Abstract Computerized reading activities were presented to 43 chronic aphasic subjects who were no longer receiving speech-language therapy in an attempt to determine the effectiveness of computer-provided treatment. Subjects were randomly assigned to one of three conditions: 78-hours of Computer Reading Treatment, 78-hours of Computer Stimulation ('non-language' activities), or No Treatment. Clinician interaction was minimal. Treatment software automatically adjusted task difficulty in response to subject performance by incorporating traditional treatment procedures, such as heirarchically arranged tasks and measurement of performance on baseline and generalization stimulus sets, in conjunction with complex branching algorithms. Three administrations of standardized tests at baseline, three months and six months revealed improved scores (p < 0·05) for the Treatment group. Additionally, the Treatment group made more improvement (p < 0·05) on the Porch Index of Communicative Ability Overall score than the other two groups. No statistically significant differences in improvement were measured between the Stimulation and No Treatment groups. Results suggest that 1) computerized reading treatment can be administered with minimal assistance from a clinician, 2) improvement on the computerized treatment tasks generalizes to improvement on non-computer language performance, 3) improvement results from the specific language content of the software and not simply the stimulation provided by the computer, and 4) chronic aphasic patients can improve performance through computerized treatment.