Perhaps the simplest mechanism for raising the amount of geniculate input required to evoke spikes would be to raise the spike threshold with increasing stimulus contrast. Based on the biophysical properties of cortical cells (see, for example, McCormick et al., 1985), it seems unlikely that the actual voltage threshold for spikes would change with contrast. Indeed, intracellular recordings from simple cells have shown directly that threshold voltage is invariant with contrast (Carandini and Ferster, 2000).
A second potential source for a contrast-dependent change in the effectiveness of the geniculate input is the frequency-dependent depression of LGN synaptic efficacy (Gil et al., 1997, Abbott et al., 1997, Markram and Tsodyks, 1996, Stratford et al., 1996). This depression increases with input firing rate, and thus would increase with contrast. While synaptic depression might partially alleviate the problem of orientation tuning widening with contrast, it is unlikely to solve the problem, for several reasons. First, the primary effect of depression is to lessen the difference between the tuning curves in Figure 4D. Depression cannot be strong enough to eliminate these differences, however, since in real simple cells spiking responses increase with increasing stimulus contrast (Figure 4A). Hence, while the problem may be alleviated, it will not be eliminated. Second, synaptic depression builds over a number of presynaptic spikes, so it would be unlikely to affect the response to transient stimuli such as a flashing bar.
A third way in which contrast could change the effectiveness of the geniculate input is for it to evoke a contrast-dependent hyperpolarization of the resting potential, thus increasing the size of the visually evoked depolarization needed to reach threshold. Contrast adaptation evokes just such a hyperpolarization (Carandini and Ferster, 1997), which may be due in part to a long-lasting potassium conductance (Sanchez-Vives et al., 1997). Such adaptation is orientation-tuned, however: it is not induced by stimuli with orientation orthogonal to the preferred (Allison and Martin, 1997). It also requires several seconds to develop fully (Carandini and Ferster, 1997, Ohzawa et al., 1985, Albrecht et al., 1984, McLean and Palmer, 1996). Thus, adaptation-induced hyperpolarization is unlikely to provide the required suppression of geniculate input.
One of the few remaining possibilities, and the one that we favor, is that the contrast-related modulation of the efficacy of the thalamic input necessary to explain contrast-invariant orientation tuning is supplied by stimulus-induced synaptic inhibition. When the feedforward model is extended to include inhibition, not only does it explain how inhibition could provide contrast invariance, but it also accounts for some of the known receptive field properties of inhibition in simple cells.