Résumé
It is widely assumed that the integration of orientation contrast across spatial gaps within the long-range regime is not selective to the contrast sign of the individual stimuli. Probabilistic models of perceptual integration, however, suggest that long-range spatial integration should be, if not selective, at least sensitive to local contrast signs. To clarify this issue, we tested predictions of a model based on conditional probabilistic weights of identical and opposite contrast signs in a simple spatial configuration of two co-linear lines. Contrast detection thresholds of the target line presented either by itself (control condition) or simultaneously with the co-linear inducer (test condition) were measured. The contrast sign of targets and inducers was varied so that all four possible combinations of signs were produced in the test conditions: (1) dark target with dark inducer, (2) dark target with bright inducer, (3) bright target with bright inducer and (4) bright target with dark inducer. The contrast intensity (Weber ratio) of dark and bright inducers was identical. The coaxial distance between target and inducer was constant in each of two experiments, testing for two distances that corresponded to an angular separation within the long-range domain of spatial integration as defined previously. It is found that targets and inducers with identical contrast signs produce significantly stronger facilitating effects on detection than stimuli with opposite signs. The data closely match predictions consistent with those of a probabilistic model of line contrast integration across spatial gaps and contrast signs within the long-range regime.