Semantic-role prominence is contingent on referent prominence in discourse: Experimental evidence from impersonals and passives in Polish

Abstract

Semantic roles are an important piece of information in sentence interpretation. While role-related effects in sentence comprehension are well established, the definition and structure of roles such as the agent are still controversially debated. One open question pertains to whether there is a general advantage for the prototypical agent or whether prototypical and atypical agents are flexibly privileged depending on the discourse function of the syntactic construction they occur in, i.e. depending on whether a construction demotes or promotes the agent referent. In two acceptability judgement tests, we investigated this open question for the Polish passive and the Polish impersonal no/-to construction. The former serves both patient promotion and agent demotion, while the latter only demotes the agent referent in discourse. We find an effect of role prominence as reflected in construction-specific acceptability clines: Both constructions show no advantage for volitional (prototypical) agents, but reveal differences for non-volitional experiencer subjects (i.e. atypical agents). This suggests that for atypical agentive arguments such as experiencers, the type of predicate (e.g. emotion vs. cognition vs. perception predicate) matters for role prominence. The experimental findings were supported by a corpus analysis, revealing that verbs with higher ratings in, e.g., the -no/-to construction also occurred more frequently in the construction. Overall, the pattern of results cannot be explained by role prototypicality that predicts a construction-inde¬pen-dent advantage for prototypical agents. Rather, the experiments provide further evidence for the model of role prominence presented by Himmelmann & Primus (2015).