Abstract
At the level of the primate LGN it is well recognized that the parvocellular (P) and magnocellular (M) neurons constitute two parallel pathways that are physiologically distinguishable from each other. Each pathway can be further divided into those cells that preferably respond to contrast increments (ON-center), and those that preferably respond to contrast decrements (OFF-center). But what are the relative contributions of ON and OFF channels in a contrast detection task? Do OFF neurons represent contrast decrements the same way that ON neurons represent contrast increments? We recorded from LGN P ON and P OFF neurons in awake monkeys while they performed two-alternative, forced choice (2-AFC) contrast increment and decrement detection tasks, and found that: 1) OFF-center neurons were significantly more sensitive in contrast decrement detections (neurometric threshold = 47% contrast), compared with ON-center neurons in contrast increment detections (neurometric threshold = 67% contrast), 2) OFF-center neurons had consistently lower choice probabilities in contrast decrement detections (0.51) than ON-center neurons in contrast increment detections (0.54), 3) OFF-center neurons had, on average, lower Fano factors (0.9) than ON-center neurons (1.1), and 4) OFF-center neurons had shorter onset latencies (0.024 s) than ON-center neurons (0.04 s), but the neural sensitivities of both ON and OFF neurons peaked at around the same time (0.07 s) after stimulus onset. Thus there exists an interesting functional asymmetry between the ON and OFF channels in contrast detection: compared with ON neurons, the faster, more sensitive OFF neurons (i.e. with shorter latency, lower threshold, and lower Fano factor) were less correlated with the behavioral choice of the animal (i.e. with lower choice probability). Different possibilities, such as perceptual learning, adaptation, task strategy, cortical readout, distinctive temporal dynamics, and special interneuronal correlation structures, were subsequently explored, and comprehensive models were proposed to explain these results.
Meeting abstract presented at VSS 2014