The Journal of Neurophysiology Vol. 86 No. 4 October 2001, pp. 2001-2010
Copyright ©2001 by the American Physiological Society

Responses to Task-Irrelevant Visual Features by Primate Prefrontal Neurons

 ¹Department of Physiology, Juntendo University, Tokyo 113-0033;  ²Brain Science Research Center, Tamagawa University, Tokyo 194-8610;  ³Department of Physiology, Nihon University, Tokyo 173-8610; and  ⁴Department of Neurology, University of Tokyo, Tokyo 113-8655, Japan

Lauwereyns, Johan, Masamichi Sakagami, Ken-Ichiro Tsutsui, Shunsuke Kobayashi, Masashi Koizumi, and Okihide Hikosaka. Responses to Task-Irrelevant Visual Features by Primate Prefrontal Neurons. J. Neurophysiol. 86: 2001-2010, 2001. The primate brain is equipped with prefrontal circuits for interpreting visual information, but how these circuits deal with competing stimulus-response (S-R) associations remains unknown. Here we show different types of responses to task-irrelevant visual features in three functionally dissociated groups of primate prefrontal neurons. Two Japanese macaques participated in a go/no-go task in which they had to discriminate either the color or the motion direction of a visual target to make a correct manual response. Prior to the experiment, the monkeys had been trained extensively so that they acquired fixed associations between visual features and required responses (e.g., "green = go"; "downward motion = no-go"). In this design, the monkey was confronted with a visual target from which it had to extract relevant information (e.g., color in the color-discrimination condition) while ignoring irrelevant information (e.g., motion direction in the color-discrimination condition). We recorded from 436 task-related prefrontal neurons while the monkey performed the multidimensional go/no-go task: 139 (32%) neurons showed go/no-go discrimination based on color as well as motion direction ("integration cells"); 192 neurons (44%) showed go/no-go discrimination only based on color ("color-feature cells"); and 105 neurons (24%) showed go/no-go discrimination only based on motion direction ("motion-feature cells"). Overall, however, 162 neurons (37%) were influenced by irrelevant information: 53 neurons (38%) among integration cells, 71 neurons (37%) among color-feature cells, and 38 neurons (36%) among motion-feature cells. Across all types of neurons, the response to an irrelevant feature was positively correlated with the response to the same feature when it was relevant, indicating that the influence from irrelevant information is a residual from S-R associations that are relevant in a different context. Temporal and anatomical differences among integration, color-feature and motion-feature cells suggested a sequential mode of information processing in prefrontal cortex, with integration cells situated toward the output of the decision-making process. In these cells, the response to irrelevant information appears as a congruency effect, with better go/no-go discrimination when both the relevant and irrelevant feature are associated with the same response than when they are associated with different responses. This congruency effect could be the result of the combined input from color- and motion-feature cells. Thus these data suggest that irrelevant features lead to partial activation of neurons even toward the output of the decision-making process in primate prefrontal cortex.