This study examined the reliability of (31)P-magnetic resonance spectroscopy (MRS) to measure parameters of muscle metabolic function in children. On separate days, 14 children (7 boys and 7 girls) completed three knee-extensor incremental tests to exhaustion inside a whole-body scanner (1.5 T, Phillips). The dynamic changes in the ratio of inorganic phosphate to phosphocreatine (Pi/PCr) and intracellular muscle pH were resolved every 30 s. Using plots of Pi/PCr and pH against power output (W), intracellular thresholds (ITs) for each variable were determined using both subjective and objective procedures. The IT(Pi/PCr) and IT(pH) were observed subjectively in 93 and 81% of their respective plots, whereas the objective method identified the IT(Pi/PCr) in 88% of the plots. The IT(pH) was undetectable using the objective method. End exercise (END) END(Pi/PCr), END(pH), IT(Pi/PCr) and IT(pH) were examined using typical error statistics expressed as a % coefficient of variation (CV) across all three exercise tests. The CVs for the power output at the subjectively determined IT(Pi/PCr) and IT(pH) were 10.6 and 10.3%, respectively. Objective identification of the IT(Pi/PCr) had a CV of 16.3%. CVs for END(pH) and END(Pi/PCr) were 0.9 and 50.0%, respectively. MRS provides a valuable window into metabolic changes during exercise in children. During knee-extensor exercise to exhaustion, END(pH) and the subjectively determined IT(Pi/PCr) and IT(pH) demonstrate good reliability and thus stable measures for the future study of developmental metabolism. However, the objectively determined IT(Pi/PCr) and END(Pi/PCr) displayed poor reliability.