This thesis presents the first detailed explanation for the entire angular function of the tilt aftereffect.
Chapter 1 explained the basic effect, including a simple demonstration. After adapting to an oriented pattern, subsequent patterns of slightly different orientation appear to have a much larger orientation difference (the direct effect). Subsequent patterns nearly perpendicular to the adaptation pattern instead appear to be tilted towards the adaptation orientation (the indirect effect).
Chapter 2 surveyed previous research on the TAE, including other computational models of the TAE. It was argued that the lateral inhibitory theory is the only viable explanation proposed so far for the direct tilt aftereffect. It was also argued that previous models have not convincingly explained the presence of the indirect tilt aftereffect.
Chapter 3 explained the RF-LISSOM system of Sirosh and Miikkulainen (1994a); Sirosh and Miikkulainen (1997) in detail, including the network architecture, activity calculation, and connection weight learning mechanisms. It also explored the biological plausibility of the model, suggesting that the current model is valid for the typical case of high-contrast inputs only. It also suggested possibilities for the role of a self-organizing model in the explanation of environmental and genetic factors observed in development.
Chapter 4 described how a realistic cortical orientation map was trained using parameters set by Sirosh (1995). It also examined possible algorithms for estimating the orientation perceived using such a map. It was shown that taking the vector average of the orientation preferences of activated neurons represents a sufficiently accurate estimate of the actual orientation.
Chapter 5 described the aftereffect experiments and results using the self-organized orientation map, and demonstrated that the results from the model are nearly identical to the psychophysical data for the tilt aftereffect in humans. It presented a quite detailed analysis of the mechanism of the TAE in RF-LISSOM, and this analysis represents predictions of phenomena that may be seen in human or monkey cortex.
Chapter 6 related psychophysical and biophysical evidence with the results found for the TAE in the RF-LISSOM model. It explored possible cortical implementations of the self-organizing processes in the model, and presented specific experimental predictions that could be used to verify or refute these proposed mechanisms. Many directions for future work were suggested, including an examination of tilt illusions and studies of aftereffects in other modalities.