Lexical differentiation in language production and comprehension

This paper presents the results of three experiments that explore the breadth of the relevant discourse context in language production and comprehension. Previous evidence from language production suggests the relevant context is quite broad, based on findings that speakers differentiate new discourse referents from similar referents discussed in past contexts (Van Der Wege, 2009). Experiment 1 replicated and extended this "lexical differentiation" effect by demonstrating that speakers used two different mechanisms, modification, and the use of subordinate level nouns, to differentiate current from past referents. In Experiments 2 and 3, we examined whether addressees expect speakers to differentiate. The results of these experiments did not support the hypothesis that listeners expect differentiation, for either lexically differentiated modified expressions (Experiment 2), nor for subordinate level nouns (Experiment 3). Taken together, the present findings suggest that the breadth of relevant discourse context differs across language production and comprehension. Speakers show more sensitivity to things they have said before, possibly due to better knowledge of the relevant context. In contrast, listeners have the task of inferring what the speaker believes is the relevant context; this inferential process may be more error-prone.