A series of calculations was performed to ascertain how well one frequency‐weighted rating, such as weighted sound level, loudness level, or perceived noise level, may be predicted from another such rating. A total of 103 average sound level spectra, measured at several distances from different types of highways, was used in these calculations. It was found that knowing a single noise rating, such as the A‐weighted sound level, enables one to predict other outdoor ratings in this set of 103 spectra with a standard deviation of the order of 1–2 dB. If, in addition, traffic speed and mix and the distance to the highway are taken into account, these standard deviations can be reduced to 0.5–1 dB, depending upon the particular noise rating of interest. Equations are given for predicting one rating from another; the associated standard deviations are presented as a measure of how well any given rating can be predicted from a single measured, or otherwise known, noise rating. It is concluded that it is not critical which frequency‐weighting procedure is used in conjunction with highway noise criteria since one descriptor can be predicted from another with small statistical uncertainty. Thus, if human response criteria, or stimulus‐response relationships, have been developed in terms of one frequency‐weighting procedure, these may be translated into equivalent criteria expressed in terms of a metric that is easier to measure or predict.