Abstract
Behavioral responses in near-threshold perceptual tasks and response times in reaction-time tasks are highly variable. Some of this behavioral variability is likely to come from variability in the relevant sensory representations. Indeed, trial-to-trial covariations between neural and behavioral responses have been observed in several extrastriate areas, but it is unclear whether they can also occur in V1. To address this question, we examined the relationship between V1 responses and the behavior of macaque monkeys performing a reaction-time visual detection task. Monkeys detected a small peripheral Gabor patch whose contrast varied from trial to trial around perceptual threshold. The target appeared on half of the trials and detection was indicated by a saccade to the target location. We used metal microelectrodes to simultaneously measure single units (SU), multi-units (MU) and local field potentials (LFP) from V1 of three monkeys. In general, our results were similar for all three measurements. We first computed the trial-to-trial covariations between behavioral choices and V1 activity during a fixed interval shortly after stimulus onset. Weak but significant covariations, or choice probabilities, were observed in target present trials (i.e., responses to targets of identical contrast tended to be higher in hits than in misses). No similar effect was found in target-absent trials. We also examined the covariations between V1 activity and reaction times (RT), using a sliding window to integrate neural responses. Starting shortly after target onset, we observed weak, but significant, negative correlations between neural responses and RT (i.e., RT tended to be shorter in trials where the response was higher). The significant RT correlations occurred earlier, and were shorter lived, in easy trials than in difficult trials. Overall, our results demonstrate a surprisingly tight link between V1 responses and behavioral performance, suggesting a central role for V1 in reaction-time visual detection tasks.
This work was supported by National Eye Institute Grant EY016454, National Science Foundation Grant ITR/IMG 0427372, and a Sloan Fellowship (E.S.). We would like to thank Shao-Ying Cheng for assistance with data collection, and W. Bosking, Y. Chen, S. Khurana, Michelson, C. and Z. Yang for assistance with experiments and for discussions.