Kanizsa's (1979) explanation of illusory contours in terms of amodal completion of perceptually incomplete elements was examined in five experiments by varying the elements partly delineating the corners and sides of a rectangle. The results from magnitude estimation of contour strength showed that (1) incomplete corner elements consistently resulted in stronger illusory contours than did complete elements, although contours nevertheless always occurred with the latter, (2) contours were always stronger with side elements than without but were not always reliably different in strength when the side elements were parts of an incomplete form, (3) incomplete octagons and slightly different but complete elements resulted in contours of the same strength, (4) the larger of two incomplete comer elements resulted in stronger contours, and (5) incomplete irregular elements resulted in stronger contours than did complete irregular ones, although weak contours again occurred with the latter. Together, the data do not strongly support the view that perceptual completion of incomplete figural elements is the basis of illusory contours. An alternative interpretation in terms of the spreading of induced contrast to partially delineated borders is proposed. It is suggested that apparent overlay occurring independently of perceptual completion enhances these borders and so strengthens the illusory contours.