Reading a word may involve the spoken language in two ways: in the conversion of letters to phonemes according to the conventions of the language's writing system and the assimilation of phonemes according to the language's constraints on speaking. If so, then words that require assimilation when uttered would require a change in the phonemes produced by grapheme-phoneme conversion when read. In two experiments, each involving 40 fluent readers, we compared visual lexical decision on Korean orthographic forms that would require such a change (C stimuli) or not (NC stimuli). We found that NC words were accepted faster than C words, and C nonwords were rejected faster than NC nonwords. The results suggest that phoneme-to-phoneme transformations involved in uttering a word may also be involved in visually identifying the word.