This paper outlines 4 assumptions behind attempts to explain the sequential organization of communication behavior during conflict. These assumptions were supported by an analysis of behavioral sequences coded from 9 hostage negotiations and 20 divorce mediations. Analyses showed that negotiators use only a small proportion of available responses to other party’s behavior, and that this proportion rapidly decreases as sequence length increases. Critical to this channeling in behavior was the triple- interact (i.e., cue-response-cue-response), which represents the maximum sequence length required to enable accurate prediction of negotiators’ future behavior. More detailed analysis showed that the triple-interact reduced uncertainty in behavior by over 70%, which compares to less than 1% from knowledge of negotiation context and approximately 10% from knowledge of individual differences.