Attempts have been made to assess the end-point temperature (EPT) of heated ®sh and shell®sh meats by near-infrared (NIR) re¯ectance spectroscopy. Since the presence of water affects NIR spectra, the in¯uence of the water content of samples on the performance of NIR spectroscopy for determining EPT was also evaluated. Blue marlin (Makaira mazara), skipjack (Katsuwonus pelamis), red sea bream (Pagrus major), kuruma prawn (Penaeus japonicus) and scallop (Patinopecten yessoensis) meats were heat-treated at different temperatures (5°C intervals between 60 and 100°C). NIR spectra were measured at 2 nm intervals between 1100 and 2500 nm. A stepwise multiple linear regression method was used to develop a calibration curve. Changes in NIR re¯ectance spectra upon heat treatment were related to the heating temperature. Moreover, the interference from the variation in water content on the prediction of EPT was eliminated by selecting appropriate wavelengths. Plotting of NIR-predicted temperatures determined from d 2 log(1/R) at four speci®c wavelengths against the actual EPT revealed a promising linear relationship with correlation coef®cients between 0.94 and 0.98. In the temperature range 60±100°C, NIR re¯ectance spectroscopy was able to detect EPT of heated ®sh and shell®sh meats with a prediction error of 1.9±3.1%.