Spatially localized phosphorus‐31 nuclear magnetic resonance (31P NMR) spectroscopy has been applied to the study of the normal canine myocardium to measure the relative content of high energy phosphates across the left ventricular wall. Transmural NMR data were acquired in five voxels spanning the wall of the left ventricle using the FLAX‐ISIS technique. The validity of the FLAX‐ISIS approach in acquiring localized spectra for transmural studies and in providing quantitative information from the localized spectra was examined rigorously by studies involving phantoms, intact rats, and the canine myocardium in vivo. The results indicated that (1) this technique yields spatially resolved spectra with partial overlap between adjacent voxels and virtually no overlap between every other voxel; (2) in the canine heart, signals from subepicardium, midwall, and sub‐endocardium can be detected separately without cross contamination; and (3) relative metabolite contents within a voxel and among voxels can be quantitated. Transmural 31P NMR spectra were acquired with cardiac gating on 29 separate animals either at early systole or late diastole, and at three different workloads with the heart rate peak systolic pressure product (RPP) increasing from 6000 mmHg/min to 35,000 mmHg/min. The data revealed that in the normal canine myocardium, the creatine phosphate (CP) content and the CP/ATP ratio was significantly lower in the subendocardium than in the subepicardium. ATP levels were transmurally constant. Both the CP content and the CP/ATP ratio measured for each voxel remained unaltered in relation to either the phase of the cardiac cycle or ∼ fourfold increase in workload. Free ADP levels calculated for each voxel showed that ADP was relatively higher in the subendocardium than the subepicardium, and in all transmural layers was higher than its apparent Km for oxidative phosphorylation. In this domain changes in ADP content with workload and MVO2 are not expected and were not observed. © 1990 Academic Press, Inc.