Although subjective appetite scores are widely used, studies on the reproducibility of this method are scarce. In the present study nine healthy, normal weight, young men recorded their subjective appetite sensations before and during 5 h after two different test meals A and B. The subjects tested each meal twice and in randomized order. Visual analogue scale (VAS) scores, 10 cm in length, were used to assess hunger, satiety, fullness, prospective food consumption and palatability of the meals. Plasma glucose and lactate concentrations were determined concomitantly. The repeatability was investigated for fasting values, Δ-mean 5 h and mean 5 h values, Δ-peak/nadir and peak/nadir values. Although the profiles of the postprandial responses were similar, the coefficients of repeatability (CR = 2SD) on the mean differences were large, ranging from 2·86 to 5.24 cm for fasting scores, 1·36 to 1·88 cm for mean scores, 2·98 to 5·42 cm for Δ-mean scores, and 3·16 to 6·44 cm for peak and Δ-peak scores. For palatability ratings the CK values varied more, ranging from 2·38 (taste) to 8·70 cm (aftertaste). Part of the difference in satiety ratings could be explained by the differences in palatability ratings. However, the low reproducibility may also be caused by a conditioned satiation or hunger due to the subjects' prior experience of the meals and therefore not just be a reflection of random noise. It is likely, however, that the variation in appetite ratings is due both to methodological day-to-day variation and to biological day-to-day variation in subjective appetite sensations.