Discipline‐related Marking Behaviour Using Percentages: a potential cause of inequity in assessment

Abstract

ABSTRACT This study set out to ascertain whether percentage mark distributions show evidence of discipline‐related marking behaviour and, if so, to consider the implications of this for equity in assessment outcomes. Performance data were obtained for 10 subjects at seven English universities for 1993–94, 1994–95 and 1995–96. The data showed that mark distributions at universities using percentages may be categorised into three Types (A,B and C). Type A distributions (English and History) are characterised by a steep‐sided negatively skewed distribution with a narrow spread. Pronounced microscale variation in the frequency is evident only in the vicinity of the mode between 55‐65%. Type B distributions (Biology, Business Studies, Fine Art, French, Law and Sociology) have a slightly broader spread and display microscale variation in frequency across a wider range of percentiles [40‐70%]. Type C distributions (Computer Studies and Mathematics) are characterised by a subdued, slightly negatively skewed distribution with a considerably wider spread. Type C distributions show conspicuous microscale variation in frequency across most of the percentile range. The types of distributions appear to reflect the relative confidence of assessors in different disciplines in awarding marks. The extremities of the percentage scale are perceived as insecure territory for the assessors of qualitative subject matter. Type A distributions are characteristic of disciplines in which the assessment is qualitative. Type B distributions may reflect the assessment of combined qualitative‐quantitative subject matter. Type C distributions are typical of disciplines in which more quantitative matter is assessed. The existence of discipline‐related patterns of distribution is a potential cause of inequity of outcomes for students on joint or multidisciplinary programmes. Data derived from institutions using grading scales show that discipline‐related patterns are less evident.