It is the usual case in cognitive modeling that a model's output is compared to the average of a number of subjects, in which case the enterprise of modeling is apparently to capture the behavior of the typical individual. Our approach is to administer wow simple tasks to each subject, using performance on those tasks as measure of individual ability. Those measure are then used as the values for parameters in an ACT-R model of more complex task, so that the model can predict individual performance on that task.