www.elsevier.com/locate/clinph

Clinical Neurophysiology 118 (2007) 177–185

Measurement of extensive auditory discrimination profiles usingthe mismatch negativity (MMN) of the auditory event-related

potential (ERP)

Satu Pakarinen a,b,*, Rika Takegata a,b, Teemu Rinne c, Minna Huotilainen a,b,d,Risto Naatanen a,b

a Cognitive Brain Research Unit, Department of Psychology, University of Helsinki, P.O. Box 9, FIN-00014, Finlandb Helsinki Brain Research Centre, Helsinki, Finland

c Department of Psychology, University of Helsinki, Finlandd Helsinki Collegium for Advanced Studies, University of Helsinki, Finland

Accepted 4 September 2006Available online 27 October 2006

Abstract

Objective: Mismatch negativity (MMN), a change-specific component of the auditory event-related potential (ERP), is sensitive to def-icits in central auditory processing associated with many clinical conditions. The aim of this study was to obtain a comprehensive multi-dimensional profile of central auditory processing by extending the recently developed fast multi-feature MMN paradigm [Naatanen R,Pakarinen S, Rinne T, Takegata R. The mismatch negativity (MMN): towards the optimal paradigm. Clin Neurophysiol 2004;115:140–144].Methods: MMN responses to changes in sound duration, frequency, intensity, and perceived sound-source location at six different mag-nitudes of deviation were recorded from healthy young adults by using the multi-feature MMN paradigm. In addition, behavioural dis-crimination accuracy and speed were measured to examine the relationship between MMN and behavioural performance.Results: All the 24 sound changes elicited significant MMNs. MMN amplitude increased and latency decreased with increasing magni-tude of sound change. Furthermore, the MMN amplitude and latency predicted the subjects’ accuracy and speed in detecting thesedeviations.Conclusions: This new paradigm provides an extensive auditory discrimination profile for several auditory attributes at different devia-tion magnitudes in a minimal recording time.Significance: The auditory discrimination profiles can offer a comprehensive view of the development, plasticity, and deficits of centralauditory processing.� 2006 International Federation of Clinical Neurophysiology. Published by Elsevier Ireland Ltd. All rights reserved.

Keywords: Mismatch negativity (MMN); Event-related potential (ERP); Central auditory processing; Sound discrimination

1. Introduction

The mismatch negativity (MMN), a component of theauditory event-related potential (ERP), is elicited by anydiscriminable violation of some regular aspect in the pre-ceding auditory stimulus sequence (Naatanen, 1990, 1992;

1388-2457/$32.00 � 2006 International Federation of Clinical Neurophysiolo

doi:10.1016/j.clinph.2006.09.001

* Corresponding author. Tel.: +358 9 191 29462; fax: +358 9 191 29450.E-mail address: satu.pakarinen@helsinki.fi (S. Pakarinen).

Naatanen et al., 2001; Naatanen and Winkler, 1999). It isthought to reflect the output of a pre-attentive memory-based comparison process, where the incoming sound iscompared with, and found to be deviant from, the neuralrepresentation (memory trace) of the regularities extractedfrom the preceding auditory input. An increasing degree ofsound deviance results in a larger MMN amplitude andshorter latency, as well as heightened behavioural detectionaccuracy and speed (Lang et al., 1990; Sams et al., 1985;

gy. Published by Elsevier Ireland Ltd. All rights reserved.

178 S. Pakarinen et al. / Clinical Neurophysiology 118 (2007) 177–185

Tiitinen et al., 1994). Moreover, the MMN emerges gradu-ally when learning to behaviourally discriminate initiallyundetectable deviations (Naatanen et al., 1993b). Thereforethe MMN is regarded as an index of auditory discrimina-tion accuracy (Novak et al., 1990; Naatanen, 1992; Tiitinenet al., 1997).

Recording of MMN requires neither subject’s behav-ioural response nor attention towards the sounds (Naata-nen et al., 1978, 1993a, 1999; Alho, 1992; Paavilainenet al., 1993). Therefore, it is less affected by attention-and task-related artefacts as compared with the behaviour-al measures of sound discrimination, which depend directlyon the cooperation and vigilance of the subject. MMN canbe measured also from subjects with problems in communi-cating or performing a discrimination task, for instance,from aphasic (Csepe et al., 2001) and comatose patients(Kane et al., 1993, 1996; Fischer et al., 1999; Morletet al., 2000), newborns (Alho et al., 1990; Cheour et al.,2002; Winkler et al., 2003), and even fetuses (Draganovaet al., 2005; Huotilainen et al., 2005). These characteristicsof MMN make it a promising tool for the clinical evalua-tion of auditory discrimination functions.

MMN is affected by various conditions including suchcommon diseases as schizophrenia (Catts et al., 1995;Michie et al., 2002, for a review, see, Michie, 2001), Alzhei-mer’s disease (Pekkonen et al., 1994, 2001; Riekkinen et al.,1997), AIDS (Schroeder et al., 1994), and diabetes (Vanha-nen et al., 1996), as well as in normal ageing (Pekkonenet al., 1996; Gaeta et al., 1998; Bertoli et al., 2002), whichindicates altered central auditory processing (despite intactperipheral hearing). The auditory impairment may be selec-tive to a specific type(s) of deviation (e.g., frequency, dura-tion, intensity, phoneme): for example, in children withdysphasia, the MMN amplitude for frequency change ismore attenuated than that to duration (Korpilahti andLang, 1994). Clinical groups may differ also in their sensi-tivity to the magnitude of stimulus deviance. Baldeweget al. (1999), for instance, found that in dyslexic adults,the MMN amplitude was attenuated only for smaller fre-quency changes (repeating standard tone ‘std’ 1000 Hz,interspersed with occasional deviant tones ‘devs’ 1015,1030, and 1060 Hz), whereas the MMN amplitude for alarger frequency change (dev 1090 Hz) and to moderate-to-large duration changes (std 200 ms, devs 160, 120, 80,and, 40 ms) were comparable to those of controls. Theyalso found that the behavioural detection accuracy of theseindividuals was relatively more impaired for smaller thanfor larger frequency changes. Most importantly, this fre-quency discrimination deficit, evident both in the MMNand the behavioural data, was correlated with the readingperformance of these individuals (Baldeweg et al., 1999).It has been suggested (Kujala and Naatanen, 2001) thatdyslexic individuals have difficulties in processing smallacoustic differences, and that this could be the cause oftheir problems in phonological processing.

In contrast to dyslexia, Javitt et al. (1998) found that theMMN amplitude of patients with schizophrenia was atten-

uated as compared with that of the controls for large (std1000 Hz, devs 1200, 2000, and 4000 Hz), but not for small(devs 1020 and 1050 Hz) frequency deviations. The authorsalso found that the MMN-amplitude difference betweenthe patients and controls increased with the magnitude offrequency change. This result, however, contradicts withsome other results obtained in patients with schizophreniausing similar experimental conditions. For example, Brock-haus-Dumke et al. (2005) found no differences betweenpatients’ and controls’ MMN amplitudes for a large fre-quency change (std 1000, dev 1200 Hz). Alain et al.(1998) also found that the MMN amplitude of the patientswas attenuated (std 1000, devs 1122 and 1414 Hz) but, incontrast to Javitt et al. (1998) findings, the MMN ampli-tude increased as a function of frequency change in a sim-ilar manner than that of the controls. On the other hand, areduction in the duration-MMN amplitude has been con-sistently reported in patients with schizophrenia (Michie,2001; Michie et al., 2000, 2002; Umbricht et al., 2003; Kir-cher et al., 2004; Light and Braff, 2005). Furthermore, theprocessing of sound-intensity changes, as indicated by theMMN, appears to be impaired in schizophrenia (Kircheret al., 2004). A recent meta-analysis (Umbricht and Krljes,2005) corroborates earlier suggestions (cf., Michie et al.,2000) that in schizophrenia, the MMN amplitude for dura-tion change might be more attenuated than that for a fre-quency change.

Even though the duration processing deficit in schizo-phrenia is rather well known, the processing of the otherauditory attributes and the effect of deviance magnitudeon the MMN of these individuals are poorly understood.Moreover, comparing results from different studies is diffi-cult due to differences in the experimental setups as well asin the subject groups examined. In order to obtain a morecomprehensive view of the cortical auditory processingcharacteristics in schizophrenia and in other clinical condi-tions, MMNs for a wide range of deviation magnitudeswithin the deviant type, as well as for different types ofdeviation, should be systematically examined from thesame individuals.

It is, however, rather time consuming to measureMMNs to several acoustic changes using the traditionalMMN paradigms (e.g., Oddball). Moreover, in the caseof clinical groups, such time-consuming measurementsare often not feasible. Recently, we developed a new mul-ti-feature paradigm (Optimum-1; Naatanen et al., 2004)in which five types of acoustic changes (frequency, intensi-ty, duration, perceived sound-source location, and gap, i.e.,introducing a silent period in the middle of the sound) arepresented in the same sound sequence so that every othertone is a standard (Pstd � 0.5) and every other is one ofthe five deviants (Pdev.type � 0.1). The paradigm is basedon the assumptions that the MMNs can be independentlyelicited for different auditory attributes, and that the devi-ant tones can still strengthen the memory trace of the stan-dard with respect to the stimulus attributes they have incommon (Nousak et al., 1996). In the multi-feature para-

S. Pakarinen et al. / Clinical Neurophysiology 118 (2007) 177–185 179

digm, the total number of standard tones presented and,thus, the duration of the experiment, are significantlyreduced. Most importantly, no difference was found inthe MMNs recorded using the new paradigm and thoseobtained in the traditional Oddball-paradigm.

The present study introduces an extension of the newparadigm in which MMNs to four different types of devia-tions (as in Optimum-1) at six deviation magnitudes (rang-ing from barely noticeable to easily detectable) arerecorded. We will show how the magnitude of deviationis reflected in the MMN amplitudes and the latencies,and further, how these MMNs well predict the subjects’detection accuracy and speed determined in a separatebehavioural detection task. Finally, we will show that anextensive auditory discrimination profile can be obtainedin one relatively short recording session.

2. Methods

2.1. Subjects

Nine healthy subjects (mean age 24 years, range 23–27years, three females) participated in the study. They gavewritten informed consent after the nature of the studywas explained to them.

2.2. Stimuli

The standard tones (Table 1) were composed of threesinusoidal partials (523, 1046 and 1569 Hz, with the funda-mental frequency corresponding to C5 in the Westernmusical scale). The second and third partials were 3 and6 dB lower in intensity, respectively, compared with thefundamental frequency. The standard tones were 75 ms(including 5 ms rise and fall times) in duration and theywere binaurally presented via headphones at an intensityof 60 dB above the subject’s hearing threshold, which wasindividually measured before the experiment using theexperimental stimulus sequence.

Table 1Stimulus parameters and probabilities

Stimulus parameters

Duration (ms) Intensity (dB) Frequency (Hz)

High (Hz) Low (Hz)

Std 75 60 523

DevL1 67 57.5 519/ 527L2 59 55 512/ 535L3 51 52.5 501/ 546L4 43 50 487/ 562L5 35 47.5 470/ 583L6 27 45 450/ 609

The rightmost column represents the magnitude of deviation ranging from thehigher and a half lower than the standard tone, and a half of the location devidenotes standard and Dev deviant tone.

The deviant tones (Table 1) differed from the standardseither in frequency (equiprobably an increment or decre-ment), intensity (decrement), duration (decrement), or per-ceived sound-source location (equiprobably left or right;abbreviated as Fre, Int, Dur, and Loc, respectively). Inaddition, the magnitude of the deviation from the standardtone varied across six levels (the smallest and largest mag-nitudes of change, L1 and L6, respectively), resulting in atotal of 24 deviants. The frequency deviants differed fromthe standards by 1/8, 3/8, 6/8, 10/8, 15/8, and 21/8 semi-tones in the Western musical scale (fundamental frequen-cies 527, 535, 546, 562, 583, 609 Hz for the higher and519, 512, 501, 487, 470, 450 Hz for the lower devianttones). A perceived change ranging from 5� to 90� in thesound-source location was produced by introducing aninter-aural time difference of 50, 100, 250, 400, 550, and700 ls between the stereo channels. For the frequency devi-ants, two tones with a corresponding magnitude (e.g., 1/8semitones) but opposite direction (high/low) of deviationwere taken as a deviant of the same level (e.g., DevFreL1).An analogous procedure was applied to the location devi-ants. The duration deviants were shorter than the stan-dards by steps of 8 ms (67, 59, 51, 43, 35, and 27 ms) andthe intensity deviants were softer than the standards bysteps of 2.5 dB (57.5, 55, 52.5, 50, 47.5, and 45 dB).

The tones were presented within the same stimulussequence, with the standard as every other tone(Pstd = 0.5), as in the Optimum-1 paradigm (Naatanenet al., 2004), and one of the 24 deviants (Pdev = 0.5/24 = 0.02) as every other tone. Each deviant type (Fre,Int, Dur, Loc) was presented once in an array of four suc-cessive deviants and two successive deviants were always ofdifferent types (e.g., DevDur � Std � DevFre � Std �DevLoc � Std � DevInt � Std � DevLoc � Std). The orderin which the different levels of deviation appeared inthe sequence was pseudo-random (e.g., DevDurL4 �Std � DevFreL2 � Std � DevLocL3 � Std � DevIntL3 �Std � DevLocL6 � Std; note that two similar levels ofdifferent attributes may follow each other). Furthermore,

Probability

Location

Semitones Time (ls) Degrees

0 0 (centre) Pstandard = 0.5

Pdeviants = 0.51/8 50 5 Ptype = 0.1253/8 100 10 Plevel = 0.0836/8 250 25

10/8 400 4015/8 550 6021/8 700 90 Plevel · type = 0.125/6

smallest (L1) to the largest (L6) one. A half of the frequency deviants wereants were perceived as coming from the left and a half from the right. Std

180 S. Pakarinen et al. / Clinical Neurophysiology 118 (2007) 177–185

the 24 different deviants were equiprobable. The soundswere presented in 5-min sequences (18 sequences in total,604 tones per sequence, each of the 24 deviants was pre-sented 225 times, all sequences beginning with four succes-sive standards), with the presentation order of thesequences varying randomly across the subjects. The stim-ulus-onset-asynchrony (SOA) was 500 ms.

2.3. ERP measurement

During the ERP recording (90 min), subjects watched asubtitled video (sound off). The EEG was recorded (0–40 Hz, sampling rate 500 Hz) from 63 channels using anelectrode cap (Ag/AgCl electrodes, attached with electrodepaste, with the electrode #12 always placed at the Fz), withan electrode on the tip of the nose as a common reference.The bipolar horizontal electro-oculogram (HEOG) wasrecorded between electrodes placed at the outer canthi ofeyes.

The EEG was filtered offline (passband 1–30 Hz).Epochs of 600 ms including a 100 ms pre-stimulus periodwere separately averaged for the different stimulus typesand levels. The mean voltage of the pre-stimulus periodserved as a baseline for the amplitude measurement.Epochs including voltage changes exceeding 75 lV at anychannel and the first eight epochs of each sequence wereomitted from the averaging in order to exclude artefactsand the large responses elicited by the first stimuli of thesequences.

To derive the MMN, the response to the standard toneswas subtracted from the response to each type and level ofdeviant tones. The responses were re-referenced to themean of the two mastoid electrodes in order to displaythe maximal MMN response at the frontal electrode sites.The MMN peak amplitudes and latencies were measuredfrom the most negative peak occurring at 100–250 ms poststimulus-onset at an electrode located at Fz. The MMNlatencies for duration changes were corrected in relationto the onset of deviation (e.g., the duration of the devianttone was subtracted from the peak latency). One-tailed t-tests were conducted to determine whether the MMNamplitudes significantly differed from zero. Two-wayrepeated measures analyses of variance (ANOVA) wereconducted to test the effects of deviant type (four: Dur,Int, Fre, Loc) and the magnitude of deviation (six: L1–L6) on both the MMN amplitude and latency. In addition,one-way ANOVAs for repeated measures were conductedto test the effect of deviation magnitude on each devianttype separately. Huynh–Feldt corrections were appliedwhere appropriate (the original degrees of freedom andcorrected p-values are reported). Newman–Keuls tests werecarried out as post hoc analyses.

2.4. Behavioural detection test

All subjects participated in the behavioural detectiontest 1–7 days (median = 1 day) after the ERP measure-

ment. In this test (without ERP recording), 16 of the 5-min stimulus sequences used in the ERP measurement werepresented to the subjects who were instructed to press abutton when hearing a change at any level in the auditoryattribute designated (e.g., frequency) and to ignore changesin the other attributes. The target (i.e., which attribute todiscriminate) remained the same for four consecutivesequences (e.g., discriminate frequency during sequences1–4, duration 5–8, intensity 9–12, and gap 13–16) and theorder of which targets to detect was counterbalancedacross the subjects. In this behavioural measurement, 50trials per each of the 24 deviants were obtained in a totalof 80-min time.

Button presses occurring 200–1200 ms from target onsetwere identified as correct responses. The mean hit rates(HR) and reaction times (RT) were separately calculatedfor the 24 deviants and the RTs for the duration deviationswere corrected in relation to the onset of deviation. Thefalse alarm (FA) rate was calculated for each attributeacross the six levels of deviation magnitude. Two-way anal-yses of variance for repeated measures were conducted forboth the HR and RT, in the same way as was described forthe ERP measurement.

To evaluate the relationship between the ERPs and thebehavioural measures, general linear mixed-model analyseswith repeated measures were separately performed for thedeviant types (Dur, Int, Fre, and Loc). The HR and theRT were assigned as dependant variables and the magni-tude of deviance as within-subject factor.

3. Results

3.1. ERPs

On the average, there were 4470 (83%) accepted epochsfor standards and 186 (83%) for each deviant per subject(Std: median 4324, range 3836–4993, Dur: median 181,range 153–214; Fre: median 183, range 155–214; Int: medi-an 183, range 155–213; Loc: median 185, range 155–214).

The upper panel of Fig. 1 presents the mastoid-refer-enced grand-average (deviant � standard) difference wave-forms at Fz. The lower panel presents the nose-referencedwaveforms (Fz and right mastoid) in order to illustratethe polarity reversal at the right mastoid. The MMNs peakbetween 100 and 220 ms from stimulus onset. As shown inTable 2, all the MMN amplitudes (measured from the mas-toid-referenced data) significantly differed from zero(t8 = 5.01–11.90, P < 0.01).

Fig. 2 (upper panel) presents the mean MMN peakamplitudes and latencies for the four deviant types as afunction of stimulus deviance. Across all deviant types,the MMN amplitude increased as a function of the magni-tude of deviance (main effect of LEVEL: F5,40 = 53.30,P < 0.001; contrast for linear increase as a function ofLEVEL: F1,8 = 74.91, P < 0.001). The MMN amplitudesfor deviations of different magnitudes differed significantlyfrom each other, except for the L3–L4 contrast (Newman–

Fz Fz-5

+5

500-100

Fz

Fz Fz-5

+5

500

500

-100

Fz

RM RM-5

+5

-100

RM

Am

plitu

de (

µV)

Time (ms)

Am

plitu

de (

µV)

Time (ms)

Am

plitu

de (

µV)

Time (ms)

Frequency MMNDuration MMN Intensity MMN Location MMNN

ose

refe

renc

eM

asto

id r

efer

ence

Fz

Fz

RM

57.5 dB55.0 dB52.5 dB50.0 dB47.5 dB45.0 dB

(deviant

(standard tone 60.0 dB

67 ms59 ms51 ms43 ms35 ms27 ms

deviant

(standard tone 75 ms

527 Hz535 Hz546 Hz562 Hz583 Hz609 Hz

(deviant

(standard tone 523 Hz

150 µs 50 µs

250 µs400 µs550 µs700 µs

(deviant

(standard tone 0 µs)

Fig. 1. Grand-average difference waveforms (nine subjects) for changes in duration, intensity, frequency, and perceived sound-source location for sixdifferent magnitudes of deviance at electrode Fz and referenced to the mean of the two mastoid electrodes (upper panel). The same waveforms referencedto the nose electrode and shown from the Fz and the right mastoid (RM; lower panel). The sound-onset is always at 0 ms.

S. Pakarinen et al. / Clinical Neurophysiology 118 (2007) 177–185 181

Keuls: P < 0.05 for all other comparisons than L3 vs. L4).Also the deviant type had an influence on the MMN ampli-tude (main effect of TYPE: F3,24 = 6.39, P < 0.01). Theduration and frequency deviants elicited larger-amplitude

Table 2The mean MMN amplitudes for the four types of deviations at six different m

Duration Intensity

Mean t Mean t

L1 �2.1 (.2) �9.6*** �1.9 (.4) �5.3L2 �2.4 (.4) �5.5** �3.4 (.5) �7.0L3 �3.6 (.4) �8.7*** �4.4 (.6) �7.1L4 �4.7 (.6) �8.1*** �4.9 (.6) �7.9L5 �5.5 (.6) �8.9*** �5.6 (.6) �9.7L6 �7.0 (.7) �10.5*** �6.7 (.6) �11.9

The rightmost column represents the magnitude of deviation ranging from the sfrom the responses referenced to the mean of the two mastoid electrodes.Standard errors of means are in parentheses.Results of one-tailed t-tests: **P < 0.01, ***P < 0.001.

MMNs than the intensity and location deviants (New-man–Keuls: P < 0.05 for Dur vs. Int, Dur vs. Loc, Frevs. Int, Fre vs. Loc, other comparisons ns.). An interactionof the deviant type and level (F15,120 = 2.48, P < 0.01),

agnitudes of deviance

Frequency Location

Mean t Mean t

** �1.7 (.3) �6.2*** �1.4 (.3) �5.0**

*** �2.5 (.4) �5.6*** �1.9 (.3) �7.6***

*** �3.5 (.4) �8.0*** �3.4 (.3) �10.8***

*** �3.7 (.6) �6.1*** �3.3 (.6) �5.3**

*** �4.0 (.5) �8.1*** �4.3 (.8) �5.7***

*** �4.9 (.5) �9.6*** �4.7 (.6) �8.0***

mallest (L1) to the largest (L6) one. The MMN amplitudes were measured

Intensity

Rea

ctio

n T

ime

(ms)

Hits

(%

)

140

160

180

1 2 3 4 5 60

-2

-4

-6

-8 220

200

Duration

Frequency

Location

1 2 3 4 5 6

1 2 3 4 5 6

400

500

600

700

1 2 3 4 5 6

MMN Peak Amplitude MMN Latency

Hit Rate Reaction Time

Smallest LargestMagnitude of Deviation

Smallest LargestMagnitude of Deviation

Smallest LargestMagnitude of Deviation

Smallest LargestMagnitude of Deviation

Am

plitu

de (

µV)

Late

ncy

(ms)

0

20

40

60

80

100

Fig. 2. The mean (±SEM) MMN amplitude (upper left panel) and latency (upper right) at electrode Fz and behavioural measures hit rate (HR; lower left)and reaction time (RT; lower right) as a function of stimulus deviance. The MMN latencies and the RTs are presented in relation to the sound onset,except for the duration deviations, which are presented in relation to the deviation onset. The ERP data were referenced to the mean of the two mastoidelectrodes.

182 S. Pakarinen et al. / Clinical Neurophysiology 118 (2007) 177–185

which was mainly due to the increased MMN amplitudefor the larger frequency and duration deviations, was alsofound. The one-way ANOVAs for repeated measureswhich were calculated separately for each deviant type con-firmed that the MMN amplitude increased with increasingdeviance for all four types of changes (main effect ofLEVEL: Dur F5,40 = 33.73, P < 0.001; Fre F5,40 = 30.48,P < 0.001; Int F5,40 = 16.91, P < .001; Loc F5,40 = 11.96,P < 0.001; linear contrasts: Dur F1,8 = 56.40, P < 0.001;Fre F1,8 = 92.12, P < 0.001; Int F1,8 = 37.30, P < 0.001;Loc F1,8 = 18.91, P < 0.01).

Two-way ANOVA on MMN latency revealed that themagnitude of deviation also had an influence on theMMN latency (main effect of LEVEL: F5,40 = 8.55,P < 0.001). The MMN latencies were shorter for largerdeviations (contrast for linear increase as a function ofLEVEL: F1,8 = 14.32, P < 0.01). The MMN latency alsodiffered between the deviant types (main effect of TYPE:F3,24 = 19.51, P < 0.001). The duration MMNs peaked ear-lier and the intensity MMNs later than the other ones,whereas the peak latencies of the frequency and location

MMNs did not differ from each other (Newman–Keuls:P < 0.05 for all comparisons except for Fre vs. Loc). Aninteraction of the deviant type and level was also observed(interaction of deviant TYPE and LEVEL: F15,120 = 2.73,P < 0.01). Separate one-way ANOVAs for the differentMMNs showed that whereas the peak latencies of frequen-cy and location MMNs decreased with increasing magni-tude of deviation (effect of LEVEL: Fre F5,40 = 12.77,P < 0.01; Loc F4,40 = 4.30, P < 0.01; linear contrasts: FreF1,8 = 34.42, P < 0.001; Loc F1,8 = 9.93, P < 0.05) the mag-nitude of intensity and duration deviations had no effect onMMN peak latency (Int F1,8 = 1.20, P < 0.31; DurF5,40 = 0.36, P < 0.87).

3.2. Behavioural responses

Fig. 2 (lower panel) shows the mean RTs and HRs as afunction of deviant magnitude. For the four types of devi-ations, the mean HR was above the chance level (12.5%)for all but the lowest level of deviation magnitude andthe mean FA rate was low (1.0–3.3%).

S. Pakarinen et al. / Clinical Neurophysiology 118 (2007) 177–185 183

Two-way ANOVA on HR showed that the HRincreased as a function of deviation magnitude (main effectof LEVEL: F5,40 = 97.33, P < 0.001). The HRs for the dif-ferent magnitudes of deviation differed from each otherexcept for L5 that did not differ from L6 (Newman–Keuls:P < 0.05 for all other comparisons except for L5 vs. L6),with the larger deviations being detected more often thanthe smaller ones (contrast for linear increase as a functionof LEVEL: F1,8 = 501.25, P < 0.001). HR was also affectedby deviant type (main effect of TYPE: F3,24 = 11.65,P < 0.001); frequency deviants were detected more oftenthan the other ones (Newman–Keuls: P < 0.001 for Frevs. Int, Dur, Loc). The HRs for the deviant types increaseddifferently with the deviation magnitude (interaction ofdeviant TYPE and LEVEL: F15,120 = 6.78, P < 0.001). Thisinteraction was mainly due to the HR for the frequencydeviants approaching ceiling level faster than the HR forthe other deviant types. Additionally, separate one-wayANOVAs for the different deviant types confirmed thatthe increasing magnitude of deviance affected HR for alltypes of sound changes (main effect of LEVEL: DurF5,40 = 57.92, P < 0.001; Fre F5,40 = 44.50, P < 0.001; IntF5,40 = 65.55, P < 0.001; Loc F5,40 = 75.37, P < 0.001; line-ar contrast: Dur F1,8 = 223.08, P < 0.001; Fre F1,8 = 82.55,P < 0.001; Int F1,8 = 600.29, P < 0.001; Loc F1,8 = 304.37,P < 0.001).

Across all deviant types, RT differed between the levelsof deviation (main effect of LEVEL: F5,15 = 57.14,P < 0.001), but not between the deviant types. RTdecreased with the magnitude of deviation (contrast forthe linear increase as a function of LEVEL:F1,3 = 120.51, P < 0.01, Newman–Keuls: P < 0.05 for allcomparisons except for L5 did not differ from L6 andL4). A significant interaction between deviant type andlevel was also found (F15,45 = 8.04, P < 0.001). The RTfor the smallest frequency deviations was longer than thatfor other deviant types. Furthermore, the one-way ANO-VAs revealed that the deviation magnitude affected RTfor all deviants (main effect of LEVEL: DurF5,20 = 16.48, P < 0.001; Fre F5,30 = 41.88, P < 0.001; IntF5,25 = 23.12, P < 0.001; linear contrasts: Dur F1,4 =123.75, P < 0.001; Fre F1,6 = 104.78, P < 0.001; IntF1,5 = 57.06, P < 0.001) except for location (LocF5,30 = 1.83, P < 0.211).

3.3. Comparison of MMN and behavioural measures

MMN amplitude predicted HR for all types of devia-tions (Dur F1,45.97 = 84.40, P < 0.001; Fre F1,49.27 = 43.83,P < 0.001; Int F1,48.70 = 51.03, P < 0.001; Loc F1,51.21 =75.37, P < 0.001), and RT for duration, frequency, andintensity (Dur F1,43.34 = 42.47, P < 0.001; FreF1,45.50 = 57.19, P < 0.001; Int F1,42.48 = 93.76, P < 0.001),but not for location changes.

MMN latency predicted HR for frequency and locationdeviations (Fre F1,49.19 = 35.56, P < 0.001; Loc

F1,29.31 = 10,39, P < 0.001) and RT for frequency devia-tions (Fre F1,42.33 = 132.43, P < 0.001) only.

RT predicted HR for other but not for location devia-tions (Dur F1,47.22 = 82.88, P < 0.001, Fre F1,44.11 =230.52, P < 0.001, Int F1,42.08 = 37.57, P < 0.001).

4. Discussion

Using an extended version of the recently introducedmulti-feature paradigm (Optimum-1, Naatanen et al.,2004), MMNs for six magnitudes of deviations in frequency,intensity, duration, and perceived sound-source locationwere obtained in one recording session. The MMN peakamplitudes increased with the stimulus deviance. Corre-spondingly, the discrimination performance improved withincreasing deviance. This suggests that the MMN amplitudeindexes not only sound change per se, but also its degree in afine-grained manner and that this information translatesinto the subject’s conscious perception of the change.

The MMN peak latency decreased with the magnitudeof frequency and location deviations, whereas the latenciesof intensity and duration MMNs were not affected. Thislack of latency effect for intensity and duration MMNs isin contrast with the prevailing theory that the MMN peaksearlier the larger the deviation is (Naatanen, 1992; Naata-nen and Winkler, 1999). However, as only intensity andduration decrements were used, the lower power of thesedeviant stimuli may have delayed the N1 and thus affectedthe accuracy of the MMN latency estimation.

Both the accuracy and the speed of behavioural detec-tion of intensity, frequency, and duration changes wereimproved when the deviation magnitude was increased.As for the location changes, the HR improved with thedeviation magnitude, but the RT did not. It should be not-ed, however, that the two smallest location changes wereonly rarely (<20%) detected, and therefore the averageRT was based on only a few observations.

Although the MMN data predicted the behaviouraldetection performance, there were some differences in theMMN and behavioural profiles. For instance, the behav-ioural detection accuracy for the large frequency changesdiffered from that for the location changes, even thoughthe MMN amplitudes for these deviations were compara-ble. On the other hand, the MMN amplitudes for the largeduration and intensity changes differed from each othereven though the hit rates did not. Corresponding resultshave been previously found by Deouell and Bentin(1998). They recorded MMNs to perceptually matched(both between dimensions and subjects) deviations in fre-quency, intensity, location and stimulus-onset-asynchrony(SOA). Despite, being detected similarly, the MMN ampli-tude was larger for the frequency than for the location,intensity and SOA changes. These differences are probablydue to the fact that the MMN mainly reflects preattentivecentral auditory processing whereas the behavioural dataincorporates also other more higher-order processes, suchas those involved in attention and decision making.

184 S. Pakarinen et al. / Clinical Neurophysiology 118 (2007) 177–185

The paradigm introduced here allows one to constructan extensive auditory discrimination profile with a mini-mum recording time: 48 MMN parameters (the amplitudeand latency measures for four auditory attributes at sixdeviance magnitudes at one electrode) were obtained in90 min. For practical use in clinical settings, the new para-digm can be easily speeded up further by using fewer levelsof deviation magnitude. Recording MMNs, for instance,for four auditory attributes at three deviation magnitudes(180 trials per each of the 12 deviants), would require nomore than 36 min recording time.

Previous studies suggest that the difference in MMNresponse between patients and controls is often observednot in general but rather in a specific auditory attributeor magnitude of change (see Introduction). Thus, theinvestigation of auditory discrimination profiles of severalauditory attributes at different deviance magnitudes couldprovide valuable insight into the affected as well as pre-served central auditory processes in schizophrenia as wellas in other clinical conditions. By comparing, for instance,the discrimination profiles of dyslexic and control sub-jects, several essential questions, such as ‘‘Does dyslexiainvolve a general deficit of auditory processing or arethese impairments specific to certain auditory attributes?’’and ‘‘Is only the processing on minor stimulus differencesimpaired?’’ could be addressed within a single experiment.Furthermore, as the MMN can be obtained from the fetalperiod (Huotilainen et al., 2005) to the old age (Pekkonenet al., 1993), the auditory discrimination profiles can pro-vide information of the development and plasticity of thecentral auditory system throughout the lifespan. Childrenat risk for developmental disorders such as familial dys-lexia, specific learning impairment or aphasia could becost-effectively screened for affected processes at a veryearly stage of development. The profiles would also beuseful in the monitoring of the effectiveness of trainingprogrammes, set up and targeted according to informa-tion obtained at the screening phase (see, for instance,Kujala et al., 2001).

Acknowledgements

This study was supported by the Academy of Finland(grants #202562, # 207180, #77322, #200522, #209973,#79820, #79821 and #204050). The authors thank TuroOas and Ilkka Vanhanen for data collection and assistancein data analysis.

References

Alain C, Hargrave R, Woods DL. Processing of auditory stimuli duringvisual attention in patients with schizophrenia. Biol Psychiatry1998;44:1151–9.

Alho K. Selective attention in auditory processing as reflected by event-related brain potentials. Psychophysiology 1992;29:247–63.

Alho K, Sainio K, Sajaniemi N, Reinikainen K, Naatanen R. Event-related brain potential of human newborns to pitch change of anacoustic stimulus. Electroenceph clin Neurophysiol 1990;77:151–5.

Baldeweg T, Richardson A, Watkins S, Foale C, Gruzelier J. Impairedauditory frequency discrimination in dyslexia detected with mismatchevoked potentials. Ann Neurol 1999;45:495–503.

Bertoli S, Smurzynski J, Probst R. Temporal resolution in young andelderly subjects as measured by mismatch negativity and a psychoa-coustic gap detection task. Clin Neurophysiol 2002;113:396–406.

Brockhaus-Dumke A, Tendolkar I, Pukrop R, Schultze-Lutter F,Klosterkotter J, Ruhrmann S. Impaired mismatch negativity genera-tion in prodromal subjects and patients with schizophrenia. SchizophRes 2005;73:297–310.

Catts SV, Shelley AM, Ward PB, Liebert B, McConaghy N, Andrews S,Michie PT. Brain potential evidence for an auditory sensory memorydeficit in schizophrenia. Am J Psychiatry 1995;152:213–9.

Cheour M, Martynova O, Naatanen R, Erkkola R, Sillanpaa M, Kero P,et al. Speech sounds learned by sleeping newborns. Nature2002;415:599–600.

Csepe V, Osman-Sagi J, Molnar M, Gosy M. Impaired speech perceptionin aphasic patients: event-related potential and neuropsychologicalassessment. Neuropsychologia 2001;39:1194–208.

Deouell LY, Bentin S. Variable cerebral responses to equally distinctdeviance in four auditory dimensions: a mismatch negativity study.Psychophysiology 1998;35:745–54.

Draganova R, Eswaran H, Murphy P, Huotilainen M, Lowery C, PreisslH. Sound frequency change detection in fetuses and newborns, amagnetoencephalographic study. Neuroimage 2005;28:354–61.

Fischer C, Morlet D, Bouchet P, Luaute J, Jourdan C, Salord F.Mismatch negativity and late auditory evoked potentials in comatosepatients. Clin Neurophysiol 1999;110:1601–10.

Gaeta H, Friedman D, Ritter W, Cheng J. An event-related potentialstudy of age-related changes in sensitivity to stimulus deviance.Neurobiol Aging 1998;19:447–59.

Huotilainen M, Kujala A, Hotakainen M, Parkkonen L, Taulu S, SimolaJ, et al. Short-term memory functions of the human fetus recordedwith magnetoencephalography. Neuroreport 2005;16:81–4.

Javitt DC, Grochowski S, Shelley AM, Ritter W. Impaired mismatchnegativity (MMN) generation in schizophrenia as a function ofstimulus deviance, probability, and interstimulus/interdeviant interval.Electroenceph clin Neurophysiol 1998;108:143–53.

Kane NM, Curry SH, Butler SR, Cummins BH. Electrophysiologicalindicator of awakening from coma. Lancet 1993;341:688.

Kane NM, Curry SH, Rowlands CA, Manara AR, Lewis T, Moss T,et al. Event-related potentials-neurophysiological tools for predictingemergence and early outcome from traumatic coma. Intensive CareMed 1996;22:39–46.

Kircher TT, Rapp A, Grodd W, Buchkremer G, Weiskopf N, Lutzen-berger W, et al. Mismatch negativity responses in schizophrenia: acombined fMRI and whole-head MEG study. Am J Psychiatry2004;161:294–304.

Korpilahti P, Lang HA. Auditory ERP components and mismatchnegativity in dysphasic children. Electroenceph clin Neurophysiol1994;91:256–64.

Kujala T, Naatanen R. The mismatch negativity in evaluating centralauditory dysfunction in dyslexia. Neurosci Biobehav Rev2001;25:535–43.

Kujala T, Karma K, Ceponiene R, Belitz S, Turkkila P, Tervaniemi M,et al. Plastic neural changes and reading improvement caused byaudiovisual training in reading-impaired children. Proc Natl Acad SciUSA 2001;98:10509–14.

Lang H, Nyrke T, Ek M, Aaltonen O, Raimo I, Naatanen R. Pitchdiscrimination performance and auditory event-related potentials. In:Brunia CHM, Gaillard AWK, Kok A, Mulder G, Verbaten MN,editors. Psychophysiological brain research. Tilburg: Tilburg Univer-sity Press; 1990. p. 294–8.

Light GA, Braff DL. Mismatch negativity deficits are associated with poorfunctioning in schizophrenia patients. Arch Gen Psychiatry2005;62:127–36.

Michie PT. What has MMN revealed about the auditory system inschizophrenia? Int J Psychophysiol 2001;42:177–94.

S. Pakarinen et al. / Clinical Neurophysiology 118 (2007) 177–185 185

Michie PT, Innes-Brown H, Todd J, Jablensky AV. Duration mismatchnegativity in biological relatives of patients with schizophreniaspectrum disorders. Biol Psychiatry 2002;52:749–58.

Michie PT, Budd TW, Todd J, Rock D, Wichmann H, Box J, et al.Duration and frequency mismatch negativity in schizophrenia. ClinNeurophysiol 2000;111:1054–65.

Morlet D, Bouchet P, Fischer C. Mismatch negativity and N100monitoring: potential clinical value and methodological advances.Audiol Neurootol 2000;5:198–206.

Naatanen R. The role of attention in auditory information processing asrevealed by event-related potentials and other brain measures ofcognitive function. Behav Brain Sci 1990;13:201–88.

Naatanen R. Attention and brain function. Hillsdale (NJ): LawrenceErlbaum; 1992.

Naatanen R, Winkler I. The concept of auditory stimulus representationin cognitive neuroscience. Psychol Bull 1999;125:826–59.

Naatanen R, Pakarinen S, Rinne T, Takegata R. The mismatch negativity(MMN): towards the optimal paradigm. Clin Neurophysiol2004;115:140–4.

Naatanen R, Gaillard AW, Mantysalo S. Early selective-attention effecton evoked potential reinterpreted. Acta Psychologica1978;42:313–29.

Naatanen R, Paavilainen P, Tiitinen H, Jiang D, Alho K. Attention andmismatch negativity. Psychophysiology 1993a;30:436–50.

Naatanen R, Schroger E, Karakas S, Tervaniemi M, Paavilainen P.Development of a memory trace for a complex sound in the humanbrain. Neuroreport 1993b;4:503–6.

Naatanen R, Tervaniemi M, Sussman E, Paavilainen P, Winkler I.‘‘Primitive intelligence’’ in the auditory cortex. Trends Neurosci2001;24:283–8.

Nousak JM, Deacon D, Ritter W, Vaughan Jr HG. Storage ofinformation in transient auditory memory. Brain Res Cogn BrainRes 1996;4:305–17.

Novak GP, Ritter W, Vaughan Jr HG, Wiznitzer ML. Differentiation ofnegative event-related potentials in an auditory discrimination task.Electroenceph clin Neurophysiol 1990;75:255–75.

Paavilainen P, Tiitinen H, Alho K, Naatanen R. Mismatch negativity toslight pitch changes outside strong attentional focus. Biol Psychol1993;37:23–41.

Pekkonen E, Jousmaki V, Partanen J, Karhu J. Mismatch negativity areaand age-related auditory memory. Electroenceph clin Neurophysiol1993;87:321–5.

Pekkonen E, Hirvonen J, Jaaskelainen IP, Kaakkola S, Huttunen J.Auditory sensory memory and the cholinergic system: implications forAlzheimer’s disease. Neuroimage 2001;14:376–82.

Pekkonen E, Jousmaki V, Kononen M, Reinikainen K, Partanen J.Auditory sensory memory impairment in Alzheimer’s disease: anevent-related potential study. Neuroreport 1994;5:2537–40.

Pekkonen E, Rinne T, Reinikainen K, Kujala T, Alho K, Naatanen R.Aging effects on auditory processing: an event-related potential study.Exp Aging Res 1996;22:171–84.

Riekkinen Jr P, Paakkonen A, Karhu J, Partanen J, Soininen H, LaaksoM, et al. THA disrupts mismatch negativity in Alzheimer disease.Psychopharmacology 1997;133:203–6.

Sams M, Paavilainen P, Alho K, Naatanen R. Auditory frequencydiscrimination and event-related potentials. Electroenceph clin Neu-rophysiol 1985;62:437–48.

Schroeder MM, Handelsman L, Torres L, Dorfman D, Rinaldi P,Jacobson J, et al. Early and late cognitive event-related potentialsmark stages of HIV-1 infection in the drug-user risk group. BiolPsychiatry 1994;35:54–69.

Tiitinen H, May P, Naatanen R. The transient 40-Hz response, mismatchnegativity, and attentional processes in humans. Prog Neuropsycho-pharmacol Biol Psychiatry 1997;21:751–71.

Tiitinen H, May P, Reinikainen K, Naatanen R. Attentive noveltydetection in humans is governed by pre-attentive sensory memory.Nature 1994;372:90–2.

Umbricht D, Koller R, Schmid L, Skrabo A, Grubel C, Huber T, et al.How specific are deficits in mismatch negativity generation toschizophrenia? Biol Psychiatry 2003;53:1120–31.

Umbricht D, Krljes S. Mismatch negativity in schizophrenia: a meta-analysis. Schizoph Res 2005;76:1–23.

Vanhanen M, Karhu J, Koivisto K, Paakkonen A, Partanen J, Laakso M,et al. ERPs reveal deficits in automatic cerebral stimulus processing inpatients with NIDDM. Neuroreport 1996;7:2767–71.

Winkler I, Kushnerenko E, Horvath J, Ceponien„ R, Fellman V,Huotilainen M, et al. Newborn infants can organize the auditoryworld. Proc Natl Acad Sci USA 2003;100:11812–5.