Twenty-five different pea extrudates were produced using a Clextral BC4S twinscrew extruder. Near infrared reflectance (NIR) spectra of the extrudates were recorded between 1100 and 2.500 nm. The application of principal component analysis (PCA) on these spectral data permitted discrimination between the extrudates produced under moderate, intermediate or severe conditions of extrusion-cooking. The study of protein quality through solubility experiments, polyacrylamide gel electrophoresis, determination of available basic amino acids and brown colour measurement showed that, under moderate conditions of extrusion-cooking, protein was only slightly affected. All protein subunits could be restored after solubilization in phosphate buffer containing a detergent and a reducer of disulphide bonds. However, under severe conditions of extrusion-cooking, the extrudates exhibited a relatively intense brown colour and all protein fractions were implicated in nondisulphide covalent bonds and underwent some macromolecular degradation. Therefore, the spectral discrimination between pea extrudates which was related to the processing parameters, could also be related to the degree of transformation of these extrudates. Thus, the method may be used to classify an extrudate produced under unknown conditions by projecting its spectrum on the similarity maps obtained by PCA on a known collection of extrudates.