The input of a waveguide probe for shell eggs was connected to a sinewave sweeper oscillator and the signal at the output was captured by a spectrum analyser. A first analysis was carried out in the range from 3 to 20 GHz with a span of 1 GHz to investigate which 1 GHz frequency range contains most information for predicting the main quality indices of eggs during 15 days of storage. Simple linear regression models were therefore set up and the coefficient of determination was calculated. The absorbance spectra in the range thus identified (from 10.5 to 11.5 GHz) were used to predict the quality indices by means of an artificial neural network (ANN). The R(2) values of the obtained ANN in validation mode were 0.918, 0.854 and 0.912 for the air cell, the thick albumen height and the yolk index, respectively. The correlations between the quality parameters and tests carried out on albumen, yolk and plastic eggs for simulating the air cell showed how one index can be indirectly predicted through another one. (C) 2010 Elsevier Ltd. All rights reserved.