要旨
Vernier performance with moving targets is accurate independent of velocity. Is this inherent in ganglion-cell responses or must central mechanisms integrate activity to achieve this invariance? Precision of responses to bars and gratings was assessed by neurometric analysis. For magnocellular (MC) cells, spatial precision with bar stimuli was independent of movement speed; this was not so for parvocellular cells. For gratings, precision was maximal for MC cells at low temporal frequencies and increased above ~6 Hz. With slow stimuli, rates are low but the many impulses give precise spatial information. With rapid movement, peak rates are high but the few impulses give noisy signals. MC-cell spatial precision resembled human performance, so these features are utilised psychophysically. This constrains central motion mechanisms. The neurometric result required analysis over several hundred milliseconds. However, near-optimal performance is achieved with brief presentations (< 100 ms). This is resolved if central spatiotemporal filters extract a motion signal by combining/comparing activity of a small sub-mosaic of cells. Secondly, in the motion tasks, human performance correlated with cell signal-to-noise ratio. In other tasks such as flicker detection, human performance correlated with peak cell rates. This difference may arise from multiple detectors, or a single more sophisticated central comparator or detector.
