Polymorphism in cryptic prey species may be ‘protective’, in the sense of reducing the rate of predation by visual predators. We tested this supposed function by presenting 80 schoolgirls (aged 12–13) with populations of square ‘prey’ on backgrounds displayed on computer screens. The backgrounds were composed of random distributions of small coloured squares and the prey were composed of the same small squares, the grid of the prey being parallel to (but usually imperfectly aligned with) that of the background. Different ‘morphs’ of prey were defined by the spatial distribution of the small squares, and prey of a given morph were always presented in the same orientation. Each subject was presented with a sequence of 40 screens, of which 36 were background containing a single randomly positioned prey, while four were background alone. Subjects ‘zapped’ prey with the mouse, or pressed the space bar if the screen was judged to be empty, and then moved automatically to the next screen in the sequence. Each subject received a single ‘treatment’: high or low crypsis combined with one of four levels of polymorphism (1, 2, 6 or 12 morphs). The results showed that hit (‘predation’) rate tended to be lower on populations with high crypsis and was inversely related to the number of morphs in the population. We thus conclude that polymorphism may indeed ‘protect’.