It is assumed that the durations of response latencies of two speakers (i. e., latencies before responding to the conversational partner's last utterance) becoming similar over the time course (i. e., convergence) influences performance evaluation within dialogues. Twelve dyads in which the participants had opposing opinions regarding a particular topic tried to reach a compromise through discussion, under audio-only conditions. Six dyads participated in each of the two experimental conditions: a decreasing transmission delay condition in which the transmission durations decrease gradually, so that response latencies tend to converge, and an increasing transmission delay condition in which the transmission durations increase gradually, so that response latencies do not converge. At the end of the 15-minute dialogue, the participants evaluated their achievement of compromise and the impression about their partners. Further, the duration of response latencies were measured from the recorded vocalizations. The results suggest that the speakers whose response latency durations became similar to those of their conversational partners over the time course evaluated that they reached a compromise. The influence of convergence on performance evaluation was further discussed.