A remaining question surrounds the type of neural coding that
subtends global property scene aftereffects. Two basic possibilities
exist: an opponent model in which global properties are coded by
two neural populations coding the extremes of the dimension, or a
multichannel model in which different neural populations code for
different but overlapping levels of the dimension. Although both
models would predict the current data, future experiments could be
designed to distinguish them. Specifically, an opponent model
would predict that adaptation magnitude would be greater for test
stimuli that are very different from the adaptor, and the multichannel
model would predict that the greatest adaptation would occur
to test stimuli more similar to the adaptors (see Robbins, McKone,
& Edwards, 2007, for details on the logic).