Changes in the selectivity of macaque inferior temporal neurons during fast shape discrimination learning

Abstract

The shape selectivity of inferior temporal neurons is thought to underlie shape discriminations. We determined whether the shape selectivity of macaque TE neurons improves while the animal learns to discriminate minute differences between shapes. Prior to recordings, 2 rhesus monkeys were trained in the discrimination of 10 shapes in a two-choice discrimination task. Responsive neurons were searched while the animal was discriminating these shapes. Once a responsive neuron was found, the animal had to learn to discriminate the familiar shape the neuron responded to from a novel shape that differed slightly from the familiar one. The correct response to the novel shape was the alternative of the one required for the familiar shape (e.g. leftward response for familiar shape .→ rightward for novel). The neural responses were recorded while the monkey learned to discriminate the shapes in a single session. The response strength of many neurons changed during learning, with 31% of 49 neurons showing a significant change in shape selectivity. Only 2 neurons showed an increase in shape selectivity with learning. The mean absolute difference between the responses to novel and familiar shapes, averaged over all neurons, was not affected by the training, although the behavioral performance measured in the same trials increased from 62% (first 40 trials) to 83% percent correct (after 200 trials). It is highly unlikely that the changes in shape selectivity observed in a minority of neurons underlie the improvement in behavioral discrimination since the neuronal changes were also observed in a 3rd monkey that was not trained in discrimination and was merely performing a fixation task. These results suggest that although single TE neurons may show changes in shape selectivity when stimulated repeatedly with the same shapes in a single session, these changes are unrelated to behavioral discrimination improvements, but reflect automatic adaptations! of shape selectivity.