Three experiments tested for effects of baseword frequency and orthographic neighborhood size with French pseudohomophones that were orthographic neighbors of the corresponding baseword. Speeded naming of pure lists of pseudohomophones gave a pattern of results that mirrored performance to the basewords themselves: effects of frequency and orthographic neighborhood that significantly interacted. When nonhomophonic nonwords were mixed with the pseudohomophones, the pattern of results mirrored that obtained with nonwords: an effect of orthographic neighborhood, no effect of frequency, and no interaction. When participants were asked to classify the same set of nonwords as sounding like real words or not, baseword frequency significantly affected correct positive responses to pseudohomophones, but orthographic neighborhood had no effect. Capturing these variations in the processing of pseudohomophone stimuli remains a challenge for any comprehensive model of reading.