Fig 3. Generalization performance.
n = 27 subjects. Drop-off in accuracy in novel-pair compared to learned-pair testing trials, as a function of training diversity. Concurrent diversity had a significant effect on this measure of generalization (pcorrected = .004, bootstrap test) whereas cumulative diversity did not significantly affect it (pcorrected = .29, bootstrap test). The plot shows individual subject accuracy (circles), group distributions of accuracy levels (violin), group means (thick lines) and standard errors (gray shading).