The capability of near infrared reflectance spectroscopy (NIRS) was evaluated to predict the content of total ash (TA), crude protein (CP), crude fiber (CF), neutral detergent fiber (NDF), acid detergent fiber (ADF) and metabolizable energy (ME); as well as pH and ammonia nitrogen content (N-NH3), in pasture silage, with and without additives. Nine hundred and twenty dried and ground samples of pasture silage, with known chemical composition, were scanned over the visible and NIR region (400 to 2500 nm) at 2 nm intervals. Calibration equations were developed by modified partial least square regression models (MPLS) with different mathematical treatments and light scatter correction as standard normal variation and Detrend (SNV & D) of the spectra. For each parameter, the optimum calibration was evaluated on the basis of the cross validation determination coefficient (1-VR) and standard error of cross validation (SECV). NIRS showed a high predictive ability, with 1-VR > 0.89 and SECV (%) of 5.14, 6.69, 9.96, 16.01 and 9.15 for A, CP, CF, NDF and ADF, respectively. NIRS showed moderate accuracy for ME, with 1-VR > 0.87, SECV: 0.07 Mcal kg-1 and low accuracy, although with feasibility as a ranking method, for pH and N-NH3, with 1-VR > 0.72 and SECV of 0.14 and 1.49, respectively. It is concluded that the equations obtained can be used to predict the nutritional composition of pasture silages.