Motivated by the particular problem of predicting defoliation based on a measure of gypsy moth egg mass density, prediction in the presence of measurement error is discussed. The measurement error variances and covariances are allowed to vary from unit to unit and are estimated by some type of within unit sampling. A general discussion is given as to when one should correct for the measurement error, and a method of estimating the prediction standard deviation is given when a correction is needed. The results are illustrated with the defoliation example.