De Valois et al. (
1986) (but see the
Introduction) and Rossi and Paradiso (
1996) both reported that when they used a temporally modulated inducer surrounding a test area, illusory modulation of the test stimulus was seen only when the inducer modulation rate was below ∼3 Hz. This cut-off frequency was found for both the brightness- and color-contrast conditions. This finding was considered evidence that simultaneous contrast is a slow process.
If the simultaneous contrast is really “slow” as these studies suggested, how could we see the strong simultaneous contrast effect in our flashed stimuli? How can we reconcile the apparent difference between these studies and our own? At some point between stimulus input and our final perception, there is a bottleneck that slows the perception of simultaneous contrast. Where exactly does the bottleneck related to this perceptual slowing emerge in visual processing? One possibility is that a stimulus with a high temporal frequency is “invisible” to a spatially local interaction mechanism and is thus incapable of affecting the brightness/color of a surrounded region in the first place. However, this possibility does not seem very plausible because previous studies, such as those by Blakeslee and McCourt (
2008) and D'Antona and Shevell (
2009), offer evidence against it. D'Antona and Shevell (
2009) used a dynamic inducer whose color was modulated in a compound waveform
f1 +
f2, where
f1 and
f2 were high frequencies well beyond the cut-off frequency reported by De Valois et al. (
1986). Induced color modulation was perceived at the frequency of |
f1 −
f2| in a static test stimulus surrounded by the inducer. This result indicates that even if the modulation of the inducer is characterized by high frequency, the modulation pattern can induce simultaneous contrast. Together with these previous studies, the present study provides evidence against the above hypothesis. Thus, the bottleneck that prevents fast changes from appearing in the perception of simultaneous contrast is not located at the very first stage of spatial interaction.
We therefore assume that there should be a sluggish process after the initial spatial interaction that causes the final perception to be slower than the first crude calculation. We argue that, at some later stage, this sluggish process acts as the low-pass filter and that, after this process, information about the fast-alternating inducing stimulus disappears. This means that even in the display with rapid alternation, quickly alternating signals for simultaneous contrast emerge at the first spatial interaction only to dissolve at a later stage. This sluggish property may have been reflected in the results of De Valois et al.'s (
1986) and Rossi and Paradiso's (
1996) studies.
Previous studies have proposed several different candidates as the neural basis of simultaneous contrast. Rossi, Rittenhouse, and Paradiso (
1996), for example, found cells responding to brightness rather than physical luminance or contrast in the primary visual cortex of cats. The responses of some cells were attenuated when the surround modulation rate was above 3 Hz, consistent with the cut-off frequency for brightness contrast. Human fMRI studies have also demonstrated activities correlated with perceived brightness or perceived color in the primary visual cortex (e.g., Pereverzeva & Murray,
2008; Sasaki & Watanabe,
2004). Additionally, one visual evoked potential study demonstrated that an early (less than 100 ms) component of visual processing was correlated with perceived brightness (McCourt & Foxe,
2004). On the other hand, neural responses to achromatic/chromatic physical flickers are known to have sensitivity to very high flicker rates that are perceptually irresolvable (Gur & Snodderly,
1997; Jiang, Zhou, & He,
2007; Lee, Martin, & Valberg,
1989). We argue that the first spatially local interaction can be activated by fast neural responses to brief stimulation. Under the 10-ms condition in our experiments, the stimulus duration itself was brief, but the rate of this flash was less than 2 Hz (roughly once in 500 ms). This is presumably below the cut-off frequency that De Valois et al. (
1986) and Rossi and Paradiso (
1996) reported. Thus, the spatial interaction process is fast enough to yield signals pertinent to simultaneous contrast, but these are subsequently averaged out during the sluggish process involved in the perception of brightness and color.
Fast spatial-interaction signals may serve an important role in detecting an object in noisy and dynamic scenes by instantaneously enhancing edges. However, when it comes to determining the brightness/color of the object, the contrast signal is not a perfect cue. The signals might inform us of a fast-changing contrast at the edge, but it is unlikely that an object's surface would change so rapidly in a realistic environment. Thus, attributing the rapidly changing contrast signal to a change in the surround, not in the object, would lead to a more reasonable estimation. The subsequent low-pass temporal filter process probably serves to prevent those signals from directly affecting the perceived brightness/color of the object (cf. D'Antona & Shevell,
2009).
In short, the apparent difference between the previous and the current studies can be explained in terms of the location of the bottleneck. If we assume a fast-responding local interaction mechanism and a slow-responding process, or a low-pass filter, that prevents the fast signal from the initial interaction from emerging into the final percept, the discrepancy between the two studies can be reconciled, as described above.