Figure 1

Contingency matrix with the four trial types that could be presented on the contingency learning task. There are four different pieces of information or trials (“a”, “b”, “c”, and “d trial) as a function of whether or not the cause and the effect are present (left panel). From this information, a measure of contingency (Δp)⁴² can be computed as the probability of the effect conditional on the cause occurrence minus the probability of the effect conditional on the cause absence (see example of these computations on the right panel). Positive values indicate a generative causal relationship between the events, and negative values indicate a preventative causal relationship. When both events are not related to each other, the index equals zero and the contingency is null. Causal illusions are usually detected on null contingency settings when the cause and the effect appear with a high probability.