This study assessed the effects of treatment designed to teach mildly aphasic individuals to monitor disfluencies--revisions, repetitions, audible pauses--in their connected speech. Treatment effects were evaluated using a multiple baseline design across 3 subjects. All subjects showed immediate and dramatic reductions in the frequency of target behaviors in a picture description task when self-monitoring was introduced. Two subjects simultaneously reduced untargeted disfluencies, and all 3 generalized their improved speech to another discourse task. Treatment effects were unrelated to the accuracy of self-monitoring. Subjects' improved speech was characterized by a slower speaking rate, but more efficient communication as reflected by longer uninterrupted utterances. Although standardized test scores were unchanged, unfamiliar listeners found perceptible improvement between baseline and posttreatment audiotapes of subjects' discourse. Self-monitoring treatment shows promise in reducing distracting and inefficient disfluencies in the speech of mildly aphasic individuals.