Quality control. a major concern in the production of all fresh tomatoes, is particularly important for 'Raf' tomatoes, a much-prized variety in Spain whose exquisite flavor commands very high market prices. Soluble solid content (SSC) and titrable acidity (TA), the two internal quality indices most contributing to tomato flavor, have traditionally been measured using destructive techniques, which enable analysis of batches but not of individual items. This study examined the feasibility of using NIRS technology to predict internal quality parameters in individual tomatoes. The results obtained using new-generation diode-array instruments, which can be adapted for 'on-site' and 'on-line' measurements, were compared with those obtained with a traditional, laboratory-based NIRS monochromator. Calibration models were developed using different regions of the spectrum and different signal pretreatments. Tomatoes picked in 2006 (N = 180) were used for the initial calibration, which was subsequently enlarged with tomatoes harvested in 2007 (N = 132). The standard error of cross validation (SECV) obtained for the prediction of SSC ranged between 0.55 and 0.74 degrees Brix, whilst SECV values for titrable acidity ranged from 0.06% to 0.1% citric acid; the coefficient of determination (r(2)) was between 0.63 and 0.82 for SSC and between 0.41 and 0.71 for TA. The regression models were tested with independent validation samples (N = 100); the resulting standard errors of prediction (SEP) and r(2) for the same parameters were (0.68,0.77) and (0.08,0.68), respectively. The results obtained suggest that NIRS can accurately predict tomato internal quality parameters, even in the variety studied, which is highly irregular in shape. Wavelengths above 1700 nm were found to be unnecessary for the applications considered, and the diode-array instrument proved sufficiently accurate. (C) 2008 Elsevier Ltd. All rights reserved.