In an experiment investigating patient mood in an out-patient setting, a Mood Adjective Checklist (Mackay et al., 1978) was administered to 189 patients before and after their consultations, obtaining 366 completed questionnaires. Both the stress and arousal scales in the Mackay et al. checklist contain unequal numbers of positive and negative items, and are thus subject to response bias. When the results were analysed, it was noted that some items in the checklist led to large numbers of question mark responses, and it was discovered that word frequency, used as an estimate of word unfamiliarity, was correlated with the use of the question mark. Furthermore, when question mark responses were treated as missing data, as recommended by Meddis (1972), the factor analytic distinction between stress and arousal disappeared. Next, items giving rise to 15 per cent or more question mark responses were omitted from analysis, and the remaining items factor analysed. The resulting checklist comprised nine high stress items, nine low stress items, four high arousal items and four low arousal items. Mackay et al.'s stress and arousal scales were reconstructed using these items; they were no longer subject to response bias, and were more accessible to low vocabulary subjects.