A study designed to determine how inducer-surround contrast and inducer polarity affect the contour clarity and the lightness of illusory figures is reported. Using magnitude estimation procedures, ten naive subjects rated both the contour clarity and the lightness of Kanizsa squares. The magnitude of the inducer-surround contrast and the inducer polarity (all-black, all-white, or black-and-white) were varied randomly on each trial. The data indicate that contour clarity increases with contrast at the same rate across polarity conditions but that contour clarity at any given contrast level depends significantly on polarity. Contour clarity judgments were significantly lower when the inducers were all-white than when the inducers were all-black or black-and-white, and significantly greater in the 'mixed' polarity case (black-and-white inducers) than in the 'same' polarity case (the average of the all-black and all-white inducer conditions). Inducer contrast and polarity significantly affected the lightness of the illusory figure in a manner consistent with simultaneous spatial contrast. Also, for a given increment in contrast, contour clarity altered significantly more than surface lightness, regardless of inducer polarity. The findings suggest that the mechanism which mediates boundary formation is sensitive to the direction of contrast, and that the boundary formation mechanism is more sensitive than the surface lightness mechanism to changes in contrast magnitude. The results are considered within the context of neural network models of form perception.