Self-regulated learning theory acknowledges the importance of an individual's metacognitive monitoring and group-level regulation for learning achievement in collaborative learning. However, very few studies have empirically investigated the interplay of these factors. This study aimed to investigate how groups' metacognitive interactions, group-level regulation, and individuals' metacognitive monitoring accuracy together predict students' learning achievement. Thirty groups of secondary school students (n = 94) attended a five-week physics course involving four 90-min lessons. Each lesson included a collaborative learning session in which the students solved physics problems in groups of three. After each session, the students filled out a quiz that evaluated their learning of the core concepts related to each lesson. The students' collaborative learning was video-recorded using 360-degree cameras. First, the groups' metacognitive interaction, co-regulation (CoRL), and socially shared regulation (SSRL) during collaborative learning were coded from the video data. Second, the students' metacognitive monitoring accuracy was determined in relation to quizzes presented at the end of each collaborative session. Each student's learning achievement was assessed at the end of the course with an individual written exam. Bayesian multilevel models were used to analyze the nested data. The results showed that the frequency of metacognitive interaction was positively related to learning achievement. However, the relation between CoRL and learning achievement was moderated by monitoring accuracy and metacognitive interaction. The relation between CoRL and achievement was positive when coupled with a high frequency of metacognitive interactions but negative under a low frequency of metacognitive interactions. In addition, in line with theory, the students with lower monitoring accuracy achieved higher learning when CoRL was frequent in a group. Due to the low occurrence of SSRL, its relation with learning achievement could not be analyzed. No reliable evidence for the relation between monitoring accuracy and learning achievement was found. The results highlighted the complex interplay of individual-and group-level factors related to collaborative learning outcomes.

自我调节学习理论承认个体元认知监测和群体水平调节对于协作学习中学习成就的重要性。然而，很少有研究对这些因素的相互作用进行实证研究。本研究旨在调查群体的元认知互动、群体层面的调节和个人的元认知监测准确性如何共同预测学生的学习成绩。中学生三十组（n = 94) 参加了为期五周的物理课程，包括四节 90 分钟的课程。每节课都包括一个协作学习课程，学生以三人一组的方式解决物理问题。每节课结束后，学生都会填写一个测验，评估他们对每节课相关核心概念的学习情况。使用 360 度摄像机对学生的协作学习进行视频录制。首先，从视频数据中编码了协作学习期间小组的元认知互动、共同调节 (CoRL) 和社会共享调节 (SSRL)。其次，学生的元认知监测准确性是根据每次协作会议结束时提出的测验来确定的。每个学生的学习成绩在课程结束时通过单独的笔试进行评估。贝叶斯多级模型用于分析嵌套数据。结果表明，元认知交互的频率与学习成绩呈正相关。然而，CoRL 与学习成绩之间的关系通过监测准确性和元认知交互来调节。CoRL 与成就之间的关系在元认知交互频率较高时呈正相关，但在元认知交互频率较低时呈负相关。此外，与理论一致，当 CoRL 在一个群体中频繁出现时，监测准确度较低的学生获得了较高的学习效果。由于SSRL的发生率较低，无法分析其与学习成绩的关系。没有发现监测准确性和学习成绩之间关系的可靠证据。