The aim of this study was to compare the n-alkanes and the ytterbium (Yb)/faecal index techniques as two methods for estimating the herbage intake of dairy cows fed indoors on different herbage : supplement ratios and feeding levels. The supplement was a mixture of maize silage and soyabean meal (ratio of 87 : 13 on a dry matter (DM) basis). In all, four treatments were studied. The herbage : supplement ratio in the diet was 25 : 75, 50 : 50, 75 : 25 and 50 : 50 for treatments 1, 2, 3 and 4, respectively. Animals were offered for treatments 1, 2 and 3, 100% of ad libitum intake measured before the experiment and 70% of ad libitum intake for treatment 4. Cows were fed herbage in the morning and supplement in the evening. A total of six lactating Holstein dairy cows were used in a 4 × 4 Latin square with four 14-day periods. Herbage and supplement intakes, faecal output (FO), in vivo organic matter (OM) digestibility and faecal recovery of markers were measured on the last 5 days of each period. Intake was estimated with the two methods and from two faecal sampling techniques, that is, total faecal collection v. grab sampling during milking. Mean herbage intake as fed, or estimated from n-alkanes or from the Yb/faecal index was 7.7, 8.1 and 10.2 kg DM, respectively. The mean prediction error, expressed as a fraction of actual herbage intake, was 0.10 and 0.50 for the n-alkanes and Yb/faecal index methods, respectively. The n-alkanes method clearly showed much better accuracy than the Yb/faecal index method for estimating intake, irrespective of the faecal sampling method, herbage : silage proportion or feeding level. For the n-alkanes method, herbage intake was slightly overestimated (7%) when herbage proportion in the diet was high, due to a ratio of faecal C33 : C32 recovery >1. The high bias for the Yb/faecal index was due to the cumulative effect of overestimation of FO (mean recovery of Yb = 0.92) and underestimation of the diet indigestible fraction (-8%). Between-treatment variations of FO were on average well estimated by Yb. Between-treatment variations of OM digestibility estimated using the faecal index technique were lower than those observed in vivo. It is concluded that intake of grazing dairy cows receiving high levels of maize silage supplement should be estimated using the n-alkanes method.