Why Polish Doesn't Like Infinitives

Katarzyna Dziwirek, University of Washington

This paper proposes an explanation of the contrast in the array of infinitive constructions in Polish, Russian, and English. Descriptively, the facts are as follows: Polish lacks constructions which involve a main verb followed by an accusative nominal and an infinitive in some way referring to this nominal (specifically, object control with accusative controllers, ECM, and object raising), Russian allows this surface configuration with some verbs ((4)-(6)), and English allows them freely. The difference between English and Polish is a source of many learner errors, which served as an inspiration for this paper, see (1)-(3).

(1) *Chcialabym poprosic cie pisac to zaproszenie.
I would like to ask you to write this invitation.

(2) *Ona poszla do niego, zeby przekonac go zostac w domu dzisiaj.
She went to him to persuade him to stay home today.

(3) *Wierzyla go byc niewinny.
She believed him to be innocent.

(4) Ona poprosila ego pomoch jej.
She asked him to help her.

(5) Oni ubedili ego prijti v sem' chasov.
They persuaded him to come at 7.

(6) On priglosil Mashu prijti v sem' chasov.
He invited Masha to come at 7.

The analysis of the contrast between Polish and English relies on three assumptions. The first two are standard in Relational Grammar, see e.g., Perlmutter and Postal (1984), Rosen (1981/88), Berinstein (1985), and Legendre (1986), the third is the key to the analysis. ASSUMPTION ONE: sentences in (1)-(3) have a structure which involves a cross-clausal 1-2 multiattachment (MA), that is, in these sentences the accusative nominal is (in some stratum) simultaneously a direct object (represented as 2) of the main clause and a subject (represented as 1) of the infinitival clause. ASSUMPTION TWO: 1-2 MAs can be resolved by a "birth" of a reflexive clitic, that is, reflexive morphology is often a marker of configurations where a single nominal is both the agent and the patient of the action (e.g. semantically reflexive clauses, unaccusatives, etc.). ASSUMPTION THREE: 1-2 MAs must be resolved in Polish with a birth of a reflexive clitic. In the case of cross-clausal MA in (1)-(3) the problem is that the nominal is the subject of one clause and a direct object of another and it is not clear which verb should receive the clitic: the matrix verb, the infinitive, or both. Since there is nothing in the grammar to help resolve this conflict between a universally well-formed structure which underlies, e.g., the English in (2) and the Russian in (5), and a Polish-specific condition on realization of 1-2 MAs, the Polish sentences like (2) crash. Or, in other words, Polish elects not to have this construction.

In English, such conflict does not arise. There frequently is no overt marking of semantically reflexive clauses (e.g., Peter shaved.), and when anaphoric relationships are overtly marked, they are marked with pronouns (NPs), not clitics. Therefore, as there is nothing in the grammar of English which rules out the universally well-formed configuration underlying (2) and (4), English makes use of it.

The paper shows that each of the three specific constructions which Polish lacks (object control with accusative controllers, ECM, and object raising) involve 1-2 MAs, presents evidence that the Polish reflexive clitic si&enasal; is a (post-)syntactic marker of 1-2 MAs, demonstrates that infinitival constructions which involve 1-1 and 1-3 MAs are well-formed, and thus argues that a single rule of Polish grammar (ASSUMPTION THREE) accounts for the absence of the relevant sentence patterns in Polish. The explanation of the contrast between Polish and Russian spans diachronic and synchronic grammar and makes a connection between two facts about Russian which have previously been unrelated, that is the existence of sentences like (4)-(6) and the fact that a few -sja marked verbs can occur with accusative nominals:

(7) Kto boitsja Tat'janu Paramonovu?
who fears-3sg-ref Tatiana-acc Paramonova-acc
Who is afraid of Tatiana Paramonova?

(8) Ja slushajus' mamu.
I-nom obey-1sg-ref mother-acc
obey my mother.

These data taken together with the sentences in (4)-(6) point to a generalization that in Russian the one to one correspondence between 1-2 MA and -sja is beginning to break down. In (7)-(8) -sja does not mark 1-2 MA anymore. In (4)-(6) -sja does not have to mark a 1-2 MA. Neither is allowed in Polish, where the connection between si&enasal; and 1-2 MA is very strong.

The paper presents a historical analysis of -sja, showing how, as -sja loses its mobility it also begins to lose its function of obligatorily marking 1-2 MA. The result is that accusative plus infinitive constructions like (4)-(6) are becoming possible. At the same time, in a few verbs, in which the meaning of -sja is not transparently reflexive, -sja becomes disassociated from the 1-2 MA marking function and becomes lexicalized as an affix ((7)-(8)). This stage of the change is going on now. Like all language change it is not immediate and proceeds in a word by word fashion. Thus the morpho-phonological changes which began to affect -sja in the 17th century continue to influence its function: as it hovers on the boundary between clitic and affix morpho-phonologically, its syntactic function vacillates.

Though the paper is couched in RG, the underlying insight that an absence of a construction in Polish is due to a conflict with a morpho-syntactic rule may potentially be translated into other frameworks.