The ratio and regression estimators of the population total based on double sampling procedures are shown to be .sensible. from the standpoint of the predictive approach; that is, they could be regarded as predictive estimators.But if the usual ratio-type (biased) estimator in double sampling is used as a predictor in respect to the unobserved units of the population, the resulting estimator is different from the usual one.The bias of this new estimator is found and compared with that of the usual one.If the new estimator is corrected for the bias, however, it turns out to be the usual unbiased ratio-type estimator based on a double sampling procedure.