<p>This thesis examines and measures differences in effciency between L1- and L2 language use for solving referential problems. Quantitative measures used in the study were time and error frequency. Qualitative measures, such as how the tasks and the participants own performance were experienced, were measured by an post interview and two questionaires. </p><p>Two different tasks, a sorting task and a construction task, in which each informant was given the role as a constructor or instructor, were used for data gathering. A total of 20 dyads, divided in two language groups (English and Swedish) participated in the study. </p><p>The results show large differences between the two language groups in respect to time for solving the two tasks. For the construction task, these were mainly due to different communication strategies used for error handling. In the sorting task a clear learning effect can be observed for the L2-groups which, in effect, makes the L2 users solve the sorting task almost as fast as the L1 groups, after an initial learning period. However, the informants in the L2 groups generally referred to the objects used in this task by describing peripheral characteristics of these objects. They also showed more signs of uncertainty and hesitation in the actual dialogue situation. However, no differences in error frequency could be found between the two language groups. </p><p>Differences in role-taking between the language groups were also made apparent in the study. These differences became especially clear for the L2 groups in the construction task.</p>

This thesis examines and measures differences in effciency between L1- and L2 language use for solving referential problems. Quantitative measures used in the study were time and error frequency. Qualitative measures, such as how the tasks and the participants own performance were experienced, were measured by an post interview and two questionaires. Two different tasks, a sorting task and a construction task, in which each informant was given the role as a constructor or instructor, were used for data gathering. A total of 20 dyads, divided in two language groups (English and Swedish) participated in the study. The results show large differences between the two language groups in respect to time for solving the two tasks. For the construction task, these were mainly due to different communication strategies used for error handling. In the sorting task a clear learning effect can be observed for the L2-groups which, in effect, makes the L2 users solve the sorting task almost as fast as the L1 groups, after an initial learning period. However, the informants in the L2 groups generally referred to the objects used in this task by describing peripheral characteristics of these objects. They also showed more signs of uncertainty and hesitation in the actual dialogue situation. However, no differences in error frequency could be found between the two language groups. Differences in role-taking between the language groups were also made apparent in the study. These differences became especially clear for the L2 groups in the construction task.