Fig 5. Learning rate results.

The learning trend versus trial number for the conditions of Experiment 2 and Experiment 3 plus an added condition in which only one shape-mapping pair was used in the training. A) Experimental Results. B) Predictions for the Mixture-of-Kalman-Filters Model. Shaded areas represent the standard error across participants. The learning rate drastically decreases with increased numbers of training pairs and the model qualitatively captures this finding.