| Summary: | Most contemporary theories of causal learning identify three primary cues to causality; temporal order, contingency and contiguity. It is well-established in the literature that a lack of temporal contiguity – a delay between cause and effect – can have an adverse effect on causal induction. However research has tended to focus almost exclusively on the extent of delay while ignoring the potential influence of delay variability. This thesis aimed to address this oversight. Since humans tend to experience causal relations repeatedly over time, we accordingly experience multiple cause-effect intervals. If intervals are constant, it becomes possible to predict when the effect will occur following the cause. Fixed delays thus confer temporal predictability, which may contribute to successful causal inference by creating an impression of a stable underlying mechanism. Five experiments confirmed the facilitatory effect of predictability in instrumental causal learning. Two experiments involving a different aspect of causal judgment found no effects of interval variability, but two further experiments demonstrated that predictability facilitates elemental causal induction from observation. These results directly conflict with findings from studies of animal conditioning, where preference for variable- interval reinforcement is routinely exhibited, and a simple associative account struggles to explain this disparity. However both a temporal coding associative account, and higher-level cognitive perspectives such as Bayesian structural inference, are compatible with these findings. Overall, this thesis indicates that causal learning involves processes above and beyond simple associations.
|
|---|
