Mask interrupts texture segregation but not grouping processes

Although texture segregation models confine border extraction to high-level processes, recent findings suggest the involvement of low-level mechanisms (Grieco et al, 2006 Vision Research 46 3526 ^ 3536). To investigate whether texture segregation is restricted to early-filtering mechanisms, or whether they involve later-levels of analysis, we measured how a backward-masking stimulus, known to interrupt re-entrant signals from high-level areas (Lamme et al, 2002 Journal of Cogni- tive Neuroscience 14 1044 ^ 1053), affects subject's accuracy in target-present (requiring texture segmentation) and target-absent (requiring distractors grouping) trials in a detection task, after a learning session without backward-masking. Here, we show that, although learning effects in texture segregation were specific to contrast polarity, indicating low-level filtering mechanisms, the improvements were abolished by a backward-mask presentation, suggesting the involvement of re-entrant signals from high-level areas. Instead, learning of distractor grouping was not specific to contrast polarity, nor was it affected by backward masking, indicating a high-level learning process. This finding raises an important challenge that must be faced in understanding the mechanisms underlying texture segregation and grouping by similarity.