The Bidirectional Influence between Coherence Establishment and Pronoun Interpretation

Stevenson et al. (1994) demonstrated an apparent asymmetry between the production and interpretation of pronouns in passage completion studies. For instance, when writing completions to Source-Goal transfer-of-possession passages with a pronoun prompt (1a), participants are equally likely to interpret the pronoun to refer to the non-subject Goal (Bill in 1a) as to the subject-position Source (John). However, in passages with a full-stop prompt (1b), participants were much more likely to re-mention the subject/Source with a pronoun than a name, whereas they were much more likely to re-mention the non-subject/Goal with a name (see also Arnold 2001). In addition to positing the existence of thematic role biases in pronoun interpretation (here, favoring Goals over Sources), Stevenson et al. argue that the full-stop condition reveals an overlaid subject assignment bias. These two biases compete in stimuli like (1a) to result in the roughly 50/50 distribution of pronoun assignments.