Many potential prey species possess features that falsely convey a risk to the predator. One form of visual signal hypothesized to have this function are the circular features commonly found on the wings of diurnal lepidopterans: ‘eyespots’. Eyespots have been found to reduce predation by intimidating predators before an attack or startling them once an attack has been initiated, and have been proposed to work because predators falsely classify the markings as the eyes of the predators' own enemies. However, this argument lacks unambiguous experimental support because the data are consistent with the hypothesis that eyespots are effective merely because they are highly conspicuous. High contrast concentric circles are liable to be highly effective in stimulating the centre-surround arrangement of receptive fields typical of the vertebrate retina. Here, we use a field technique involving artificial moth-like stimuli, with or without different forms of grey-scale eyespots, to test these two opposing hypotheses. Across five experiments, the consistent effect is that predation risk is reduced by patterns that have high internal pattern contrast and also contrast highly with the target background, irrespective of the specific pattern arrangement. Circles are more effective than less eye-like shapes, but this constitutes only limited support for the eye-mimicry hypothesis because such patterns will also be highly conspicuous to animals with circular receptive fields in their retinae. We conclude that, contrary to popular belief, ‘eyespots’ can be effective predator deterrents without mimicking eyes.