We report a study that uses retrospective analysis to understand the relationships between American and Chinese participants' utterances during a conversation and the moment-by-moment feelings and reactions they subsequently described. Intercultural and intracultural pairs of Chinese and American participants talked about a fictional crime story and then individually watched and reflected on an audio-video recording of the interaction. We analyzed three types of utterances produced during the interactions: task-related utterances, relational utterances, and back channel responses. We found that American participants tended to increase their frequency of back channel responses when talking to a cross-culture partner, whereas Chinese participants did not. The number of back channel responses and task-related utterances produced by the partners were correlated with the participants' subsequent self-reported level of understanding, involvement, and negative emotions during their conversations. Our results contribute to the understanding of how communication problems may emerge during dyadic conversations, especially in intercultural interactions.