In poultry, the determination of protein content in excreta in digestibility trials is complicated since faeces and urine are excreted together. Utilization of total nitrogen as estimation of protein is therefore not possible because urine contains high amount of non-protein nitrogen as uric acid. The classical method used involves the dissolution of uric acid and precipitation of proteins. This method has some drawbacks: it is quite long to perform, it introduces an approximation of protein (use of a fixed calculation ratio between precipitable protein and "true protein") and uses a high amount of lead (2.5g lead acetate / sample resulting in nearly 16g pure lead per digestibility trial) which is harmful to environment. The potential of NIRS for true protein determination was therefore evaluated through two strategies: 1- prediction of the result of the classical method, and 2- prediction of uric acid to estimate true protein content by difference between total nitrogen and uric acid nitrogen. A preliminary study with pure uric acid diluted in wheat flour indicated that the prediction of uric acid content by NIRS was reliable, with SEP=0.21 for uric acid contents ranging between 0 and 30%. A total number of 344 excreta samples coming from many digestibility trials in a wide range of experimental conditions were used for equation building and validation. They were all analysed by reference methods for total nitrogen, true protein determination and uric acid determination. Spectra were collected in reflectance mode on a Foss 6500 spin cell equipment. Data treatment was made with PLS regression. Reliability of prediction models was calculation by both cross validation (with 4 subgroups) and prediction on a subset of 50 samples randomly chosen and not used for equation building. The SECV, SEP, SD/SECV and SD/SEP ratio were respectively 0.22%, 0.26%, 5.10 and 4.30 for total nitrogen, 0.66%, 0.85%, 4.40 and 3.40 for true protein determination and 0.47%, 0.60%, 5.58 and 4