Chemical methods for the analysis of connective tissue components of beef in relation with meat quality are quite heavy and expensive. However, near infrared reflectance spectroscopy (NIRS) is a simple and fast technique allowing estimation of a large number of different parameters simultaneously. This study presents preliminary results for the use of NIRS in the prediction of the main chemical constituents of beef meat connective tissue (collagen, cross-links (CLs), proteoglycans (PGs)). In order to create variability, a set of 120 freeze-dried samples obtained from three muscles and three breeds were used. The NIRS prediction models obtained for total collagen (expressed in mg/g of dry matter) and total PGs (expressed relative to collagen) showed a R2 of calibration of 0.85. Insoluble collagen, CLs and PGs expressed relative to dry matter showed an R2 of calibration lower than 0.60. These results suggest that the development of the NIRS calibration model is possible for predicting beef collagen content but needs more development both for collagen content and for the other components.