In animal production digestibility is an essential factor of the valorization of diets. The main factor of variation of digestibility is the characteristics of the feed itself. In recent experiments we have shown that there are also differences between animals for the ability to digest diets. Such genetic experiments require individual digestibility trials, involving a considerable number of faeces analyses. Therefore we now use routinely NIRS for prediction of faeces chemical composition: DM (Dry matter), GE (Gross energy), Protein, Starch , Fat. A step further in the use of NIRS would be the direct prediction of digestibility by NIRS, because it would avoid the precise registration of feed intake and excreta quantities, which would considerably simplify experimental work. In the framework of an experiment on genetic selection in poultry, we measured DM and GE digestibility values, and the corresponding faeces samples were kept for NIRS analysis. This resulted in 240 faeces samples of known DM (dDM) and 173 of known GE (dGE) digestibilities. Diffuse reflectance NIRS spectra of samples were measured on a FOSS NIRSystem 6500 spectrometer. Calibration equations were built after mathematical pre-processing of spectral data (SNV and detrend with 2nd derivative of spectra) without visible wavelengths. Partial Least Squares regression (modified PLS in WINISI software) was found to be the most efficient method for calibrations. Equations obtained had R2, SEC and SECV values of 0.97, 2.11, 2.67 for dDM and 0.99, 1.31, 1.93 for dGE. The corresponding SD/SECV rations were 4.6 and 5.8 respectively. A validation was run by removing 60 samples (randomly chosen) from the database and predict them with equations developed on remaining samples. SEP values obtained were 3.82 and 2.18 respectively, which shows the robustness of equations. These results confirm that the digestibility of DM and GE can be predicted with a sufficient accuracy directly from poultry faeces in our experimen