Inference Generation during Auditory Language Comprehension

It has become nearly axiomatic (and even likely true) to hold that inference generation during language comprehension is a function of two things: (1) the nature of the inference under consideration and (2) the conditions under which the inference is examined. Thus, in the first case, we find the literature rife with observations that while there is little evidence of "automatic" generation of certain "forward-looking" or "elaborative" inferences (e.g., Alba & Hasher, 1983; McKoon & Ratcliff, 1981; Singer & Ferreira, 1983), inferences involving certain likely causal consequences may be "weakly generated" (e.g., Keenan, Baillet, & Brown, 1984; McKoon & Ratcliff,1986). Further, it has been observed that certain "backward-looking" inferences such as bridging inferences or the inference linking an overt anaphor with its antecedent may be reliably and automatically generated (Clark & Sengul, 1979; Corbett & Chang, 1983; McKoon & Ratcliff, 1981, 1986). Indeed, the literature on text processing details an amazingly variable array of different elaborative functions for different "types" of inferences. However, it has also been amply demonstrated that inference elaboration appears to be a function of the conditions under which these inferences were examined. Roughly speaking, this latter issue has three aspects: the beliefs the listener or reader generates about the task to be