In this study, we examined how the patterns of emoticons and emotional word marks appeared in CMC, especially in e-mail exchanges. A text-mining analysis was used in the study. Eighteen university students were requested to exchange e-mails over a period of two weeks with unknown correspondents who were, in fact, cooperating in the experiment. One hundred and forty-one e-mails thus obtained were analyzed. A text-mining software "True Teller" extracted 4,125 words of 257 kinds for further analysis. As a result, various emotional expressions that appeared in the CMC as well as in Face-to-Face communication were clarified. In addition, emoticons were seen to appear frequently in the early stages of communication and had the effect of softening tense relationships between e-mailers.