6.3 Specific predictions for experimental verification

  If tilt aftereffects occur in the human visual system based on the same mechanisms as in the LISSOM orientation map model, then there are certain features we would expect to see in the human data. Many of these have already been discussed here and in chapter 5. However, several other consequences of the model have not yet been investigated specifically in humans, and RF-LISSOM has provided much more detailed information than that so far available for humans or animals.

The RF-LISSOM model predicts that for the indirect effect, the number of V1 neurons responding to a test pattern should increase as adaptation progresses. The net response level should also increase. Other models such as satiation, fatigue, and the lateral axis theory (described in chapter 2 and section 6.2.4) would predict a decrease or no change in activity levels in V1 for that stimulus. In particular, the lateral axis theory would predict that changes would occur only at higher levels than V1. The RF-LISSOM prediction could be verified or refuted in humans by visualizing activity levels using optical imaging, visual-evoked-potential (VEP), or similar techniques. One would simply compare the response to a stimulus before and after adaptation. If cortical activity does indeed decrease for such a pattern, as it does in RF-LISSOM, it would be difficult to explain how this could arise from fatigued neurons or neurons with a buildup of some inhibitory substance.

If sufficient temporal and spatial resolution is available from the imaging process, plots like those presented for RF-LISSOM in section 5.4 could be computed in monkeys for comparison with the model. First, an orientation map would be computed using standard means (Blasdel and Salama, 1986; Blasdel, 1992a; Grinvald et al., 1994; Ts'o et al., 1990; Weliky and Kandler, 1995). Next, the response of the cortex would be measured for test patterns at orientations typical of direct and indirect effects, (e.g. 10° and 60° from the adaptation angle). The imaging technique would need very high temporal resolution to make this measurement without prompting significant adaptation to the test pattern. Test pattern presentations vary in different psychophysical experiments, but they are typically on the order of a few seconds, so an image with the required spatial resolution (see below) would need to be obtainable within that time. Methods like PET (positron emission tomography) which require integration over long time periods would be unsuitable (Sejnowski and Churchland, 1989).

Next, the cortex would be adapted to a stimulus of a particular orientation. Finally, the test patterns would be presented again, measuring the cortical response once more. The before-and-after activity plots could then be matched with the orientation map, showing exactly which areas change activity as a result of adaptation. In order to determine how the activity changes modify the overall orientation, the activity plots would need spatial resolution at least sufficient to resolve individual orientation columns, i.e. down to at least 0.1mm. When such a before plot is subtracted from an after plot, there should be a net decrease in activity for orientation detectors near the orientation used during adaptation, a net increase in activity of those with more distant orientations, and no change for very distant orientations.

If this type of spatial resolution is available, one would also be able to calculate perceived orientations as in section 4.5. The difference between the perceived orientations before and after should be within the range of TAE seen for human subjects. If this is found to be the case, it would represent strong support for the lateral inhibition theory of direct and indirect tilt aftereffects.

A further test of the model could be made by blocking intracortical inhibition mediated by GABA receptors with bicuculline (Sillito, 1979) and that mediated by GABA receptors with phaclofen (Pfleger and Bonds, 1995). Without inhibition, the sign of the TAE should be reversed and it should have a smaller magnitude (perhaps not even being measurable, if afferent plasticity is insignificant). Adaptation under those conditions should result in the afferent-only effects shown in figure 5.3. Note that blocking the inhibitory receptors does not necessarily prevent inhibitory adaptation (Vidyasagar, 1990), since changes could still occur in the excitatory connections to inhibitory interneurons. However, the functional consequences of inhibitory adaptation would no longer be apparent. It is not possible to perform such tests in humans, since the chemicals disrupt brain function. It may be possible to devise suitable protocols for testing the TAE in animals, at which point such tests could be performed. If similar effects are found in these animals as in the model, it would also represent strong support for the lateral inhibition theory of tilt aftereffects.