We describe here a modification of Sokoloff's deoxyglucose method which requires fewer blood samples but does not diminish accuracy. To calculate the local cerebral metabolism for glucose (LCMRglu), the input function containing the initial peak must be known. This requires many blood samples for accurate definition. To estimate the influence of the peak on the LCMRglu, a dummy input function was generated by extrapolating the data points after 2 min back to 0 min. Next, a biexponential curve, fitted to a reduced number of samples, was applied for the input function. The error introduced into the LCMRglu by using these different approximations was analyzed with an error function derived from an operational equation. The results indicate that the error of LCMRglu caused by neglecting the peak is less than 3%. Furthermore, the error of LCMRglu resulting from use of the biexponential approximation is a function of LCMRglu: the error decreases as the LCMRglu increases and remains at less than 3% even when only seven data points are used (half the usual number of the standard method). Finally, the LCMRglu was calculated by the biexponential curve-fitting, using the experimental data, and compared with that calculated by Sokoloff's computer program. Both methods were in good agreement. This modified method offers great advantage for simultaneous measurement of CMRglu, CBF, and other physiological parameters by means of the multiple tracer autoradiographic technique in the small laboratory animal.