Constraints on full NP object shift and pronominal object shift in Scandinavian

Maia Andréasson

Abstract

This paper presents a unified analysis of mainland Scandinavian pronominal object shift and Icelandic full NP shift, the latter a phenomenon that has been compared to scrambling in German. I demonstrate the impact of accessibility on object placement in Scandinavian languages, and propose an Optimality Theoretic analysis that includes both pronominal and full NP object shift, and where constraints involving different pragmatic factors (such as accessibility and contrastive focus) interact, but are ranked lower than syntactic constraints on for example verb placement.

Ever since Holmberg (1986) it has been well known that object shift is subject to both structural and pragmatic constraints. For instance, the lexical verb must be in the V2 position for an object to be licensed in a position preceding the negation, i.e. shifted, see example (1).

(1) a. Jag såg henne inte. / *Jag har henne inte sett. [SWEDISH]
I saw her not / I have her not seen
'I didn't see her.'/'I have not seen her.'
    b. Jón las bækurnar ekki. / *Jón hefur bækurnar ekki lesið. [ICELANDIC]
Jon read the books not / Jon have the books not read
'Jon didn't read the books.'/'Jon has not read the books.'

The pragmatic constraint on pronominal object shift most commonly discussed in previous studies is that only "non-stressed" pronominal objects shift, whereas pronouns with contrast interpretation and hence - in speech - contrastive stress must appear following the negation, i.e. in situ, see the Danish example in (2) (where capital letters mark contrast intonation).

(2) a. *Jeg HENDE ikke, men jeg den ANDEN pige. [DANISH]
I saw her not, but I saw the other girl
'I didn't see HER, but I saw the OTHER girl.'
    b. Jeg ikke HENDE, men jeg den ANDEN pige . [DANISH]
I saw not her but I saw the other girl
'I didn't see HER, but I saw the OTHER girl.'

This information dynamical restriction has given rise to analyses of object shift suggesting that an unstressed pronoun "escapes" from a focus domain (cf. Holmberg 1999:23). Shifting non-contrasted pronouns is considered to be obligatory in standard Danish and in Icelandic, but more optional in for example Swedish.

Nevertheless, Andréasson (2008; 2009; 2010) shows that this pragmatic restriction on pronominal object shift does not explain all the word order variation where pronominal objects are involved. An extensive corpus study of written Swedish and Danish reveals that non-contrasted pronominal complements with sentence/VP antecedents, see example (3), appear in situ to a greater extent than those with NP antecedents, both in Swedish and, more surprisingly, in Danish, see example (4).

(3) a. Jag tror inte det. [SWEDISH]
I think not that
'I don't think so.'
    b. Jag kan inte det. [SWEDISH]
I ca not that
'I can't.'
(4) a. Hvorfor tror hun ikke det? [DANISH]
why think she not that
'Why doesn't she think so?'
    b. Tror du ikke det? [DANISH]
think you not that
'Don't you think so?'

This difference in distribution is not related to the dichotomy contrast vs. non-contrast, but to a difference in cognitive status (cf. Gundel et al 1993; Gundel et al 2003) between on the one hand pronouns with NP antecedents and pronominal complements of factive verbs, and on the other hand pronominal complements of non-factive verbs. The former are at the highest level of accessibility (Gundel et al's in focus) and are licensed in the shifted position both in Swedish and in Danish. The latter are normally not at the highest level (activated and below), and thus may not shift, see (5).

(5)Swedish/Danish
   in focus   activated familiar identifiable referential type identifiable
SHIFTED  IN SITU

This analysis is shown to hold also for Icelandic full NP shift. In Icelandic the dividing line between shifted and in situ placement goes further down on the accessibility scale, between definite (identifiable) and non-definite (referential) full NPs, see (6). The fact that Icelandic is even more restrictive than Danish, in not allowing non-contrasted object pronouns in situ, also supports this claim.

(6)Icelandic
   in focus activated familiar identifiable     referential type identifiable
SHIFTED IN SITU

An optimality theoretic analysis of these facts is proposed, introducing a family of alignment constraints for the different levels of accessibility, see (7). The ranking of the constraints express a universal tendency for elements with a higher level of accessibility to appear more to the left in a domain, than elements of lower levels of accessibility.

(7) GIVINF-L >> GIVACT-L >> GIVFAM-L >> GIVID-L >> GIVREF-L >> GIVTYPE-L

These constraints are ranked in relation to other constraints on word order. The differences between Mainland Scandinavian and Icelandic - i.e. between pronominal and full object shift - is illustrated by differences in ranking. For Icelandic GIVINF-L, GIVACT-L, GIVFAM-L and GIVID-L outrank for example ICON-MOD, a constraint on modifier placement, rewarding structures where modifiers (here sentence adverbials) precede the domain modified (eg. the object). In Danish and Swedish only GIVINF-L outranks ICON-MOD.

Finally, another finding that supports a joint analysis of full NP object shift and pronominal object shift is presented, namely the fact that contrast on another element in the clause overrides the accessibility constraint on object placement, in Swedish and Danish - as well as in Icelandic. As mentioned above, contrast on the object constrains otherwise preferably shifted objects to remain in situ. Still, contrast on another element also affects the word order; it has the effect that objects with lower cognitive status shift, a phenomenon mentioned already by Diesing (1997) for Icelandic, see (8) (for Swedish and Danish, see Andréasson 2010).

(8) Ég les bækur ekki ...
I read books not
'I don't READ books (, but only BUY them).' (Diesing 1997:412)

Selected references