A study of the effect of dietary variations on muscular efficiency in 3 healthy subjects, using the bicycle ergometer of Sevringhaus (1926). "Net efficiency" is calculated as calories of work/working minus resting heat production. No significant differences between figures found with diet of ordinary mixed type, and limiting regimes of pure carbohydrate and high fat (carbohydrate-free to marked ketosis). Evidence of "training effect" was found only in first few rides, whereafter figures average 22-24%. The findings vary from those of Krogh and Lindhard (1920), but agree better with those of Benedict and Cathcart (1913).