Two experiments explored boundary conditions for the ubiquitous left-hemisphere advantage in visual-word recognition. Subjects perceptually identified words presented directly to the left or right hemisphere. Strong left-hemisphere advantages were observed for UPPERCASE and lowercase words. However, only a weak effect was observed for AlTeRnAtInG-cAsE words, and a numerical reversal of the typical left-hemisphere advantage was observed for words in a visual prototype font (a very unfamiliar word format). Results support the theory that dissociable abstract and specific neural subsystems underlie visual-form recognition and fail to support the theory that a visual lexicon operates in the left hemisphere.