To evaluate the accuracy of the 1970 census' occupational classification system, a comparison was made between that year's responses to census questions and the replies by the same population group to the 1972 Postcensal Manpower Survey (PMS). Each of these questionnaires relied on a different methodology; in the PMS the respondent classified himself directly while in the census a clerical coder translated the individual's response into an occupation category. The study sample consisting of 34,938 participants was chosen from the engineering, scientific, and technical fields. Mismatches were found in 19,620 cases, and were ascribed to the following causes: coding or processing errors, structural or methodological differences, and insufficient or conflicting answers. Recommendations made to improve the census classification system included the following: (1) coders should read all the answers in the occupational section before making an assignment; (2) use of the lowest-code rule should be examined; (3) certain occupational titles need to be more clearly defined; (4) responses indicating dual occupations should be placed in a residual category, not assigned to one of the two fields; and (5) more occupational titles should be added to the list. Three recommendations for PMS were proposed: the present categorization of university or college teachers of engineering or science should be changed, examples of appropriate occupational titles should be given for each PMS category, and an investigation should be made into respondents' use of residual categories. (ELG)