1. The dynamics of the subunit mechanism of individual cat Y retinal ganglion cells is investigated. In order to isolate the response of the nonlinear subunit mechanism, the visual stimuli were sine gratings of a spatial frequency sufficiently high so that contrast-reversal of the grating elicited no fundamental response in any spatial phase. For study of the nonlinear subunit mechanism, the contrast of the spatial sine grating was varied in time by a temporal modulation signal, consisting of either a square-wave or a sum of sinusoids.
2. The responses of 23 Y ganglion cells (16 on-centre, 7 off-centre) to these two stimulus types were measured at a range of contrasts. Responses to the sum-of-sinusoids signal were characterized by the second-order frequency kernel. The overall size of the second-order frequency kernel was approximately proportional to contrast. The deviation from proportionality suggested a power-law scaling, with a power in the range 0.8 to 0.9.
3. Square-wave responses, as characterized by the post-stimulus histogram, demonstrated identical responses at both reversals of the grating. A similar contrast-dependence was observed in the overall size of the square-wave responses.
4. In order to attempt to predict the square-wave responses from the sum-of-sinusoids responses, the second-order frequency kernel measured at each contrast level was fitted with a lumped linear-static nonlinear-linear model. In 18 of 23 cells (11 on-centre, 7 off-centre), this model provided an adequate description of the response to the sum-of-sinusoids stimulus. In these cells, the linear-static nonlinear-linear model accurately reproduced the square-wave response.
5. In the remaining 5 ganglion cells (all on-centre), the second-order frequency kernel could not be fit by a linear-static nonlinear-linear model. This diversity of dynamical properties among Y cells was not apparent from the responses of these Y cells to the square-wave temporal stimulus.
6. In the 18 Y ganglion cells that were fit well with the linear-static nonlinear-linear model, substantial variation of the dynamical parameters was found. However, there were systematic differences between the dynamics of the typical on-centre and off-centre ganglion cells. These differences relate to both linear stages of the model, and are not merely consequences of the lower firing rate of the off-centre cells.
7. In these ganglion cells, the dynamics of the first linear filter was similar to the linear dynamics of the X cell centre. There was a tendency for greater transience in the first linear filter of the Y cell compared with that of the X cell. However, most of the greater transience of the Y cell response originates in the second linear filter. The nonlinearity per se does not contribute to the transience of the Y cell.