Compression of visual information in redundancy masking follows grouping and segmentation

In typical scenes, the available visual information exceeds the processing limits of the visual system.To reduce information, the visual system organizes visual input according to various principles. Forexample, in visual grouping local elements are grouped into larger wholes. Recently, it was shown howthe visual system compresses redundant information by ‘redundancy masking’ which reduces thenumber of perceived items in repeating patterns. For example, when three identical items are presentedin the periphery, observers often report perceiving only two items. In the current study, we examinedthe role of grouping and segmentation in redundancy masking. In particular, we asked whether visualgrouping determines –and precedes— the units in which information is compressed in redundancymasking. Stimuli consisted of arrays of 3-5 vertical lines, briefly presented in the left or right visualfield. We varied the contrast polarity of the lines creating the following patterns: uniform arrays,alternating arrays, and subgroups of identical lines with a different line at the stimulus edge. Observersreported the number of lines and then indicated the perceived feature values for each reported line. Ourresults revealed redundancy masking in all conditions. Importantly, redundancy masking always cameafter segmentation and grouping: lines were masked predominantly within segmented subgroups ofidentical lines. The reported patterns were highly systematic, even when redundancy masking occurred:The presence/absence of the two contrast polarities (Contrast), the stimulus edge (Edge) and theirordinal frequencies (Ratio) were largely accurately reported. Hence, important stimulus informationremained largely intact.We suggest that redundancy masking comes after grouping and segmentation, and that thevisual system compresses redundant information in the visual periphery by reducing the number ofperceived identical items while preserving key stimulus features.