A NEW APPROACH TO EXPLORING BIASES IN EDUCATIONAL-ASSESSMENT

This paper presents and illustrates an approach to the study of markin g biases based on multi-sample confirmatory factor analysis. This is a pplied to the marks awarded by two independent markers to the final ye ar dissertations of 197 female and 58 male psychology undergraduates. One of the two markers had supervised the work on which the dissertati on was based on a one-to-one basis. The results suggest that about 30 per cent of the variance in the supervisor's mark is attributable to i nfluences which are specific to the supervisor, orthogonal to the meri t of the project as assessed by the two markers jointly, and general a cross each of the four marks awarded by the supervisor. The most plaus ible interpretation of these influences is that they represent a conta mination of the supervisor's mark by personal knowledge of the student . These biases in the supervisor's marking were found to have more inf luence for male than for female students and to elevate the marks of m ales relative to those of females to a small but significant extent. I t would be unwise to overgeneralize from these findings, but they demo nstrate the potential value of this method of studying marking biases.