Leaders are often identified in empirical studies by either their position in an organizationally defined hierarchy or by sociometric survey. However, such methods conflate behavioral antecedents and outcomes, including subjective co-participant attributions, with assessment of behaviors themselves. In this study, we propose the “interruption network” as a model of small group structure based on a nonverbal behavior—conversational interruptions—that has been previously validated as a correlate of social status. A directed, weighted edge exists in an interruption network i → j representing the number of times j interrupted i (Figure 1A, B); a “leadership” or “vote” network based on co-participant leader nominations is shown in Figure 1C. In simulation studies, randomizing edge directions within node pairs suggests that interruption network edge direction accounts for substantial variance in leader emergence while retaining the association between speaking time and leadership. Network motif analysis reveals hierarchical, nontransitive mesoscale structures (triad 021C: i → j → k; p = 0.00). Comparison of exponential random graph models finds that leadership attributions contain a male bias (vote network in-degree; β = 1.061, p < 0.001) not found in the interruption networks (weighted in-degree; β = 0.075, p = 0.136). We conclude that while it is often reported that male group members emerge more as leaders than female group members, the difference may be more attributional than behavioral.
