A critical piece of information in the selection process for academic positions is the letter of recommendation. A study was conducted to determine if differences exist in the letters written for female versus male candidates for academic positions. Using letters which were retyped to avoid personal, institutional, and sex-specific identification, the letters were coded on a number of coding categories, including length of letter, reference to personal and professional attributes, use of "masculine" versus "feminine" words, and proportion of words devoted to research and teaching. Sex differences in file variables which might account for differences in letters of recommendation, such as number of publications, grades, research and teaching assistantships, and teaching experience, were assessed. Sex of recommender, quality of institution, academic area, and year of recruitment (1975-76 and 1977-78) were used as major control variables. Results favored males in two variables: presence of a negative statement and proportion of highest intensity modifers. Many variables hypothesized to favor males showed no sex differences (words denoting ability, masculine terms, length of letter). Stereotypically feminine references to non-job-related qualities, such as appearance, were infrequently found. A number of indicators favored females: emphasis on research versus teaching, formality of address, and clinical skills. Distinct area patterns were found, particularly in clinical psychology. Changes from 1975-76 to 1977-78 all favored females. Interpretation of data suggests no clear sex bias against females but instead a tendency to avoid bias against females. (Author/BM)