Search

We investigate the pronunciation of English compounds ending in -man (e.g. policeman, clergyman, mailman), some of which are usually pronounced with a full vowel in -man (mailman), some usually with a reduced vowel (policeman), while other words vary between the two (businessman, gunman). Is it possible to predict when the vowel is reduced or not? We investigate a number of factors that could play a role in this issue: (i) the age of introduction of the word into the English language, (ii) the frequency of the word in the synchronic language, (iii) the meaning of -man in the specific compound, (iv) number of syllables, (v) stress pattern and (vi) spelling variation. Although some of these points have been discussed in the literature before, we present the first large full-scale quantitative investigation of the question whether any of these factors are reliable predictors of the pronunciation (and if so, which one predicts the pronunciation best), as well as the interaction between these factors.

Many linguistic analyses rely on phonetic transcriptions. Two factors that might affect the accuracy of transcriptions are perceived accent and the listener's linguistic knowledge about that accent. This article focuses on the susceptibility of transcribers to such accent-induced bias, based on speaker's accent, transcriber's variety, and the transcriber's knowledge about the language variety they are transcribing. We offer a method to statistically detect accent-induced biases in linguistic transcriptions. This article investigates accent-induced bias in transcriptions of Reference French and Laurentian French (in Canada). In Reference French, high front vowels are phonologically tense /i y/, but they may be phonetically realised as lax [ ɪ ʏ ]. In Laurentian French, lax [ ɪ ʏ ] occur as allophones of /i y/. We compare transcriptions of different groups of coders: native speakers of Reference French, native speakers of Laurentian French, and advanced L2 learners of French who are native speakers of Dutch. Half of the Dutch group was informed about the native variety of the speakers and the other half was not. Results show that perceived accent among the native transcribers is a relatively strong predictor for transcribed sounds. In addition, the Dutch transcribers who were informed about the speaker's variety transcribed more lax vowels in Laurentian French and more tense vowels in Reference French than the Dutch transcribers who were not informed. Native transcribers also showed contextual effects related to the allophonic distribution of the vowels in Laurentian French. These results are indicative of accent-induced bias. The results are analysed in a predictive top-down model to explain the effect of prior experience and information about the language variety in speech perception.

Believed dialect influences speech perception by linguistically naïve speakers. How much accent-induced bias affects perception of linguistically trained speakers is still unclear. This study experimentally investigates the influence of believed dialect on plosive perception by subjects who were phonetically and phonologically trained. Identical syllables were presented twice to each subject. In one session, the subjects were informed that the variety was a Mandarin dialect which has voiceless unaspirated and aspirated voiceless stops; and in the other session that it was a Wu dialect, which has voiceless unaspirated, voiceless aspirated, and breathy stops. More breathy stops were reported if Wu was the believed dialect. Plosive phonation in Wu is related to lexical tone, and we show that lexical tone causes another bias to plosive perception. This suggests that linguistically trained transcribers are susceptible to higher order linguistic knowledge and it demonstrates the difficulty of avoiding biased perception when the coder forms a belief about the variety that he/she transcribes. We also advocate speech perception models which include a component that accounts for the role ofexpectedsounds.

This article proposes a measure of the competitive strength of two rival patterns in the domain of a subgroup of irregular verbs in English. There is competition between simple pasts built on the vowels /æ/ and /ʌ/, and the same competition is found in the domain of past participles. As a result of such competition, the past tensestang(fromsting) was replaced withstung. The /ʌ/ forms are more competitive than the /æ/ forms (Bybee & Slobin 2007,Bybee & Moder 2007). To understand this, we counted the number of types for /æ/ (such assang, rang) and /ʌ/ (such asstung, stuck) in the irregular simple past and did the same in the irregular participle (such assat, hadandsung, done). We calculated a measure of competitiveness for these two patterns incorporating type frequency and token frequency. This measure was used to explain why /ʌ/ forms are more competitive than /æ/ forms.

In this paper we discuss the four tones of Standard (Beijing) Mandarin Chinese. First, we will suggest a proposal for their phonological representation. Then, we discuss the order in which they are acquired in first language acquisition, relating this both to the representations we propose and to the relative frequency with which these tones appear in the most frequently used Chinese words. It turns out that the former predicts the order of acquisition more closely than the latter: We provide an explanation for this based on the early stage at which the tones are acquired.

In dialectology, it is often necessary to obtain a measure for the level of dialectal accent shown by individual speakers, especially if statistical analysis is needed. This also applies to studies on standard variants which are “coloured” by regiolects or dialects. In this paper I explore the feasibility of letting native speakers judge the degree of accentedness in Low-Alemannic German. Specifically, I investigate whether listeners who speak a similar dialect as the speakers who are evaluated assign different judgements than listeners who do not. A novel methodology is applied, which involves an on-line elicitation task using audio files. This experiment shows that listeners who speak different varieties of German form a homogeneous group, with respect to rating the level of accent., In der Dialektologie wird häufig ein Maß für den Grad des dialektalen Akzents einzelner Sprecher benötigt, insbesondere, wenn eine statistische Analyse benötigt wird. Das gilt auch für Studien über Standardvarietäten, die durch Regiolekte oder Dialekte "gefärbt" sind. In diesem Artikel prüfe ich, inwieweit es möglich ist, Muttersprachler den dialektalen Akzent in nieder-alemannischem Deutsch einschätzen zu lassen. Insbesondere untersuche ich, ob Hörer, die einen ähnlichen Dialekt wie die bewerteten Sprecher sprechen, zu anderen Urteilen kommen als Hörer, die keinen ähnlichen Dialekt sprechen. Eine neuartige Methode mit einer Online-Elizitations-Aufgabe anhand von Audio-Dateien kommt zum Einsatz. Das Experiment zeigt, dass Hörer, die verschiedene Varietäten des Deutschen sprechen, den erhobenen Bewertungen nach eine homogene Gruppe bilden.