Most evidence indicates that with supportive context, predictive inferences do become activated. The present experiments were designed to investigate whether the characteristics of the protagonist can mitigate against activation of a predictive inference, even when the immediately preceding context supports it. Participants read passages containing a detailed description of characteristics of the protagonist. The characteristics were either consistent, inconsistent, or neutral with respect to a subsequent predictive inference. This character information was followed by a context that supported the predictive inference. Experiment 1 demonstrated that predictive inferences were activated, except when they conflicted with the characteristics of the protagonist. Experiment 2 demonstrated that when an inference was activated, it was also instantiated into the representation of the text. These results are discussed in terms of the memory-based view of text processing.
