中文摘要：

共享心智模型指团队成员关于团队作业、策略、团队情境、团队互动等的共同的知识结构，根据内容不同可以分为团队作业模型和团队互动模型两类。该研究通过实验室实验探讨两类共享心智模型的发展特征。132名大学生组成44个3人团队，随机分为团队绩效反馈组和无团队绩效反馈组。采用相似性评定法在团队执行作业前、中、后三次测量两类共享心智模型。结果发现：（1）共享心智模型随团队运作时间增加而发展；（2）团队互动模型形成和发展的难度高于团队作业模型；（3）团队绩效反馈对两类心智模型发展的作用不同，反馈加速了团队作业模型的发展速度，但却是团队互动模型能否得以发展的决定性因素。两类共享心智模型的不同发展特征应归因于两者特异性高低的不同。

英文摘要：

Shared mental models （SMMs） were defined as team members＇shared understandings about team task, task context and strategies, team interaction patterns, and teammates＇traits. Team task models and team interaction models were distinguished as two types of SMMs. It has been showed that SMMs can enhance the efficiency of team processes and team performance. Although researchers have focused on antecedents and consequences of SMMs, little research has been conducted to examine the development of SMMs. The current study aimed to investigate how two types of SMMs developed from low to high level. Totally 132 student participants formed 44 3-person teams to perform on a simulation task, which was developed particularly for the current study. Among them, 23 teams were provided with team performance feedback during task execution, the other 21 teams were under no-feedback condition. Participants were randomly assigned to feedback conditions and to positions within teams. Mental models （MMs） were elicited using participants＂ individual paired ratings of the relatedness among key concepts of team task or interaction behavior before, during and after task execution. Pathfinder, a network-analysis computer program, could represent each member＇s MMs based on the matrix of paired ratings, and could offer a closeness index （ C were averaged to form the index） to reflect the similarity of MMs between each pair of teammates. Three ＂ C ＂ indexes team similarity score. Results of mixed-model ANOVA showed that： a） similarity of mental models increased significantly over time; b） similarity of task models was significantly higher than that of interaction models; and c） similarity of task models in- creased faster under feedback condition than under no-feedback condition, whereas similarity of interaction models in- creased significantly only under feedback condition. Differences in the development of these two models were discussed in terms of whether the models were task-specific in natureThe results