Questionnaires were administered to 410 inmates (the entire inmate population) of an adult, medium-security prison located in a midwestern State. This procedure took 6 evening sessions spanning a 2-week period. Completed questionnaires were received from 364 inmates for a response rate of 88 percent. Previous interviews with selected inmates revealed five major social types within the inmate group. A social type is defined by a pattern of behavior which is recognized and labeled by members of a group according to the group's lines of interest. These lines of interest center around the maxims of the inmate normative system, the inmate code. The five types are snitches (informers), wheeler-dealers (those who exploit fellow cons economically), punks (those who demonstrate unmasculine behavior), hardrocks (those who pick fights and are antistaff), and the solid inmate (the ideal inmate from the con's perspective). Nominations of social type role incumbents were made by 13 correctional officers who were in continuous daily contact with the inmates. Four hypotheses were tested. The main hypothesis stated that deviants' self-reactions are more reflective of their perception of others' reactions to their behavior than to the actual reactions to this behavior. Following detailed analysis, the research found that labeled deviants generally underestimate the severity of the actual group reaction. Moreover, study results strongly support a major proposition of labeling theory; that labeled deviants maintain conceptions of self which are consistent with the labels which have been assigned these individuals by the group. Members of the group react to deviants by showing dislike, rejection, and condemnation. Thus, occupants of the most stigmatized roles of inmate society (snitch and punk) are the most insecure in interpersonal relationships with others and maintain the greatest sense of inferiority of any group in the institution. On the other hand, the solid inmate group felt the most secure and confident in interpersonal relationships of the five groups. Thus, this research has furnished moderate, and in some cases, strong support for hypotheses designed to examine the interactional basis of labeling theory. Tables, footnotes, and 25 references are supplied.