No pooling, no averaging: How varying the number of identical Gabors modulates orientation discrimination in the periphery

In crowding, flankers interfere with target perception. Pooling models suggest that features of the target and the flankers are averaged. A typical example is the (compulsory) averaging of orientation signals of peripherally presented Gabors. Here, we show that averaging accounts fail when the Gabors have identical orientations. We measured orientation discrimination thresholds for peripherally presented single and identical Gabors arranged in ‘snake-like’ and ‘ladder-like’ arrays (1, 2, 3, 5 or 7 Gabors). Gabor arrays were arranged radially or tangentially, typically yielding strong and weak crowding, respectively. In Experiment 1, participants were not informed about the identical orientations; in Experiment 2 they were informed prior to the experiment. In Experiment 3, participants reported the orientation of the innermost Gabor, and indicated whether all Gabors had the same orientation. In the radial conditions, where strong crowding was expected, we found a strong increase of thresholds with increasing numbers of Gabors in the radial snake conditions in both Experiments 1 and 2. Thresholds for radial ladders did not vary with the number of Gabors in Experiment 1, but increased slightly for larger numbers of Gabors in Experiment 2. In the tangential condition, thresholds did not vary significantly with the number of Gabors. In Experiment 3, observers frequently reported varying orientations of the presented (identical) Gabors. Reported orientation variation was higher in the radial snake (50% of the trials) than ladder condition (20%). A control experiment where participants adjusted the perceived orientation of each Gabor confirmed larger variance in the snake than in the ladder condition. Our results show that neither the number nor the orientation of Gabor arrays alone predicts performance when the Gabors have identical orientations. Observers perceived varying orientations in uniform displays. Averaging accounts fail when targets have identical orientations.