Figure 4.

Example predicted time-evolving d-prime (left panel) and psychometric function threshold (right panel) from each learning function fit to one participant for each task. The x-axis is overall trial number, including both initial training and generalization (the disjunction in the curves occurs on the trial in which participants switch from initial training to the generalization task). In the left panel, the y-axis is d-prime, or sensitivity, with greater d-prime indicating superior performance. In the right panel, the y-axis is a Weibull psychometric function threshold indicating the number of seconds needed to achieve 75% accuracy. Lower threshold values indicate better performance. Learning curves display characteristic differences between the power family (Power_3pars and Power_4pars) and the exponential family (Exponential_3pars, Exponential_4pars, and Weibull). In the left panel, the Weibull function is strikingly different from the other four functions, however, by allowing for an initial acceleration of learning and a sigmoid shape. Note that overlapping predictions, arising from models from the same family, may be difficult to discern.