Collaborative learning is a widely used teaching model in programming education. A deeper understanding of the roles and behavior patterns within collaborative learning could improve its performance. In this study, an emergent role configuration and behavioral pattern are analyzed using audio and video data from 10 groups in a 7th-grade programming course. First, the study applies content analysis to construct the emergent role framework for collaborative programming learning. Then, learners' role behavioral events and behavioral sequences among the high- and low-performance groups are identified using lag sequential analysis. A total of 1555 emergent role behaviors are generated. The results show that focus on task and process is a special emergent role of collaborative programming; and compared to the low-performance group, the high-performance group exhibited a more balanced distribution of emergent roles and more orderly behavior patterns. Specifically, high-performance groups demonstrated better team spirit, feedback consciousness, participation enthusiasm, a friendly and equal discussion environment, and effective resolution strategies for negative emotional roles. These findings will guide teachers in designing high-quality collaborative programming activities and assist students in developing programming competency.