Analysis by near-infrared spectroscopy (NIRS) was investigated as a means of predicting quality parameters of ovine milk. Calibration equations were developed with samples of ovine milk obtained from a flock of Manchega and Lacaune dairy ewes at different stages of lactation for a wide variation in milk composition. Prediction equations for milk protein, fat, and total solids content were developed by use of reflection or transflection methods to measure absorbance values. Accuracies of measurements were compared. R2 (squared multiple correlation coefficient) values were satisfactory in most cases. The highest R2 value for milk protein content (0.92) was obtained in transflectance mode with unhomogenized milk. The highest R2 values for fat (0.99) and total solids (0.98-0.96) content were obtained in both a transflectance mode without sample conditioning and in a transflectance mode with milk homogenized at 40 degrees C. To validate the calibration, an independent set of 40 milk samples was used. The best r2 (simple correlation coefficient) values for protein, fat, and total solids were 0.92, 0.97, and 0.92, respectively. The study showed that NIRS is a potentially useful technique for evaluating the composition of unhomogenized ovine milk.