In a previous publication the theory, procedure, and results of a method were described for making two sequential measurements of cerebral metabolic rate for glucose (CMRglc), within a 2-hr period, using [18F]fluorodeoxyglucose. The error that is specific to this technique was estimated using computer simulations. CMRglc for the second state was sensitive to errors in (a) the values of the rate constants, (b) alignment of PET slices between the two scans, and (c) subtraction of one PET image from another. The root mean square of the average error from each error source was 6.4%, which gives the theoretical reliability of this method. The measured reproducibility, taken from our previous publication, was 4.2-6.2%, which is in good agreement with the present result. This method contributes a small additional error above that expected for two independent scans. However, independent scans done on different days are likely to be subject to larger physiological variations in CMRglc than would occur using this method.