The purpose of this study was to explore the characteristics of a good public speech performance in terms of phonetic and non-verbal gesture aspects. The video recordings and the scripts of the commencement speech delivered by four different speakers, during the graduation ceremonies at Harvard or Stanford University, were obtained from the official websites. First, we analyzed a distribution of speech pauses based on their types and duration. In the following analysis, we analyzed patterns of facial movements during performance, focusing on the frequency of lifting the head. The present analysis identified the following findings: 1) Average durations of pauses at period were longer than those of pauses at comma and other pauses for all four speakers, but individual difference was found for the average durations of pauses at period. 2) All speakers showed rhythmical patterns of facial movements, and there was a correlation between the speech rate and the frequency of lifting the head. These results are expected to be useful for improving the teaching of English public-speaking skills in EFL classrooms.