A great need exists for increased attention to standards of rigor in collecting sequential interaction data. As in any other research strategy, powerful statistics are not sufficient to organize data meaningfully. A knowledge of individuals, their relational histories, and communication are all necessary to insure proper conclusions about the conceptual nature of the interaction. Coding systems must preserve function; statistical structure is less important than the qualitative nature of a coding scheme. Research must observe the following standards of quality. Before a coding scheme is used it should reflect considerable agreement about the presence and definability of a category. Coders should not be aware of the research purposes. Demographics and compositional variables are important, as is the setting where people interact. Researchers should be wary of comparisons made across units of analysis, and must give special attention to within-group versus across-group comparisons. There should always be a match between concept and method. (References are included.) (DF)