This paper presents the results of a trial conducted during the development of The Graduate Australian Medical School Admission Test (GAMSAT) 2006, where two versions of a cartoon unit were trialled; in both cases the cartoon and the question itself were identical. The only difference was in the way the cartoon was introduced within the test. Psychometrically, across the 650 candidates involved in the trial, one version ‘worked’ very well, the other not at all.