Object coreference in German: Grammar without exposure?

Sam Featherston
sam.featherston@uni-tuebingen.de
University of Tuebingen

While object coreference structures are possible in English, the equivalent German construction is only marginally grammatical (1). Additionally, there appears to be little grammaticality difference between a reflexive (1)a and a pronoun (1)b here, which has been long been regarded as a problem for binding theory (eg Reis 1976, Grewendorf 1988).

(1)

a.

?Der Friseur zeigte Paul_i sich_i im Spiegel
the barber showed Paul himself in.the mirror

b.

?Der Friseur zeigte Paul_i ihn_i im Spiegel
the barber showed Paul him in.the mirror

In spite of considerable attention, little progress has been made in explaining why these structures should be so weakly grammatical in German. on this phenomenon. One factor which has obstructed progress on this question is the lack of agreement on the data. In order to clarify this situation we carried out a corpus search for the relevant constructions and elicited grammaticality judgements. The results of our corpus searches produced no single relevant example of object coreference in 530 million word-forms of German (COSMAS corpus) nor in 145 million word-forms of English (British National Corpus, COBUILD). In the light of this we applied the methodology of magnitude estimation (Bard et al 1996) to carefully matched and counterbalanced sets of materials. This experimental approach to grammaticality judgements captures very fine grammaticality differences and produces hard, replicable data. We tested in sixteen conditions, varying the factors Case Order (Dative > Accusative, Accusative > Dative), Bindee (Pronoun, Reflexive), Binder (full NP, Pronoun), and Selbst (with or without reflexivity marker "selbst"). The results revealed clear effects for all of these factors. Each had a preferred and a dispreferred value, and the sentences showed very fine distinctions in grammaticality; each dispreferred value was associated with a cumulative degradation in grammaticality.

We may summarise that this elicitation data shows that subjects have clear consistent intuitions about these structures, even though, as the corpus data makes clear, it is highly unlikely that they form any part of the subjects' exposure to language. We argue that this finding of clear grammaticality intuitions even among quite ungrammatical structures undermines models of syntactic processing based merely on linguistic experience, and supports the existence of a real mental grammar.

References

Bard E., Robertson D. & Sorace A. 1996 Magnitude Estimation of Linguistic Acceptability. Language 72 (1), 32-68.

Grewendorf G. 1988 Aspekte der deutschen Syntax: Eine Rektions-Bindungs-Analyse. Studien zur deutschen Grammatik 33. Tübingen: Narr.

Reis M. 1976 Reflexivierung in deutschen ACI-Konstruktionen: Ein transformations- grammatisches Dilemma. Papiere zur Linguistik 9, 5-82.