Madeleine Schlag-Rey

A model for response latency in paired-associate learning☆

Abstract A 3-state model is proposed which is designed to account for both choice and latency data in paired-associate learning. Some properties of this model are derived, and some of its uses are illustrated with reference to a paired-associate experiment with 3 response alternatives and 12 stimulus items. Some empirical results not accounted for by the model are also described.

Observable Changes of Hypotheses under Positive Reinforcement

In mathematical models of concept learning it has consistently been assumed that positive reinforcement cannot lead to a change of the hypothesis determining the overt response. When hypotheses are experimentally identified and recorded along with positive and negative reinforcements of stimulus-response pairs, it can be shown that hypotheses may change after a positive reinforcement. Positive reinforcement has an information content for subjects that has not yet been adequately recognized in concept formation studies.

Interface of Visual Input and Oculomotor Command for Directing the Gaze on Target

Looking at a target seems to be a trivial task for the nervous system. It is performed almost automatically every second of active life. However, when one tries to explain in terms of pathways and nerve impulses how the gaze can be brought on target, a major difficulty becomes immediately apparent: How is the target selected? It is not a priori obvious in those circumstances in which looking serves to explore, to search, to anticipate something that may not yet be there. The whole scene has to be analyzed, patterns must be isolated and identified, and their significance assessed as a function of various contingencies. This involves a number of processes that are beyond the scope of neurophysiological studies today.

Latencies on last error in paired-associate learning

Response latency was measured in a paired-associate experiment with 3 response alternatives and 12 stimulus items. The mean error latency showed a rapid initial increase compared to the mean success latency. Further analysis revealed that the mean latency on the trial of last error was considerably larger than mean latencies on trials before or after the trial of last error.

Higher-order dimensions in concept identification'

In a complex concept identification task, solvers and nonsolvers tended to diverge in overall but not in trial-by-trial changes of hypothesis. Neither error frequency nor specific content of feedback was a major factor in concept attainment. The results suggest that Ss sample from subsets of hypotheses structured by higher-order dimensions.