The apparent physical speed of an object in the field of view remains constant despite variations in retinal velocity due to viewing conditions (velocity constancy). For example, people and cars appear to move across the field of view at the same objective speed regardless of distance. In this study a series of experiments investigated the visual processes underpinning judgements of objective speed using an adaptation paradigm and video recordings of natural human locomotion. Viewing a video played in slow-motion for 30 seconds caused participants to perceive subsequently viewed clips played at standard speed as too fast, so playback had to be slowed down in order for it to appear natural; conversely after viewing fast-forward videos for 30 seconds, playback had to be speeded up in order to appear natural. The perceived speed of locomotion shifted towards the speed depicted in the adapting video (‘re-normalisation’). Results were qualitatively different from those obtained in previously reported studies of retinal velocity adaptation. Adapting videos that were scrambled to remove recognizable human figures or coherent motion caused significant, though smaller shifts in apparent locomotion speed, indicating that both low-level and high-level visual properties of the adapting stimulus contributed to the changes in apparent speed.

A great deal is known about how the visual system of the brain responds to stimuli received by the eye. The bulk of this research has used artificial stimulus sets such as sine-wave gratings, which allow precise parametric control of the visual properties driving lower level processing in striate and extrastriate cortex. For example, early cortical areas are known to contain neurons that respond selectively to the local retinal orientation and direction of simple visual patterns¹. However the visual system evolved to process images depicting more complex natural scenes, and neural circuits at higher levels of analysis in the cortex may be largely unresponsive to these artificial stimulus sets^2,3,4. For example, the perceived stability of visual properties such as size, shape, lightness and colour (the perceptual constancies) cannot be explained solely by responses in early visual areas that vary with retinal image parameters^{5, 6}, but may require high-level processes operating over extended areas of the visual field, involving large ensembles of neurons⁷.

In motion perception, the apparent speed of an object in the field of view remains constant despite variations in retinal velocity due to viewing conditions. For example, visual objects such as people and cars appear to move at the same objective speed regardless of viewing distance (velocity constancy^{8, 9}). Some researchers have viewed velocity constancy as an extension of size constancy, while others have suggested that the temporal dynamics of the image are important for maintaining velocity constancy^8,9,10,11. Little is known about how the responses of neurons in early visual areas of the cortex contribute to velocity constancy. The present experiments addressed this issue using a novel motion adaptation paradigm in which participants judged the speed of a common real-world action, human locomotion, after exposure to different kinds of adapting pattern. The speed of human locomotion was selected for study because it is particularly important for social interactions and is known to support subtle judgements of meaning, emotion and intent^{12,13,14,15,16,17}. The first two experiments show that prior viewing of speeded-up or slowed-down video recordings of locomotion causes changes in the perceived speed of locomotion in subsequently viewed video clips¹⁸. Later experiments investigate whether this adaptation effect can be explained in terms of known changes in the responsiveness of low-level neurons, or implicates higher-level processes involved in velocity constancy¹⁹. We tested whether adaptation depends on playback speed per se or on retinal speed, and then investigated whether image flicker plays a role. Although retinal stimulus parameters were found to be important, results were qualitatively different from those obtained in previously reported studies of low-level retinal velocity adaptation, and indicated that image temporal frequency properties contribute to maintaining speed constancy in perception.

Experiments 1 and 2: Adaptation to walking and running

In each test trial experimental participants viewed a short video excerpt taken from a recording of people walking along a local High Street, or running in a sports event (London Marathon). The videos contained moving figures at a range of distances, speeds and directions, as is typical in everyday scenes. They were shown at playback speeds ranging from slow-motion (0.48x) to fast-forward (1.44x) relative to standard playback speed (1x, which represents real-life speed). After viewing each clip the participant made a binary judgement as to whether the action in the clip appeared to be performed at a slower or faster pace than natural pace. From the pattern of responses to different test speeds we were able to estimate the playback speed which was judged as natural by participants (full details of experimental procedures are given in Methods).

We found that viewing of a slow-motion (SM) video for 30 seconds caused participants to perceive subsequently viewed clips played at standard speed as too fast, so playback had to be slowed down in order to appear natural. Conversely, after viewing fast-forward (FF) videos for 30 seconds, playback speed had to be increased in order to appear natural.

https://www.nature.com/articles/s41598-017-06841-5