Multisentence production was examined in three nonfluent aphasic patients who had undergone a single sentence production training program using a computerized visual communication system (C-VIC). Patients were required to describe videotaped vignettes in English and using C-VIC. C-VIC allowed for an investigation of production abilities previously impossible to measure in severely aphasic patients, since C-VIC does not require internal generation of appropriate lexical items or phonological and articulatory realization. Results are discussed in the context of language production models in an attempt to determine the breakdown(s) in the production system that result in difficulty in trying to produce multiple sentences.