Henry and Arrow (under review) suggest that groups be categorized by thermodynamic states (i.e. fixed, fluid, chaotic, and complex). The four group dynamic states are defined as: (1) fixed state groups: low mutual influence, energy drain; (2) fluid state groups: active communication, rhythm and flow in interactions; (3) chaotic state groups: uncoordinated action, interruption; and (4) complex state groups: high mutual influence, self-organization, energizing. The goal of the present study was to distinguish the group dynamic states empirically. Group thermodynamic states require a coding system that has temporal flexibility, which previous research (e.g., Gersick, 1989; Keyton, 1999l; Wheelan, Verdi, & McKeage, 1994; Worchel, 1994) has not examined. Those existing coding systems focus on groups moving through a fixed developmental sequence rather than states that groups could enter, leave and return to. Thus, we aimed to develop a new coding system. The present study describes that system, the reliability achieved using it, and how the coding patterns relate to three of the four thermodynamic states.

Member experience is integral to thermodynamic states and can be intuitively judged. As this coding system was in its infancy, we began with intuition as our starting point. Groups that appeared enjoyable and energizing to members were labeled either fluid or complex. Groups that appeared draining to their members were labeled fixed. Although we did not observe a chaotic group in this study, the member experiences that would have identified such a state would have been confusing and frustrating. Interaction patterns in thermodynamic states vary, ranging from large amounts of interaction with elaborated role structure (complex) to independent, isolated work (fixed). We proposed four patterns that might be used to distinguish the four states using a volleyball metaphor: (1) bump-set-spike: the majority of group members interact with each member’s contribution reflecting a unique role; (2) over-the-net: two group members interact with each other while others watch; (3) ace: one person attempts to interact with the group, yet no one responds; and (4) interruptions: a group member is speaking and another interjects an unrelated thought.

Hypothesis 1: Groups in the complex state should have a proportionally higher number of bump-set-spike interactions and more interaction overall (as measured by transcript length).

Hypothesis 2: Groups in the fluid state should have proportionally higher number of over-the-net interactions and less interaction than groups in the complex state.Hypothesis 3: Groups in the fixed state should have a proportionally higher number of aces and the lowest levels of interaction of all groups

Hypothesis 4: Groups in the chaotic state (which we did not observe) should have proportionally higher numbers of interruptions.

Hypothesis 5: Bump-set-spike interactions will be positively related to group performance.

Hypothesis 6: Aces will be negatively related to group performance.

Participants

This study consisted of 15 undergraduate level students enrolled in Introduction to Psychology. These 15 students composed three groups that interacted over the course of 10 weeks for one-half hour per week, providing a total of 30 hours of interaction.

Measures

The groups were videotaped over a span of ten weeks performing class-related tasks, for which they received a grade. These grades were used as performance scores. Transcripts of weekly interactions served a dual purpose. The transcripts provided the means to unitize and code interaction as well as measure amount of interaction overall (average number of pages of conversation per weekly interaction).

Procedure

Pairs of members unitized the transcripts together. They then independently coded the group interaction (each unit of interaction was given one of the three codes). Overall reliability of 86.74% was achieved across all pairs. Disagreements were then resolved via discussion until 100% agreement was subsequently achieved.

To determine if our intuitive categorization of groups as “fixed”, “fluid” and “complex” might correspond to empirical hypotheses, we tested the assumption that fixed groups would have the lowest levels of interaction, followed by fluid groups, and complex groups the highest. The fixed group did have the shortest interactions of all, with their transcript lengths averaging 9.66 pages. The fluid group had more interaction than the fixed group, with an average transcript length of 11.4 pages. The complex group had the greatest amount of interaction, reflected in an average transcript length of 11.7 pages. These findings support the latter half of Hypotheses 1, 2 and 3. The general conversational patterns of these groups reveal that the complex group has a proportionally higher number of bump-set-spike interactions, while the fluid group has a proportionally higher number of over-the-net interactions. These findings support the former half of Hypotheses 1 and 2. Contrary to predictions, the fixed group did not have a proportionally higher number aces; resulting in a lack of support for the latter half of Hypothesis 3.

The proportions of bump-set-spike and ace interactions were then examined for the high-performing versus low-performing weeks. The bump-set-spike interactions were proportionally higher for the high performing weeks and proportionally lower for the low performing weeks (supporting Hypothesis 5). However, Hypothesis 6 was not supported as ace interactions did not seem to vary with performance levels.

The purpose of our study was to develop a coding system that would measure the group dynamic states empirically. Both length of group interaction and proportion of varying types of interaction patterns were helpful in distinguishing the state of the three groups studied. The evaluation of the three groups over a span of ten weeks was very rich and detailed, and sufficient reliability was achieved using the novel coding system we have proposed. Future research may continue to examine interaction patterns of additional groups so that inferential statistical tests might more rigorously validate the new coding system we have developed.

Gersick, C. J. G. (1989). Marking time: Predictable transitions in task groups. Academy of Management Journal, 32, 274-309.

Henry, K. B., & Arrow, H. (under review). Making heat work: A thermodynamic model of groups. Personality and Social Psychology Review.

Keyton, J. (1999). Analyzing interaction patterns in dysfunctional teams. Small Group Research, 30, 491-518.

Wheelan, S. A., Verdi, A. F., & McKeage, R. L. (1994). The Group Development Observation System: Origins and applications. Provincetown, MA: GDQ Associates.

Worchel, S. (1994). You can go home again: Returning group research to the group context with an eye on developmental issues. Small Group Research, 25, 205-223.