The most frequent complaints in the mediations were harassment and assault. The study hypothesis was that the greater the number of instrumental (fact-related) and expressive (feeling-related) exchanges among the participants, the greater the likelihoods of reaching an agreement, of making an agreement that would last, and of high satisfaction levels among the disputants. The four observers were knowledgeable in mediation techniques and trained in the observation of social interaction, but they were not aware of the research hypothesis. They recorded both the amounts and nature of verbal communication. Findings showed that the length of the session had no clear, consistent positive influence on the outcome measures. In addition, the verbal expression of feelings showed little relation to the outcome measures. Moreover, complainants and respondents reacted differently. Complainants reacted more positively to more communication as long as feelings were not explicitly involved, while respondents consistently reacted more negatively to more communication from the mediator, particularly when feelings were involved. Discussion of implications for mediator training, tables, and 25 references.