1. Louise Stringer and Paul Iverson
UCL, Speech, Hearing and Phonetic Sciences, London, UK
Accent intelligibility across native and non-native accent pairings:
investigating links with electrophysiological measures of word recognition
Funded by the EU FP7 Marie
Curie Initial Training Network
INSPIRE For further information, contact l.stringer.11@ucl.ac.uk
Results
Methods
Accent intelligibility task
• Stimuli: 412 neutral sentences in three accents
• Noise: In quiet or mixed with speech-shaped noise at +3dB, 0dB or -3dB SNR
• Listeners repeated the words they understood; the proportion of words identified
correctly was averaged over each noise level for each accent
EEG task
• Stimuli: 206 predictable and 206 anomalous sentences in three accents
• Extracted 800ms epochs time-locked to onset of final word and calculated average
amplitude in each accent/sentence type combination across the 200-350ms window
(for PMN) and 350-500ms window (for N400 effect)
• PMN and N400 calculated for each accent by subtracting responses to predictable
sentences from responses to anomalous sentences and averaging over midline
electrodes
References
6−
4−
2−
0
2
200− 0 200 400 600 800
Time(ms)
Amplitude(µV)
Accent
SE
GE
Sp
SpanishListeners
6−
4−
2−
0
2
200− 0 200 400 600 800
Time (ms)
Amplitude(µV)
Accent
SE
GE
Sp
Spanish Listeners
6−
4−
2−
0
2
200− 0 200 400 600 800
Time (ms)
Amplitude(µV)
Accent
SE
GE
Sp
English Listeners
6−
4−
2−
0
2
200− 0 200 400 600 800
Time(ms)
Amplitude(µV)
Accent
SE
GE
Sp
SpanishListeners
10−
5−
0
5
English Spanish
Listeners
Amplitude(µV)
Accent
SE
GE
Sp
N400
10−
5−
0
5
English Spanish
Listeners
Amplitude(µV)
Accent
SE
GE
Sp
PMN
**
*
*
*
Adank, P., Evans, B. G., Stuart-Smith, J., & Scott, S. K. (2009). Comprehension of familiar and unfamiliar native accents under adverse listening conditions. Journal of experimental psychology. Human perception and performance, 35(2), 520–9.
doi:10.1037/a0013552
Bent, T., & Bradlow, A. R. (2003). The interlanguage speech intelligibility benefit. The Journal of the Acoustical Society of America, 114(3), 1600. doi:10.1121/1.1603234
Brunellière, A., & Soto-Faraco, S. (2013). The speakers’ accent shapes the listeners' phonological predictions during speech perception. Brain and language, 125(1), 82–93. doi:10.1016/j.bandl.2013.01.007
Goslin, J., Duffy, H., & Floccia, C. (2012). An ERP investigation of regional and foreign accent processing. Brain and language, 122(2), 92–102. doi:10.1016/j.bandl.2012.04.017
Pinet, M., Iverson, P., & Huckvale, M. (2011). Second-language experience and speech-in-noise recognition: effects of talker-listener accent similarity. The Journal of the Acoustical Society of America, 130(3), 1653–62. doi:10.1121/1.3613698
0.00
0.25
0.50
0.75
1.00
English Spanish
Listeners
Proportioncorrect
Accent
SE
GE
Sp
***
***
***
*** ***
For all figures: * = p<0.05, ** = p<0.01, *** = p<0.001
Background
Accented speech can be hard to understand
•Regional and non-native accents can be much less intelligible in noise than a standard
accent (Pinet et al., 2011).
•The intelligibility of an accent depends on the combination of the talker and listener’s
language backgrounds (Adank et al., 2009; Bent & Bradlow, 2003).
•Unsure if accents cause processing difficulties in general, or only in noisy conditions
because of ceiling effects in behavioural tasks in quiet.
Accented speech and word recognition processes
•Some EEG studies suggest the PMN and N400 effects, which are related to
phonological and lexical integration, may be influenced by regional and non-native
accents, even in quiet conditions (Brunellière & Soto-Faraco, 2013; Goslin et al., 2012).
•So far, this has only been investigated for native listeners
Aims
1. Compare the influence of accent on speech-in-noise intelligibility and PMN and
N400 effects to investigate whether EEG methods reveal accent-related processing
difficulties in quiet conditions
2. Extend research into the influence of accent on PMN and N400 effects to non-
native listeners
Standard
SSBE
Regional
Glaswegian
Non-native
Spanish
Talker Accents
Native
English
Non-native
Spanish
Listener groups
Significant Effects
Linear mixed effect models with accent and listener
groups as fixed effects and by-participant random
slopes found significant effects of:
•Accent intelligibility: accent, listener group and
accent*group interaction (all p<0.001)
•PMN: accent (p=0.019), listener group (p=0.015)
•N400: accent (p=0.014), listener group (p=0.005)
For the N400 effect, word recognition scores in noise
are a better predictor of N400 amplitude than scores
in quiet
Summary
•Regional and non-native accents show some
influence on PMN and N400, even in quiet
•More intelligible accents in noise seem to show larger
PMN and N400 responses
•Results suggest that at least some of the
processing difficulties caused by accented speech
in noise are also present in quiet conditions
•Accents may influence listeners’ expectations about
upcoming words
Future work
•Also look at measures of acoustic-phonetic similarity
between talker and listener accents
•Test non-native listeners that are more familiar with
native accents
Fig. 1: Psychometric functions of the proportion of words correctly identified Fig. 2: Word recognition accuracy in noise
Fig. 3: Grand mean responses across the epoch, averaged over midline electrodes
Fig. 4: ERPs for the three accents, averaged over midline electrodes
English Listeners
Signal to noise r− − atio (dB)
Proportioncorrect
8− 6− 4− 2− 0 2 4
00.51
SE
GE
Sp
Spanish Listeners
Signal to noise r− − atio (dB)
Proportioncorrect
8− 6− 4− 2− 0 2 4
00.51
6−
4−
2−
0
2
200− 0 200 400 600 800
Time (ms)
Amplitude(µV)
Accent
SE
GE
Sp
Spanish Listeners
6−
4−
2−
0
2
200− 0 200 400 600 800
Time (ms)
Amplitude(µV)
Accent
SE
GE
Sp
English Listeners
6−
4−
2−
0
2
200− 0 200 400 600 800
Time (ms)
Amplitude(µV)
Accent
SE
GE
Sp
English Listeners
6−
4−
2−
0
2
200− 0 200 400 600 800
Time (ms)
Amplitude(µV)
Accent
SE
GE
Sp
English Listeners