Predict or bridge? Investigating inference making strategies in skilled readers

Inference-making is one of the crucial skills necessary to achieve good reading comprehension. This thesis explored which inference-making strategies are employed by skilled readers when they integrate new information in their mental text representation. Specifically, I examined how context affected readersâ choice of inference-making strategy and its time-course in the word-to-text integration process. Four eye-tracking experiments were carried out to observe inference-making as reading happens. By manipulating the amount of supporting context, the first three experiments studied how different properties of target word referents presented in the previous context influenced generation of bridging inferences. In particular, the effect of having lexically explicit, direct referent as opposed to a contextually-related but indirect referent was examined. The results of the first three experiments conflicted suggesting that readers either used predictive inferences instead of bridging to integrate the target words, or that context offered no advantage for the drawing of bridging inferences. To address this question, Experiment 4 examined whether participants integrated the target words by making predictive inferences even in weakly constraining contexts. Participants were presented with texts where context provided either strong or only weak indication of future text development, which was then confirmed or disconfirmed by the target word. The results suggested that even if context provided only limited indication of future development, skilled readers would generate predictive inferences to integrate new concepts. Only when predictive inferences fail to integrate new words, bridging inferences would be employed.