Purpose. To evaluate training induced metabolic changes noninvasively with 31P magnetic resonance spectroscopy ( 31P-MRS) for measuring muscle fibre type adaptation. Methods. Eleven volunteers underwent a 24-week training, consisting of speed-strength, endurance, and detraining (each 8 weeks). Prior to and following each training period, needle biopsies and 31P-MRS of the resting gastrocnemius muscle were performed. Fibre type distribution was analyzed histologically and tested for correlation with the ratios of high energy phosphates ([PCr]/[P i ], [PCr]/[ β ATP] and [PCr + P i ]/[ β ATP]). The correlation between the changes of the 31P-MRS parameters during training and the resulting changes in fibre composition were also analysed. Results. We observed an increased type-II-fibre proportion after speed-strength and detraining. After endurance training the percentage of fast-twitch fibres was reduced. The progression of the [PCr]/[P i ]-ratio was similar to that of the fast-twitch fibres during the training. We found a correlation between the type-II-fibre proportion and [PCr]/[P i ] ( r = 0.70, P < 0.01) or [PCr]/[ β ATP] ( r = 0.69, P < 0.01); the correlations between its changes (delta) and the fibre-shift were significant as well (delta[PCr]/[P i ] r = 0.66, delta[PCr]/[ β ATP] r = 0.55, P < 0.01). Conclusion. Shifts in fibre type composition and high energy phosphate metabolite content covary in human gastrocnemius muscle. Therefore 31P-MRS might be a feasible method for noninvasive monitoring of exercise-induced fibre type transformation.