brain research 1470 (2012) 52–58

Available online at www.sciencedirect.com

www.elsevier.com/locate/brainres

Research Report

Hyperarticulation of vowels enhances phonetic change

responses in both native and non-native speakers of
English: Evidence from an auditory event-related
potential study

Maria Uthera,b,n, Anastasia Giannakopouloua, Paul Iversonb

a
Centre for Cognition and Neuroimaging, School of Social Sciences, Brunel University, Uxbridge, Middlesex UB8 3PH, United Kingdom
b
Department of Speech, Hearing and Phonetics Sciences, University College London, United Kingdom

ar t ic l e in f o abs tra ct

Article history: The finding that hyperarticulation of vowel sounds occurs in certain speech registers (e.g.,
Accepted 26 June 2012 infant- and foreigner-directed speech) suggests that hyperarticulation may have a didactic
Available online 4 July 2012 function in facilitating acquisition of new phonetic categories in language learners. This

Keywords: event-related potential study tested whether hyperarticulation of vowels elicits larger

Mismatch negativity phonetic change responses, as indexed by the mismatch negativity (MMN) component of

Cross-language perception the auditory event-related potential (ERP) and tested native and non-native speakers of

Hyperarticulation English. Data from 11 native English-speaking and 10 native Greek-speaking participants

Phonetic categorisation
showed that Greek speakers in general had smaller MMNs compared to English speakers,
confirming previous studies demonstrating sensitivity of the MMN to language back-
ground. In terms of the effect of hyperarticulation, hyperarticulated stimuli elicited larger
MMNs for both language groups, suggesting vowel space expansion does elicit larger pre-
attentive phonetic change responses. Interestingly Greek native speakers showed some
P3a activity that was not present in the English native speakers, raising the possibility that
additional attentional switch mechanisms are activated in non-native speakers compared
to native speakers. These results give general support for models of speech learning such
as Kuhl’s Native Language Magnet enhanced (NLM-e) theory.
Crown Copyright & 2012 Published by Elsevier B.V. All rights reserved.

1. Introduction
There is a growing body of evidence suggesting that language
directed at language learners is acoustically different from
language directed at fluent, native speakers. Data from infant
studies show that infant-directed speech hyperarticulates
(exaggerates) contrasts in F1/F2 space as compared to normal
adult speech (Burnham et al., 2002; Kuhl et al., 1997; Uther
et al., 2007). Kuhl et al. (1997) first demonstrated this hyper-
articulation effect in infant-directed speech by contrasting
acoustic properties of speech directed at infants and adults in
three language populations (American English, Russian and
Swedish). In all three cases, mothers spoke to their infants
with acoustically more extreme vowels compared to when

n
Corresponding Author. Fax: þ44 1895 269724.
E-mail address: maria_uther@hotmail.co.uk (M. Uther).

0006-8993/$ - see front matter Crown Copyright & 2012 Published by Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.brainres.2012.06.041
 brain research 1470 (2012) 52–58
they were speaking to adults, resulting in a ‘‘stretching’’ of
vowel space. Interestingly, it has been subsequently shown
that vowel space expansion is a function of the age of the
child, with the largest hyperarticulation effects with younger
infants (Liu et al., 2009). Although infant-directed speech also
contains a number of other acoustical modifications (heigh-
tened pitch for example), hyperarticulation of vowels appears
to be uniquely didactically-oriented, as suggested by a series
of related studies conducted by Burnham et al. (2002) and
later, Uther et al. (2007). In the study by Burnham et al. (2002),
they investigated the parallels between infant and pet-direc-
ted speech (a speech register that although sounds qualita-
tively similar to infant-directed speech and yet requires no
didactic modification). They replicated the findings of vowel
hyperarticulation in infant-directed speech, but found that
pet-directed speech did not contain vowel hyperarticulation
at all. However, they did find equivalent pitch and affect
changes in pet-directed and infant-directed speech as com-
pared to normal, adult-directed speech.
In a follow-up study, Uther et al. (2007) found that mothers
hyperarticulated their vowels in an equivalent manner when
talking to adult foreigners (L2 speakers of English) as they did
when talking to their own infants. This hyperarticulation of
vowels did not occur when the mothers were speaking to
native adult speakers of English. By contrast to the pet-
directed speech, the foreigner-directed speech was unique
in that it did not contain the heightened pitch or affect
patterns that were evident in the infant-directed speech.
Together these results provide strong evidence for a didactic
role of hyperarticulation which occurs both when speaking to
L1 (infants) and L2 (foreigners) learners of (in that case),
English. These results suggest that the hyperarticulation
found in both infant- and foreigner-directed speech are not
only didactically oriented, but are independent of vocal pitch
and affective valence. Such a notion of a didactic role of
hypearticulation has been incorporated into models of
speech learning such as Kuhl’s Native Language Magnet (e)
theory (Iverson and Kuhl, 1995; Kuhl and Iverson, 1995; Kuhl,
1991, 2004; Kuhl et al., 1992). There are of course other models
that also account for or are compatible with these data (e.g.
Best et al., 2001; Escudero and Boersma, 2004; Flege, 1995,
2003; Strange, 2011; Vallabha and McClelland, 2007), but
arguably one of the most detailed models with respect to
language development, vowel hyperarticulation effects and
the brain is Kuhl’s theory.
Although hyperarticulation therefore appears to have a
didactic function, the question then arises as to whether
hyperarticulation actually assists the learner in distinguish-
ing vowel contrasts. Several studies have attempted to look at
this issue in different forms. On the one hand, there is some
support for this idea: infant studies show a correlation
between infant discrimination and maternal hyperarticula-
tion (Liu et al., 2003) and a later study, Song et al. (2010) also
showed that vowel hyperarticulation significantly improved
infants’ ability to recognise words. Similarly, effects have also
been shown in studies with adults, with Liu et al. (2005)
finding that increased vowel space was significantly corre-
lated with vowel and word intelligibility and Bradlow et al.
(1996) found that a hyperarticulated vowel space was a good
indicator of overall intelligibility. On the other hand, other
53
studies that have attempted to enhance acoustic cues in a
manner not unlike natural hyperarticulated speech have
shown no advantage in being used as training stimuli
(Iverson et al., 2005).
With respect to neural processing, however, there are very
few studies looking at whether phonetic processing in the
brain is enhanced by hyperarticulated stimuli. Matsuda et al.
(2011) showed that only mothers (as opposed to fathers or
even non-parents) with preverbal infants showed enhanced
activity in the language areas of the auditory dorsal pathway
IDS, although Matsuda et al.’s study did not directly investi-
gate the phonetic processing of IDS. Zhang et al.(2009) used
an IDS-motivated training programme and measured MEG
responses to phonetic change. Although the trainees did not
achieve native-like improvements, training did have signifi-
cant identification improvement compared to baseline con-
ditions. Zangl and Mills (2007) also studied infant responses
to familiar and unfamiliar words using IDS and ADS condi-
tions and measured the N200–N400 (word meaning)
responses and the N600–N900 (attentional processing). IDS
was found to increase activity for both 6 month and 13
month-old participants but only the 13 month-old partici-
pants showed increased activity to both familiar and unfa-
miliar words. Again, Zangl and Mills’ study did not look
specifically at phonetic processing. This study therefore
sought to investigate this area further by looking at the effect
of hyperarticulation (a property of IDS) on the brain’s pre-
attentive detection of phonetic change.
To this end, we used the mismatch negativity (MMN)
component of the auditory event-related potential (ERP) as
an index of phonetic processing. The MMN component of the
auditory ERP is elicited when a repetitive sequence of ‘stan-
dard’ stimuli are presented interspersed with occasional
‘deviant’ stimuli which differ from the standard stimuli in
some way. This ‘deviant’ stimuli can differ in simple acous-
tical features (e.g. duration, frequency, intensity) or could also
differ from the standard stimuli in some more abstract sense
(e.g. reversal of an ascending sequence into a descending
sequence; see Garrido et al., 2009; Naatanen, 2011 for a
review). Traditionally, the MMN response has been inter-
preted within a ‘memory trace’ (Naatanen, 1992) or a later
variant, the ‘model adjustment’ hypothesis (Winkler, 2007),
although more recent work would posit an alternative view
suggesting that it represents the activity from an ‘adaptation’
system based on afferent neuronal activity (where MMN is
posited as a refractory effect from the N1 component) being
attenuated (May and Tiitinen, 2010). Generally speaking
however, researchers have lately assumed that the MMN at
least reflects a combination of two kinds of activity (N1
refractoriness and a sensory memory representation) (e.g.
Garrido et al., 2009). Another important observation about the
MMN is that it is elicited in the absence of conscious,
focussed attention to the auditory stimuli. It is therefore
suggested to reflect a preattentive mechanism that may
trigger an attentional switch to particularly novel or surpris-
ing stimuli (Naatanen et al., 2011).
With respect to processing of speech sounds, an interesting
effect observed on the MMN is seen when examining responses
to phonetic changes. The earliest studies (Naatanen et al.,
1997) showed that MMN to phonetic changes was specific to
54 brain research 1470 (2012) 52–58
native-language. In that study, they demonstrated that a lan-
guage-specific Estonian phonetic contrast elicited a significantly
reduced MMN response when presented to Finnish speakers, as
compared to their own Finnish contrast (despite the fact that
the acoustic difference in the non-native contrast was greater).
The fact that the same acoustical difference can be perceived
differently by individuals depending on their language environ-
ment suggests the existence of language-specific representa-
tions. Interestingly, Cheour et al. (1998) further demonstrated
maturational language-specific effects on the MMN. When
tested at six months, the MMN appeared to reflect only the
acoustical difference between the contrasts. On the other hand,
at one year of age, the MMN amplitude was dependant on native
language. For the Finnish speaking infants, the MMN was
considerably attenuated for the non-native (Estonian) vowel
contrast. In a later study with adults, Winkler et al. (1999)
showed that the language-specific representations can also be
affected by training. In that study, it was shown that Hungarian
adult participants who were fluent in Finnish produced larger
MMN responses (virtually equivalent to that shown by native
Finns) compared to naive Hungarian adult participants who
never learnt Finnish.
It is therefore clear that the preattentive neural processing
as indexed by MMN is affected by perceptual processes
involved in learning another language. In this respect, the
MMN component is a useful tool to investigate language-
specific phonetic processing. Our aim was therefore to use
the MMN to measure whether hyperarticulated vs. non-
hyperarticulated vowels elicited an equivalent MMN response
in both native and non-native speakers of a language (in this
case, English). If MMN was enhanced in hyperarticulated
conditions (particularly for non-native speakers), then this
would provide good evidence that hyperarticulation does
indeed facilitate discrimination of unfamiliar phonetic con-
trasts for a language learner.
2. Results
2.1. Mismatch negativity (MMN) data
MMN amplitudes were tested for significance using a t-test
against a value of zero at Fz to determine if a statistically
significant response was elicited. A significant MMN was
elicited in all conditions except for the matched duration
conditions in the Greek speakers (t9 ¼2.107, p¼ 0.64 for non-
hyperarticulated matched duration condition and t9 ¼ 1.796,
p¼ 0.106 for the hyperarticulated matched duration condition).
Natural duration stimuli
Bot
Bought
Fig. 1 – Diagram of different stimuli duration conditions. The matched duration conditions had vowels that were identical
duration. The natural duration condition had stimuli with a longer vowel in ‘bought’ and a shorter vowel, in ‘bot’.
An ANOVA was performed including Fz, FCz and maximal
frontal lateral electrodes (F1 and F2). These analyses showed
that hyperarticulated stimuli elicited larger MMNs (F1,19 ¼5.146,
po0.05) and MMN was more frontally distributed (F1,19 ¼9.750,
po0.05) and the English L1 speakers had larger MMNs com-
pared to the Greek L1 speakers (F1,19 ¼ 9.750, po0.05, see Figs. 2
and 3). There was no significant lateralization of the MMN
response. Analysis of the MMN peak latency showed that
hyperarticulated stimuli elicited earlier MMNs overall com-
pared to the non-hyperarticulated stimuli (F1,19 ¼ 34.365,
po0.01). The matched duration stimuli also elicited later MMNs
compared to the natural duration stimuli (F1,19 ¼ 9.741, po0.01).
Interestingly, the English L1 speakers also showed a later peak
latency compared to the Greek speakers (F1,19 ¼ 5.146, po0.05).
The duration  hyperarticulation interaction was also signifi-
cant (F1,19 ¼12.576, po0.05) with the natural duration condition
showing an earlier peak latency for the hyperarticulated
stimuli compared to non-hyperarticulated stimuli. On the
other hand, the peak latency for matched duration stimuli
did not show much of a difference between hyperarticulated
and non-hyperarticulated stimuli. No other main effects or
interactions were significant.
2.2. P3a data
A t-test was used to determine whether a significant P3a was
elicited. These analyses showed that there was some kind of
activity in the P3a latency range (see Fig. 4) for the L1 Greek
speakers. The amplitude analyses was conducted on the
centro-parietal midline electrode (CPz) as P3a is maximal in
Difference ERP waves (deviant minus standard)
hyperarticulated hyperarticulated
natural duration matched duration
L1 English
L1 Greek
non-hyperarticulated non-hyperarticulated
natural duration matched duration
Fig. 2 – Difference (deviant minus standard) ERP waveforms
at the Fz electrode.
Matched duration stimuli
Bot
Bought
 brain research 1470 (2012) 52–58 55

the centro-parietal range. The L1 English speakers did not
show any significant P3a activity. The L1 Greek speakers only
showed a reliable P3a for the natural duration conditions
(hyperarticulated natural mean¼ 1.22 mV, t9 ¼ 4.644, po0.05;
non-hyperarticulated natural mean¼ 1.357 mV, t9 ¼ 2.563,
po0.05) and the non-hyperarticulated matched condition
(mean¼ 0.608 mV, t9 ¼ 2.563, po0.05). The P3a amplitude in
the hyperarticulated matched condition approached, but did
not reach statistical significance at 0.05 level (t9 ¼ 2.211,
p¼ 0.054). The amplitudes in conditions where there was
statistically significant P3a activity in the Greek speakers
were submitted to a repeated measures ANOVA. This did
not show any significant difference between the 3 conditions.
However, descriptively, one can see that the smallest P3a
activity was within the hyperarticulated matched duration
-3.5
English as L1
-3
Greek as L1
µV)
condition. It was interesting that only the native Greek-
speaking group showed any P3a effects. The fact that the
P3a effects also tended to be larger in the natural duration
condition is in a sense not entirely surprising, as the duration
condition conditions also tended to elicit larger MMN
responses.
3. Discussion
These data show a robust effect of hyperarticulation on MMN
amplitude. Generally hyperarticulation elicits larger MMN
responses regardless of language group. Therefore, hyperar-
ticulation does appear to provide a benefit for the listener,
regardless of their L1 background. There was no significant
hyperarticulation by group interaction which might have
been expected although there are a couple of good possible
explanations. Firstly, it could be argued that the Greek speak-
ers of English were so proficient in English to the extent that

-2.5 they might be more ‘native-like’ in their perception (given

Amplitude (µ

-2 they had an average of 9 years instruction in English).

However, the fact that the native Greek speakers had sig-
-1.5
nificantly smaller MMN responses compared to the native
-1 speakers of English would, to a large extent, mitigate this as a
-0.5 complete explanation. Another possibility could be that the
hyperarticulated stimuli elicited MMN responses that were
0
hyper hyper natural non-hyper non-hyper near ceiling regardless of L1 background. The fact that the
matched duration matched natural Greek speakers elicited very clear P3a responses in the
duration duration duration natural duration conditions would certainly suggest that the
Stimulus condition MMN response was likely at saturation for that group in those
conditions. There was also a P3a effect in hyperarticulated
Fig. 3 – Mean MMN amplitude at Fz across different matched duration condition. To further definitively tease out
stimulus conditions for each English and Greek L1 the effects of these two possible contributing factors, more
speakers. All MMN mean amplitudes in every condition testing (possibly using a more naive L2 group) would be
were statistically significant when tested against zero, useful. It is of course entirely possible that one might not
except for the native Greek-speakers’ non-hyperarticulated necessarily expect that hyperarticulation would only aid
matched duration condition. second language learners, so it is of theoretical interest to

Native English speakers Native Greek speakers

Hyperarticulated Non-hyperarticulated Hyperarticulated Non-hyperarticulated

natural natural
duration duration

matched matched
duration duration

Fig. 4 – Standard and deviant ERP waveforms at selected frontal, fronto-central, central and parietal sites.
56 brain research 1470 (2012) 52–58
note that hyperarticulated stimuli elicit enhanced MMN
responses even in native listeners. In this manner, the
hypothesis that vowel hyperarticulation in F1/F2 space facil-
itates phonetic processing is supported.
In terms of MMN latency, the results generally mirrored the
MMN amplitude results, with the MMN responses to hyper-
articulated stimuli tending to peak earlier compared to
responses to non-hyperarticulated stimuli, which concurs
with the generally larger MMN responses to hyperarticulated
stimuli. Despite there being no difference in size of MMN
response to matched vs. natural duration stimuli, it was
certainly the case that the matched duration stimuli elicited
later MMNs. Interestingly, there were MMN latency effects on
that if anything, might be the opposite to what might have
been expected. Despite larger MMN responses for the English
L1 speakers, their responses peaked later compared to the
native Greek speakers. Looking at the morphology of the
waveforms, it looks as though the more ‘peaked’ nature of
the MMN in the native English speakers may be the reason
for this delay.
Despite the absence of a group by hyperarticulation effect,
this does not necessarily rule out that hyperarticulation
would be useful for the non-native speaker in a training
context. The only way to really test this idea would be to
examine MMN responses to hyperarticulated stimuli as a
function of directly training with hyperarticulated stimuli.
Presumably with a period of training, one would expect that
non-native speakers might start to process the hyperarticu-
lated stimuli in a speeded manner more like the native
speakers. Nonetheless, the overall effects of hyperarticula-
tion on the MMN response regardless of language group are
supportive of Kuhl’s (2004) view of hyperarticulation as a
didactic tool to aid the speech learner.
Stimulus duration had surprising little effect on the MMN
response. None of the duration main effects or interactions
were significant. In a sense this is somewhat surprising given
that we have previously shown that duration is more readily
used as an acoustic cue in both Greek (Giannakopoulou et al.,
2011) and Finnish speakers of English (Ylinen et al., 2010).
Looking at the data more closely, it looks as though there was
a tendency for the matched duration stimuli elicit smaller
MMNs compared to the natural duration stimuli, but this was
not statistically significant at the 0.05 level. This may well be
due to the English language proficiency of our participants. In
this study, the Greek speakers were already in an immersive
English environment at university, whereas in our previous
study of Greek speakers, they were all tested in their home
countries. Unfortunately, we do not have corresponding
behavioural data on these participants to be able to compare
the two Greek native speaker groups. Future work would
therefore be focussed to test lower proficiency Greek speakers
of English and also collect corresponding behavioural data.
Future work might also test the correlation between MMN
and how ‘good’ an exemplar it is to each individual, given
their language proficiency.
It was also interesting to observe that P3a (or ‘novelty P3’)
responses were elicited in the native Greek speakers of
English, which tended to be more prominent in the natural
duration condition. In terms of theoretical significance, P3a
activity is thought to index involuntary switches as a result of
surprising or particularly novel changes in the auditory
environment (see Polich, 2007 for a review), and is distinct
in topography and morphology from other P3 waves. It was
interesting that P3a effects were observed in the Greek
speakers but not the English speakers. It is possible that for
the Greek speakers, the most prominent changes resulted in
an attentional switch and engagement of attentional pro-
cesses because that they were listening to English-language
contrasts. To the native English speakers, it is possible that
the vowel changes do not actually sound so unusual or
‘surprising’ to a native speaker. Of course, speculation about
the attentional processing would necessitate more experi-
ments to definitively test this. One useful way of investigating
this further might be, for example, to compare active attend
conditions as well as passive listening conditions.
In summary, these results suggest that hyperarticulated
stimuli do generate more efficient phonetic processing in
both native and non-native speakers, even in the presence of
significant native language main effects. In this way, the
hypothesis that hyperarticulated stimuli does aid phonetic
processing is supported. Further studies should focus on the
potential benefit of directly training the language learner
with hyperarticulated stimuli as well as and also testing its
possible effect on different language learner proficiency
levels. Certainly, these data would suggest that the time is
ripe to further study the effects of hyperarticulation on the
brain’s phonetic processing and its ability to change with
training.
4. Experimental procedures
4.1. Participants
Participants with their L1 as either English or Greek were
recruited from Brunel University, UK. The native English-
speaking participants (6 male and 5 female) had a mean age
of 22 (range 19–30), were born in Britain and were studying at
Brunel University. The native Greek-speaking participants
(6 male and 4 female) had a mean age of 26 (range 22–30)
and were born and spent most of their lives in Greece, but
were studying at Brunel University on exchange or tempora-
rily. The Greek participants had all studied English as L2 in
school (normally 9 years duration). The participants were
either given course credit for their participation or offered
small incentives (chocolates and a chance to enter a prize
draw for an mp3 player).
4.2. Stimuli and procedure
The hyperarticulated stimuli were synthesised based on best
exemplars in Iverson and Evans (2007) (bot: F1¼ 535 Hz,
F2 ¼762 Hz, vowel duration¼ 70 ms (entire word measuring
320 ms); bought: F1 ¼392 Hz, F2 ¼ 636 Hz, vowel duration¼ 216
ms (entire word measuring 460 ms)). The non-hyperarticu-
lated versions were 29% closer to the centre of the vowel
space, simulating a 50% reduction of the total area of the
vowel space (bot: F1¼ 525 Hz, F2 ¼976 Hz; bought: F1¼ 423 Hz,
F2 ¼887 Hz). We included matched-duration versions of the
 brain research 1470 (2012) 52–58
stimuli, in which vowel durations were equated to 143 ms
(entire word measuring 393 ms), see Fig. 1.
The ‘bot’ and ‘bought’ stimuli were presented in a passive
oddball paradigm that contained a repetitive standard stimu-
lus and occasional deviant stimuli. Stimuli were presented
in blocks of 400 stimuli, each including 60 deviant stimuli
(15% probability). There were 4 different conditions: hyperarti-
culated matched duration, non-hyperarticulated matched
duration, hyperarticulated natural duration and non-hyperar-
ticulated matched duration. In the matched duration condi-
tion, the standard and deviant stimuli were identical. Each
condition was presented twice in a block of 400, so that there
was a total of 800 stimuli and 120 deviants and the block order
was randomized. The ISI (from offset to onset) in these blocks
was 500 ms.
4.3. EEG recording and analysis
For the EEG recording, the participants sat in an electrically
shielded chamber in a quiet room and watched a muted
video documentary with subtitles in their native language.
The auditory stimuli were presented via headphones at the
intensity of 65 dB SPL. The participants were instructed to
ignore stimuli and to concentrate on watching the documen-
tary and were informed that they were going to be given a
questionnaire on the documentary afterwards.
The EEG was recorded with a NeuroScan system and
Synamps amplifier using a 64-channel electrode cap with
Ag/AgCl electrodes and a 250-Hz sampling rate. Eye move-
ments were monitored with a bipolar EOG recorded from the
canthi of both eyes and below and above the left eye. After
the recording, the data were filtered offline with a 1.0- to 30-
Hz band-pass filter (24 dB/octave). Artefacts exceeding
7100 mV at any channel were rejected, and the EEG was
averaged into event-related potentials (ERPs) using epochs
from 100 to 500 ms from stimulus onset. ERPs were aver-
aged separately for each stimulus type (standards and devi-
ants), and the baseline was set to -100ms to 0ms from the
stimulus onset.
The MMN component was measured from the difference
waveforms (i.e. subtracting the standard stimulus response
from the deviant response) and the P3a was also measured
from the difference ERP waveforms. Each ERP component was
measured using a mean amplitude measurement centred on a
40 ms time window that was chosen from the grand average
maximum peak in the expected latency region (around 200 ms
for MMN and 300 ms for P3a). Mean amplitude and latency
analysis comparisons were made using a multivariate ANOVA
with Bonferroni correction.
r e f e r e n c e s
Best, C., McRoberts, G., Goodell, E., 2001. American listeners’
perception of nonnative consonant contrasts varying in per-
ceptual assimilation to English phonology. J. Acoust. Soc. Am.
1097, 775–794.
Bradlow, A.R., Torretta, G.M., Pisoni, D.B., 1996. Intelligibility of
normal speech I: global and fine-grained acoustic-phonetic
talker characteristics. Speech Commun. 20 (3–4), 255–272.
57
Burnham, D., Kitamura, C., Vollmer-Conna, U., 2002. What’s new,
pussycat? On talking to babies and animals. Science 296
(5572), 1435.
Cheour, M., Ceponiene, R., Lehtokoski, A., Luuk, A., Allik, J., Alho,
K. Development of language-specific phoneme representa-
tions in the infant brain. Nat. Neurosci. 1 (5), 351–353.
Escudero, P., Boersma, P., 2004. Bridging the gap between L2
speech perception research and phonological theory. Studies
in Second Language Acquisition 26 (04), 551–585 (Cambridge
University Press).
Flege, J.E., 1995. Second language speech learning: theory, finding
and problems. In: Strange, W. (Ed.), Speech Perception and
Linguistic Experience Theoretical and Methodological Issues.
York Press, pp. 233–277.
Flege, J.E., 2003. Assessing constraints on second-language seg-
mental production and perception. In: Meyer, A., Schiller, N.
(Eds.), Phonetics and Phonology in Language Comprehension
and Production: Differences and Similiarities. Mouton de
Gruyter, Berlin, pp. 319–355.
Garrido, M.I., Kilner, J.M., Stephan, K.E., Friston, K.J., 2009. The
mismatch negativity: a review of underlying mechanisms.
Clin. Neurophysiol. 120 (3), 453–463.
Giannakopoulou, A., Uther, M., Ylinen, S., 2011. Phonetic cue-
weighting in the acquisition of a second language (L2):
evidence from Greek speakers of English. In: Wrembel, M.,
Kul, M., Dziubalska-Kolaczyk, K. (Eds.), Achievements and
Perspectives in SLA of Speech: New Sounds 2010, 1. Peter
Lang Verlag, Frankfurt am Main, pp. 91–102.
Iverson, P., Evans, B.G., 2007. Learning English vowels with
different first-language vowel systems: perception of formant
targets, formant movement, and duration. J. Acoust. Soc. Am.
122 (5), 2842–2854.
Iverson, P., Kuhl, P.K., 1995. Mapping the perceptual magnet effect
for speech using signal detection theory and multidimen-
sional scaling. J. Acoust. Soc. Am. 97 (1), 553–562.
Iverson, P., Hazan, V., Bannister, K., 2005. Phonetic training with
acoustic cue manipulations: a comparison of methods for
teaching English vertical bar r vertical bar-vertical bar l
vertical bar to Japanese adults. J. Acoust. Soc. Am. 118 (5),
3267–3278.
Kuhl, P., Iverson, P., 1995. Linguistic experience and the ‘‘percep-
tual magnet effect’’. In: Strange, Winifred (Ed.), Speech Percep-
tion and Linguistic Experience: Issues in Cross-Language
Research. York Press, Baltimore, pp. 121–154.
Kuhl, P.K., Andruski, J.E., Chistovich, I.A., Chistovich, L.A.,
Kozhevnikova, E.V., Ryskina, V.L. Cross-language analysis of
phonetic units in language addressed to infants. Science 277
(5326), 684–686.
Kuhl, P.K., 1991. Human adults and human infants show a
‘‘perceptual magnet effect’’ for the prototypes of speech
categories, monkeys do not. Percept. Psychophys. 50 (2),
93–107.
Kuhl, P.K., 2004. Early language acquisition: cracking the speech
code. Nat. Neurosci. Rev. 5 (11), 831–843.
Kuhl, P.K., Williams, K.A., Lacerda, F., Stevens, K.N., Lindblom, B.,
1992. Linguistic experience alters phonetic perception in
infants by 6 months of age. Science 255 (5044), 606–608.
Liu, H.-M., Kuhl, P.K., Tsao, F.-M., 2003. An association between
mothers’ speech clarity and infants’ speech discrimination
skills. Dev. Sci. 6 (3), F1–F10.
Liu, H.-M., Tsao, F.-M., Kuhl, P.K., 2005. The effect of reduced
vowel working space on speech intelligibility in Mandarin-
speaking young adults with cerebral palsy. J. Acoust. Soc. Am.
117 (6), 3879–3889.
Liu, H.-M., Tsao, F.-M., Kuhl, P.K., 2009. Age-related changes in
acoustic modifications of Mandarin maternal speech to pre-
verbal infants and five-year-old children: a longitudinal study.
J. Child Lang. 36 (4), 909–922.
58 brain research 1470 (2012) 52–58
Matsuda, Y.-T., Ueno, K., Waggoner, R.A., Erickson, D., Shimura, Y.,
Tanaka, K. Processing of infant-directed speech by adults.
NeuroImage 54 (1), 611–621.
May, P.J.C., Tiitinen, H., 2010. Mismatch negativity (MMN), the
deviance-elicited auditory deflection, explained. Psychophy-
siology 47 (1), 66–122.
Naatanen, R., Lehtokoski, A., Lennes, M., Cheour, M., Huotilainen,
M., Iivonen, A. Language-specific phoneme representations
revealed by electric and magnetic brain responses. Nature 385
(6615), 432–434.
Näätänen, R., 1992. In: Attention and Brain Function. Erlbaum, NJ.
Näätänen, R., Kujala, T., Winkler, I., 2011. Auditory processing
that leads to conscious perception: a unique window to
central auditory processing opened by the mismatch negativity
and related responses. Psychophysiology 48 (1), 4–22.
Polich, J., 2007. Updating P300: an integrative theory of P3a and
P3b. Clin. Neurophysiol. 118 (10), 2128–2148.
Song, J.Y., Demuth, K., Morgan, J., 2010. Effects of the acoustic
properties of infant-directed speech on infant word recogni-
tion. J. Acoust. Soc. Am. 128 (1), 389–400.
Strange, W., 2011. Automatic selective perception (ASP) of first
and second language speech: a working model. J.Phonetics 39
(4), 456–466.
Uther, M., Knoll, M., Burnham, D., 2007. Do you speak E-NG-L-I-
SH? A comparison of foreigner- and infant-directed speech.
Speech Commun. 49 (1), 2–7 (Elsevier Science Publishers B.V.).
Vallabha, G.K., McClelland, J.L., 2007. Success and failure of new
speech category learning in adulthood: consequences of
learned Hebbian attractors in topographic maps. Cogn., Affec-
tive, Behav. Neurosci. 7 (1), 53–73.
Winkler, I., Kujala, T., Tiitinen, H., Sivonen, P., Alku, P., Lehtokoski,
A. Brain responses reveal the learning of foreign language
phonemes. Psychophysiology 36 (5), 638–642.
Winkler, I., 2007. Interpreting the mismatch negativity. J. Psycho-
physiol. 21 (3), 147–163.
Ylinen, S., Uther, M., Latvala, A., Vepsäläinen, S., Iverson, Paul,
Akahane-Yamada, Reiko Training the brain to weight speech
cues differently: a study of Finnish second-language users of
English. J. Cogn. Neurosci. 22 (6), 1319–1332.
Zangl, R., Mills, D.L., 2007. Increased brain activity to infant-
directed speech in 6- and 13-month-old infants. Infancy 11 (1),
31–62.
Zhang, Y., Kuhl, Patricia K., Imada, T., Iverson, Paul, Pruitt, J. Neural
signatures of phonetic learning in adulthood: a magnetoence-
phalography study. NeuroImage 46 (1), 226–240.