ABSTRACT
The general consensus among inference researchers is that elaborative inferences are made only under certain circumstances, such as highly predictive contexts, if they are made at all (e.g., Garrod, O'Brien, Morris, & Rayner, 1990; Graesser, Singer, & Trabasso, 1994; McKoon & Ratcliff, 1992; Murray, Klin, & Myers, 1993; Potts, Keenan, & Golding, 1988; van den Broek, 1994). In the present experiment we tested: (1) whether elaborative inferences within moderately predictive contexts were made during reading, and (2) whether individual differences in working memory capacity were associated with differences in the length of time the elaborative inferences were maintained. Three and four sentence length paragraphs were designed so that one particular outcome of the story was more likely than other potential outcomes. The outcomes were 70% probable given the context of the paragraphs, based on normative data collected prior to running the study. Typically, outcomes in other studies of elaborative or predictive inferences are between 90% and 100% predictable within the given context. In the present experiment, the final sentence of all of the paragraphs explicitly stated the inference readers may have made. There were four conditions: an Inference condition and a Control condition, with two possible times of test, Immediate or Delayed. Forty-four sets of paragraphs were created, with each set containing a version corresponding to each condition. In the Immediate condition (three sentences in length), the sentence explicitly stating the inference immediately followed the sentence which may have elicited the inference (sentence two). In the Delay condition (four sentences in length), there was a filler sentence related to the story between the eliciting sentence (two) and explicit statement of the inference (sentence four), which neither confirmed nor precluded the inference. A "word-based priming" Control condition was created using the relevant content words from the experimental paragraphs in a slightly different context so that they did not elicit the inference. Immediate and Delay conditions were created for the Control as well as for the Inference condition.
