The frequency of sounds is coded partly by phase locking to the temporal fine structure (TFS) of the waveform evoked on the basilar membrane. On the basis of data obtained using sinusoids, it is usually assumed that in mammals, including humans, TFS information becomes unusable for frequencies above about 5000 Hz. Here, sensitivity to the TFS of complex sounds up to much higher frequencies is demonstrated. Subjects discriminated a harmonic complex tone, with a fundamental frequency F0, from a tone in which all harmonics were shifted upwards by the same amount in hertz. The phases of the components were selected randomly for every stimulus. Both tones had an envelope repetition rate equal to F0, but the tones differed in their TFS. To prevent discrimination based on spectral cues, the tones were passed through a fixed bandpass filter, centered at 14F0. A background noise was used to mask combination tones. Performance was well above chance for most subjects when F0 was 800 or 1000 Hz and all audible components were above 8000 Hz. Supplementary experiments confirmed that performance was not based on changes in the excitation pattern or on the discrimination of partially resolved components.