The reflectance spectra of bruised and unbruised tubers was measured with a scanning monochromator and bifurcated fibre optic light guide from 250 to 1750 nm. The spectra were converted into reflectance relative and first derivative format and divided into two groups: known and trial spectra. Discriminant analysis selected wavelengths from the known spectra that were most sensitive to differences between unbruised and bruised tubers and to generate linear classification functions to predict whether a tuber was bruised or unbruised. It was found that 75 to 95% of tubers were correctly identified as bruised, with wavelengths selec- ted in the visible and NIR for unpeeled and peeled tubers. When the classification functions were used on the trial spectra to determine the robustness of the method, none of the unpeeled tubers were correctly classified, while 68 to 85% of peeled tubers were.