /slept. NEU me.DAT się well/ “ I

slept well” is a statement about how I slept. This truth value /logical form effect depends in

Polish on the pronominal status and transmission properties for the clitic, and what awaits

development is a precise analysis for middle/passive se in this case.

One (semantic) suggestion in the case of the transitive predicate in Bl (43) could be that

se indicates an agentless Event as open sentence predicated of the Dative. This EVENT se may

bind a situation variable, which is reminiscent of Kański’s approach using a free variable (1986).

By contrast, in (42b) impersonal się as expletive corresponds in my view to a lambda operator

with the sentence interpreted as a property predicated of the Dative Phrase. In this case the

variable has as value the External Argument of the predicate describing the Event, and its

semantic content is human/animate. That is, impersonal się in (42b), which Kański does not

discuss, is not a free variable in his sense, and seems to pose a problem for his analysis. As

35

expletive, however, this się can translate as a lambda operator that binds a variable inside the

sentence, which is the Agent. The transmission procedure for the non-selected Dative appended

to impersonal się in Polish seems language particular. We saw in section 2 that Slovenian

displays an impersonal se similar to Polish, and this language has Dative constructions with

impersonal se. However, sentences similar to (42b) in Slovenian do not describe an eventuality,

and their non-selected Dative is interpreted as an Experiencer, which is what happens with

passive and intransitive se in the Slavic languages without the impersonal. Thus, if viewed from

the Slavic perspective, the characteristics of the Polish Dative construction now under discussion

are unique.

Now let us consider Romance, which also makes Polish look unique. Contrasts with

Romance languages with the impersonal also depend on the referential status of the pronoun, but

have a different source. Polish się may be a (human/animate) expletive in (42b) in the way

discussed above, but its Romance counterpart is in similar cases referential in the sense of being

an indefinite pronoun that is not expletive. Compare (42) to Sp. (44a-b), as the parallel

morphosyntactic patterns.

(44) a. Antes se leía estos libros con placer. Sp

Before se read3S these books with pleasure

“In the past {one/they} read these books with pleasure.”

b. Antes se me leía estos libros con placer.

Before se meDAT read3S these books with pleasure

“In the past {one/they} read these books to me with pleasure.”

(44a) is very much like (42a), and should be analyzed via the SE-anaphor of sect. 3. While (44b)

is like (42b) in morphosyntax, the added item is interpreted here as an Ethical Dative/ Dative of

Interest, not as the Agent defined by the argument structure of the predicate. The Ethical reading

can be forced in Polish, but with difficulty. What separates the two languages, then, is that the

Polish pattern corresponds to I read this book with pleasure while in Spanish it is They read me

these books with pleasure. Thus, I propose that Sp se is referential in the sense of sect. 3, and

cannot be treated as an expletive pronoun, a difference which I attribute to an additional

transmission procedure present in Polish and absent from Spanish. In sum, the Dative

36

clitic/phrase in these impersonal constructions is not an argument of read in either language, but

has different semantic effects. In Polish the (much preferred) option is to treat the Impersonal

clitic as an expletive that transmits properties to the Dative: i.e. an open sentence with się the

variable for Agent predicated of the Dative. This option is absent from other languages, and

should be considered language specific. Thus, Sp. impersonal se remains referential, “is” the

Agent as in sect. 3, and retains all properties.

Datives on an argument position of the predicate have a different effect. They receive

their semantic role in the usual way, so fail to affect the argument status of się, which behaves in

their presence as referential. One consequence is a lack of contrast between Polish and Spanish

with this type of predicate: both languages involve the SE-anaphor of sect. 3, as in (45) where

the Dative is the Goal and się/se the Source, or equivalent.

(45) a. Wysłało się Ewie paczke.

sentNEU się EvaDAT packageACC

“Eva was sent a package. Someone sent a package to Eva” (Dziwirek 1994)

b. Se envió a Eva unos paquetes. Sp

se sent 3S to Eva some packages

“Eva was sent some packages./Someone sent some packages to Eva.”

The Polish example in (45a) is problematic for Dziwirek’s analysis: it must necessarily lack the

(covert) Dative that in her view contributes the human flavor to się, since she argues that

sentences can only have one Dative. However, since this example is an ordinary impersonal

sentence, the conclusion is that there can be no covert/overt Dative in the construction with just

the bare impersonal się.

An interesting and often unmentioned characteristic of the unselected Dative with się

pointed out to me by E. Willim is that it need not have inherent human reference, as shown in

(46a-b). Note, however, that inanimates are excluded as unselected Datives: (46c).

(46) a. Koniom dziś się <le ciagnęło wóz.

horsesDAT today się badly pullNEUT cartACC

“The horses had a hard time pulling the cart today.”

37

b. Psu się niewygodnie śpi w obroży.

dogDAT się uncomfortably sleep3S in collar

“The dog must be uncomfortable sleeping in a collar.”

c. *Półce ciężko się wisiało na ścianie.

shelfDAT hard się hangNEUT on wall

Intended: “The shelf had difficulty hanging on the wall.”

I already mentioned that many have observed that the bare impersonal construction in Polish and

other languages seems to display intrinsic human denotation. However, (46) shows that the

Polish Dative construction need not, which is another reason to distinguish the two constructions.

From my perspective, (46) suggests that the transmission procedure, which eliminates

indefiniteness and makes (human) się an expletive, is based on the animate feature as intrinsic

part of the feature composition of the impersonal. That is, in the expletive strategy that wipes out

arbitrary/indefinite character, the crucial feature present in się that must be shared by the Dative

is animacy.

Spanish counterparts of (46a) exist, and one with horses and cart as definites appears in

(47). This Spanish sentence has the (odd) interpretation that some person experienced difficulty

when pulling the cart for the benefit of the horses. In other words, the expletive strategy is not

available, so impersonal se retains its human flavor as indefinite pronoun, and the Dative phrase

is a benefactive.

(47) A los caballos hoy se les tiró mal del carro. Sp

To the horses today se theyDAT pulled3S bad of.the cart

‘Today someone had a hard time pulling the cart for the horses.’

Expletive się is in Polish more restricted as to predicate classes than its referential

counterpart, still another reason not to equate the dativeless impersonal construction with the

more limited Dative construction. The expletive use is not found with copulas, passives, or

modals, so the patterns repeated in (48a-50a) disallow the unselected Dative -- Markowi-- as

shown in the b. sentences. I suggest that the transmission procedure, seen here as language

specific, is sensitive to the semantic relation transmitted, with Agents high in the scale

(reminiscent of Passives), and other roles more difficult to transmit.

38

(48) a. Jest się wysokim.

be3S się tallINS

“One is tall.”

b. *Markowi jest się wysokim.

MarkDAT be3S się tallINS

Intended: “Mark is tall.”

(49) a. Jest się karanym przez przyjaciół.

be3S się punishedINS by friends

“One is punished by friends.”

b. *Markowi jest się karanym przez przyjaciół.

MarkDAT be 3S się punishedINS by friends

Intended: “Mark is punished by friends.”

(50) a. Powinno się być lysym.

shouldNEU się beINF baldINST

“One should be bald.”

b. *Markowi powinno się być lysym.

The arbitrary object use called NO-się in sect. 2 does not participate in the Dative construction,

which follows if the transmission procedure is restricted to some Th-roles.

Finally, I noted above that expletive się may display variation in the properties it

transmits. In general, it seems to transmit a restricted number of Th-roles, and binding properties

to the Dative. However, some speakers reject as ungrammatical the Dative as controller for

adjuncts, as in (51) that is borrowed from Dziwirek (1994:76).

(51) Siedząc przy kominku, Jankowi czytało się

sitPART by fireplace JohnDAT readNEUT się

tę książke z przyjemnościa..

this bookACC with pleasure

“Sitting by the fireplace, John read the book with pleasure.”

That sentences such as (51) are ungrammatical for some speakers suggests that while Th-roles

and binding transmission are not open to variation, control is. A precise analysis for this

39

phenomenon awaits development, but it constitutes still another reason to distinguish the more

restricted Dative+ się construction from the bare impersonal construction, where the reflexive

clitic can always function as controller in Polish and the other languages.

As we just saw, my proposal accounts for the contrast in (42) in areas where previous

analyses fail. In addition, with its comparative approach, it offers the advantage of identifying

previously unnoticed differences with other Slavic languages, and extensive similarities and one

difference with Romance languages with the impersonal.

5. Summary and conclusions

In impersonal subject and object uses, Polish się represents a type of SE-anaphor that had not

been identified in previous literature: one involving a clitic. Like the SE-anaphors of languages

like Dutch, się indicates a NP without phi-features that occupies an argument position of the

predicate and must raise to repair formal deficiency. However, this NP moves to a clitic, which

makes it differ from the SE-anaphors discussed up to now in the literature, which move to an

ordinary subject. Like a regular pronoun, się as SE-anaphor is a non-reflexivizer and can be free

(principle B). It does not inherit phi-features from another argument, which distinguishes it from

non-clitic SE-anaphors, as in Dutch, and makes its pronominal character evident. Also like a

pronoun, impersonal się can be referential (an indefinite) or expletive, as in the presence of an

unselected Dative. This last aspect distinguishes it from its very close counterparts, the SE-

anaphors of Romance, impersonal se/si in Spanish and Italian, which cannot be expletives in the

same sense.

References.

Babby, Leonard H. (1975). “A transformational analysis of transitive -sja verbs in Russian”. Lingua 35. 297-332.

Babby, Leonard H. (1998). “Voice and Diathesis in Slavic.” Position Paper, Conference on Comparative Slavic Morphosyntax, Bloomington, Indiana.

Belletti, Adriana. (1982). “Morphological passive and pro-drop: The impersonal construction in Italian.” Journal of Linguistic Research 2:1-34.

Borsley, Robert D. and María Luisa Rivero. (1994). “Clitic Auxiliaries and Incorporation in Polish“. Natural Language and Linguistic Theory 12. 373-422.

Chierchia, Gennaro (1995). “The Variability of Impersonal Subjects” In Emmon Bach, Eloise Jelinek, Angelika Kratzer, and Barbara Hall Partee, eds. Quantification in Natural

40

Language, 107-143. Kluwer, Dordrecht. Chomsky, Noam. (1981). Lectures on Government and Binding. Foris, Dordrecht. Chomsky, Noam. (1998). “Minimalist Inquiries: the Framework”. Preliminary version in MIT

Working Papers in Linguistics 15. To appear in R. Martin, D. Michaels, and J.Uriagereka, eds. Step by Step: Essays on Minimalist Syntax in Honor of Howard Lasnik. MIT Press, Cambridge.

Cinque, Guglielmo. (1988). “On Si Constructions and the Theory of Arb” .Linguistic Inquiry 19.52.1-583.

Dimitrova-Vulchanova, Mila. (1996). Verb Semantics, Diathesis, and Aspect. Doctoral Dissertation. University of Trondheim.

Dyla, Stefan. (1983). “Some Further Evidence against an Impersonal Passive Analysis of Polish Impersonal Constructions”. Studia Anglica Posnaniensia 25. 123-128.

Dobrovie-Sorin, Carmen. (1998). “Impersonal se Constructions in Romance and the Passivization of Unergatives.” Linguistic Inquiry 29. 399-438.

Dziwirek, Katarzyna. (1990). “Default Agreement in Polish”. In Katarzyna Dziwirek, Patrick Farrell, & Errapel Mejías-Bikandi, eds. Grammatical Relations: A cross-theoretical perspective, 147-161. CLSI, Stanford.

Dziwirek, Katarzyna. (1994). Polish Subjects. New York: Garland. Everaert, Martin. (1991). “Contextual Determination of the Anaphor/Pronominal Distinction”.

In Jan Koster & Eric Reuland, eds. Long-Distance Anaphors, 77-118. Cambridge: Cambridge University Press.

Golab, Zbigniew. (1975). “Endocentricity and endocentrization of verbal predicates: illustrated with Latin and Slavic material”. General Linguistics 15. 1-35.

Grimshaw, Jane. (1982). “On the Lexical Representation of Romance Reflexive Clitics.” In Jane Bresnan, ed. The Mental Representation of Grammatical Representations, 87-148. Cambridge, Mass: MIT Press.

Jaeggli, Osvaldo. (1986). “Passive.” Linguistic Inquiry 17: 587-622. Kański, Zbigniew. (1986). Arbitrary Reference and Reflexivity: A Generative Study of the Polish

pronoun się and its English equivalents. Uniwersytet Slaski, Katowice. Kański, Zbigniew. (1992). “Impersonal constructions as a Strategy for Second Order

Predication”. In M. Kefer and J. van der Auwera, eds. Meaning and Grammar: Cross-Linguistic Perspectives, 95-121. Mouton de Gruyter: Berlin.

Kardela, Henryk. (1981). “Some Remarks on Binding in Polish”. WCCFL 1. 158-167. Kardela, Henryk. (1985). A Grammar of English and Polish Reflexives. Uniwersitet Marii Curie-

Sklodowskiej, Lublin. Kipka, Peter F. (1989). “Impersonals and Inflection in Polish.” MIT Working Papers in

Linguistics 10. 135-150. Krapova, Iliyana. (1998). “Subjunctive Complements, Null Subjects and Case Checking in

Bulgarian” Paper read at CLITE 1, Szeged, Hungary. Kubiński, Wojciech. (1982). “Polish się constructions and Their English Equivalents”. Papers

41

and Studies in Contrastive Linguistics 15.55-67. Kubiński, Wojciech. (1987). Reflexivization in English and Polish: an Arc Pair Grammar

Analysis. Linguistische Arbeiten 179. Tübingen: Niemeyer. Kupść, Anna. (1999). “Haplology of the Polish reflexive marker.” In Robert D. Borsley &

Adam Przepiórkowski, eds. Slavic in HPSG. CSLI, Stanford. Mendikoetxea, Amaya. (1992). On the Nature of Agreement: the Syntax of Arb Se in Spanish.

Unpublished Doctoral Dissertation, University of York, York. Moore, John. (1993). “Reflexive, Reflexive Passives, and Generalized Chains”. WCCFL 12.389-

404. Progovać, Ljiljiana. (1998). “Events in Serbian”. Papers from Second Conference on Formal

Approaches to South Slavic Languages. University of Trondheim Working Papers in Linguistics 31. 79-116.

Reinders-Machowska, Ewa. (1991). “Binding in Polish”. In Jan Koster & Eric Reuland, eds. Long-Distance Anaphors, 137-150. Cambridge: CUP.

Reinhart, Tanya & Eric Reuland. (1991). “Anaphors and Logophors: an Argument Structure Perspective”. In Jan Koster & Eric Reuland, eds. Long-Distance Anaphors, 283-321. Cambridge: CUP.

Reinhart, Tanya and Eric Reuland. (1993). “Reflexivity”. Linguistic Inquiry 24. 657-720. Rivero, María Luisa . (1997). “Two locations for complement clitic pronouns: Bulgarian, Serbo-

Croatian and Old Spanish.” In Ans van Kemenade, & Nigel Vincent, eds. Parameters of Morphosyntactic Change. 170-206. Cambridge: CUP.

Rivero, María Luisa. (1998). “On Impersonal SE in Romance and Slavic”. Paper read at Going Romance. December 1998, Utrecht, Holland.

Rivero, María Luisa and Milena Sheppard. (1999a). “On Impersonal SE in Slovenian”. Paper read at the Workshop on Pronouns in Generative Grammar, Societas Linguistica Europeae, July 1999, Ljubljana, Slovenia.

Rivero, María Luisa and Milena Sheppard. (1999b). “On Impersonal SE/SIE in Slavic”.Paper read at the Formal Description of Slavic Languages 3 Meeting. December 1999. Leipzig, Germany.

Rizzi, Luigi. (1986). “Null Objects in Italian and the Theory of pro”. Linguistic Inquiry 17. 501-557.

Siewierska, Anna. (1988). “The Passive in Slavic”. In M. Shibatani, ed. Passive and Voice, 243-289. Amsterdam: Benjamins.

Sportiche, Dominique. (1995). “Clitic constructions.” Johan Rooryck & Laurie Zaring, eds. Phrase Structure and the Lexicon, 213-76. Dordrecht: Kluwer.

Szymańska, Izabela. (1998). “Polish reflexiva tantum- a construction-based interpretation”. In Piotr Stalmaszczyk, ed. Projections and Mapping. Studies in Syntax, 99-118. Lublin: PASE.

Szymańska, Izabela and Grzegorz Spiewak. (1998). “Some remarks on the role of się in organizing the argument structure of Polish verbal predicates”. In Piotr Stalmaszczyk, ed.

42

Projections and Mapping. Studies in Syntax, 120-136. Lublin: PASE.

Willim, Ewa. (1998). “On the Grammar of Polish Nominals.” To appear in R. Martin, D.Michaels, and J. Uriagereka, eds. Step by Step: Essays on Minimalist Syntax in Honor of Howard Lasnik. MIT Press, Cambridge.