Multidimensional functional learning may be studied by means of a general class of tasks in which a multidimensional stimulus pattern is functionally related to a criterion. Subjects learned to predict a numerical criterion value from three numerical cues, varying in importance. In addition to simple outcome feedback, several sophisticated methods of computer-controlled feedback were employed. Results showed that a task with nonlinear cue-criterion relationships was not learned by discrete-trial outcome feedback whether or not the learners had active control over the stimulus sequence, and whether or not the cues were varied randomly or systematically over trials. Computer displays showing both the task characteristics and the learner's response characteristics produced effective learning. So did partial differential feedback (PDF), a continuous form of feedback in which the stimulus characteristics are controlled by the subject and the numerical criterion is replaced by a tonal frequency. These results suggest that appropriate feedback might enable people to learn far more complex functional relationships than have been thought possible.
Discussion(0)
No comments yet. Be the first to comment.