This study evaluated the feasibility of using different doses of polyethylene glycol (PEG) as an external marker of faecal output in sheep fed permanent grasslands fodder and compared two near-infrared reflectance spectroscopy (NIRS) calibration strategies for determining faecal PEG content. Three levels of PEG (0.25%, 0.75% and 1.5% of total daily intake) were administered to eight wethers, with each level dosed twice daily. Animals were fed forage obtained from two permanent grasslands cut at two dates during the first cycle of growth. Polyethylene glycol recovery rate was higher (P < 0.001) at the highest dose (0.78) and decreased as dose level decreased (0.61 and 0.30 for PEG levels of 0.75% and 0.25% of total daily intake, respectively). NIRS calibration equations established on PEG data dosed directly on the faecal samples (0.61) gave higher (P < 0.001) PEG recovery rates than NIRS calibration equations performed on mixtures of faeces with different PEG concentrations (0.49). Finally, faecal output estimates were more accurate (P < 0.001) when faeces were sampled at 8:00 (0.61) than at 16:00 (0.51). The highest PEG recovery rate (0.88) was achieved using the highest dose on morning samples when PEG content was estimated by NIRS using turbidimetric results as reference values. We conclude that the usefulness of PEG as an external marker for estimating faecal output on permanent grasslands is limited at PEG doses lower than 1.5% of intake. (C) 2013 Elsevier B.V. All rights reserved.