Recent clinical observations, in the absence of experimental data, appear to suggest that written expression in conduction aphasics parallels their speech (Goodglass, 1992). The current study undertakes an analysis of word level writing in two conduction aphasics, and attempts to explore the posited 'parallel' relationship between speech production deficits and deficits in written expression. JL, a 66-year-old female with left posterior parietal lobe lesion and PP, a 65-year-old female with a left posterior temporo-parietal lobe lesion served as subjects of this study. Their response patterns on Boston Naming Test (BNT) and written naming task (John Hopkins Dysgraphia Battery) were utilized to verify the parallel hypothesis. Although both cases have exhibited phonological and semantic paraphasias on BNT, PP's overall performance was far superior to that of JL. JL produced numerous multiple responses to stimuli compared to PP's occasional multiple responses. PP's performance on the written naming task was far inferior to that of JL. JL's predominant error pattern in writing was the production of phonologically similar words to the target words. This paper argues that such seemingly contradictory, unpredicted patterns can be parsimoniously better explained, not by the parallel hypothesis but by current cognitive-neuropsychological models of writing.