To the Editor: Whatever the method used to select medical students (whether academic, psychometric, or interview), the basic problem in assessing the method’s predictive capability is that only candidates who perform at the higher levels in the assessment will be admitted. The only way to test the predictive validity of an assessment is to admit candidates from a much wider band of performance, creating a much less compressed score range for comparison. By definition, candidates with lower scores are excluded, thus making this analysis impossible.

The study by Groves et al had an overall response rate of 13.6%,1 a rate at which no conclusions could, or should, be drawn.

Entry to medicine remains a highly charged and emotional subject. It is unfortunate that the press has drawn conclusions from a study from which conclusions cannot be drawn.

In reply: Marley makes an important point about the conclusions that can be drawn from our study on the effectiveness of medical school admissions tests. As acknowledged in our article, the size of the study and the restriction of range to which he refers do limit the strength of the findings. Nevertheless, it is precisely because admissions tests elicit such an emotional response from the general community, as well as being costly and stressful for both aspiring students and medical schools, that there is an urgent need for a large-scale multi-institutional evaluation of the process — a need that our results highlight.

It is reassuring that the Australian Council for Educational Research, the developers of the Graduate Australian Medical School Admissions Test (GAMSAT), is currently conducting one such study in conjunction with several Australian graduate-entry medical schools.

We hope that our study serves to stimulate still more discussion and analysis of medical student selection processes.