This dissertation looks at the people use causal coherence information is used to determine the referent of a pronoun in sentence pairs, such as John accused Mark of stealing a car. He called the police. The experiments reported here test the idea that the process of pronoun disambiguation is informed by the causal connection between the person denoted by the pronoun (e.g., the one who called the police) and each of its potential referents (John and Mark).
Experiment 1 varied how likely it is that the event of the first sentence (e.g., the accusation) would cause the event of the second (the notification) for each of the two individuals, and required participants to choose the appropriate referent. Participants selected the antecedent that had high causal strength or split their votes if neither did. Experiment 2 used explicit connectives that were either causal (and so, because) or temporal (before, after) and demonstrated that participants were more confident in their use of causal information when the connective was causal than when it was temporal. Experiment 3 explored the influence of negation on participants’ choice of referent. Participants’ ratings were affected by negation, but their choices of referent and their confidence in their choices were not. This indicates that participants use causal information when choosing the appropriate referent rather than relying on likelihood of occurrence. Finally, Experiment 4 used a self-paced reading paradigm to explore the effect pronoun ambiguity and reference has on participants reading time of two types of causal coherence relations – result and explanation. Participants read sentence pairs connected by result relations fastest when an unambiguous pronoun referred to the object. They read sentence pairs connected by explanation fastest when an unambiguous pronoun referred to the subject. The results of these four experiments suggest that the process of causal coherence information and the process of pronoun disambiguation interact. These findings are in line with models of intersentential coherence in which discourse coherence is achieved by relating adjacent discourse segments by means of coherence relations (e.g., Hobbs, 1985; Kehler, 2002).