Fast multi-feature paradigm for recording several mismatch negativities (MMNs) to phonetic and acoustic changes in speech sounds.
ABSTRACT In this study, we addressed whether a new fast multi-feature mismatch negativity (MMN) paradigm can be used for determining the central auditory discrimination accuracy for several acoustic and phonetic changes in speech sounds. We recorded the MMNs in the multi-feature paradigm to changes in syllable intensity, frequency, and vowel length, as well as for consonant and vowel change, and compared these MMNs to those obtained with the traditional oddball paradigm. In addition, we examined the reliability of the multi-feature paradigm by repeating the recordings with the same subjects 1-7 days after the first recordings. The MMNs recorded with the multi-feature paradigm were similar to those obtained with the oddball paradigm. Furthermore, only minor differences were observed in the MMN amplitudes across the two recording sessions. Thus, this new multi-feature paradigm with speech stimuli provides similar results as the oddball paradigm, and the MMNs recorded with the new paradigm were reproducible.
[show abstract] [hide abstract]
ABSTRACT: For the perception of timbre of a musical instrument, the attack time is known to hold crucial information. The first 50 to 150 ms of sound onset reflect the excitation mechanism, which generates the sound. Since auditory processing and music perception in particular are known to be hampered in cochlear implant (CI) users, we conducted an electroencephalography (EEG) study with an oddball paradigm to evaluate the processing of small differences in musical sound onset. The first 60 ms of a cornet sound were manipulated in order to examine whether these differences are detected by CI users and normal-hearing controls (NH controls), as revealed by auditory evoked potentials (AEPs). Our analysis focused on the N1 as an exogenous component known to reflect physical stimuli properties as well as on the P2 and the Mismatch Negativity (MMN). Our results revealed different N1 latencies as well as P2 amplitudes and latencies for the onset manipulations in both groups. An MMN could be elicited only in the NH control group. Together with additional findings that suggest an impact of musical training on CI users' AEPs, our findings support the view that impaired timbre perception in CI users is at partly due to altered sound onset feature detection.PLoS ONE 01/2012; 7(9):e45375. · 4.09 Impact Factor
Article: The use of acoustic cues for phonetic identification: effects of spectral degradation and electric hearing.[show abstract] [hide abstract]
ABSTRACT: Although some cochlear implant (CI) listeners can show good word recognition accuracy, it is not clear how they perceive and use the various acoustic cues that contribute to phonetic perceptions. In this study, the use of acoustic cues was assessed for normal-hearing (NH) listeners in optimal and spectrally degraded conditions, and also for CI listeners. Two experiments tested the tense/lax vowel contrast (varying in formant structure, vowel-inherent spectral change, and vowel duration) and the word-final fricative voicing contrast (varying in F1 transition, vowel duration, consonant duration, and consonant voicing). Identification results were modeled using mixed-effects logistic regression. These experiments suggested that under spectrally-degraded conditions, NH listeners decrease their use of formant cues and increase their use of durational cues. Compared to NH listeners, CI listeners showed decreased use of spectral cues like formant structure and formant change and consonant voicing, and showed greater use of durational cues (especially for the fricative contrast). The results suggest that although NH and CI listeners may show similar accuracy on basic tests of word, phoneme or feature recognition, they may be using different perceptual strategies in the process.The Journal of the Acoustical Society of America 02/2012; 131(2):1465-79. · 1.55 Impact Factor
Fast multi-feature paradigm for recording several mismatch negativities
(MMNs) to phonetic and acoustic changes in speech sounds
Satu Pakarinena,b,*, Riikka Lovioa,b, Minna Huotilainena,b,d, Paavo Alkue,
Risto Na ¨a ¨ta ¨nena,b,f,g, Teija Kujalaa,b,c
aCognitive Brain Research Unit, University of Helsinki, Helsinki, Finland
bHelsinki Brain Research Centre, Helsinki, Finland
cDepartment of Psychology, University of Turku, Finland
dHelsinki Collegium for Advanced Studies, University of Helsinki, Helsinki, Finland
eHelsinki University of Technology, Laboratory of Acoustics and Audio Signal Processing, Espoo, Finland
fDepartment of Psychology, University of Tartu, Tartu, Estonia
gCenter of Integrative Neuroscience (CFIN), University of Arhus, Arhus, Denmark
The mismatch negativity (MMN) component of the event-
related potential (ERP) is elicited by any discriminable change
(‘deviant’) in some regular aspect (‘standard’) of the auditory
environment. According to the traditional theory, the current
auditory input is compared with the neural representation of the
standard and if a difference is found, the MMN response is
generated (Na ¨a ¨ta ¨nen, 1990, 1992). More recent views stress the
role of this process in extracting regularities of the current and
building predictions about the forthcoming auditory input
(Na ¨a ¨ta ¨nen and Winkler, 1999; Winkler, 2007). Both the magnitude
and the timing of the MMN depend on the degree of sound change:
the larger the deviation, the earlier the latency and higher the
amplitude (Tiitinen et al., 1994; Pakarinen et al., 2007). This
suggests that the MMN indexes not only sound change per se, but
also its degree in a fine-grained manner. Furthermore, the MMN
parameters correlate well with the behavioural discrimination
ability of the subject (Lang et al., 1990; Amenedo and Escera, 2000;
Pakarinen et al., 2007). This, together with the observation that a
sound change is usually not detected in the absence of MMN
(Winkler et al., 1999) indicates that the information carried by the
MMN plays an essential role in preattentive processing and also in
conscious detection of sound changes in our environment.
The principal neural generators of the MMN, which are
assumed to be responsible for the actual comparison process
between the input and the memory trace, lie bilaterally in the
auditory cortices (Hari et al., 1984; Alho et al., 1993; Cse ´pe, 1995;
Shalgi and Deouell, 2007) and reach their peak of activityat around
100–250 ms after the deviation (Rinne et al., 2000). Another
generator process, originating in the frontal areas peaks slightly
later (Giard et al., 1990; Rinne et al., 2000) and is assumed to
initiate a chain of reactions enabling an involuntary attention
switch to the deviation initially detected by the auditory cortex
(Na ¨a ¨ta ¨nen, 1990, 1992). For changes in phonemes of ones own
native (Na ¨a ¨ta ¨nen et al., 1997) or later-learned foreign language
(Winkler et al., 1999; Shestakova et al., 2003), also a third
subcomponent, overlapping with the two main ones can be
Biological Psychology 82 (2009) 219–226
A R T I C L EI N F O
Received 27 June 2008
Accepted 21 July 2009
Available online 29 July 2009
Mismatch negativity (MMN)
Event-related potential (ERP)
Central auditory processing
A B S T R A C T
In this study, we addressed whether a new fast multi-feature mismatch negativity (MMN) paradigm can
be used for determining the central auditory discrimination accuracy for several acoustic and phonetic
changes in speech sounds. We recorded the MMNs in the multi-feature paradigm to changes in syllable
intensity, frequency, and vowel length, as well as for consonant and vowel change, and compared these
MMNs to those obtained with the traditional oddball paradigm. In addition, we examined the reliability
of the multi-feature paradigm by repeating the recordings with the same subjects 1–7 days after the first
recordings. The MMNs recorded with the multi-feature paradigm were similar to those obtained with
the oddball paradigm. Furthermore, only minor differences were observed in the MMN amplitudes
across the two recording sessions. Thus, this new multi-feature paradigm with speech stimuli provides
similar results as the oddball paradigm, and the MMNs recorded with the new paradigm were
? 2009 Elsevier B.V. All rights reserved.
* Corresponding author at: Cognitive Brain Research Unit, Department of
Psychology, P.O. Box 9, University of Helsinki, FIN-00014, Finland.
Tel.: +358 9 191 29462; fax: +358 9 191 29450.
E-mail address: firstname.lastname@example.org (S. Pakarinen).
Contents lists available at ScienceDirect
journal homepage: www.elsevier.com/locate/biopsycho
0301-0511/$ – see front matter ? 2009 Elsevier B.V. All rights reserved.
observed. It is usually lateralized to the language-dominant
hemisphere (typically left-lateralized in right-handed individuals)
and it appears to represent the activation of long-term memory
traces for these speech sounds (Na ¨a ¨ta ¨nen et al., 1997; Winkler
et al., 1999; Shestakova et al., 2002, 2003). Thus, the MMN offers a
means to probe the neural representations of sounds and the
cortical accuracy of sound discrimination.
Recently, the MMN has been increasingly used also in clinical
research, for instance,inthe field of developmentaldyslexia.So far,
these studies have shown that dyslexic adults have problems both
in processing phonetic changes (Schulte-Ko ¨rne et al., 1998;
Lachmann et al., 2005) and certain simple acoustic changes, such
as deviations in frequency (Baldeweg et al., 1999; Renvall and Hari,
2003). The MMN also reflects plastic changes in auditory
discrimination after intervention in dyslexia (Kujala et al.,
2001a). Moreover, different subtypes of dyslexia may show
different patterns of deficient auditory processing as compared
with each other, as well as with controls (Lachmann et al., 2005;
Shankarnarayan and Maruthy, 2007).
Since the MMN can be recorded also from infants and children
before birth (Huotilainen et al., 2005; Draganova et al., 2005, 2007)
it allows one to determine which aspects of the auditory
information are deficiently processed at a very early stage. One
could, by recording MMNs at different developmental stages,
define when these problems of central sound processing arise or
even predict the future development. For instance, in the JLD
project (Jyva ¨skyla ¨ longitudinal study on dyslexia, Jyva ¨skyla ¨,
Finland) children with genetic risk for dyslexia have been followed
from the birth to the school age, and it appears that even the very
early recordings of these children predict their future linguistic
abilities. For instance, larger responses in infancy to vowel-
duration deviant (std /ka:/, dev /ka/) in the left hemisphere
predicted better receptive language skills at 2.5 years and verbal
memory at 3.5 years, whereas similar pattern in the right
hemisphere was associated with poorer verbal memory at 5 years
as compared with controls (Lyytinen et al., 2004). An early
identification of the markers of the risk for dyslexia, as well as
dyslexia subtypes would be especially beneficial because of
elevated plasticity of the neural networks at young age. Effective
training programs (cf. Kujala et al., 2001a) could be assigned even
before learning to read to avoid the cumulating effects of the
disability to academic and social performance.
The so-called oddball paradigm has traditionally been used for
recording the MMN. With this approach the recording sessions
tend to be long and provide usually information on cortical
discrimination of 1–2 sound features. Especially in clinical studies
and in studies with children and infants, short recording times are
of major importance. Children have a limited patience to sit still
and in long recording sessions, the signal-to-noise ratio tends to
become poor. Recently, new multi-feature MMN paradigms have
been developed to allow a very fast assessment of extensive
auditory discrimination profiles, covering the central auditory
discrimination for several auditory attributes (e.g., frequency,
intensity, location and duration; Na ¨a ¨ta ¨nen et al., 2004), even at
different difficulty levels (from barely noticeable to easily
detectable; Pakarinen et al., 2007). These profiles can be useful
in evaluating specific impairments of central auditory processing,
as well as the development and plasticity of the system (for
reviews, see Na ¨a ¨ta ¨nen, 2003; Na ¨a ¨ta ¨nen and Escera, 2000;
Na ¨a ¨ta ¨nen et al., 2007).
So far these new fast paradigms have used harmonic sounds.
Detecting minor but relevant changes in speech sounds, however,
is a far more demanding task for the central auditory system than
detecting changes in harmonic sounds. There is also evidence that
the processing of phonological information differs from that of
non-linguistic auditory information already at the preattentive
level (Na ¨a ¨ta ¨nen et al., 1997; Shtyrov et al., 1998). Further, some
clinical studies suggest that central auditory processing may be
differentially affected in speech and non-speech contexts, for
instance, in dyslexia (Schulte-Ko ¨rne et al., 1998) and aphasia
(Cse ´pe et al., 2001; Ilvonen et al., 2004). Thus, recording MMNs
from subjects with difficulties in phonological processing may
reveal different profiles of central auditory processing for speech
and non-speech sounds.
The aim of this study was to develop a fast multi-feature
paradigm for the evaluation of central auditory processing of
speech sounds. We recorded MMNs to five deviations in Finnish-
language semi-synthetic CV syllables in 20 min. The deviations
were changes in syllable fundamental frequency and intensity, a
change in vowel-duration, and consonant and vowel changes, all of
which are generally relevant features in speech sounds (vowel-
duration being especially relevant in quantity languages such as
Finnish). The replicability of the responses recorded in the new
paradigm was estimated by repeating the recordings with the
same subjects and comparing the responses over the two sessions.
For the sake of comparison, we also recorded the MMNs to the
same deviations in the oddball paradigm traditionally used in
Fifteen healthy subjects (mean age 25 years, range 20–40 years, 10 males)
participated in the study. They gave a written informed consent after the nature of
the studywas explained tothem. Thestudy was approved by theEthical Committee
of the Department of Psychology, University of Helsinki.
2.2. Stimuli and procedure
The stimuli were semi-synthetic Finnish-language consonant-vowel (CV)
syllables. The standard stimuli were /te:/ and /pi:/. The fundamental frequency
(F0)was101 Hz,andthesyllableduration170 ms.Thedeviantstimulidifferedfrom
the standards either in syllable frequency (F0 ?8%; 93/109 Hz), syllable intensity
(?6 dB), vowel-duration (?70 ms; 100 ms /te/ and /pi/), consonant (/pe:/ and /ti:/) or
vowel (/ti:/ and /pe:/; abbreviated as FRE/te:/, FRE/pi:/,INT/te:/, INT/pi:/, VOWDUR/te/,
VOWDUR/pi/, CONS/pe:/, CONS/ti:/, VOWEL/ti:/, and VOWEL/pe:/, respectively).
In the duration control condition, the durations of the stimuli were reversed so
that all other stimuli were 100 ms in duration, except for the vowel-duration
deviant which was 170 ms. Thus, the standard stimuli were 100 ms long syllables /
te/ and /pi/, and the corresponding frequency, intensity, consonant and vowel
deviants were also 100 ms (abbreviated as FRE/te/, FRE/pi/, INT/te/, INT/pi/, CONS/
pe/, CONS/ti/, VOWEL/ti/, and VOWEL/pe/, respectively), whereas the vowel-
duration deviant was 170 ms in duration (abbreviated as VOWDUR/te:/, and
The syllables were created using the Semisynthetic Speech Generation method
(SSG) (Alku et al., 1999). A male Finnish speaker pronounced long isolated vowels
(/e:/ and /i:/) and short words (/pito/ and /pe:ti/) in an anechoic chamber. From the
isolated vowels, the SSG method first estimated the glottal excitation waveform
from the vowel /e:/, and then vocal-tract models of both vowels /e:/ and /i:/. The
fundamental frequency (F0) of the estimated waveform was 101 Hz and the lowest
four formant frequencies of the vocal-tract models were 410, 2045, 2260, and
3320 Hzfor/e:/and320,2240,2690,and3275 Hzfor/i:/.Theestimatedglottalflow
was then modified both in duration and F0 in order to obtain semi-synthetic vowels
with two durations (/e:/ and /e/ as well as /i:/ and /i/) and three F0 values (93, 101,
and 109 Hz). The plosive /t/ and /p/ waveforms were added in the beginning of the
each of the produced semi-synthetic vowel sounds, to obtain words /te/, /te:/, /ti/, /
ti:/, /pe/, /pe:/, /pi/, and /pi:/ with different F0 values. The intensities of these words
were normalised by adjusting the square sums of the digital sound waveforms to
the samevalue. However, twoadditional intensityvariants, a 6 dB louderand a 6 dB
softer ones, were generated for the /te/, /te:/, /pi/, and /pi:/ words.
2.2.1. Multi-feature condition
The sounds were presented using the same stimulus sequence as was used in the
Optimum-1 paradigm (Na ¨a ¨ta ¨nen et al., 2004): every other syllable was a standard
(P = 0.5) and every other one of the 5 deviant syllables (P = 0.1, each). There were
four 5-min sequences including 465 syllables, of which the first 5 were always
standards. In two of the sequences the standard syllable was /te:/ and the deviants
were FRE/te:/, INT/te:/, VOWDUR/te/, CONS/pe:/, and VOWEL/ti:/, whereas in the
other two sequences the standard was /pi:/ and the deviants were FRE/pi:/, INT/pi:/,
S. Pakarinen et al./Biological Psychology 82 (2009) 219–226
VOWDUR/pi/, CONS/ti:/, and VOWEL/pe:/. The occurrence of the deviants was
pseudo-randomized in a way that all 5 deviants appeared once in an array of 10
successive stimuli and the same deviant was never repeated after the standard
following it. For half of the subjects, the two sequences with the standard syllable /
/pi:/ were presented fist. Each deviant type FRE, INT, VOWDUR, CONS, and VOWEL
was presented 184 times, the stimulus-onset asynchrony (SOA, onset to onset) was
650 ms, and the total recording time of this condition was 20 min.
2.2.2. Duration control condition for the multi-feature condition
In order to dissociate the effects of processing physically different stimuli (e.g.,
the exogenous responses, as the sustained negative response to duration and
negative offset response differ between the ERPs of different durations) from the
genuine MMN, a control condition was employed (cf. Jacobsen and Schro ¨ger, 2003).
This condition was identical to the multi-feature condition, except for the duration
of the syllables. Whereas the vowel-duration deviant of the multi-feature condition
was 170 ms and the rest of the syllables 100 ms, the sound durations of this
condition were reversed so that the deviant syllable was 170 ms and all the other
ones 100 ms. Thus, in two sequences the standard syllable was /te/ and the deviants
were FRE/te/, INT/te/, VOWDUR/te:/, CONS/pe/, and VOWEL/ti/, whereas in the
other two the standard was /pi/ and the deviants FRE/pi/, INT/pi/, VOWDUR/pi:/,
CONS/ti/, and VOWEL/pe/.
2.2.3. Oddball condition
The sequences for oddball condition were created using those of the multi-
feature condition. These sequences included one deviant type (P = 0.1; e.g., change
in frequency) only, with all the rest being replaced with standard syllables (P = 0.9).
Thus, there were four 5-min oddball-sequences for each deviant type (FRE, INT,
VOWDUR, CONS, and VOWEL).
The ERP recordings were carried out in two separate sessions. The second
recording session took place 1–7 (median = 1) days after the first session. The
different conditions were presented in a pseudorandom order, with the following
constraints. For a measure of reliability, the MMNs for the five deviations in the
multi-feature conditionand theMMN forthe vowel changein the oddball condition
wererecorded twice,onceonbothsessions.Toavoidanyconfoundingeffects dueto
the MMN attenuation during the recording (McGee et al., 2001), these two blocks
were recorded at the same stages of the recording session on both days, i.e., if the
multi-feature condition had been presented as the third block on the first session, it
was presented as the third on the second session also. Further, the multi-feature
duration control condition was always presented as the last condition of the second
recording session. Consequently, the multi-feature and vowel change oddball
conditions were always presented as the first, second or third block. Furthermore,
the order of presentation of the rest of the conditions, as well as that of the
presentation of the multi-feature and vowel-duration oddball conditions, was
counterbalanced between subjects.
In order to study how well the MMN response reflects the degree of perceived
sound change magnitude of the stimuli used, a behavioural matching-to-sample
test was performed and the correlation between these test results and the MMN
was tested. The behavioural test was always performed after the 2nd ERP recording
only so that subjects remained as naı ¨ve of the stimuli as possible during the ERP
recordings in order to avoid carry-over effects of attention or learning. There were
one it was /pi:/, with the same five deviants as described for the ERP recordings. The
stimuli were presented in pairs with a within-pair SOA of 650 ms and an inter-pair
SOAof2650 ms.Thefirstsyllableofthepair wasalways thestandardsyllable,andit
was followed by a probe syllable. In 50% of the trials the probe syllable was a
standard and in remaining 50% of the trials it was one of the 5 deviants
(Pdevtype= 0.10). The order of the deviants was randomly assigned, so that each of
the five deviants was presented 20 times within the sequence of the 200 syllable
pairs. The subjects’ task was to judge whether the probe syllable was the same as or
different from the standard syllable. The response had to be given by pressing one
mouse button as fast as possible when the probe sound differed from the standard
syllable and the other when the two sounds were the same. The responses were
given with the forefinger and middle finger of the subject’s preferred hand, and the
association between the finger (forefinger or middle finger) and the type of
response (same or different) was counterbalanced across subjects.
During the ERP recordings, subjects watched a self-selected, subtitled film and
were instructed to ignore the stimuli. Neither was electroencephalogram (EEG)
recorded nor film presented during the behavioural task. The stimuli were always
binaurally delivered via headphones at an intensity of 50 dB above the subject’s
hearing threshold measured using the multi-feature condition stimulus sequence.
2.3. ERP recording
The EEG (0–100 Hz, sampling rate 500 Hz) was recorded with 30 channels using
an electrode cap placed according to an individually measured location of the Fz.
The reference electrode was placed on the tip of the nose. The horizontal and
vertical electro-oculograms (HEOG and VEOG) were recorded between two
electrodes placed at the outer canthi of eyes (HEOG) and above and below the
right eye (VEOG).
The continuous EEG was offline filtered (passband 1–20 Hz). Epochs of 600 ms
including a 100 ms pre-stimulus period were separately averaged for the standards
and deviants in each condition. The mean voltage of the pre-stimulus period served
as a baseline for the amplitude measurement. Epochs including voltage changes
exceeding 75 mV and those for the first 8 stimuli of each sequence were omitted
from the averaging.
To delineate the MMN, the responses to the standard syllables were subtracted
from those to the corresponding deviant syllables. In addition, the response to the
standard (100 ms in duration) syllable in the duration control condition was
subtracted from that to the physically identical duration deviant of the multi-
feature condition (100 ms). Similarly, the response to the standard syllable in the
multi-feature condition (170 ms) was subtracted from that to the deviant in the
duration control condition (170 ms). The responses were re-referenced to the mean
of the two mastoid electrodes in order to pool the strength and timing of MMN
activity to frontal-electrode sites. As the signal was found to be strongest at Fz in
most subjects, it was chosen for the statistical analysis. This kind of analysis of
mastoid-referenced fontal-electrode (Fz) MMN data is fast and simple, and thus
more applicable to clinical settings as compared with more complex multi-
electrode analysis methods, and it is shown to provide a good estimate of both
mastoidal and frontal MMN activity (Na ¨a ¨ta ¨nen, 1990; Schro ¨ger, 1997).
The MMN mean amplitudes were calculated as a mean voltage at the 50-ms
period centred at the peak latency in the grand-average response. One-tailed t-tests
were conducted to determine whether the MMN mean amplitudes differed
significantly from zero on the group level. The MMN peak amplitudes and latencies
were measured from the most negative peak occurring at 120–250 ms after
stimulus-onset, except for the consonant change which was measured at 150–
250 ms after stimulus-onset. The peak latencies of the duration MMNs were
corrected in relation to the deviation onset (latencies of other MMNs remained in
relation to the stimulus-onset).
Two two-way analyses of variance (ANOVA) for repeated measures were
conducted to test the effects of different paradigms and deviant types on both the
multi-feature (1st recording), duration control for multi-feature and oddball
(paradigm 3 levels: multi-feature 1st recording, multi-feature duration control,
oddball; deviant type 5 levels: Fre, Int, Vowdur, Cons, Vowel). The second analysis
compared the MMNs obtained with the multi-feature paradigm across the two
successive recording sessions (session 2 levels: multi-feature 1st recording, multi-
feature 2nd recording; deviant type 5 levels: Fre, Int, Vowdur, Cons, Vowel). In
addition, a third two-way ANOVA for repeated measures was carried out to test the
effects of paradigm (2 levels: multi-feature, oddball) and recording session (2
levels: 1st recording, 2nd recording) on the vowel MMN peak amplitude and
latency. Greenhouse-Geisser corrections were applied where appropriate (the
original degrees of freedom and p-values after the correction are reported).
Newman–Keuls tests were carried out as post hoc analyses.
In addition, paired samples t-tests were carried out to compare the peak
amplitude of the vowel-duration MMN in the multi-feature condition (1st
recording) with that of the vowel-duration MMN obtained by subtracting the
100 ms standard of the duration control condition from the 100 ms deviant of the
multi-feature condition, as well as with the peak amplitude of the vowel-duration
MMN obtained by subtracting the 170 ms standard of the multi-feature condition
from the 170 ms deviant of the duration control condition.
2.4. Behavioural discrimination test
In the separate behavioural discrimination test, button presses occurring 200–
1200 ms after the probe onset were identified as responses. The correct responses
were separately calculated for the stimuli that served as standards and deviants in
the ERP session. As the deviants appeared randomly in the same sequence, an
average false alarm rate was calculated separately for the standard and across the
five deviants. Two-way ANOVA for repeated measures was conducted to test the
effects of stimulus sequence (2 levels: /te:/, /pi:/) and stimulus type (6 levels: Std,
Fre, Int, Vowdur, Cons, Vowel) on percent of correct responses.
The relationship between the behavioural and brain responses was evaluated by
calculating the Pearson product moment correlation coefficients between the
measures for the five deviant types, i.e., the percent of correct responses and false
alarms were correlated to the MMN peak amplitudes and latencies measured in the
multi-feature condition (1st recording).
Fig. 1 presents the average of mastoid-referenced deviant-
minus-standard subtraction signals of 15 subjects for each deviant
type in the multi-feature paradigm (1st recording), in the duration
S. Pakarinen et al./Biological Psychology 82 (2009) 219–226
control multi-feature paradigm and in the traditional oddball
paradigm. The MMNs peaked between 120 and 250 ms from
deviation onset. Syllable frequency, intensity, vowel-duration, and
vowel changes elicited statistically significant MMN mean
amplitudes in all conditions (t14= ?12.9 to ?2.7, P < 0.001),
whereas the MMN mean amplitude for the consonant change was
significant only in the 1st multi-feature paradigm recording
(t14= ?3.2, P < 0.001; Table 1).
Two-way ANOVA on MMN peak amplitude for multi-feature
(1st recording), multi-feature duration control, and oddball
Fig. 1. Grand-average difference signals (15 subjects) for changes in syllable frequency, syllable intensity, vowel-duration, consonant change, and vowel change at all
electrodes (electrode Fz enlarged in the insert) in the multi-feature condition (1st recording session; solid line), oddball condition (thick dashed line) and multi-feature
duration control condition (thin dashed line). The data were referenced to the mean of the two mastoid electrodes (LM and RM). Sound onset is always at 0 ms.
S. Pakarinen et al./Biological Psychology 82 (2009) 219–226
paradigms revealed that the paradigm had an influence on the
P < 0.001). The MMN amplitude was smaller in the duration
control multi-feature paradigm than in the multi-feature or
oddballparadigms(Newman–Keuls:P < 0.001fordurationcontrol
paradigm vs. multi-feature and oddball, other comparisons ns.).
The MMN amplitude differed also between the different deviant
types (main effect of DEVIANT TYPE: F4,56= 34.15, P < 0.001). The
consonant change elicited smaller and the vowel change larger
MMNsthan the other soundchanges (Newman–Keuls: P < 0.05for
Cons vs. other changes and for Vowel vs. other changes, other
Corresponding ANOVA for the MMN peak latency revealed that
the paradigm had an influence on the MMN latency also (main
effect of PARADIGM: F2,28= 4.83, P < 0.05). The MMN peaked
earlier in the duration control multi-feature paradigm than in the
multi-feature or oddball paradigms (Newman–Keuls: P < 0.05 for
duration control paradigm vs. multi-feature and oddball, other
comparisons ns.). Also the deviant type had an influence on the
MMN latency (main effect of DEVIANT TYPE: F4,56= 48.27,
P < 0.001). The MMN peaked later for the consonant change and
earlier for vowel-duration change than for the other changes.
Furthermore, the latency for the frequency change was later than
that for the vowel change (Newman–Keuls: P < 0.05 for all
comparisons except for Int vs. Vowel and Int vs. Fre).
The peak amplitude of the vowel-duration MMN recorded with
the multi-feature paradigm (1st recording session; std 170 ms, dev
100 ms) significantly differed from the peak amplitude of the
control MMN constructed from the responses to the physically
identical 100 ms (deviant from the multi-feature condition minus
standard from the control condition; t14= 3.3, P < 0.01) stimuli,
but not from that to the corresponding 170 ms stimuli (t14= ?1.3,
ns.; Fig. 2.).
3.2. Replicability of the MMN
Fig. 3 presents the MMN responses for changes in the syllable
fundamental frequency, intensity, vowel-duration and for the
consonant and vowel changes in the first and the second recording
sessions with the multi-feature paradigm (panel A) and the MMNs
for thevowel changesinthefirstand thesecondrecordingsessions
in the oddball paradigm (panel B).
recording, and multi-feature 2nd recording revealed that the
session had an influence on the MMN amplitude (main effect of
SESSION: F1,14= 6.17, P < 0.05). The responses were larger in
amplitude on the first than on the second recording session. Also
the deviant type had an influence on the MMN amplitude (main
effect of DEVIANT TYPE: F4,56= 23.96, P < 0.001). The MMN
amplitude was smaller for the consonant than for the other
changes and the MMN for the vowel change was larger than for the
consonant, frequency and intensity changes (Newman–Keuls:
P < 0.01 for Cons vs. Vowel, Vowdur, Fre, Int, and for Vowel vs.
Cons, Fre, Int, other comparisons ns.).
Corresponding ANOVA for the MMN peak latency revealed that
the deviant type had an influence on the MMN latency (main effect
of DEVIANT TYPE: F4,56= 54.27, P < 0.001). The latency for the
consonant change was longer and the latency for the vowel-
duration change shorter than for the other changes. Moreover, the
latency for the frequency change was longer than for the intensity
and vowel changes (Newman–Keuls: P < 0.05 for all, except for Int
The mean MMN amplitudes of 15 subjects (with standard errors of means in parentheses) at electrode Fz and t-values in the multi-feature condition, multi-feature duration
control condition, and oddball condition as well as for the vowel-duration MMNs constructed from responses to the physically identical 100ms and 170ms long stimuli.
Paradigm Frequency IntensityVowel durationConsonantVowel
Meant Meant MeantMeant Meant
Results of one-tailed t-tests *P<0.05, **P<0.01, ***P<0.001.
Fig. 2. Grand-average difference signals referenced to the mean of the 2 mastoid
electrodes(15 subjects) forchangesin vowel-duration atFzelectrode. Thesolid line
denotes MMN response recorded with the multi-feature condition (1st recording
session; deviant 100 ms minus standard 170 ms), the thick dashed line denotes the
subtraction signal constructed from the responses to the 100 ms long stimuli
(deviant in the multi-feature condition minus standard in the duration multi-
featurecondition)and thethindashed linedenotesthose tothe170 mslong stimuli
(deviant in the duration control multi-feature condition minus standard in the
multi-feature condition). The data were referenced to the mean of the two mastoid
electrodes. Sound onset is always at 0 ms.
S. Pakarinen et al./Biological Psychology 82 (2009) 219–226
Two-way ANOVAs on both the MMN peak amplitude and
latency for the vowel change in the two multi-feature and oddball
recording sessions revealed neither effects of paradigm (multi-
feature vs. oddball) nor recording session (1st vs. 2nd session).
3.3. Behavioural responses
The standards were correctly identified as same as the probe
in 96% of the trials (range 91–98%). Moreover, the intensity,
frequency, vowel-duration, and vowel change deviants were
correctly identified as being different from the probes in more
than 90% of trials (92%, range 73–100%; 96%, range 73–100%;
91%, range 53–100%; 97%, 85–100%, respectively), whereas the
consonant change was identified as different from the probe in
only 75% of the trials (range 5–100%). The false alarm (FA) rate
for the standards was 7.5% and for deviants 10% on the average.
The standards were erroneously regarded as different from the
probe in 8% (range 1–20%) in the /te:/-sequence trials and in 7%
(range 0–20%) in the /pi:/-sequence trials. The deviants in the /
te:/ sequence were again erroneously regarded the same as the
probe in 7% (range 0–25%) and the deviants in the /pi:/ sequence
were regarded the same as the probe in 13% (range 0–50%) of
Two-way ANOVA on percent of correct responses revealed that
the stimulus type had an influence on the number of correct
responses (main effect of STIMULUS TYPE: F5,70= 4.95, P < 0.05).
There were less correct responses for the consonant than for the
other stimuli (Newman–Keuls: P < 0.001 for Con vs. all other
3.4. Correlation between the ERP and behavioural measures
Only the MMN latency for the intensity deviations correlated
with the percent of correct responses (R = 0.523, P < 0.01) and
misses (R = 0.662, P < 0.001) for the same deviations. All other
correlations were not significant. This was probably due to lack of
statistical power, which resulted from the low variation (low task
difficulty) and the relatively small sample size.
When recorded with the new fast multi-feature paradigm, the
five different sound deviations: syllable frequency (F0), syllable
intensity, vowel-duration, consonant change, and vowel change
elicited statistically significant MMN responses (Table 1). More-
over, the responses were comparable with those obtained in the
traditional oddball paradigm. As can be seen from Fig. 1, the shape
of the MMN responses, as well as the peak amplitudes and
latencies, were highly similar between the multi-feature and
The MMN responses were also well replicated between the
repeated recording sessions with the same subjects. When the
MMNs were compared between the two multi-feature paradigm
recording sessions (Fig. 3, panel A), the MMN peak amplitude was
slightly smaller in the second than in the first recording. The
consonant change elicited a significant MMN response in the first
recording session only (Table 1). There were no differences in the
MMN peak latencies, however. Moreover, when the vowel change
MMN, which was recorded twice both with the multi-feature and
with the oddball paradigm (Fig. 3, panel B), was examined, no
differences were found between the paradigms or the recording
Regardless of the paradigm (multi-feature, oddball, duration
control multi-feature), the MMN was the smallest and latest (if
existent) for the consonant change. In fact, the MMN for the
consonant change was reliably observed only in the first recording
session with the multi-feature paradigm: in the oddball, multi-
feature duration control condition, and in the second recording
with the multi-feature paradigm the MMN mean amplitude failed
to reach statistical significance (Table 1). Also the behavioural
discrimination test showed that the consonant change was more
difficult to discriminate as compared with the other sound
changes. Only 75% of the consonant changes were correctly
perceived as different from the probe, as compared to 91–97%
was largest for the vowel change MMN as compared with the
MMNs for the other sound changes, and the vowel change was also
Fig. 3. Grand-average difference signals (15 subjects) at Fz electrode over the two repeated recording sessions for changes in syllable frequency, syllable intensity, vowel-
duration, consonantchange, and vowel changein themulti-featurecondition (panelA)and forvowel changein theoddball condition (panelB).Thesolid linedenotes thefirst
recording session and the dashed line the second recording session. The data were referenced to the mean of the two mastoid electrodes. Sound onset is always at 0 ms.
S. Pakarinen et al./Biological Psychology 82 (2009) 219–226
most accurately behaviourally discriminated (97% of the trials on
average). This is consistent with previous findings (cf. Tiitinen
et al., 1994; Jaramillo et al., 2000; Pakarinen et al., 2007) showing
that the MMN response (especially the MMN amplitude) reflects
the degree of perceived sound change magnitude. The result is also
in line with previous ones showing that the replicability of the
MMN declines with decreasing degree of deviance (and thus lower
signal-to-noise ratio; Joutsiniemi et al., 1998; Tervaniemi et al.,
1999; Kujala et al., 2001b).
The responses measured in the multi-feature duration control
condition were shorter in latency and smaller in amplitude than
those in the multi-feature and oddball paradigms. The only
difference between the multi-feature and duration multi-feature
condition was in the stimulus durations: the multi-feature
condition’s stimuli were 170 ms in duration (e.g., /te:/) whereas
the multi-feature duration control condition’s stimuli were only
100 in duration (e.g., /te/). Thus the amplitude difference between
the conditions (also that between the oddball and multi-feature
duration control) is likely to reflect the fact that the longer-
duration 170 ms stimuli provided more information for the
deviant detection, and thus a better signal-to-noise ratio. The
latency difference instead may be more closely related to the mere
duration of the tones; the processing of shorter sounds, containing
less information, might be faster than that of longer sounds.
The main purpose for the multi-feature duration control
condition in this study, however, was to dissociate the effects of
physically different stimuli from the genuine vowel-duration
MMN. Fig. 2, comparing the vowel-duration MMN obtained in the
multi-feature paradigm (deviant of 100 ms minus standard of
170 ms) and the vowel-duration MMNs constructed form the
physically identical stimuli (deviant of 170 ms minus standard of
170 ms and deviant of 100 ms minus standard of 100 ms) shows
that the MMN obtained solely in the multi-feature condition falls
between the other two control MMN traces. The peak amplitude of
the multi-feature condition MMN was smaller than the one
constructed from the responses to the 100 ms tones, but did not
differ from that constructed from those to the 170 ms tones. From
this one can conclude that this simple comparison between
responses to physically different tones in the multi-feature
condition in adult participants leads to a rather accurate or only
slightly underestimated vowel-duration MMN amplitude.
As the present results suggest, this new paradigm offers a fast
(20 min) and reliable method for evaluating the central auditory
discrimination of acoustic and linguistic sound features [syllable
frequency (F0), syllable intensity, vowel-duration, consonant
change and vowel change] in a speech sound context. With
linguistic stimuli, one may evaluate the functioning and develop-
ment of language-related long-term memory traces (Na ¨a ¨ta ¨nen
et al., 1997; Shestakova et al., 2002). One could, for instance, follow
the development of enhanced processing of those features that are
relevant in the native language (cf. Nenonen et al., 2003) and
determine its time course with repeated recordings in early
childhood. Also, one could compare this development between
different languagesorin the course of second-languageacquisition
(Winkler et al., 1999; Shestakova et al., 2003; Kirmse et al., 2008).
In clinical studies, the version of the multi-feature paradigm
including harmonical tones has been already successfully used in
research of, for instance, dyslexia (Kujala et al., 2006), Asperger’s
syndrome (Kujala et al., 2007), and schizophrenia (Fisher et al.,
2008). However, as the central auditory processing of phonological
information may differ from the processing of non-linguistic
auditory information (Na ¨a ¨ta ¨nen et al., 1997; Shtyrov et al., 1998)
language context is likely to provide important new information,
especiallyregardinglanguage-relateddisorders. Thereisalso some
evidence that the responses measured in more challenging
paradigms, as compared with the traditional oddball paradigm
(see, for instance, Baldeweg et al., 2004; Kujala et al., 2006), may
more appropriately characterize the processes underlying subtle
cognitive impairments compared with more static ERP measures,
which, in turn, may be appropriate for very severe impairments.
In addition, the fastness of the paradigm makes it highly
practical in clinical settings, where the time efficiency is of major
importance. Short recording time is also practical when investi-
gatingsubjects, such as small children whocan sitstill for a limited
time only. Shorter-duration recordings are also less prone to long-
term habituation and amplitude fluctuation effects as compared
with long continuous recordings. The MMN for the speech sounds
begins to fluctuate and increasingly attenuate as soon as after
15 min of continuous recording (McGee et al., 2001). These effects
can be somewhat compensated with resting periods between the
stimulus presentations (McGee et al., 2001), but this will of course
further increase the time required for data acquisition.
In all, with this new paradigm alone, or in combination with the
non-linguistic version (Na ¨a ¨ta ¨nen et al., 2004; Pakarinen et al.,
2007), one can obtain an extensive profile of central auditory
processing and its deficits. More subtle abnormalities might be
revealed and a more fine-grained insight into deficits on central
auditory processing, for instance in language-related disorders
suchasdyslexia,specificlanguage impairmentoraphasia, couldbe
This study was supported by the Academy of Finland (grants
#1213933 and 1211486) and the Finnish Cultural Foundation. The
collection and assistance in data analysis.
Alku, P., Tiitinen, H., Na ¨a ¨ta ¨nen, R., 1999. A method for generating natural-sounding
speech stimuli for cognitive brain research. Clinical Neurophysiology 110,
Alho, K.,Huotilainen, M., Tiitinen, H., Ilmoniemi, R.J., Knuutila, J., Na ¨a ¨ta ¨nen, R., 1993.
Memory-related processing of complex sound patterns in human auditory
cortex: a MEG study. Neuroreport 4, 391–394.
Alho, K., Sainio, K., Sajaniemi, N., Reinikainen, K., Na ¨a ¨ta ¨nen, R., 1990. Event-related
brain potential of human newborns to pitch change of an acoustic stimulus.
Electroencephalography and Clinical Neurophysiology 77, 151–155.
Amenedo, E., Escera, C., 2000. The accuracy of sound duration representation in the
human brain determines the accuracy of behavioural perception. European
Journal of Neuroscience 12, 2570–2574.
Baldeweg, T., Klugman, A., Gruzelier, J., Hirch, S.R., 2004. Mismatch negativity
potentials and cognitive impairment in schizophrenia. Schizophrenia Research
Baldeweg, T., Richardson, A., Watkins, S., Foale, C., Gruzelier, J., 1999. Impaired
auditory frequency discrimination in dyslexia detected with mismatch evoked
potentials. Annals of Neurology 45, 495–503.
Cse ´pe, V., 1995. On the origin and development of the mismatch negativity. Ear and
Hearing 16, 91–104.
Cse ´pe, V., Osman-Sa ´gi, J., Molna ´r, M., Go ´sy, M., 2001. Impaired speech perception in
aphasic patients: event-related potential and neuropsychological assessment.
Neuropsychologia 39, 1194–1208.
Draganova, R., Eswaran, H., Murphy, P., Lowery, C., Preissl, H., 2005. Sound fre-
quency change detection in fetuses and newborns, a magnetoencephalographic
study. Neuroimage 28, 354–361.
Draganova, R., Eswaran, H., Murphy, P., Lowery, C., Preissl, H., 2007. Serial magne-
toencephalographic study of fetal and newborn auditory discriminative evoked
responses. Early Human Development 83, 199–207.
Fisher, D.J., Labelle, A., Knott, V.J., 2008. The right profile: mismatch negativity in
schizophrenia with and without auditory hallucinations as measured by a
multi-feature paradigm. Clinical Neurophysiology 119, 909–921.
Giard, M.H., Perrin, F., Pernier, J., Bouchet, P., 1990. Brain generators implicated in
the processing of auditory stimulus deviance: a topographic event-related
potential study. Psychophysiology 27, 627–640.
Hari,R.,Ha ¨ma ¨la ¨inen,M.,Ilmoniemi,R.J.,Kaukoranta,E.,Reinikainen,K.,Salminen,J.,
Alho, K., Na ¨a ¨ta ¨nen, R., Sams, M., 1984. Responses of the primary auditory cortex
to pitch changes in a sequence of tone pips: neuromagnetic recordings in man.
Neuroscience Letters 50, 127–132.
Huotilainen, M., Kujala, A., Hotakainen, M., Parkkonen, L., Taulu, S., Simola, J.,
Nenonen, J., Karjalainen, M., Na ¨a ¨ta ¨nen, R., 2005. Short-term memory functions
S. Pakarinen et al./Biological Psychology 82 (2009) 219–226
of the human fetus recorded with magnetoencephalography. Neuroreport 16,
Ilvonen, T., Kujala, T., Kozou, H., Kiesila ¨inen, A., Salonen, O., Alku, P., Na ¨a ¨ta ¨nen, R.,
2004. The processing of speech and non-speech sounds in aphasic patients as
reflected by the mismatch negativity. Neuroscience Letters 366, 235–240.
Jacobsen, T., Schro ¨ger, E., 2003. Measuring duration mismatch negativity. Clinical
Neurophysiology 114, 1133–1143.
Jaramillo, M., Paavilainen, P., Na ¨a ¨ta ¨nen, R., 2000. Mismatch negativity and beha-
vioural discrimination in humans as a function of the magnitude of change in
sound duration. Neuroscience Letters 290, 101–104.
Joutsiniemi, S.-L., Ilvonen, T., Sinkkonen, J., Huotilainen, M., Tervaniemi, M., Lehto-
koski, A., Rinne, T., Na ¨a ¨ta ¨nen, R., 1998. The mismatch negativity for duration
decrement of auditory stimuli in healthy subjects. Electroencephalography and
Clinical Neurophysiology 108, 154–159.
Kirmse, U., Ylinen, S., Tervaniemi, M., Vainio, M., Schro ¨ger, E., Jacobsen, T., 2008.
Modulation of the mismatch negativity (MMN) to vowel duration changes in
native speakers of Finnish and German as a result of language experience.
International Journal of Psychophysiology 67, 131–143.
Kujala, T., Aho, E., Lepisto ¨, T., Jansson-Verkasalo, E., Nieminen-von Wendt, T., von
Wendt, L., Na ¨a ¨ta ¨nen, R., 2007. Atypical pattern of discriminating sound features
in adults with Asperger syndrome as reflected by the mismatch negativity.
Biological Psychology 75, 109–114.
Kujala, T., Kallio, J., Tervaniemi, M., Na ¨a ¨ta ¨nen, R., 2001a. The mismatch negativity as
an index of temporal processing in audition. Clinical Neurophysiology 112,
Kujala, T., Karma, K., Ceponiene, R., Belitz, S., Turkkila, P., Tervaniemi, M., Na ¨a ¨ta ¨nen,
R., 2001b. Plastic neural changes and reading improvement caused by audio–
visual training in reading-impaired children. Proceedings of the National Acad-
emy of Sciences of the United States of America 98, 10509–10514.
Kujala,T.,Lovio,R.,Lepisto ¨,T.,Laasonen,M.,Na ¨a ¨ta ¨nen,R.,2006.Evaluationofmulti-
attribute auditory discrimination in dyslexia with the mismatch negativity.
Clinical Neurophysiology 117, 885–893.
Kushnerenko, E., Cheour, M., Ceponiene, R., Fellman, V., Renlund, M., Soininen, K.,
Alku, P., Koskinen, M., Sainio, K., Na ¨a ¨ta ¨nen, R., 2001. Central auditory processing
of durational changes in complex speech patterns by newborns: an event-
related brain potential study. Developmental Neuropsychology 19, 83–97.
Lachmann, T., Berti, S., Kujala, T., Schro ¨ger, E., 2005. Diagnostic subgroups of
developmental dyslexia have different deficits in neural processing of tones
and phonemes. International Journal of Psychophysiology 56, 105–120.
Lang, H.A., Nyrke, T., Ek, M., Aaltonen, O., Raimo, I., Na ¨a ¨ta ¨nen, R., 1990. Pitch
discrimination performance and auditory event-related potentials. In: Brunia,
C.H.M., Gaillard, A.W.K., Kok, A., Mulder, G., Verbaten, M.N. (Eds.), Psychophy-
siological Brain Research, vol. 1. Tilburg University Press, Tilburg, the Nether-
lands, pp. 294–298.
Lyytinen, H., Ahonen, T., Eklund, K., Guttorm, T., Kulju, P., Laakso, M.L., Leiwo, M.,
Leppa ¨nen, P., Lyytinen, P., Poikkeus, A.M., Richardson, U., Torppa, M., Viholai-
nen, H., 2004. Early development of children at familial risk for dyslexia—
follow-up from birth to school age. Dyslexia: the Journal of the British Dyslexia
Association 10, 146–178.
McGee, T.J., King, C., Tremblay, K., Nicol, T.G., Cunningham, J., Kraus, N., 2001. Long-
term habituation of the speech-elicited mismatch negativity. Psychophysiology
Na ¨a ¨ta ¨nen, R., 1990. The role of attention in auditory information processing as
revealed by event-related potentials and other brain measures of cognitive
function. Behavioral and Brain Sciences 13, 201–288.
Na ¨a ¨ta ¨nen, R., 1992. Attention and Brain Function. Hillsdale, Erlbaum.
Na ¨a ¨ta ¨nen, R., 2003. Mismatch negativity: clinical research and possible applica-
tions. International Journal of Psychophysiology 48, 179–188.
Na ¨a ¨ta ¨nen, R., Escera, C., 2000. Mismatch negativity (MMN): clinical and other
applications. Audiology and Neuro-otology 5, 105–110.
Na ¨a ¨ta ¨nen, R., Lehtokoski, A., Lennes, M., Cheour, M., Huotilainen, M., Iivonen, A.,
Vainio, M., Alku, P., Ilmoniemi, R.J., Luuk, A., Allik, J., Sinkkonen, J., Alho, K., 1997.
Language-specific phoneme representations revealed by electric and magnetic
brain responses. Nature 385, 432–434.
Na ¨a ¨ta ¨nen, R., Paavilainen, P., Rinne, T., Alho, K., 2007. The mismatch negativity
(MMN) in basic research of central auditory processing: a review. Clinical
Neurophysiology 118, 2544–2590.
Na ¨a ¨ta ¨nen, R., Pakarinen, S., Rinne, T., Takegata, R., 2004. The mismatch negativity
(MMN)—towards the optimal paradigm. Clinical Neurophysiology 115, 140–
Na ¨a ¨ta ¨nen, R., Winkler, I., 1999. The concept of auditory stimulus representation in
neuroscience. Psychological Bulletin 125, 826–859.
Nenonen, S., Shestakova, A., Huotilainen, M., Na ¨a ¨ta ¨nen, R., 2003. Linguistic
relevance of duration within the native language determines the accuracy
of speech-sound duration processing. Cognitive Brain Research 16, 492–
Pakarinen, S., Takegata, R., Rinne, T., Huotilainen, M., Na ¨a ¨ta ¨nen, R., 2007. Measure-
ment of extensive auditory discrimination profiles using mismatch negativity
(MMN) of the auditory event-related potential. Clinical Neurophysiology 118,
Renvall, H., Hari, R., 2003. Diminished auditory mismatch fields in dyslexic adults.
Annals of Neurology 53, 551–557.
Rinne, T., Alho, K., Ilmoniemi, R.J., Virtanen, J., Na ¨a ¨ta ¨nen, R., 2000. Separate time
behaviors of the temporal and frontal MMN sources. NeuroImage 12, 14–19.
Schulte-Ko ¨rne, G., Deimel, W., Bartling, J., Remschmidt, H., 1998. Auditory proces-
sing and dyslexia: evidence for a specific speech processing deficit. Neuroreport
Shalgi, S., Deouell, L.Y., 2007. Direct evidence for differential roles of temporal and
frontal components of auditory change detection. Neuropsychologia 45, 1878–
Shankarnarayan, V.C., Maruthy, S., 2007. Mismatch negativity in children with
dyslexia speaking Indian languages. Behavioral and Brain Functions 3, 36.
Schro ¨ger, E., 1997. On the detection of auditory deviations: a pre-attentive activa-
tion model. Psychophysiology 34, 245–257.
Shestakova, A., Brattico, E., Huotilainen, M., Galunov, V., Soloviev, A., Sams, M.,
Ilmoniemi,R.J.,Na ¨a ¨ta ¨nen,R.,2002.Abstractphonemerepresentationsintheleft
temporal cortex: magnetic mismatch negativity study. Neuroreport 13, 1813–
Shestakova, A., Huotilainen, M., Ceponiene, R., Cheour, M., 2003. Event-related
potentials associated with second language learning in children. Clinical Neu-
rophysiology 114, 1507–1512.
Shtyrov, Y., Kujala, T., Ahveninen, J., Tervaniemi, M., Alku, P., Ilmoniemi, R.J.,
Na ¨a ¨ta ¨nen, R., 1998. Background acoustic noise and the hemispheric lateraliza-
tion of speech processing in the human brain: magnetic mismatch negativity
study. Neuroscience Letters 251, 141–144.
Tervaniemi, M., Lehtokoski, A., Sinkkonen, J., Virtanen, J., Ilmoniemi, R.J., Na ¨a ¨ta ¨nen,
R., 1999. Test–retest reliability of mismatch negativity for duration, frequency,
and intensity changes. Clinical Neurophysiology 110, 1388–1393.
Tiitinen, H., May, P., Reinikainen, K., Na ¨a ¨ta ¨nen, R., 1994. Attentive novelty detection
in humans is governed by pre-attentive sensory memory. Nature 370,
Winkler, I., 2007. Interpreting the mismatch negativity. Journal of Psychophysiol-
ogy 21, 147–163.
Winkler, I., Kujala, T., Tiitinen, H., Sivonen, P., Alku, P., Lehtokoski, A., Czigler, I.,
Cse ´pe, V., Ilmoniemi, R.J., Na ¨a ¨ta ¨nen, R., 1999. Brain responses reveal the learn-
ing of foreign language phonemes. Psychophysiology 36, 638–642.
S. Pakarinen et al./Biological Psychology 82 (2009) 219–226