Role of duration and quality in nonnative perception

and production of English front vowels


Šárka Šimáčková


Vowel quantity rather than quality is identified as a core feature of EIL, essential for its intelligibility and acceptability. This concerns both duration signalling contrast and contextually variable duration linked to obstruent voicing (and also to prominence of nuclear stress). In perception experiments, duration is found to be a more effective cue in nonnative discrimination of English vowel contrasts than vowel quality. Acquiring appropriate timing properties of vowels thus appears to be more important and possibly also easier for an EFL/EIL learner than acquiring spectral properties.


Duration of English front vowels in relation to their quality is the subject of the present study. The goal is to evaluate the relative importance of duration in identification and production of English [i:, I, e, ae] by nonnative speakers. The study was conducted with 36 EFL learners, prospective language professionals, in the context of a Czech higher education institution. In a free choice identification task, learners identified target vowels in bead/beat, bid/bit, bet/bed, bat/bad pairs. The stimuli were used once in the natural form (as spoken by two GA speakers) and once with edited duration (duration of [I] and [e] in bid/bit and bed/bet was increased to that of [i:] and [ae] in bead/beat and bad/bat; duration of [i:] and [ae] was shortened in the opposite direction.)


Learners identified [i:] and [I] reliably in both voicing contexts and in both the natural and edited conditions. An insignificant tendency to mislabel shortened [i:] in beat was noted.  When [i:] and [I] were produced by the learners, they sufficiently differed from each other both in quantity and in quality (i.e. in midpoint value of F1and F2). This is attributed to positive transfer of L1 categories since an equivalent pair of vowels exists in Czech.


Learners relied on duration in identifying [ae] and did not compensate for the contextual effect of obstruent voicing. [ae] was accurately identified in the natural tokens of bad where it had increased duration due to the following voiced /d/. In natural bat and in edited bad, the identification rate dropped to chance. Vowel [e] was identified with equal accuracy in both natural bed and bet. However, editing had a differential effect. Increasing duration of [e] resulted in more [e]->[ae] misidentifications when [e] of bed was levelled with [ae] of bad.  Though the learners relied on duration in the perception task, they produced only a small durational difference between [e] and [ae]. Duration ratio [ae]:[e] is 1,2 for bad:bed and 1,3 for bat:bet. In terms of quality, there is a statistical difference in F2 but the difference is small and learners’ [ae] completely overlaps with [e] of native speakers. In other words, in nonnative speech vowels [ae] and [e] merge.


Learners’ productions did not show any effect of the voicing context on the duration of any of the four vowels.


A follow up experiment focuses on the problematic [e]-[ae] contrasts. It asks whether classroom pronunciation training in an EFL setting can improve identification and discrimination of the vowels and affect quality and quantity properties of the vowels in learners’ production.