The present study determined the intra-individual variation of BMR measurements, using a standard out-patient protocol, with the subjects transporting themselves to the laboratory for the BMR measurements after spending the night at home. The effect of a non-fasting state and variation in daily habitual physical activity the day before testing was evaluated. Eight male and eleven female subjects participated in three BMR measurements with 2-week intervals. Physical activity was estimated with a tri-axial accelerometer for movement registration, during the 3 d before each BMR measurement. There were no significant differences in estimated BMR (ANOVA repeated measures, P=0·88) and in physical activity (ANOVA repeated measures, P=0·21). Mean within-subject CV in BMR was found to be 3·3 (SD 2·1) %, ranging from 0·4 to 7·2 %. Differences between BMR measurements could not be explained by differences in physical activity the day before; however the mean within-subject CV in BMR changed from 5·7 to 5·2 % after correcting for within-machine variability and from 5·2 to 3·3 % after excluding five measurements because of non-compliance to the protocol including fasting. In conclusion, BMR values measured with a standard out-patient protocol are sufficiently reproducible for most practical purposes despite the within-subject variability in physical activity the day before the measurement. For this purpose, however, non-fasting subjects must be excluded and a regular function check of the ventilated-hood system is recommendable.