A methodological note on content analysis: estimates of reliability.

Abstract

The reliability of measurements obtained from a dictionary-based form of content analysis was investigated in this experiment by tape-recording subjects as they spoke about a predetermined topic in two separate sessions 1 week apart. Subjects were randomly assigned to speak in one of two contexts. The content of transcripts of these sessions was analyzed using adaptations of the Harvard III Psychosociological Dictionary and the General Inquirer content analysis program (Stone, Dunphy, Smith, & Ogilvie, 1966). Reliability was high over time, both within and between sessions, and there were few differences observed between contexts. Results are discussed in terms of their implications for the use of content analysis in personality research.