Vowel representation: Some observations on temporal and spectral properties of the first formant frequency

Acoustic analysis of the vocalic portion of consonant–vowel–consonant (CVC) syllables (where V is one of the five vowels [I,q,1,a,v] of American English) spoken by three speakers (two males and one female) in the sentence frame ‘‘The — again’’ is presented. Results of acoustic measurements show that ambiguities between vowels, for each speaker, occur if the vowels are represented by the values of F1 and F2 sampled at the time where F1 reaches its maximum. These ambiguities occur primarily in the F1 dimension. Examination of the F1 trajectories of the vowels for which confusion occurs shows variations in the way F1 reaches its maximum among different vowels. In particular, if two different vowels such as [I] and [q] have the same maximum F1, then F1 for the lower vowel reaches its maximum value earlier. In addition, results show that the F1 onset frequency also might be important in determining vowel height. The implication is that the spectral characteristics at a particular ‘‘target,’’ represented by the time at which F1 reaches its maximum, are not invariant attributes of the vowel. The results support a hypothesis that time and/or frequency variations of the first formant must be taken into account if an invariant property is to be associated with a vowel.