Background: The graphemic buffer was originally conceived as a component dedicated to the temporary storage of abstract orthographic representations prior to their format-specific expression in spelling (Caramazza, Miceli, Villa, & Romani, 1987). Since then, it has been argued that the graphemic buffer is also involved in reading (Caramazza, Capasso, & Miceli, 1996). Aims: The goal of this study was to test the hypothesis that a single graphemic buffer is used in reading and spelling. In addition, we examined the hypothesis that, under normal reading circumstances, words and pseudowords place different demands on the graphemic buffer because of differences in the degree to which they can be processed globally. Methods & procedures: We performed detailed analyses of the reading and spelling performance of MC who presented with dyslexia (restricted to pseudowords) and dysgraphia following resection of a left parietal lobe tumour. Additionally, we compared the effect of presentation format (standard print, mirror reversed, and orally spelled stimuli) on word and pseudoword naming. Outcomes & results: First, MC's spelling impairment showed all the characteristics of a graphemic buffer deficit. Second, MC's spelling of words and pseudowords as well as his reading of pseudowords were remarkably similar, both quantitatively and qualitatively (quasi-identical distribution of errors, length effect, and error position curves). Third, MC's reading of words (but not pseudowords) was disrupted under conditions that interfered with global (i.e., whole-word) processing. Conclusions: This study supports the claim that a single graphemic buffer is used in reading and spelling. It also suggests that reading nonwords places greater demands on the graphemic buffer than reading words.