Repassivization is the passivization of a passive predicate. It is the process whereby a derived subject is “demoted” and thus appears in a by-phrase. In most languages the result is completely incomprehensible:
(1) Hannah was deceived by her sisters.
(2) *There was (been) deceived by Hannah (by her sisters).
Based on the English alone, it appears that repassivization should not be countenanced by any theory of grammar. Under the standard GB view of passive, the passive morpheme is assigned the predicate’s external theta role. This ensures that only transitives (and unergatives) will be passivized. Other frameworks have put forth their own principles to ensure the same distribution of passive (cf. RG’s 1AEX Law).
At first blush, Lithuanian appears to flout all such bans on repassivization. The paradigm in (3a-c), adapted from Timberlake 1982, appears to be one such case. In (3c), the passive morpheme (PASS) occurs twice (a “double passive”), attaching both to a non-finite auxiliary and to the passive verb, now a predicative complement:
(3) a. Aš pirkau laikrodį.
I:NOM bought watch:ACC ‘I bought a watch’
b. Laikrodis buvo ( mano) pirktas.
watch:NOM.M.SG AUX:PAST me:GEN bought:PASS.M.SG ‘A watch was purchased (by me).’
c. Laikrodžio būta ( mano) pirkto.
watch:GEN.M.SG AUX:TA me:GEN bought:PASS.GEN.M.SG ‘A watch was apparently purchased (by me).’
While we might imagine the function of causativizing a causative, what possible functional explanation could repassivization have? We will see that structures such as (3c) involve a contrast in mood (evidential), rather than voice.
Note that the finite form of the auxiliary does not occur in the evidential, in contrast to the corresponding passive in (3b) (Ambrazas et al. 1997:280-284). See, e.g., (4):
(4) Ten šuns (*buvo) bėgta. [*buvo on the evidential reading only]
there dog:GEN AUX:PAST ran:-TA ‘A dog must have run there (there are footmarks).’
[Ambrazas et al. 1997: 282-283]
That is, the evidential occurs with non-finite būta, or with no auxiliary at all. The evidential semantics is encoded syntactically by the combination of a pre-posed genitiveNP and /-ma/-ta/ predicate-final morphology.
We find a similar ban on the cooccurrence of passive morphology and finite auxiliaries in the Polish -no/-to construction (see Lavine 2004). The Lithuanian /-ma/-ta/ morpheme shares with cognate Polish /-no/-to/ its extra-paradigmatic status: both sets of morphemes are frozen vestigial forms. I will argue that /-ma/-ta/, like Polish /-no/-to/, has been reanalyzed as an (affixal) auxiliary in the modern language, thus accounting for its complementary distribution with finite auxiliaries. As an affixal auxiliary, /-ma/-ta/ is generated not where it is pronounced, motivating a “separationist” view of morphology(Beard 1995), whereby the syntactic representation of an object is potentially separated from its morphophonological representation. Thus, apparent repassivization in Lithuanian involves a rich interplay between morphology, syntax, and semantics.
Ambrazas, V. et al. 1997. Lithuanian grammar. Vilnius: Baltos lankos.
Beard, R. 1995. Lexeme-Morpheme Base Morphology. Albany: SUNY Press.
Lavine, J. 2004. “The morphosyntax of Polish and Ukrainian -no/-to.” JSL 12.
Timberlake, A. 1982. “The impersonal passive in Lithuanian.” BLS 8: 508-524