We investigated the role of antecedent focus and conceptual distance in NP anaphor resolution by using four antecedent manipulations: focus status, typicality, category specificity, and adjective modification. The Informational Load Hypothesis (ILH, Almor, 1999) proposed that increased conceptual content of an anaphor with respect to its antecedent causes an increase in processing load which may only be justified when the antecedent is more difficult to access (i.e. when it is not in focus). Thus, focused antecedents are best referred back to by more general anaphors that have a greater conceptual distance from their antecedents. However, when the antecedent is not in focus, and is more difficult to access, then greater conceptual content and specificity is justified.
The conceptual distance between an antecedent and its anaphor can be manipulated via typicality: typical exemplars (``robin'') of a categorical NP anaphor (``bird'') have less conceptual distance than atypical exemplars (``penguin''). In keeping with this, Almor (1999) reported an inverse typicality effect for focused antecedents which was predicted by the ILH.
In Experiment 1, we manipulated conceptual distance in a different way by using three-level hierarchies in which the least specific thing (vehicle) was the anaphor and the antecedent could be either from the next level of the hierarchy (car) or from the most specific level (hatchback). Our results support the ILH's prediction that when focused, more conceptually distant antecedents (hatchback) cause longer reading times at the anaphor than conceptually closer antecedents (car).
Experiments 2a and 2b tested the possibility that additional conceptual distance introduced via pre-nominal adjective modification could result in effects similar to those seen for typicality and category specificity. From the ILH, we predicted that for category anaphors, an adjective-modified antecedent would have a greater conceptual distance than an unmodified antecedent and thus when the antecedent was focused, anaphors would be read faster when the antecedent was modified by an adjective. Experiment 2a failed to produce this result, even though a norming study found that the adjective-modified antecedents were more atypical than unmodified antecedents. In Experiment 2b, new adjectives were used that had been shown to make the modified antecedents significantly more atypical than those used in Experiment 2a. Under these conditions, ILH-predicted effects were found only for the focus condition. The combined results of Experiments 2a and 2a suggest that the interaction of conceptual distance and focus may be much more complicated than was first suggested by the results of Almor (1999) and Experiment 1.
Almor, A. (1999) Noun-phrase anaphora and focus: The informational load hypothesis. Psychological Review, 106 (4), 748-765.