The perception of dialect
Julia Fischer-Weppler
HS Speaker Characteristics Venice International University
17.10.2007
Perception of dialect
Introduction
Sources of variability are natural
consequences of language variation
Different forms of variability including the impact of regional dialect have to be
included in speech perception research
Perception of dialect
Introduction
Dialect variation is perceived and encoded in everyday language situations
The process of speech perception
includes dealing with those variations
Adank and McQueen (2007)
Goals of the Study
To determine how variability due to
regional accents affects the processing of words spoken in isolation
To determine if short-term exposure to an unfamiliar accent affects the speed of
processing words spoken in that accent
Adank and McQueen (2007)
Experiment
30 participants, divided into two exposure groups: familiar accent (“Local Dutch”) and unfamiliar accent (Dutch spoken in East
Flanders)
Stimuli for animacy decision tests: 120 Dutch nouns spoken by two females of each accent
Stimuli for exposure phase: 50 declarative sentences from six female speakers of each accent
Adank and McQueen (2007)
Experiment
Test 1: Listeners accomplished an animacy decision task for 30 words spoken from all four speakers
The exposure phase lasted about 23 minutes participants performed a distracter task
Test 2: Listeners repeated the animacy decision task
Adank and McQueen (2007)
Results
Performance was similar for both groups
Performance across tests was alike for both groups
Short-term exposure did not affect the speed of word processing
But: for all participants speed of word comprehension was slower for words spoken in the unfamiliar accent
Clopper and Pisoni (2006)
Goals of the Study
To evaluate the perceptual similarity
structure of regional dialect variation in the USA
To further explore how residential history affects dialect perception
Clopper and Pisoni (2006)
Hypotheses
1. Naïve listeners are predicted to produce a relatively small number of groups of talkers
2. Geographic mobility and location are expected to affect performance
3. Mobile listeners are presumed to have developed more perceptual dialect categories and are
therefore expected to better distinguish different dialects and to create more groups of talkers
Clopper and Pisoni (2006)
Experiment 1
66 talkers from six dialect regions in the US
One (different) sentence per talker containing dialect-specific vowel shifts
22 listeners with different residential histories
Listeners should group talkers in as many
groups with as many members in each group as they wanted; no time limit was presented
Clopper and Pisoni (2006)
Experiment 1
On average:10 groups of talkers, with a range from 3-30 and a median of 7 and 9.36 talkers per group with a range from 1-34 and a
median of 4.
Three main perceptual clusters: New England, South and Midwest/West
Relevant dimensions for perceptual similarity:
linguistic markedness and geography
Clopper and Pisoni (2006)
Experiment 2
48 talkers, even number of males and
females from six dialect regions in the US
One novel sentence per speaker
87 Listeners, split up in 4 groups based on residential history (non-mobile Midland, non- mobile North, mobile Midland, mobile North)
The task was the same as in Experiment 1
Clopper and Pisoni (2006)
Experiment 2
On average:8.48 groups of talkers, with a range from 3-23 and a median of 8 and 7.08 talkers per group with a range from 1-38 and a median of 4.
Significantly more groups for mobile listeners
No significant difference in the ability to correctly group the talkers by dialect
Relevant dimensions for perceptual similarity:
markedness, gender, geography
Adank et al. (2007)
Goals of the Study
Eliciting regional variation patterns in the vowel system of Standard Dutch spoken in the Netherlands and Flanders
Improving the language’s vowel system description by including regional varieties
Providing an overview of the extent of regional variation of the Dutch vowel system
Adank et al. (2007)
Experiment
160 Dutch teachers (professional
language users) from four different regions of each country
Target vowels were produced in carrier sentences
Measurements of duration and formant frequencies of F1 and F2 for the 4800 vowel tokens were analyzed
Adank et al. (2007)
Results
Enough regional information was present in the steady-state formant frequency
measurements of vowels produced by professional users of the standard
language to correctly classify the majority of the speakers into the appropriate
speech community
Discussion and Conclusion
Research on the relationship between regional, social and ethnic language variation is rapidly growing
It shows that listeners are able to make judgments about residential background and social characteristics based on the speech signal