Two experiments examined the acoustical correlates of traffic sound aversiveness. In experiment I all pairs of 13 45‐s traffic sound recordings were presented; subjects chose the sound of each pair to which they would rather be exposed and judged the relative similarity of the sounds. The energy equivalent sound level, Leq, accounted for choice and similarity judgments better than any other noise index, and there was no substantial improvement in ability to account for subjective judgments when other indices were used in conjunction with Leq. In experiment II subjects judged the same sounds which had been made equal in Leqlevel. Responses were nonrandom, showing that Leqis not the only subjectively important information in traffic sounds, but no acoustical measure employed in the study captured that information. Instead, two subjective measures of information content accounted best for the data of experiment II. Response data for the equal Leq sounds were less structured than for the sounds differing in Leq. The data of the two experiments were statistically related; it is argued that the data sets were related via the measure of information content. Experiment I replicated a previously reported study, while experiment II failed to replicate the results of a second study.