In this study coarticulatory effects on the formant frequencies and on the duration of the Dutch schwa were investigated, both in open syllables and in closed syllables, by using nonsense words of the form C 1ə C 2 V and VC 1ə C 2 . In these nonsense words C 1 and C 2 could be any of the consonants /p, t, k, f, s, χ, m, n, η, r, l, j, ν/ and V was taken from the vowel set /i, a:, u/. Consonants and vowels were systematically varied in all possible combinations which gave a total of 897 test words that were read aloud by three male speakers. It appeared that the coarticulatory effects on the schwa could be successfully described with a simple linear model. Especially for F 2-tracks of schwas, the model fit turned out to be very good. The model for F 2-tracks could also be successfully applied to schwas in meaningful words. We believe that the schwa should be interpreted as a vowel without articulatory target that is completely assimilated with its phonemic context. The widespread view of a schwa position in the centre of the vowel triangle, that the formant patterns of reduced vowels are shifting to, is not very accurate. In our interpretation vowel reduction results in a shift of formant frequencies to a schwa position that can be almost anywhere in the vowel plane, dependent on the phonemic context.