Alkanes can be used as natural markers for estimating diet composition, but a factor should be used to correct for incomplete recovery in faeces. Faecal alkane recovery rates may be influenced by diet and animal factors. However, little research has been conducted to evaluate the effects of herbage species and live weight of animals on faecal alkane recoveries. In the current study, faecal recoveries of alkanes were determined in sheep in four different live weight groups (from 20 to 40 kg) fed three plant species (Elymus sibiricum, Leymus chinensis and L. dasystachys). In a second experiment, the accuracy of intake and diet composition estimates, using alkanes as faecal markers, was assessed by feeding known amounts of the same plant species as a three-component mixture. The results showed that faecal alkane recoveries were influenced significantly by herbage species (P<0·01), but no effect of live weight of animals was observed. Total dry matter intake was estimated correctly based on either C31:C32 or C29:C32 alkane pairs. With respect to estimators of E. sibiricum intake, reasonable results could only be obtained if the faecal alkane concentration was corrected based on diet-specific faecal recovery. More accurate estimations were obtained only if the alkanes found in relatively higher concentrations were used in diet composition estimates instead of using all available alkanes. Due to lower alkane concentrations or similar alkane patterns of L. chinensis and L. dasystachys in the diet, estimates of diet composition of these two herbage species were significantly different from the actual ones (P<0·05), implying that other markers need to be used for accurate estimation.