Abstract
Simoncelli & Heeger (1998) proposed a model of responses in MT that incorporates two stages corresponding to cortical areas V1 and MT; each stage includes a divisive contrast gain control or normalization. The model accounts well for many properties of MT neurons, but direct physiological evidence for a second stage of contrast gain control is lacking. To examine this question, we recorded from MT neurons in anesthetized, paralyzed macaques. We identified two responsive regions within each neuron s receptive field (RF), each 25–50% of the RF diameter, separated by 50–75% of the RF diameter. We studied responses to otherwise optimal pairs of gratings, of independently varying contrast, confined to the two regions. The separation between the regions minimized interactions in V1, allowing us to study MT neurons' summation behavior directly. We wished to know in particular whether contrast signals in one patch would modify the gain of contrast signals from the other. We modelled MT responses as the sum of the two regions' responses, each independently normalized, and asked whether a subsequent stage of normalization after summation was needed to describe the data. The normalization stages were represented by Naka-Rushton nonlinearities. For all of our MT cells, responses were better described by a model incorporating a second-stage normalization than by a linear sum of independently normalized signals. This was the case even when we only measured the first 200 msecs of each cell's response. We take this as evidence for a separate rapidly-acting stage of contrast gain control within area MT.
SimoncelliEPHeegerDJ(1998), A model of neuronal responses in visual area MT. Vision Res 38(5):743–761.