On each trial, observers viewed a Gabor patch for 1.5 s, created by presenting a sine wave grating (Michelson contrast one, mean luminance same as background luminance) within a Gaussian envelope upon a gray background (35 cd/m
2). The relation between the grating's spatial frequency (0.25 c/dva) and the Gaussian's standard deviation (1 dva) was such that less than one grating cycle was visible, resulting in the half dark, half light appearance. The afterimage of such an adapter also looks like a Gabor patch but one that is bright where the adapter was dark and vice versa. We assessed the strength of the afterimage on each trial with a nulling method, in which the display after adaptation immediately changes into a pattern that is a photo negative of the afterimage or, in other words, a low-contrast version of the original adapting pattern itself. By varying the contrast of such a more null pattern across trials and asking observers which side of the display appears lighter, one obtains an index of afterimage strength in terms of the more null contrast required to cancel it out (Leguire & Blake,
1982; Georgeson & Turner,
1985; Kelly & Martinez-Uriegas,
1993). Our nulling procedure had one twist relative to this conventional procedure, motivated by the fact that some observers during pilot work had trouble focusing on the relatively weak and fleeting perceptual impression during the more null period and, instead, reported the appearance of the adapting image itself. To avoid this confusion, we modified the stimulus used during the nulling period. The purpose was to achieve a percept during nulling that could not be confused with the adapting Gabor patch, thus making it easier to instruct observers as to which time period to respond. Our stimulus during the nulling period consisted of a superimposed combination of a Gabor patch that would conventionally be used to null our observers' afterimages and a second Gabor patch that was similar but orthogonal to the more null. This second Gabor patch played no role in canceling the afterimage, and its Michelson contrast was fixed at 0.3. Importantly, however, the inclusion of this second component resulted in a percept, during nulling, of a type of dipole that could not be confused for the adapter stimulus. This dipole was characterized by a dark corner and a bright corner, and the perceived location of the bright corner (or, equivalently, of the dark corner) on a given trial was now diagnostic of whether the more null overpowered the afterimage or not. Accordingly, observers were asked to report on each trial which of four corners (top left, top right, bottom left, or bottom right) appeared brightest. Note that, on any given trial, only two out of these four options could possibly be correct because the orthogonal stimulus unambiguously ruled out the other two. Across consecutive trials, the adapting stimulus was alternately oriented vertically or horizontally (because of our particular more null method, this did not impact the observer's task), and more null contrast was chosen quasi-randomly from a preset range of values. During Experiment 1, these were 0, 0.12, 0.24, 0.36, 0.48, and 0.60; during Experiment 2, they were 0, 0.10, 0.20, 0.30, 0.40, 0.50, and 0.60. The contrast polarities of the adapter as well as of the orthogonal component during nulling were independently randomized across trials. In both experiments, each more null contrast was presented 12 times. The more null display lasted 0.5 s, immediately followed by a checker pattern aimed at masking any remaining afterimage. The intertrial interval was 3 s, and observers were given a self-timed break after every sequence of 20 trials. Trials were distributed equally across three sessions. Observers were instructed to fixate their gaze at the center of the display throughout each trial, and this was facilitated by a circle (radius 2.2 dva) that framed the stimulus area and a circular fixation mark at its center (radius 0.15 dva).