In this study, the traditional claim that loudness is a weak cue in the perception of linguistic stress is reconsidered. This claim is based on perception experiments in which loudness was varied in a naive way: all parts of the spectrum were increased by the same amount of energy. In an earlier study we found that if a speaker produces stressed syllables higher frequencies increase more than lower frequencies. Varying loudness in this way would therefore be more realistic, and should bring its true cue value to the surface. Results of a perception experiment bear out that realistic loudness manipulations (i.e., concentrated in the higher frequency bands) provide stronger stress cues than traditional intensity differences, and are close in strength to duration differences.