Task analyses have traditionally been framed in terms of overt behaviors performed in accomplishing tasks and goals. Pioneering work at the Learning Research and Development Center looked at what contribution a cognitive analysis might make to current task analysis procedures, since traditional task analysis methods neither elicit nor capture cognitive components of the job. This paper describes an approach to cognitive task analysis, and reports on initial case studies trying out the methods used. Specifically, the approach attempts to capture two aspects of a job that have been generally overlooked by traditional task analysis: (1) describe job performance not only in terms of overt behaviors, but also in terms of the underlying knowledge content and thinking processes, and (2) capture not only the formal but also the informal aspects of the job. Knowledge about jobs is analyzed into four knowledge representation types: factual knowledge, imagistic knowledge, procedural knowledge and mechanism (or mental model-type) knowledge collectively referred to as FIPM. This FIPM model of cognitive task analysis is found to be effective in capturing rich knowledge content both in relation to the performance of specific tasks, and in relation to the general work setting. (Contains 13 references.) (Author/AEF)