| The activation level of bridging and expectation inferences was measured during the comprehension of short scientific texts. The texts were presented via a rapid serial visual presentation (RSVP) procedure. Subjects performed a lexical decision on either a nonword or a true word after the RSVP presentation of each sentence. True words included both bridging and expectation inference words. The inference words associated with the texts were collected by a question/answering procedure (Millis, Morgan & Graesser, in press). Lexical decision times for each inference word were gathered under three types of context: when the text was ordered, scrambled, and unrelated to the inference word. The influence of pre-established semantic associations between the inference words and words in the text was statistically controlled for by multiple regression analyses. The pattern of results indicated that bridging inferences were generated during comprehension, whereas expectation inferences were not generated during comprehension. The results confirmed Bridging models of text comprehension and were discussed in regard to previous research on inference generation. |
