Abstract
The one-dimensional nature of contours complicates the analysis of moving objects, because local velocity measurements become confounded with contour orientation. This “aperture problem” is relevant to neurons in the primary visual cortex, where small receptive fields permit only local measurements. The endpoints of contours, however, yield two-dimensional information, so they are not subject to the aperture problem. While the responses of cortical neurons to contours have been well studied, the processing of endpoints is more poorly understood. We recorded the responses of V1 neurons to sequences of flashed bars. Because the bars were much longer than the receptive fields, they generated potentially ambiguous motion signals (i.e. many more one-dimensional than two-dimensional motion signals). Using reverse correlation analysis, we obtained the neurons' responses to the positions of single bars and to the directions of two-bar sequences. The single bar responses indicated that many neurons clearly responded only to the endpoints of the bars. These endpoint-responsive neurons were generally end-stopped: they responded poorly to long bars centered on their receptive fields. For these cells, the directional responses to two-bar sequences were only modestly affected by the ambiguous contour signals. Furthermore, these directional responses were largely independent of stimulus orientation. End-stopped neurons also exhibited a characteristic time-course: they initially responded to a bar positioned anywhere in their receptive fields, but after 20–30 ms they responded mainly to the endpoints of the bars. A similar time-course was observed in the directional responses of MT neurons to the 2-bar sequences. MT neurons initially responded to the ambiguous contour signals, but later responses encoded the correct two-dimensional motion. We conclude that end-stopped V1 neurons accurately measure the direction of moving edges, and that they probably provide two-dimensional motion information to MT neurons.