Behavioral and neural correlates of vowel length in German and of its interaction with the tense/lax contrast

Abstract

The purpose of the present doctoral thesis was to investigate whether and how the theoretical account of the German vowel length contrast is mirrored by psychoacoustic characteristics and neural correlates. German vowels are contrasted by vowel duration (+/- long) that interacts with vowel quality (+/- tense).

The questions arise i) which phonetic cue, i.e. duration or quality, is used primarily for the phonological distinction between short and long vowels ii) and to what extent the theoretical account is mirrored in the psychoacoustic and neural reality. This aim was accomplished by combining behavioral psychoacoustic tests and whole-head magnetoencephalography (MEG).

Mainly two behavioral tests were performed in various experiments in order to assess the effects of vowel duration and quality on the perception using temporal and spectral continua: an identification test to assess the category boundary and an adaptive discrimination test, which tests the sensitivity to temporal and spectral change. It was found that the vowel length contrast matched the psychoacoustic criteria for categorical perception: Characteristics exhibited a sharp boundary between category short and long. In addition, identification response times were slower for durations at the boundary in contrast to within-categorical durations. Finally, the discrimination sensitivity to temporal change increased at the boundary as opposed to within-categorical durations. An analysis of the relationship between duration and quality found that listeners have an ambiguous spectral area whose identification depends on vowel length: Short instances from that area were perceived as high back lax vowels (e.g. [T]), long instances as mid back tense vowels (e.g. [n9]). This result was interpreted to be the effect of a strong magnet effect in case of short vowel durations of the high back vowel /u/: Sensitivity to spectral change decreased significantly the lower and shorter the vowel was, extending the vowel space for /u/.

Simultaneously, differences between lax and tense /u/ instances were neglected. The findings of the behavioral tests were supported by two neural experiments in the MEG. The first neural experiment investigated whether and how the category boundary was reflected by auditory evoked responses. At this aim, a passive listening task was performed using a vowel duration continuum embedded in a disyllabic nonsense word. It was found that each syllable of the word was reflected by a typical M50/M100 deflection. The phonological boundary between category short and long, however, was by a reflected neural correlate that emerged between the first and second syllable which has so far not been described. The new correlate occurred for categorically long vowels only, indicating additional neural activity for phonologically long vowels.

The second neural experiment set out to show whether and to what extent the auditory system is sensitive to minimal changes in vowel quality. A priming study was performed for this purpose, examining the effects of within-categorical and across-categorical quality changes in a probe word on the M100 in response to a target word. By these means it was found that the auditory cortex can differentiate between lax and tense instances during the processing of vowel identity, even in short vowels.

By taking the findings in the behavioral and neural experiments into consideration, it was concluded that vowel duration is the primary distinctive phonetic cue for the phonological vowel length contrast in German. Vowel quality serves as a secondary – supporting – cue which enhances the perception of categorical durations as short or long vowels.

Die vorliegende Doktorarbeit untersuchte, ob und inwiefern die theoretischen Annahmen über den deutschen Vokallängenkontrast sich in psychoakustischen Charakteristika und neuronalen Korrelaten widerspiegeln. Deutsche Vokale werden durch Vokaldauer (+/- lang) unterschieden, ein Kontrast der mit Vokalqualität (+/- gespannt) interagiert.

Folgende Fragen haben sich gestellt: 1) Welches phonetische Signal, also Dauer oder Qualität, unterscheidet primär zwischen kurzen und langen Vokalen? 2) Inwiefern spiegelt sich diese Interaktion in der psychoakustischen und neuronalen Realität wider? Diese Fragen wurde beantwortet, indem verhaltensbasierte psychoakustische Tests und Ganzkopf-Magnetenzephalographie miteinander kombiniert wurden.

Hauptsächlich zwei Verhaltenstests wurden in mehreren Experimenten angewandt: Ein Identifikationstest, der die Kategoriegrenzen erfasst und ein adaptiver Diskriminationstest, der die Sensitivität im Hinblick auf temporale und spektrale Veränderungen feststellt.

Zusätzlich wurden zwei neuronale Experimente durchgeführt. Das erste untersuchte, inwiefern die Kategoriengrenze zwischen kurzen und langen Vokalen sich in auditorisch evozierten Potenzialen niederschlägt. Das zweite untersuchte, wie sensitiv das auditorische System auf minimale Veränderungen der Vokalqualität reagiert.

Die Resultate der Experimentenreihen zeigen, dass Vokaldauer dasjenige phonetische Signal ist, das primär zwischen kurzen und langen Vokalen unterscheidet. Vokalqualität dient als sekundäres Signal, das die Wahrnehmung der kategorialen Dauer als kurz oder lang unterstützt.