When isolated vowels were produced beyond F0 of speech (F0>150 Hz for men, F0>300 Hz for women and children), the related formant patterns were found to deviate substantially from the values of formant statistics, and the formant frequency variations proved to be nonsystematic (for details, see poster Klinkert and Maurer). Moreover, formant frequency estimation for vocalizations with F0>400 Hz is highly problematical, although many vocalizations remain unambiguous in the vowel identity up to F0=700 Hz and even above. The nonsystematic formant variations and the methodological problem of formant frequency estimation hardly allow for a normalization of formant patterns. Dynamic spectral properties do often not appear in isolated vowels. Thus, for the acoustic theory of vowel sounds, the question arises as to whether there is a concept to determine the physical correlates of the sounds apart from formants. A first approach toward such a concept is described for voiced vowels: A hypothesis is presented which predicts characteristic differences in the harmonic spectra of different vowels. Implications for vowel perception are discussed.