Markus Steinbach, Edgar Onea, Derya Nuhbalaoglu and Anne Wienholz (University of Göttingen): How to Distinguishing Discourse Referents - A Cross-modal Discourse Semantic Analysis

In this paper, we propose a unified analysis of referential loci (R-loci) in sign languages (German Sign Language, DGS) and grammatical gender in spoken languages (German) within a modified version of Discourse Representation Theory (DRT, Kamp 1981), building on ideas from Barberà (2012). The core idea of our approach is that spoken and sign languages, apart from modality driven differences, use similar strategies to distinguish Discourse Referents (DRs) in the discourse semantic representation.

One obvious strategy to introduce DRs in sign languages is the INDEX sign. In the first sentence of example (1) the first DR is introduced in the ipsilateral area (glossed as R) and the second DR is localized in the contralateral area (glossed as L). Crucially, the pronominal INDEX in the second sentence singles out the R-region thus unambiguously establishing an anaphoric link to the first DR. By contrast, the German translation of (1) remains ambiguous. However, sentence (2) with a similar structure, is non-ambiguous due to the MASC feature on the pronoun.

(1) P-E-T-E-R INDEX-R J-O-H-N INDEX-L LOVE. INDEX-R LAUGH.
'Peter loves John. He (i.e. Peter) is laughing.'

(2) Der-MASC Roman-MASC liegt auf der-FEM Couch-FEM. Er-MASC war teuer.
'The book is on the couch. It (i.e. the book) was expensive.'

(3) Der-MASC Roman-MASC liegt auf dem-MASC Tisch-MASC. Der-MASC war teuer.
'The book is on the table. It (i.e. the table) was expensive.'

We claim that the discourse semantic function of R-loci in sign languages and gender in spoken languages as well as their morphosyntactic realization is similar: both contribute to anaphoric disambiguation. We argue, partly following Barberà (2012), that INDEX is composed of an abstract INDEX-sign and a cliticized R-locus. Similarly, the German determiners are composed of a determiner root and a cliticized gender marker. This is also confirmed by the fact that determiners in German and INDEX in DGS can also be used as pronominal elements (cf. second sentence in (1) and (3)).

We follow Onea/Steinbach?s (2013) modified version of DRT model with so-called referent structures (RSs), which directly reflect the geometric properties of the signing space. For the analysis of gender in German, we use exactly the same mechanism with two differences: (i) RSs are not recursive and (ii) instead of binary distinctions, gender in German introduces a fixed tripartite structure. Crucially, in both modalities, RSs are used in the language of Discourse Representation Structures but do not receive a semantic interpretation. This is in line with Geurts/Maier (2013), who label DRs with gender-features suggesting that gender is not a truth conditional but rather a procedural meaning component.

Syntactically, we suggest, following Alexiadou (2004), that gender may either stem from the lexical entry of the corresponding head noun or from natural features of the respective referent. The gender feature of the noun agrees with the determiner hence, gender cannot be freely assigned in German. As opposed to this, the assignment of R-loci is partly regulated by default constraints e.g., ?link the first DR to the ipsilateral area?. Hence, R-loci are not lexically determined but only assigned in discourse.

In sum, we show interesting parallels between discourse structure and grammar in sign and spoken languages. Our approach can be seen as a first theoretical import from sign language semantics to spoken language semantics. In addition, it contributes to the view that even obvious modality-specific grammatical implementations can be analyzed with the same underlying model.