Abstract
We describe and quantify an unusual motion aftereffect (MAE), induced by flickering the adapting pattern, not by moving it. Quite unlike the classic MAE, it reverses direction when the test contrast is inverted. Background. Adapting to an image patch ramping from dark to light over time (‘brightening’) makes a steady test patch appear to be dimming, and vice-versa (Anstis, 1967). Superimposing this temporal afterimage on a test edge makes the test image appear to move (Anstis, 1990). Methods. We adapted to sine-wave gratings (0.3 c/deg) with sawtooth contrast modulation (1–15 Hz), and we nulled the motion seen in a stationary test grating whose spatial phase was offset ±900 from the adapter. We used a 3-choice task to minimize response bias, and found the null point by maximum likelihood. Results. As an index of aftereffect strength, nulling contrast (i) increased nearly linearly with adapting contrast, (ii) varied little with test contrast, and (iii) was a bandpass function of sawtooth flicker rate, peaking at 4 Hz. (iv) It doubled when test duration doubled from 0.5 to 1s, implying that the key factor in nulling was the temporal gradient of the nulling stimulus. (v) In dichoptic viewing, aftereffect strength for the non-adapted eye was about 20% of the adapted eye, implying weak interocular transfer. Conclusions. Models for direction selectivity combine input filters that encode spatial & temporal luminance gradients. In our model these non-directional filters also adapt, and evoke negative afterimages that represent illusory temporal gradients. These combine with spatial gradients of the test image to form illusory motion signals. This model fits the data accurately, and suggests a fast temporal-derivative filter (peak 15 Hz) with tuning similar to M-cells in the primate retina or LGN. The site(s) of this non-directional adaptation may be largely in monocular pathways, before binocular combination.
Acknowledgement: The Leverhulme Trust