We have found that the timing of suppression within the receptive field from orthogonal masks is very fast. Cross-orientation suppression appears to act on the neuron even before the onset response from the CRF (by 6.4 ms on average). As we would expect, it does occur significantly after the offset response (by 13.5 ms on average), which is known to be a fast and reliable measure of a cell's earliest latency (Bair et al., 2002). In comparison with surround suppression in the same cells, suppression from within the receptive field occurs approximately 12 ms earlier than that from the surround. Moreover, the relationship between suppression strength and latency that is quite strong for the surround is either absent or very weak for cross-orientation suppression. These new results about cross-orientation timing have implications for the possible circuits involved.
Allison et al. (2001) suggested a possible extrastriate origin for cross-orientation suppression in the cat based on its preference for high temporal frequencies. Although connections between V1 and extrastriate cortex can be quite fast (Nowak and Bullier, 1997; Hupé et al., 2001; Girard et al., 2001; Movshon and Newsome, 1996), it seems unlikely that such a circuit could be fast enough to make a round trip to another cortical area before the onset response of most cells. The variation in response latency could possibly mask this circuit delay if the cells with the shortest latencies send signals to extrastriate cortex, which in turn project back to inhibit other cells in V1 with longer response latencies. If such a system were in place, however, we would expect neurons with the shortest response latencies to have some delay for cross-orientation suppression, while only those with longer latencies could show no delay or have suppression occur before response onset. We found that this was not the case - cross-orientation suppression was evident before response onset even in cells with the shortest response latencies. Allison et al. (2001) suggested the possible extrastriate origin based on their results from areas 17 and 18 in the cat, which both receive direct LGN input. Area V2 in the macaque does not receive a substantial input from the LGN, and it therefore seems highly improbably that signals from macaque V2 could feedback on V1 neurons fast enough to account for our results. We suspect that an earlier (i.e., feedforward) mechanism may be responsible for this type of suppression.
Another possible mechanism for cross-orientation suppression involves a pooled ``normalization'' signal that arises from cells within striate cortex (Heeger, 1992; Bonds, 1989). This type of mechanism has been suggested to underly multiple suppressive phenomena within striate cortex, including cross-orientation suppression and surround suppression (Heeger, 1992). If the same pool of inhibitory neurons were to underly both types of suppression, it would be expected that their response properties in terms of time course, gain control, and stimulus selectivity might match. This does not appear to be the case based on recent evidence. Sengpiel et al. (1997) found that the effect of cross-orientation suppression was best characterized as contrast-gain control (a rightward shift of the contrast response function), while the effect of surround suppression was better described as response-gain control (a downward shift of the contrast response function). Furthermore, most studies recognize that surround suppression is selective for the orientation of the grating in the surround (DeAngelis et al., 1994; Sengpiel et al., 1997), while cross-orientation suppression can be accounted for by a mechanism lacking orientation selectivity (DeAngelis et al., 1992). Finally, our results show that the time course of cross-orientation suppression and surround suppression is quite different. We therefore believe that cross-orientation suppression and surround suppression must arrive from quite distinct mechanisms.
Recently, Freeman et al. (2002) suggested that synaptic depression in thalamocortical synapses might underly cross-orientation suppression. They propose that a cortical basis is unlikely because the suppression is largely immune to visual adaptation and is engaged by gratings drifting faster than 20 Hz, which is above the cutoff frequency for most V1 cells. These two findings argue against a cortical basis for cross-orientation suppression, unless it is mediated by inhibitory interneurons which might have these properties. This is difficult to establish because these neurons might be rarely encountered with extracellular electrodes. Nonetheless, the Freeman et al. (2002) model is consistent with our findings on the timing of cross-orientation suppression. An alternative model using local cortical circuitry has been proposed by Lauritzen et al. (2001). They find that the addition of synaptic depression to their model does not substantially alter its behavior, nor change the success of their model in matching cross-orientation suppression data within a cortical inhibition framework. It appears that distinguishing between these two models, and in turn between the two proposed circuits, will require more knowledge about the properties of inhibitory interneurons in striate cortex. In addition, intracellular recordings and more theoretical work is necessary to prove or disprove that synaptic depression is the mechanism that mediates cross-orientation suppression. Our time course measurements might be an essential piece of data to constrain these models and ultimately pin down the mechanism for cross-orientation suppression.