When “AA” is long but “A” is not short: speakers who distinguish short and long vowels in production do not necessarily encode a short–long contrast in their phonological lexicon

Abstract

In some languages (such as Dutch), speakers produce duration differences between vowels, but it is unclear whether they also encode short versus long speech sounds into different phonological categories. To examine whether they have abstract representations for ‘short’ versus ‘long’ contrasts, we assessed Dutch listeners’ perceptual sensitivity to duration in two vowel qualities: [a] and [ɑ], as in the words maan ‘moon’ and man ‘man,’ which are realized with long and short duration respectively. If Dutch represents this phonetic durational difference as a ‘short’–‘long’ contrast in its phonology, duration changes in [a] and [ɑ] should elicit similar neural responses [specifically, the mismatch negativity (MMN)]. However, we found that duration changes evoked larger MMN amplitude for [a] than for [ɑ]. This finding indicates that duration is phonemically relevant for the maan-vowel that is represented as ‘long,’ while it is not phonemically specified for the man-vowel. We argue that speakers who in speech production distinguish a given vowel pair on the basis of duration may not necessarily encode this durational distinction as a binary ‘short’–‘long’ contrast in their phonological lexicon.

A Cross‐linguistic Examination of Toddlers’ Interpretation of Vowel Duration

“…This asymmetry was also revealed in an electrophysiological study of Dutch adults by Chládková et al. ().…”

“…Dutch children matched Dutch adults in being less strongly affected by vowel lengthening than vowel shortening (Chládková et al., ; Nooteboom & Doodeman, ; Van der Feest & Swingley, ). Why might this be?…”

“…Like English, Dutch is not generally considered to be a language that has durational oppositions in its vowel system, unlike languages such as Japanese, Thai, or Hungarian, which have phonologically distinct long and short versions of several vowels that are similar in quality. Still, as with English, Dutch adults’ judgments of vowel identity can be affected by manipulations of vowel duration (e.g., Chládková, Escudero, & Lipski, ; Tillman, Benders, Brown, & van Ravenzwaaij, ). In particular, Nooteboom and Doodeman () found that for the word pair /tɑk/‐/taːk/, Dutch listeners perceived artificially shortened [aː] as /ɑ/, but not lengthened [ɑ] as /aː/.…”

Perceptual Sensitivity to Spectral Change in Australian English Close Front Vowels: An Electroencephalographic Investigation

“…If listeners detected a change from the Standard stimulus (a vowel with a static F1 × F2 trajectory) to the Deviant types (vowels containing dynamic F1 × F2 trajectories), reliable differences between the Oddball and Control conditions, i.e., MMN responses, will emerge according to F1 × F2 trajectory directions and lengths. Statistical analyses of MMN effects are often performed on the differences between each subject's mean amplitudes from the Oddball and Control conditions per Deviant type (e.g., [12], [18]). Limitations of prior averaging include the inability to consider the variance within each subject's pool of epochs [20] and across randomly-assigned blocks [21], and the assumption that experimental manipulations have uniform effects across subjects [22].…”

“…We were interested the MMN response for the four Deviant types, i.e., amplitude differences between the Oddball and Control conditions. The following procedure identified the latencies at which this MMN amplitude peaks (cf., [12] and [18]). For each subject, their frontocentral midline electrode (FCz) waveforms were averaged across each block (Figure 2).…”