Howard Allen implies that if teaching is effective, distributions of grades should be heavily skewed towards firsts and upper seconds (Teaching, THES, February 9). For too long, fetishes have been made of grade distributions that are normal in the statistical sense. Further, it is time assessors appreciated the law of small numbers - that distributions in a small sample are unlikely to resemble those in a large one - and that this could explain why, in a cohort of, say, 120 students, six different tutors might each have marked 20 assignments with divergent distributions of grades. This makes one question whether checking cross-tutor parity in grade distribution really is part of quality assurance?
Anthony Edwards
University College Northampton