Accurate determination of relaxation times has become increasingly important in efforts to determine the diagnostic specificity of nuclear magnetic resonance (NMR) imaging. Techniques used in NMR imaging, not routinely employed in conventional NMR spectroscopy, can significantly affect the resulting relaxation time determinations. For the saturation recovery (SR) approach of T1 measurement used in our laboratory, these include selective excitation to define the image plane and magnetization refocusing for NMR signal acquisition. Computer modeling of the Bloch equations shows that errors well over 50% can be made in image derived T1 measurements if the conventional SR relation between signal intensity and the 90-90 degrees interpulse delay, tr, is used. However, corrected expressions can be derived for the actual pulse and gradient sequence used by our imaging system, and phantom data acquired in imaging experiments have verified the validity of these equations. This allows for the correction of T1 data to compensate for systematic bias introduced during imaging procedures and suggests a means whereby errors introduced by radio frequency inhomogeneities across the imaging volume can be reduced.