Few validated studies have tested the n-alkane method for diet estimation in non-ruminants and the technique has rarely been validated for estimating diet composition for more than two dietary components. The arboreal marsupial Trichosurus vulpecula was fed leaf and fruit diets with up to five component species in two trials. In Trial 1, alkane recovery was estimated, and in Trial 2 these estimates were used to predict digestibility, diet composition and dry matter intake. Alkane recovery increased non-linearly (convex up) with increasing n-alkane chain length. Recovery was linearly and negatively correlated with diet digestibility and appeared to explain a progressive bias with observed digestibility in the alkane estimates of digestibility (slope = 0.37). Diet composition was successfully estimated for up to five leaf dietary components without correction for recovery. Correcting for recovery gave less reliable diet-composition predictions, indicating that the recovery estimates could not be extrapolated to the second trial. Dry matter intake appeared to be relatively robust to variation in alkane recovery between individuals (mean bias =1.6%). If recovery is shown to vary with digestibility in other taxa, calibration of the alkane technique will require diet- and taxon-specific calibration trials to give reliable estimates of diet composition and dry matter digestibility.