Abstract
We have previously reported the involvement of area MT in a coarse depth discrimination task: single MT neurons exhibit sensitivity comparable to that of monkeys, and trial-to-trial fluctuations in MT responses are correlated with monkeys' perceptual choices (VSS 2001). Here we asked whether MT is engaged in a stereo task with different information processing demands: we recorded from 98 MT neurons in two monkeys while they performed a fine depth discrimination (stereoacuity) task.
Visual stimuli were bi-partite (center/surround) random-dot stereograms tailored to match the receptive field properties of each neuron. Monkeys discriminated the relative depth between the two patches. The disparity of the surround patch was chosen to lie at the steepest slope of each neuron's disparity-tuning curve. Task difficulty was titrated around psychophysical threshold by finely varying the disparity of the center patch.
Neuronal and psychophysical thresholds were calculated using ROC analysis (e.g. Britten et al. 1992). Fifty seven neurons had thresholds significantly larger than the corresponding psychophysical thresholds, and we only found 13 neurons with thresholds lower than the monkeys'. The average neuronal to psychophysical threshold ratio was 1.76. We also tested whether MT responses were correlated with perceptual reports by computing “choice probabilities” (Britten et al. 1996). The average choice probability (0.52) was not significantly above chance (p>0.91), and equal numbers of neurons had significant choice probabilities above and below 0.5. Thus, we found no consistent relationship between neuronal responses and behavioral choice in this task.
Results are consistent with the idea that MT does not contribute to fine judgments of relative depth, which may be due to MT neurons not encoding relative disparity (VSS 2002). This contrasts with our previous findings in the coarse depth discrimination task, and argues for a task-specific role of MT in stereopsis.