Abstract.Our research provided empirical evidence about the alternative means of communication used by 25 members of a resehrch group who had available to them: unscheduled face-to-face encounters, scheduled face-to-face meetings, electronic mail, telephone, fax, and desktop videoconferencing. The intent of our research is to learn whether there are elements in existing group communication patterns that suggest how future communication systems can be designed or selected to fit the actual work relationships of a group. A detailed social network survey provided information about what members of the group communicated about, how they communicated, and with whom they communicated. Most communication was done through a combination of media, but predominately through unscheduled encounters, electronic mail, and scheduled meetings; people rarely vldeoconferenced, telephoned, or faxed. Factor analysis reduced the 24 work relationships to six distinct dimensions: receiving work. giving work, collaborative writing, major emotional support, sociabihty, and computer programming. The proportion in which the three main media were used varied according to the nature of the work dimension. Our findings suggest that a multivariate perspective that considers group norms and practices, social networks, and work dimensions is necessary to analyze media use.Key Words: computer-mediated communication, electronic mail, information technology, collaborative work, media choice, multimedia integration, network analysis, social networks, videoconferencing, work communication Can we discover which kinds of communication media are used, and presumably preferred, for different types of work? To do so, we need to study an organization which has available a wide range of media, and we need to use a method that can identify which media are used for which kinds of communication. In order to examine this issue, we asked members of a university research group to complete a social network survey about their choice of communication media for different work relationships, that is, for specific kinds of interaction at work, such as giving instruction, demonstrating work, or disagreeing with others. The 25 computer scientists provided information about their use of unscheduled face-toface encounters, scheduled face-to-face meetings, e-mail, telephone, fax, and desktop videoconferencing for each of the work relationships.