Abstract
Because vowels in English are largely distinguished by the frequencies of their first two formants (F1, F2), the division of the (F2, F1) plane is an important and quantifiable component of accents. We report results of a web-based study into some of the many accents of English. Participants identified the vowel in h[vowel]d words produced by synthesis from a large set of possible values of F1, F2 and F3, using two different fundamental frequencies and two different durations. Compared to analysing spoken utterances, this approach has a number of obvious disadvantages, which we discuss. It has the significant advantages, however, of low cost, large scale and wide-ranging international participation. It is then possible to use the same experimental protocol to characterise the (perceptual) vowel plane of a substantial number of subjects and accents, thus allowing simple comparisons. From the large data base thus acquired, we present four examples of vowel maps for different Anglophone countries and regions therein. Knowledge of local variations in the perceptual (F2, F1) map, and the way in which these depend on fundamental frequency f(0), is not only of phonetic interest, but may be useful to those who use synthetic speech in automated communication systems.