Perceived Swedish vowel quantity: effects of postvocalic consonant duration

In the production of Swedish, vowel quantity is known to be realized in the vowel, but also affects duration of a postvocalic consonant. The goal of this study is to examine the use of postvocalic consonant duration as a perceptual cue to vowel quantity. Listeners ́ responses and reaction times were recorded for synthesized materials in which the vowel spectra and duration were kept constant and the postvocalic consonant duration was adjusted. Results show no indication that listeners actively used the duration of a postvocalic consonant to identify vowel quantity. These findings suggest that adjustments in postvocalic consonant duration in Swedish productions may be temporal artifacts of the preceding vowel quantity rather than reflecting linguistically relevant information. 1. SWEDISH VOWEL QUANTITY The vowel system of Swedish has traditionally been described as having contrastive vowel quantities, with "short" (e.g., [a] in [tak] tack "thanks”) and "long" (e.g., [A:] in [tA:k] tak "roof”) vowels characterized as similiar in quality but phonologically distinguished by their relative lengths (e.g., [1]). Phonotactically in Swedish, a short vowel quantity will only occur in a closed syllable (e.g., [a] in [tak:] tack "thanks”), whereas a long vowel quantity can occur in either an open or closed syllable (e.g., [A:] in [tA:] ta "take”, or in [tA:k] tak "roof”) (e.g., [1]). This pattern is highly consistant in Swedish. In speech production, the distinction between short and long vowel quantities is realized in the acoustic signal corresponding to both the vowel and, if present, a postvocalic consonant. The duration of a long vowel quantity extends over more time than a short vowel quantity which, in turn, may allow more time for an articulation using greater extremes of the vocal space, and consequently may also affect the vowel spectrum. In particular, long vowel quantities are generally known to be articulated with more closure than short vowels, with the open articulation of [A:] and [a] being a possible exception. Swedish vowel quantity is also known to affect a postvocalic consonant in speech production. The duration of a postvocalic consonant has been regularly observed [1] to have an inverse relationship to the vowel duration, so that following a short vowel the duration of a postvocalic consonant tends to be relatively long (e.g., [k:] in [tak:] tack "thanks”), whereas following a long vowel the postvocalic consonant duration tends to be shorter (e.g., [k] in [tA:k] tak "roof”). In a classic study on Swedish, Hadding-Koch and Abramson [2] investigated whether vowel duration or spectral attributes of a vowel had the more dominant perceptual role in distinguishing short and long vowel quantities. For three vowel pairs, tape recordings were carefully spliced with differences of 10-15 ms, resulting in approximately 5-8 steps from the long to the short vowel quantitiy. Although the role of spectral characteristics could not be excluded from being an important perceptual cue, their results show that length is the primary parameter distinguishing Swedish vowels. Recent studies[3][4] reexamined the effects of vowel duration and spectrum on perceived vowel quantity identification with the goal of identifying thr role of F1 and F2 for identifying vowel quanity. Based on natural Swedish productions, three sets of 100 /kVd/ materials were synthesized, one set for each of three "long-short" vowel pairs [4]. Within each set 10 stepwise adjustments of vowel duration and 10 stepwise adjustments of the first two vowel formants (F1 and F2) were made from the long vowel quantity to the short one, resulting in a total of 100 synthesized /kVd/ items for each of the three vowel pairs. The closure duration of the postvocalic /d/ was kept constant. Subjects ́ identification are presented again here in Figure 1. The results illustrate that Swedish listeners use vowel duration more than spectral information to identify the quantity of the non-low vowel pairs ([i:]-[I] and [o:][O]), whereas for low vowel pair ([A:]-[a]), they use both vowel duration and spectral attributes of the vowel.