Grape seed and skin sensory parameters are important characteristics in making decisions concerning the optimal harvest time and producing high-quality red wines. The potential of near infrared spectroscopy to determine several sensory parameters of seeds and skins was evaluated. Taste (sourness), texture (astringency, tannic intensity, dryness and hardness), visual (colour) and olfactory (intensity and type of aroma) attributes were considered. Calibration models were performed by modified partial least squares regression. The differences between sensory analysis and NIRS analysis in external validation were lower in the case of seeds; they were between 4.5% for hardness and 8.7% for colour. For the skins, differences in external validation were between 9.8% for tannic intensity and 13.7% for astringency. The results obtained show that NIRS technology has considerable potential for predicting the above sensory attributes. (C) 2013 Elsevier Ltd. All rights reserved.