It is widely held that computer‐mediated communication (CMC) filters out many of the social and affective cues associated with human interaction with consequent effects on communication outcomes and the medium's suitability for interpersonal tasks. The relationship between paralanguage and social perception in CMC in different social contexts is investigated in two experiments. In Study 1, it was hypothesized that there would be significant differences in subjects’ perceptions of anonymous communicators as a function of the paralinguistic content of the electronic mail messages they received. Subjects read three sets of messages containing different types of paralinguistic cues and a fourth set of control messages. They also completed a set of person‐perception rating scales in respect of each message sender. The hypothesis was supported for both novice electronic mail users and for experienced users drawn from a large telecommunications organization. In Study 2, subjects participated in group discussions over a CMCS under four conditions, manipulated in a 2 × 2 between‐subjects design. The salience of the task group was either high or low, and subjects were either de‐individuated (physically isolated and visually anonymous) or individuated (physically copresent and visually identifiable). From social identity theory, it was hypothesized that de‐individuated subjects for whom group identity had been made salient would evaluate users of paralanguage more positively than when group salience was low, in accordance with a social attraction response associated with perceptions of group identity. The hypothesis was supported. The results suggest that paralanguage is one means by which social information is communicated in CMC and that the meaning of paralinguistic marks is dependent on the group or individual context that is pre‐established for the communication. The studies, therefore, question earlier assumptions that the social context is dramatically reduced or eliminated in this medium. The implications of contextual effects for the use of CMC by work groups in organizations are discussed.