A study was conducted to determine the effects of mildly self-disparaging humor on audiences' perceptions when it is used by speakers perceived to be of differing credibility (ethos) levels. Responses by 27 students who were potential subjects of the experiment were used to construct a credibility scale. The most credible speaker, a university psychologist, and the speaker of average credibility, a high school senior intending to major in pscyhology, were the ostensible speakers used in the study. A speech used in an earlier experiment was modified to yield four different conditions: (1) high ethos/humor, (2) low ethos/humor, (3) high ethos/no humor, and (4) low ethos/no humor. Subjects, students in a speech course, were told that the speeches were designed to enlighten students about majoring in psychology and had been broadcast on the radio in Atlanta. Approximately 25 useable responses resulted for each condition. Subjects rated the speeches according to measures included to reinforce the study's "cover story," as well as on James C. McCroskey's differential ethos scale and Holly Stocking and Dolf Zillman's seven-step scales. Results indicated that the ethos manipulation worked only for the "authoritativeness" factor and that the hypothesis that speakers using mildly self-disparaging humor would be rated higher on "sense of humor" and "wittiness" regardless of ethos level was correct. There was no interaction between humor and ethos. Although not significant, character and honesty scores did favor humorous speakers, indicating that text book writers' advice to use such humor is sound. (JL)