This special issue is the major outcome of an invited seminar titled "Collaborative Learning: What can children learn together, specific skills or general concepts?" held in Oxford, England. Key themes ranged from conceptual and methodological issues to the role of task environment in promoting different learning processes and outcomes. The conceptual issues discussed include: * What theoretical models are the most appropriate? * What do we mean by collaborative learning? * Does collaborative learning differ from cooperative learning and peer tutoring? * How do we explain the outcomes of collaborative learning? * What are the influences on group functioning of students" cognitive development, goals, intentions, and task-specific expertise? The methodological issues addressed include: * What data should we collect, analyze, and report? * Which unit of analysis -- individual, dyad, small group -- is the most useful? * How do we assess the "success" of collaborative learning? Discussion of the role of task environment, both material and social, involved the following issues: * How do students learn from feedback -- from other students, the teacher, an experimental task, and/or a computer? * When computers are used in a group setting, how should their role be conceptualized? The aim behind the identification of these themes was to advance the field in one or more of the following areas: * the clarification of concepts concerning collaborative learning; * the identification of new approaches to the study of communication and cognition;and * the analysis of the interface among task demands, the feedback available to the learner, and cognitive processes and outcomes. These themes are apparent in the modified and expanded articles presented in this issue, and new issues have emerged as well. Reflecting concerns with preexisting theoretical and empirical models, the articles move away from simple experimental contrasts between the learning outcomes of students exposed to different interactional conditions. All of the articles include detailed analyses of on-task dialogue, while rejecting analyses that merely correlate interactional behaviors with learning outcomes. Instead, the need to focus on the psychological processes and task factors that mediate both learning and interaction are recognized.