17 August: Shira Tal

Many languages exhibit differential case marking (DCM), where marking targets only certain types of arguments. In such systems, atypical arguments (e.g., patients that are high on the animacy scale) are more likely to be marked. Recent work using artificial language learning experiments suggests that DCM is driven by a bias for efficient communication. For example, in a language with variable word order, case marking only animate objects prevents ambiguity in an efficient way, since marking is restricted to events that are potentially ambiguous (e.g., events with two animate participants). However, an alternative account suggests that DCM instead reflects the pragmatics of discourse. Agents of transitive sentences tend to be discursively old (given) while patients tend to be new. Atypical discourse associations (i.e., new agents or given objects) are often marked with additional linguistic material. For instance, in Catalan, only given objects are case marked. Marking of atypical associations between argument structure and information structure–rather than ambiguity avoidance–has therefore been argued to be the main source of DCM. Here, in two experiments, we test whether discourse status drives case marking during learning of an artificial language. In the first experiment the language has flexible word order whereas in the second experiment it has a fixed word order. Our results suggest a larger impact on word order than on case marking. We relate these findings to historical accounts of development of DCM.