The alkane technique allows estimates of intake and diet composition. For the latter a correction for incomplete recovery of faecal alkane concentrations has to be made. Feeding level and diet proportions have been discussed as factors possibly influencing recovery. A balance trial was conducted to study these effects on faecal alkane recovery rates and their consequences on estimates of roughage intake in sheep. The diets consisted of force-dried pelleted alfalfa meal (A) and wheat whole meal (W) in three levels each (360, 480, 600 and 270, 360, 450 g/day, respectively); the wheat was partly labelled with beeswax to provide a distinctive alkane pattern. Each diet was fed to four wethers. Despite significant effects (P<0·05) of A and W on organic matter digestibility, no such effects were observed for alkane recovery rates. Estimates of alfalfa intake were better if alkanes with a higher coefficient of variation of recovery were not included in diet composition estimates. Furthermore, in spite of the lack of significance of the dietary factors A, W and A×W, faecal alkane concentration correction based upon diet-specific faecal recoveries gave better estimates of alfalfa intake than based on overall recoveries. The estimate of alfalfa intake differed from known intake by −1·7 to −2·6%, depending on the method of estimation. The relative mean discrepancy was regarded as a tool to assess estimated intakes of individual animals. It ranged from 4·4 to 14·2% for dietary treatments and was 7·6% for all animals. The poor quality of estimates for individual animals may result from the variation of faecal recovery rates and needs further investigation.