The purpose of the study was to identify the relative influence nonachievement factors have on secondary school teachers' grading practices. Specifically, an attempt was made to predict on the basis of non-achievement related factors the discrepancy between the grades teachers actually award to hypothetical student profiles and the grades that would result if based on academic achievement alone.

Teachers learn in their professional training to base grades on student achievement, however in actual practice they do not always follow such an approach. They openly acknowledge including non-achievement factors in awarding students grades, explaining that there are psychosocial consequences in grades that must be considered.

The study involved 192 secondary school teachers from twelve school districts who taught high school academic subjects. Teachers were asked to award grades to hypothetical student profiles in simulated grading situations. Each simulation, or vignette, presented a hypothetical student profile and included a unique combination of factors which had been shown in previous research to influence teachers' grading decisions. A technique referred to as policy capturing was used to evaluate the extent to which assigned grades reflect student gender, effort, behavior, ability, and test score improvement.

In predicting the grade discrepancies of teachers, it was judged necessary to restrict the number of grade discrepancies for teachers to five or more, which represented approximately one-sixth of the possible 32 discrepancies provided by the vignettes, reducing the number of teachers in the sample to 99.

Regression models predicting the 99 teachers' grade discrepancies, which were differences between the letter grade awarded by teachers and the grade implied by the test scores in the vignettes, and policy capturing analysis revealed that student test score improvement and student behavior influenced teachers most in their grading. However, teachers seemed to place more importance on students' quiz average grades as a measure of academic performance, than on students' six-week test grades. In addition, secondary analyses of the data revealed that in predicting the importance placed by the 99 teachers on students' test score improvement and on behavior, teachers with more years of experience placed less importance on either of the factors in grading than their less experienced colleagues.