Camouflaged animals remain inconspicuous only insofar as they remain static. This demonstrates that motion is a powerful cue for figure–ground segregation, allowing detection of moving objects even when their luminance and texture characteristics are matched to the background. We investigated the neural processes underlying this phenomenon by testing the responses of neurons in the middle temporal area (MT) to ‘camouflaged’ bars, which were rendered visible by motion. These responses were compared with those elicited by ‘solid’ bars, which also differed from background in terms of their mean luminance. Most MT neurons responded strongly to camouflaged bars, and signaled their direction of motion with precision, with direction-tuning curves being only slightly wider than those measured with solid bars. However, the tuning of most MT cells to stimulus length and speed depended on the type of stimulus – in comparison with solid bars, responses to camouflaged bars typically showed more extensive length summation, weak end-inhibition, and stronger attenuation at high speeds. Moreover, the emergence of direction selectivity was delayed in trials involving camouflaged bars, relative to solid bars. Comparison with results obtained in the first (V1) and second (V2) visual areas, using similar stimuli, indicates that neural computations performed in MT result in significantly stronger and more accurate signals about camouflaged objects, particularly in situations in which these are relatively large and slow moving. These computations are likely to represent an important step in enabling cue-invariant perception of moving objects, particularly in biologically relevant situations.