Voicing effect on vowel duration is not an absolute universal

A familiar model of the linguistic analysis of speech tends to describe speech in terms of the segment units of phonetic transcription under the assumption that nondiscrete physical properties of speech can be derived from the discrete elements by universal rules of performance. Thus it is widely accepted that vowels are universally longer before voiced consonants than before voiceless ones; for example, in English tab and tap, the vowel is obviously longer in the first. A spectrographic test of Arabic minimal pairs by eight Arabs, however, revealed that Arabic did not exhibit a difference in vowel duration as a function of the segmental “voicing” feature. Yet to further support this finding another spectrographic test of English minimal pairs spoken by eight Arabs revealed that they had considerable difficulty learning the novel temporal implementation for the English “voicing” feature. These results disprove the assumption of current linguistic analysis that nondiscrete physical properties of speech sou...