Generalisation results for Experiments 2 and 3.

A) Results for the 2-Pair Condition of Experiment 2. B) Results for the 5-Pair Condition of Experiment 2. C) Results for Experiment 3 in which 3 training pairs were used. In each graph the numbers in the lower right corner indicate how many participants learned and the total number of participants for that condition. Axes and colour coding are the same as for Fig. 3. The results indicate that extrapolation of learning does not occur in a linear manner. The Mixture-of-Kalman-Filters performs better at capturing the generalisation results.