In order to identify the cortical areas involved in the reading process and to spare them during surgery, we systematically studied cortical areas by direct cortical stimulation in patients operated on for brain tumours. Seventy-six cortical stimulation mapping studies for language were performed in 35 monolingual and 19 bi- or multilingual patients over a 5-year period. We systematically searched for reading interference areas in addition to standard naming areas using an 'awake surgery' technique for brain mapping. A 'reading aloud' task (translated into different languages in multilingual patients) was used. Brain mapping was performed in left (44 patients) and right (10 patients) hemispheres. Cortical areas involved in reading were identified according to the type of interference, location and distinctness from naming areas. Stimulation of several major hemispheric regions resulted in significant interference with reading aloud: (i) the lower part of the pre- and postcentral gyri (P < 0.00001); (ii) the dominant supramarginal, angular and the posterior part of the superior temporal gyri (P < 0.00001); (iii) in the dominant inferior and middle frontal gyri (P < 0.001); and (iv) in the posterior part of the dominant middle temporal gyrus (P < 0.05). Interferences in reading were generally found in small cortical areas, with intervening areas evoking no reading interferences. Only partial overlap between reading and naming sites was found. Reading-specific sites were preferentially found when stimulating dominant inferior parietal or posterior temporal areas. Different types of reading interferences were noted. While 'articulatory' interferences were found in pre- and postcentral gyri bilaterally, and ocular-induced movements in bilateral middle frontal gyri, paraphasias were found mainly in the dominant supramarginal and posterior superior temporal gyri. Reading arrest sites were found in many regions. Reading interference sites were also occasionally found in the non-dominant hemisphere. In bilingual patients, if common cortical areas could be found, language- and reading-specific areas were sometimes detected, lending support to the concept that bilinguals can have relatively distinct cortical representation of their language skills. Finally, in this series, the location of reading interference sites and their relative specialization showed considerable individual variability.