The U.S. Army’s synthetic training environment (STE) will combine live, virtual, and constructive technologies to deliver a comprehensive collective, multi-echelon training experience to form agile, adaptive teams. To optimize training effectiveness, the STE will automate the collection and assessment of team processes (e.g., cohesion) and performance outcomes (e.g., mission accomplishment) and use the results in automated after action reviews. In their recent meta-analysis Sottilare et al. (2017) demonstrated team cohesion, collective efficacy, and communication are significant antecedents of effective team performance, however, they determined the diagnostic effectiveness of such self-report measures as cohesion are limited (e.g., judgment errors and social desirability bias) due to the dynamic emergence of team effectiveness during training. Therefore, they provided a detailed analysis of potential behavioral markers of cohesion, efficacy, trust, etc., which they proposed could improve team measurement under dynamic conditions. Toward this end, in this paper, we report on a qualitative study in which we identified and collected behavioral markers of cohesion and efficacy communications in conventional and special operations combat fire teams in training situations. Pattern matching methods were developed with subject matter experts (SMEs) to identify the markers, and then the SMEs applied the assessments to video and audio recordings of 64 Soldiers (21 fire teams) performing training scenarios.  We found that patterns of team cohesion and efficacy can be systematically observed and assessed during dynamic training situations. We also identify how these patterns could relate to mission outcomes, and then discuss lessons learned on the technical challenges, such as speech recognition, that need to be addressed to inform the STE.