Observable Interaction Patterns in Collaborative Learning: Evidence from Four Small Groups Across Two Speaking Tasks

This study adopts a lightweight, process-oriented behavioral analysis framework to examine groupinteraction differences in collaborative learning through observable language-based interaction. Using tworounds of small-group speaking tasks (Story 1 and Story 2), we conducted sentence-level coding of sharedunderstanding (Q2), constructive feedback (Q3), consensus building (Q4), and sustained motivation (Q6),and examined these behaviors in relation to five group-level interaction structure dimensions: turn-takingsmoothness, topic continuity, interactional rhythm, interruption-free flow, and role complementarity.Across four naturally formed groups, radar profiles, interaction trajectories, and cross-task comparisonsrevealed markedly different interactional trajectories. The results show that even under identical task designconditions, groups may follow divergent interactional pathways, including maintaining stable interactionstructures, exhibiting gradual structural change, or failing to develop stable interaction patterns despitesustained participation. In particular, Group 4 demonstrated a clear shift in interaction structure in thesecond task, whereas Group 3 did not exhibit corresponding changes following increased task demands.Based on cross-group contrasts and process-level observations, this study offers an exploratory conclusionthat the stabilization of interaction structure is typically accompanied by the sustained co-occurrence andtemporal organization of multiple interactional behaviors, rather than being triggered by any single behavioror event. By foregrounding observable interaction patterns and their process-level evolution, this workprovides a reproducible, process-oriented analytical approach for examining collaborative learningprocesses and offers empirical grounding for future process-based learning analytics and AI-supportedcollaborative research.