Experiments of genetic selection on animals require the registration of individual performance parameters on a large number of animals. These measurements become very heavy to carry out when chemical analyses are involved. This study on the potential of Near Infrared Spectroscopy (NIRS) for prediction of excreta composition in poultry was part of an experiment on the evaluation of the genetic contribution to the variation in digestibility in broilers. A total number of 432 broilers were raised under controlled conditions, with a low digestibility diet. Growth performance and feed intake were measured throughout the trial. Digestibility trials on all individuals were carried out between 19 and 22 days of age, resulting in 432 freeze-dried excreta samples. These samples needed to be analysed for DM, gross energy, crude fat, starch, total nitrogen and true protein in order to calculate data for metabolizable energy and nutrient digestibilities. As this represents a very huge amount of chemical analytical work, it was decided to analyse only 100 samples by reference methods and predict the remainder with NIRS. For total nitrogen and true proteins, equations developed previously on another set of samples were updated with 61 measurements performed on samples selected in this series. Spectra were collected in reflectance mode on a Foss 6500 spin cell equipment. Data treatment was made with PLS regression. The 100 samples to be analysed by reference methods were selected to be representative of spectral variability. Reliability of prediction models was estimated by cross validation with 4 sub-groups. The SECV and SD/SECV values obtained were respectively 111.05 kJ/kg and 10.05 for gross energy, 0.33% and 11.28 for crude fat, 0.59% and 10.25 for starch. The overall models obtained for total nitrogen and true protein had SECV and SD/SECV values of 0.26% and 4.32 (total nitrogen) and 0.66% and 4.68 (true protein). These results allowed the calculation of digestibility paramet