This study was designed to obtain information on predicting intake and digestibility from near infrared reflectance spectroscopy (NIRS) scans of faeces in dairy cows given different diets and levels of intake. Comparisons were made between using NIRS to predict alkanes in faeces and direct calibration of NIRS for intake and digestibility. Faecal samples were obtained from 91 cows in five experiments where dry-matter intake (DMI) had been measured for individual cows. All samples were scanned by NIRS and concentrations of alkanes C32, C33 and C36 were determined in 32 samples. DMI (mean 19-4, s.d. 5-06 kg/day) was estimated with standard errors of 0-36 and 0-44 kg/day from C32 and C36 alkanes determined by gas chromatography, and with standard errors of 1-17 and 1-42 kg/day when DMI was estimated from C32 and C36 alkanes predicted by NIRS. When DMI was predicted directly by NIRS, the standard error of prediction was 0-48 kg/day (R2 = 0-97). Prediction of dry matter digestibility by NIRS was not accurate (standard error of cross validation = 0-032; R2= 0-68), probably because of the limited variation in digestibility values within the data set. It is concluded that using NIRS to predict alkane concentrations of faeces does not give accurate estimates of DMI but direct prediction of DMI by NIRS gives estimates with similar accuracy to estimates derived from the traditional alkane technique.