This study assessed the efficacy of the n-alkane technique to estimate intake and diet composition in animals given single foods or a choice of two. In the first experiment intakes of pelleted ryegrass and lucerne, given either alone or as a choice, were measured in lambs housed indoors in individual pens. Each of the three feeding treatments was given to 12 lambs at two degrees of maturity (0·30 and 0·45 of estimated mature sizes). The 12 lambs were constituted as three replicates of the two sexes of each of two breeds. The measured intakes were compared with those estimated using the n-alkanes C31 and C33, found naturally in the foods, and C32, which was given as a dose. On the choice treatment diet composition was estimated using a non-negative least squares procedure and data on C31 and C33 alone. The agreement between actual and predicted intake was good: R2 of 0·938 for log-linear regression with a residual standard deviation of 0·0845. Intake of lucerne when offered alone was slightly yet significantly over-predicted. The proportion of ryegrass in the diet was also accurately predicted (R2 of 0·950 and residual s.d. of 0·0398). Using the data on C27 and C29, in addition to that on C31 and C33, gave a poorer agreement with the observed diet compositions. The low and similar levels of C27 in the two foods meant that this n-alkane provided little extra information that could be used to estimate diet composition. In a second experiment faecal samples were collected every 4 hours over a 24-h period in six lambs on ad libitum, and in six lambs on a restricted quantity, of pelleted ryegrass. There was no significant diurnal variation in the ratios of either C31 or C33 to C32 on either ad libitum or restricted feeding. The time of faecal collection within a day should not therefore affect the reliability of the predictions. The study confirmed the value of using n-alkanes in methods to determine the intake of forages by sheep, and that the time of faecal collection within the day does not affect the reliability of these predictions. The results also confirm the utility of the n-alkane method for estimating diet choice, at least with two-component mixtures.