Acoustic impulse resonance frequency sensor and miniaturized VIS/NIR spectrometer were applied on apple fruit Malus domestica 'Idared' and 'Golden Delicious' (n = 800) to predict fruit flesh firmness and soluble solids content (SSC) on tree and in shelf life. Partial least-squares calibration models on acoustic data and VIS spectra of 'Golden Delicious'/'Idared' apple fruits on tree were built for predicting the fruit flesh firmness: coefficients of determination (R-2) and standard errors of cross-validation (SECV) of R-2 = 0.93/0.81 and SECV = 7.73/10.50 [N/cm(2)] were calculated. SSC prediction of freshly harvested apples using NIR spectrometry was obtained with R-2 = 0.20/0.41 and SECV = 1.29/0.94 [degrees Brix]. Prediction of SSC and fruit flesh firmness of stored 'Golden Delicious'/'Idared' apple fruits showed high errors or was not possible. The fruit maturity stage on tree was predicted as classes based on calendar weeks for 'Golden Delicious'/'Idared' apple fruits with 64%/66% correct classification and 92%/84% correct plus neighboring class with SECV = 0.9/0.9 [weeks]. Classes of 'Golden Delicious'/'Idared' apple fruit at different quality levels due to different storage conditions were non-destructively discriminated with 77%/84% correctly classified fruits and 93%/99% correct plus neighboring class with SECV = 0.8/0.5 [classes]. The results show the potential of non-destructive sensors for predicting accepted fruit parameters enabling the determination of optimum harvest date and fruit quality in shelf life. (c) 2005 Elsevier Ltd. All rights reserved.