Abstract
Surround suppression can reduce the perceived contrast of a stimulus when the neighboring surround is visually similar due to inhibitory neural interactions. While human observers are much better at discriminating small differences in orientation that are near cardinal rather than oblique (i.e., the oblique effect), it is unknown whether surround suppression might reveal a similar orientation-tuned anisotropy. To address this question, we conducted an intensive psychophysical study (5280 trials/observer, N=12) in which participants reported which of two identically oriented circular figures appeared higher in contrast: one was isolated while the other was surrounded by a bandpass-filtered pattern that differed in mean orientation (range, 0-90˚). Surround suppression at each relative orientation was measured using an adaptive staircase to adjust the contrast of the isolated figure until it matched the perceived contrast of the figure with surround. The orientation tuning of surround suppression was then estimated by fitting a Gaussian function to these psychophysical data. Surprisingly, we discovered that surround suppression was more broadly tuned around cardinal (SD=18.4°) than oblique orientations (SD=11.1°). To investigate how these findings might be accounted for under a divisive normalization framework, we evaluated multiple models by simulating neuronal populations with either greater representation or sharper tuning for near-cardinal orientations. The only model that could account for our findings included three key components: more near-cardinal than near-oblique units, local non-orientation-selective normalization of responses within the center population and within the surround population, followed by orientation-selective normalization across the center and surround populations to induce surround suppression. Models without local non-selective normalization could not account for our data, nor could models that only assumed sharper orientation tuning for units preferring near-cardinal orientations. Overall, our results reveal that surround suppression leads to a counterintuitive oblique effect that likely arises from multiple stages of divisive normalization in the human visual system.