Gummerman, Hill, and Chastain (1974) suggest that figural characteristics of the mask are unimportant in determining the course of backward recognition masking in vision. This conclusion is potentially very important because it contradicts several recent models of various aspects of visual information processing (e.g., Estes, 1975; Smith, Haviland, Reder, Brownell, & Adams, 1976; Turvey, 1973). The present note provides a critical examination of the Gummerman et al. conclusion and presents additional demonstration data using the Gummerman et al. paradigm that indicates that figural relationships between target and mask can, indeed, be important determinants of the amount of backward masking. Gummerman et al. employed a Landolt C with the opening in either the top or the bottom as the target. Each of three masks consisted of Xs placed in the visual field surrounding the target. A no-overlap mask consisted only of these surrounding Xs, a partial-overlap mask contained an additional X overlapping the target but oriented so as not to block the opening, and a total-overlap mask had the additional X oriented so that it blocked the top and bottom openings in the target. When the target and mask were presented simultaneously in the same viewing field, recognition of the location of the target opening was quite good with both the noand partial-overlap masks and at chance performance with the total-overlap mask. However, at all positive stimulus onset asynchronies (SOAs), the three masks were equally disruptive of target recognition. From these results, Gummerman et al. conclude that the mask interrupts processing of the target so that figural characteristics of the mask are unimportant in determining the amount of backward masking. Furthermore, the authors conclude that the results argue against a model of backward recognition masking that proposes that the target and mask are integrated into a single preperceptual image, because such an integration model would predict that the