Contour integration and contour segmentation are two opposite requirements of visual perception that presumably involve competitive processes. To study this issue, we used two collinear line-segments separated by a gap (35′ and 8.8′ of arc), translating clockwise or counterclockwise along a circular path, and measured the ability of human observers (n=13) to recover the direction of line-ends that signal contour discontinuities. Occlusion by invisible masks that partially covered the line-segments was used to measure direction discrimination independently for inner line-ends (ILE) and outer line-ends (OLE). The results —percent correct and response time— indicate that processing ILEs' motion takes longer (by 60 msec; p<0.005) and yields more errors (by 18%; p<0.005) than processing OLE's motion. These differences are greatly reduced when line-segments are spatially offsets or are at an angle (60°) to each other, suggesting they strongly depend on segments' alignment. Psycho-pharmalogical testing was then used to assess the contribution of inhibition to these effects, using Lorazepam, a benzodiazepine that facilitates the fixation of GABA on GABA-A receptors. Observers (n=16) were tested in the ILE and OLE conditions before, during and after an intake of Lorazepam (0.038 mg/kg). In top of an expected global sedative effect of Lorazepam —yielding lengthened response times and increased error rates—, processing the motion of ILE and OLE was differentially affected, suggesting that Lorazepam boosted inhibition biases the competition between contour integration and segmentation. A simple model, in which facilitation through long-range horizontal connections (contour integration) competes with short-range end-stopping (contour segmentation), is proposed to account for the data.