Troxler fading. In experiment 5, with a low adaptation contrast, a stimulus would periodically fade (Troxler fading), even without any CFS stimulus present. A clear, positive correlation between visibility duration and afterimage duration was again observed (R² = 0.8, P < 0.005, slope = 0.25). Error bars SEM.

Experiment 2: Changing perceptual load. The amount of attention paid to the afterimage-inducing stimulus was manipulated by using a difficult or easy RSVP task at fixation, leading to low- and high-attention conditions respectively. (A) Increased attention to the inducer stimulus led to decreased afterimage durations, whereas increased visibility led to increased afterimage durations (both P < 0.03, two-way repeated-measures ANOVA). (B) A table with mean ± SEM afterimage durations, and P values of statistical comparisons based on paired one-tailed t tests.

Dependence on visibility with constant stimuli. (A) In experiment 4, with low CFS contrast (∼10%), about half of the trials were reported to be completely invisible. When we divided the data into completely invisible, and partly/completely visible (n = 10 subjects), visible trials produced longer afterimages than invisible trials (1.79 ± 0.23 s vs. 1.52 ± 0.27 s, P = 0.02 one-tailed t test). (B) When divided into 10 bins (per subject and then averaged over subjects) according to the actual reported visibility duration, there was a clear correlation (R² = 0.96, P < 0.001, slope = 0.13). Error bars SEM.

Effects of CFS contrast. (A) The CFS mask started to have a suppressing effect on afterimage durations at contrasts higher than ∼6%. This is also when the mask started to have effects on the visibility of the afterimage inducer, which was kept at a fixed contrast of 34%. (B) The difference in afterimage duration between high- and low-attention conditions. The effects of attention were present without suppressor CFS (i.e., zero contrast), and for CFS contrasts larger or equal to 12%. Stars represent significance at the 0.05 level for one-tailed t tests, e.g., P < 0.025 for 0% and 100% contrast. Bars are SEM.

Experiment 1. (A) Our study used a 2 × 2 full-factorial design, independently manipulating the levels of attention and visibility. (B) Each trial started with 4-s peripheral adaptation of a Gabor patch. Attention was modulated by having observers report on the perceptual (dis)appearance of the inducer (high-attention condition) or count the number of Xs that appeared in a central RSVP task (low-attention condition). Perceptual visibility was modulated by showing, or not showing, a rapidly flickering and rotating high-contrast stimulus in the eye contralateral to the afterimage inducer, while the adaptor was physically present throughout the trial (23). After the induction phase, observers indicated how long their afterimage lasted, by pressing and releasing a button. Finally, they entered how many Xs were counted (or they ignored the question in the high-attention conditions). (C) Afterimage durations depend on visibility and attention levels (n = 13 subjects; each subject has a different color). When the inducer stimulus is highly attended, afterimages are shorter than when the stimulus is not/slightly attended (in both visible and invisible conditions). Visible trials have longer afterimage durations than invisible trials, for both high-attention and low-attention conditions. (D) A table with mean ± SEM afterimage durations, and P values of statistical comparisons based on paired one-tailed t tests.