2-Keto-3-methylvaleric acid (KMVA) has been found not to undergo spontaneous keto-enol tautomerization in neutral aqueous solution, alone or in the presence of large concentrations of pyridoxamine or pyridoxamine-5-phosphate. This finding denies the commonly held suppositions that 3R-KMVA is derived spontaneously from 3S-KMVA in vivo, and that L-alloisoleucine is the product of the reamination of this 3R-KMVA. Evidence presented here suggests that racemization of the 3-carbon of L-isoleucine occurs during transamination, that L-alloisoleucine is an inherently unavoidable by-product of L-isoleucine transamination (and vice versa), and that a KMVA enol is not obligate in this racemization. The four stereoisomers of 2-hydroxy-3-methylvaleric acid have been synthesized and the mass spectra of their trimethylsilyl derivatives recorded. An achiral methylsilicone column was used to separate the diastereomeric pairs and to determine their relative ratios in plasma and urine from normal controls and two maple syrup urine disease (MSUD) patients. The urinary ratio of the two diastereomers is different from that for plasma, both in normals and in MSUD patients. The plasma ratios may provide a rapid and simple measure of residual branched chain 2-keto acid dehydrogenase activity in MSUD patients.