Measuring the quality of computer-mediated communication

There is a growing literature of experiments whose purpose is to compare different configurations for computer-mediated communication. If the results of these experiments are to be useful they must: ( i ) use the right experimental tasks; and ( ii ) measure the right dependent variables. This paper is concerned with the latter problem which is illustrated using data collected in experimental comparisons of three configurations of a text-based conferencing system. No significant differences were found using a measure of task outcome. This accords with numerous previous findings. However, a numberof process-related dependent variables were devised that did show significant effects. These included common ground, as measured by shared recall, and references to the topic of one message in the next available turn. Another, the use of first and second person pronouns in conversation approached significance. Finally, an approach to the selection of measures for use in studies of computer-mediated communication is commended.