Prediction error and cognitive control in human causal learning

Abstract

Prediction error, the extent to which an outcome is unexpected, is thought to be an important determinant of the strength of learning. The blocking effect in human causal learning is consistent with this notion as it suggests that people fail to encode an association between a cue and outcome if that outcome is already expected on the basis of other predictive cues. In forming causal judgments, humans are clearly capable of drawing on other knowledge, such as information provided by testimony, or inferences derived from more complex reasoning processes like deduction. Nonetheless, it is plausible that associative memory processes and reasoning processes might both contribute to the expression of causal judgements in humans. If we take the assumption that both processes are involved, then this raises the question, to what extent do these processes operate independently—or alternatively, interact—to determine causal learning? This thesis examined one possible way in which these interactions may occur: that the predictions important for prediction error learning are subject to control by non-associative information. To that end, I provided instructions that manipulated the relevance of associative history to deriving predictions in a prediction error causal learning task. In Chapter 2, such instructions reduced the magnitude of the blocking effect in causal judgements, yet they did not appear to affect prediction error. That is, associative memory encoding was surprisingly resistant to instructed control. Chapter 3 examined the effect of providing a more explicit source of predictions which were either consistent with or conflicted with associative history. Finally, Chapter 4 investigated the influence of motivation on the efficacy of instructions that manipulate the relevance of associative history. The theoretical and practical implications of the findings in this thesis are discussed in the final chapter.