| The utility of sensory information is characterized by how well it can be put to use in the control of behavior. Sensory information is represented by the activity of sensory cortical neurons. The response of these neurons engenders uncertainty. For the brain to efficiently decode an uncertain neural response, it should compute how likely different stimuli could have elicited that response. We present a simple and biologically plausible model that integrates the responses of sensory neurons and computes the desired stimulus likelihoods. The model provides the basis for optimal detection, discrimination, and identification of stimuli, and directly predicts the relative contributions that different sensory neurons make to different perceptual judgments.; We evaluate the implications of our model in the context of decoding the direction of visual motion from the responses of direction-selective neurons in area MT/V5 of the extra-striate cortex. Our model predicts that for coarse direction discriminations, neurons tuned to the directions of interest provide the most reliable information, whereas for fine discriminations, neurons with preferred directions off to the sides of the target directions (i.e. off-optimal neurons) are more reliable.; To test these predictions, in a psychophysical paradigm, we evaluate the effects of perturbing the pattern of activation across direction-selective neurons on human observers' choice behavior in a coarse versus a fine direction discrimination task. Our results suggest that the observers' perceptual judgments are based on the activity of neurons with a wide range of direction preferences and that the decoding strategy they adopt is consistent with the model's prediction in that in each tasks, observers seem to integrate neural responses according to their reliability for that task.; Finally, we show that when observers discriminate between nearby directions, their perception of direction of motion is shifted away from the discrimination boundary and towards the preferred direction of the off-optimal neurons that best discriminate those directions. We evaluate this misperception in the context of our decoding model and discuss why it is likely a direct consequence of the decoding strategy observers adopt during fine discriminations. |
