Reduced summation with common features in causal judgments

In three experiments human participants received training in a causal judgment task. After learning which patterns were associated with an outcome, participants rated the likelihood of the outcome in the presence of a novel combination of the patterns. The first two experiments used two conditions in which two visual patterns were associated with the outcome. In one condition these patterns shared a common feature. The third experiment only used the common feature condition. According to an elemental theory (Rescorla & Wagner, 1972) the response to the novel test pattern should have exceeded that made to the individual training patterns, a summation effect, and this effect should have been reduced by the addition of a common feature. Summation was observed but since the common feature condition abolished, rather than merely reduced, summation the results were not consistent with the Rescorla-Wagner Model (RWM) nor with a configural alternative (Pearce, 1994). Instead, it is necessary to consider models which allow the possibility of both elemental and configural strategies in causal learning. The Replaced Elements Model (Wagner, 2003) is a development of the RWM which can best predict the patterns of summation and summation failure in these experiments.