Robotics education is valuable for developing students' 21st century competencies. It is of significant importance for teachers and researchers to explore its effective pedagogies. Robotics learning is interdisciplinary, and the most effective pedagogy is problem-oriented instruction. This type of instruction requires students to independently identify heuristic problems which can be categorized into four types: paradox, critical reflection, anomaly, and practice problems. School students are too young to develop sophisticated self-regulated learning, thus requiring guidance from teachers. However, how teachers can best guide students to solve these four types of problems remains unclear. Accordingly, this study aimed to investigate teaching behavioral patterns in the instruction of a robotics course. The participants were 30 seventh-grade students and a teacher. We proposed a Teaching Behavior Coding Scheme and used Lag Sequential Analysis to analyze the course videos to identify various major teaching behaviors and their sequences in the heuristic problems. The results showed that the behaviors included indirect influence, direct influence and wait time. Indirect influence occurred more often than direct influence, and it is argued that enough wait time should be provided in the instruction. Moreover, the results further suggest that paradox is missing from this instruction. The teaching behavior patterns for the three types of problems are discussed. The findings propose a new method to analyze teaching behavior patterns, and also recommend how teachers can support students learning in interdisciplinary robotics courses. The results contribute to the literature by designing the Teaching Behavior Coding Scheme and confirming its use in a robotics course.