Forward inferences: From activation to long‐term memory

Abstract

Past research has been inconsistent regarding the extent to which forward inferences are activated and encoded during reading. To investigate the prevalence and the time course of forward inferences, 3 different tasks were employed. In Experiment 1, participants’ naming times were facilitated to a probe word when it represented a predicted action, both when that action was highly predictable and when it was only moderately predictable. In Experiments 2 and 3, participants were slow to read a subsequent sentence that contradicted the intended inference, indicating that the inference had been encoded and retained in working memory in both the high‐and low‐predictability conditions. In Experiment 4, the results of a recall task suggest that the high‐predictability forward inferences were encoded into the long‐term memory representation of the text. These findings suggest that forward inferences may be more prevalent and more persistent than has been indicated previously.