Figure 7. Parallels between Behavioral Acquisition of Learned Associations between Pairs of Visual Stimuli (top panel) and Changes in the Selectivity of Cortical Neurons for the Paired Stimuli (bottom panel).

Top: Average performance as a function of trial number on a task in which subjects were instructed to choose a stimulus that we arbitrarily paired with a cue. For example, if subjects were shown stimulus A, they would be rewarded only for choosing stimulus B in the presence of distracters. Initial performance in this task is always at chance because the subject has no way to predict the pairings we impose. After a few dozen trials, performance begins to increase and ultimately reaches a high level.

Bottom: We predicted that concomitant with behavioral evidence of learning, we would see changes in the visual responses of neurons in the inferior temporal (it) cortex (see Figure 2) to the paired stimuli. Plotted here is the trial-by-trial change in a measure of the similarity of response magnitudes to the paired stimuli (A and B). As learning proceeded, the responses to paired stimuli became more similar to one another and the responses to unpaired stimuli became less similar. We believe that these neuronal response changes reflect ongoing modifications to the local circuitry in the it cortex, through which long-term memory for the learned pairings is achieved.

All data are averaged over many learning sessions, experimental subjects, and cortical neurons. Figure data: Adam Messinger, Larry R. Squire, Stuart M. Zola, and Thomas D. Albright, “Neuronal Representations of Stimulus Associations Develop in the Temporal Lobe during Learning,” Proceedings of the National Academy of Sciences 98 2001): 12239–12244.