The visual system is constantly faced with the problem of identifying parti
ally occluded objects from incomplete images cast on the retinae. Phenomeno
logically, the visual system seems to fill in missing information by interp
olating illusory and occluded contours at points of occlusion, so that we p
erceive complete objects. Previous behavioural [1-7] and physiological [8-1
2] studies suggest that the visual system treats illusory and occluded cont
ours like luminance-defined contours in many respects. None of these studie
s has, however, directly shown that illusory and occluded contours are actu
ally used to perform perceptual tasks. Here, we use a response-classificati
on technique [13-20] to answer this question directly. This technique provi
des pictorial representations - 'classification images' - that show which p
arts of a stimulus observers use to make perceptual decisions, effectively
deriving behavioural receptive fields. Here we show that illusory and occlu
ded contours appear in observers' classification images, providing the firs
t direct evidence that observers use perceptually interpolated contours to
recognize objects. These results offer a compelling demonstration of how vi
sual processing acts on completed representations, and illustrate a powerfu
l new technique for constraining models of visual completion.