G. Wolford's perturbation model of letter identification is designed to account for identification errors of briefly presented characters. Its chief assumptions are that features are extracted in parallel, that some of these features become perturbed or mislocalized, and that mislocalizations are more likely to occur in the direction of the fovea for parafoveally presented characters. C. L. Krumhansl and E. A. C. Thomas used this model to account for their findings that placing a nontarget character to the peripheral side of a parafoveally presented target character disrupted target identification more than did placing one to the target's foveal side. Features from a peripheral nontarget were assumed to be mislocalized to mix with those of the target, making target identification difficult. To test the feature mixing explanation, 20 subjects participated in an experiment that involved identifying a gap in a Landolt C in a left and then right order and then reporting whether a nontarget character (either a circle or a capital H rotated 90 degrees) had appeared between two targets in a series. The experience produced no evidence that the nontarget that was similar was foveally mislocalized to become confused with the inner target. The nontarget's adverse influence, which was greater when the nontarget was less similar to the targets, affected both the inner and outer targets. (FL)