Visual figure–ground discrimination in the honeybee: the role of motion parallax at boundaries

Abstract

Free flying bees were trained to collect a reward of sugar-water from a structured figure, placed at a randomly varying location on a sheet of transparent Perspex, positioned 5 cm above a structured Background. During subsequent tests, done in the absence of a reward, the bees’ landings on the boundaries of the figure, as well as within the figure and outside it, were recorded. The same bees were also tested with the figure placed directly on the background, thus eliminating the difference in height between the figure and the background. The results of both types of tests were then compared to identify and investigate the cues that bees use to detect a structured figure, when presented over a structured background. The structure of both the figure and the background were varied in a series of experiments, training a fresh group of bees in each experiment. A randomly structured figure presented against a randomly structured background cannot be detected by the bees unless it is raised above the background. A height difference of 2 cm is sufficient to elicit a rate of landings on the figure that is significantly higher than the chance level. The detectability of the figure does not depend upon the shape of the figure or on differences in density between the structures of the figure and the background. Thus, in detecting the raised figure, the only cue used by the bees appears to be the apparent motion of the figure relative to the background. The majority of landings on a raised figure occur at its boundaries. This shows that the visual stimulus that is crucial in detecting the figure is the local discontinuity in apparent motion that occurs at the boundary. We refer to this as ‘boundary parallax ’. In a series of experiments that used a striped background and a variety of structured figures, three different types of boundary parallax were offered to the bees. These were: (i) ‘covering parallax’, at a boundary in which stripes on either side of the boundary are parallel to the boundary; (ii) ‘shearing parallax’, in which stripes on either side are perpendicular to the boundary and (iii) ‘orthogonal parallax’, in which the stripes on one side are perpendicular to those on the other side. The bees performed very well at detecting raised boundaries that offered covering or shearing parallax, despite the fact that such boundaries are not readily discernible on the basis of their static geometry. On the other hand, bees performed poorly in detecting raised boundaries that offered orthogonal parallax, despite the fact that such boundaries are geometrically quite vivid to the human eye. We propose two neural models for the detection of boundary parallax that account for the sensitivity of bees to covering and shearing parallax and their insensitivity to orthogonal parallax.