Differentiating ERAN and MMN: An ERP study

NEUROPHYSIOLOGY, BASIC AND CLINICAL Differentiating ERAN and MMN: An ERP tudy Stefan Koelch, 1,CA Thoma C. Gunter, 1 Erich SchroÈger, 2 Mari Tervaniemi, 3 Daniela Sammler 1,2 and Angela D. Friederici 1 1 Max Planck Intitute of Cognitive Neurocience, Stephantr. 1a, D-04103 Leipzig; 2 Intitute of General Pychology, Leipzig, Germany; 3 Cognitive Brain Reearch Unit, Helinki, Finland CA Correponding Author Received 14 February 2001; accepted 27 February 2001 In the preent tudy, the early right-anterior negativity (ERAN) elicited by harmonically inappropriate chord during litening to muic wa compared to the frequency mimatch negativity (MMN) and the abtract-feature MMN. Reult revealed that the amplitude of the ERAN, in contrat to the MMN, i peci cally dependent on the degree of harmonic appropriatene. Thu, the ERAN i correlated with the cognitive proceing of complex rule-baed information, i.e. with the application of muic±yntactic rule. Moreover, reult howed that the ERAN, compared to the abtract-feature MMN, had both a longer latency, and a larger amplitude. The combined nding indicate that ERAN and MMN re ect different mechanim of pre-attentive irregularity detection, and that, although both component have everal feature in common, the ERAN doe not eaily t into the claical MMN framework. The preent ERP thu provide evidence for a differentiation of cognitive procee underlying the fat and pre-attentive proceing of auditory information. NeuroReport 12:1385±1389 & 2001 Lippincott William & Wilkin. Key word: Auditory proceing; EEG; ERAN; ERP; MMN; Muic INTRODUCTION Accurate pitch perception i a prerequiite for the proceing of melodic, harmonic, and proodic apect of both language and muic. Recently, neural dynamic underlying pitch proceing within a muical context have been extenively invetigated by recording the mimatch negativity (MMN) [1,2], the right anterior-temporal negativity (RATN) [3], and the early right-anterior negativity (ERAN) [4±6], which are component of the auditory event-related potential (ERP). The preent tudy aimed at differentiating MMN and ERAN. In previou tudie invetigating the perception of major±minor (i.e. Wetern) tonal muic, harmonically inappropriate chord preented within a muical chord equence elicited a negativity in the ERP which wa maximal around 200 m and right-anteriorly predominant [4,5]. Due to it early latency with repect to the RATN, and due to it imilarity to the early left anterior negativity (ELAN) [7], thi ERP component wa termed the early right-anterior negativity, or ERAN. With repect to their functional igni cance, the RATN and the ERAN are uggeted to re ect the proceing of muical yntax, wherea the ELAN i known to be elicited by word category violation and therefore taken to re ect yntactic language proceing. In the tudie of Koelch et al. [4±6], harmonically inappropriate chord did not repreent a phyical deviancy with repect to the preceding chord. Thu, the ERAN i not a frequency MMN which i known to be elicited only when a deviant tone, or chord, i preceded by a few tandard tone or chord with identical frequency [8,9]. Notably, both harmonically appropriate and inappropriate chord were cononant, major chord. It wa only the urrounding muical context that made ome chord with repect to principle and rule decribed by the theory of harmony inappropriate (ee below). Neverthele, the ERAN i reminicent of the MMN: (a) both ERAN and MMN have a imilar time-coure and calp-ditribution, (b) the amplitude of both ERAN and MMN increae with the amount of violation and are connected to behavioral dicrimination performance, and (c) both MMN and ERAN can be elicited pre-attentively [5,8,9]. Moreover, in previou experiment the ERAN wa elicited in a paradigm which ha imilaritie to the auditory oddball paradigm, a paradigm often ued to elicit an MMN. In the preent tudy, chord equence were preented to the participant, each equence coniting of ve chord, and one equence directly following the other (middle of Fig. 1). Mot of the equence conited of in-key chord only, but infrequently chord at the third or at the fth poition of the equence were Neapolitan ixth chord. Neapolitan chord (in C major: f±a at±d at; Fig. 1) contain out-of-key note (in C major: a at and d at) with repect to the harmonic context etablihed by the preceding in-key chord. Following the theory of harmony, 0959-4965 & Lippincott William & Wilkin Vol 12 No 7 25 May 2001 1385

S. KOELSCH ET AL. abtract feature MMN: chord deviation: frequency MMN Fig. 1. Example of timuli. In all block, timuli were preented with the ame time-coure, loudne, and probability of deviant event (deviant event are indicated by the arrow). etc. Neapolitan may be conidered a a variation of the ubdominant (a minor ubdominant with a minor ixth intead of a fth). Neapolitan at the fth poition are perceived a more inappropriate than when they are preented at the third poition for two reaon: (a) becaue of the muical context buildup, the tonal expectancie of litener are more peci c at the end of a equence [10,11], and (b) ince Neapolitan function in muic- theory a ubdominant variation, a Neapolitan at the third poition i fairly uitable (ince a ubdominant i appropriate), wherea a Neapolitan at the fth poition (where a tonic chord i appropriate only) i inappropriate; it ha been hown that harmonically appropriate chord function are expected to a higher degree compared to inappropriate chord [10,11]. Notably, harmonic expectancie of litener follow principle which correpond to the harmonic relation of different chord (and key, repectively). Thee relationhip form the bai of the major±minor tonal ytem and are decribed by the theory of harmony. The principle which govern harmonic expectancie have been decribed in detail a a hierarchy of harmonic tability [11]. In previou muic experiment employing an experimental paradigm imilar to that of the preent tudy [4±6] it could be hown that the degree of harmonic incongruity (i.e. the degree of harmonic expectancy violation) i re ected in the amplitude of the ERAN: The ERAN i maller when elicited by Neapolitan at the third poition of a chord equence compared to the fth poition. That i, the elicitation and amplitude modulation of the ERAN can be explained on the bai of muic theory, or, in other word, the ERAN may re ect cognitive procee which refer to a complex rule ytem (which may be taken a muical yntax [4,6,10±12]). Thi would contrat the MMN, which i known to be elicited either by phyical deviance, or by rather imple abtract feature deviance which do not refer to a ytem of complex rule. However, although the ERAN cannot be a frequency MMN (ee above), the ERAN could in principle be an MMN elicited by the abtract feature `harmonically appropriate/inappropriate' [13±16]. If o, the amplitude modulation of the ERAN (third veru fth poition of a chord equence) would not be due to any complex rule-baed proceing of harmonic information, but merely be dependent on the poition within a timulu train (i.e. within a equence of acoutic event), regardle of the buildup of a muical context. In order to tet whether the ERAN i an abtract feature MMN, participant of the preent tudy were preented with acoutic event containing an abtract feature (tone pair raiing (tandard) or falling (deviant) in pitch, ee Fig. 1), but not building up a context toward the end of each timulu equence. If the procee re ected in the ERAN are the ame a thoe re ected in the MMN, the amplitude of the abtract feature MMN hould, like the amplitude of the ERAN, increae toward the end of a timulu equence. Three block were conducted: an abtract feature MMN block, a block with chord equence, and a frequency MMN block (with ingle tone, Fig. 1). In all three block, timuli were preented with the ame time-coure and the ame probability of deviant event. MATERIALS AND METHODS Subject: Twenty-eight right-handed and normal-hearing ubject (aged 19±28 year, mean 23.4, 15 of them female) participated in the experiment. Subject were non-muician, i.e. they had never learned to play an intrument or profeional inging, and they did not have any pecial muical education beide normal chool education. Stimuli: All timuli were played under computerized control via MIDI with a piano ound on a Roland JV-2080 yntheizer. In all block, each timulu equence conited of ve event, preentation time (PT) of event 1±4 wa 600 m, of the fth event 1200 m (in the rt block, the rt tone of each tone-pair had a PT of 100 m, i.e. PT of the econd tone wa 500 m (1100 m at the fth poition) repectively). Deviant event (in the rt block falling tonepair, in the econd block Neapolitan chord, in the third block tone with different frequency) occurred in ome equence randomly at either the third ( p ˆ 0.2) or fth poition ( p ˆ 0.2). All chord and tone had the ame loundne and the ame decay of loudne, there wa no ilent period between event or equence; one equence directly followed the other (Fig. 1). In each block, the ame deviant event were employed at the third and fth poition, i.e. deviant event were in each block on average phyically identical. Sequence containing a deviant event were alway preceded by a equence excluively coniting of tandard. Chord and tone were played with 55 db SPL. In each block, 255 equence were preented, reulting in a block duration of 15 min. Tak: In all block, ubject were playing a video-game under the intruction to ignore all acoutic timuli. In Block 1 (abtract feature MMN), tandard timuli were ingle-tone pair raiing in pitch, deviant pair were falling in pitch (top of Fig. 1). The pitch difference of two tone of a pair wa one emitone (6% frequency difference). Three different tone pair could occur at the rt poition of a equence, ix at the econd, ix tandard at the third, three deviant at the third, three at the fourth, 1386 Vol 12 No 7 25 May 2001

DIFFERENTIATING ERAN AND MMN: AN ERP STUDY two tandard at the fth, and three deviant at the fth poition (3/6/6(3)/3/2(3)). The MMN wa meaured from the onet of the econd tone of a pair. In block 2 (ERAN), all equence conited of ve chord that began with a tonic-chord. Chord at the econd poition were tonic, ubdominant, mediant, or ubmediant; at the third poition: ubdominant, Neapolitan chord, dominant, or dominant ix-four chord; at the fourth poition: dominant eventh chord; at the fth poition: tonic or Neapolitan chord [17]. All chord were preented in different chording (e.g. with the root, the third, the fth, and the eventh in the top voice), leading to a pool of 108 different chord equence. Importantly, the number of phyically different chord poible at each poition of a equence equalled the number of different tone pair of Block 1 (3/6/6(deviant: 3 different Neapolitan)/3/2(deviant: 3 different Neapolitan)). Part-writing wa according to the claical rule of harmony [17]. In block 3 (frequency MMN), Standard were ingle tone with a frequency of 440 Hz, deviant tone had a frequency of 496 Hz. Data analyi: The EEG wa recorded with noe-reference from 41 electrode of the extended 10-20 ytem (ampling rate 250 Hz). All EEG data were ltered off-line with a bandpa lter (0.25±25 Hz, 1001 point, FIR). Artifact caued by drift or body movement were eliminated by rejecting EEG data of all block whenever the tandard deviation within any 600 m or 200 m interval of all data. 25 ìv at any electrode. Eye artifact were rejected whenever the.d. within any 200 m interval of all data exceeded 30 ìv at either the vertical or the horizontal EOG. Baeline of ERP wa 50 to 0 m relative to timulu onet. ERP were analyzed by repeated meaure ANOVA a univariate tet of hypothee for within ubject effect. Two anterior region of interet (ROI) were computed: left (mean of, FC3, F5, FC5, A, AF7) and right (mean of, FC4, F6, FC6, A, AF8). If not eparately indicated, ANOVA were conducted with factor condition (tandard 3 deviant), poition within the equence (third 3 fth poition), and hemiphere (left 3 right ROI). RESULTS In the abtract feature MMN block, deviant tone pair preented at both the third and the fth poition elicited an abtract feature MMN (Fig. 2, left). The latency of the MMN wa at both the third and the fth poition around 160 m (meaured from the onet of the econd tone of the abtract feature MMN ERAN to chord devation frequency MMN 3rd 5th 3rd 5th 3rd 5th V tandard deviant Fig. 2. ERP elicited at frontal electrode ite by timuli at the third (top row) and fth (econd row) poition, eparately for the abtract feature MMN block, the block with chord equence, and the frequency MMN block. Vertical line indicate the onet of the deviant timulu (in the abtract feature MMN block: econd tone of a tone pair). Bottom row: Potential map of abtract feature MMN (tandard ubtracted from deviant), ERAN (harmonic appropriate chord ubtracted from Neapolitan chord), and frequency MMN (tandard ubtracted from deviant), eparately for third and fth poition. Map were calculated uing the data from all 41 electrode and interpolated over time window from 125 to 185 m (abtract feature MMN), 170±230 m (ERAN) and 90±150 m (frequency MMN), polarity inverion are bordered by thick line. In contrat to the ERAN, the MMN did not differ in amplitude between poition 3 and 5, indicating that the ERAN re ect context-dependent muical proceing. Moreover, the ERAN elicited at the fth poition had a later latency, but larger amplitude than the abtract feature MMN. Vol 12 No 7 25 May 2001 1387

S. KOELSCH ET AL. tone pair), the amplitude of the MMN did not differ between both poition (ee alo Fig. 3); no polarity inverion wa viible at matoidal ite. An ANOVA conducted for a time window from 125 to 185 m revealed an effect of condition (F(1,26) ˆ 15.54, p, 0.0005), an interaction between factor condition and hemiphere (F(1,26) ˆ 9.12, p, 0.006), and no interaction between factor condition and poition. In the econd block (chord-equence), Neapolitan chord at both the third and the fth poition elicited an ERAN with a latency around 200 m, the ERAN at the fth poition howed a clear polarity inverion at matoidal ite (Fig. 2, middle, the polarity inverion i indicated in the potential map). The ERAN wa ditinctly larger at the fth than at the third poition (ee alo Fig. 3). The amplitude of the ERAN elicited at the fth poition wa clearly larger than the amplitude of the abtract feature MMN elicited at the fth poition in the rt block. An ANOVA for the data of the econd block (time window 170±230 m) revealed an effect of condition (F(1,26) ˆ 38.53, p, 0.0001), an interaction between factor condition and hemiphere (F(1,26) ˆ 10.24, p, 0.005), and an interaction between factor condition and poition (F(1,26) ˆ 9.32, p, 0.006). ANOVA with factor condition conducted eparately for the third and the fth poition revealed an effect of condition at both third and fth poition (third poition: F(1,26) ˆ 4.74, p, 0.05; fth poition: F(1,26) ˆ 36.33, p, 0.0001). An ANOVA of the data from the fth poition from block 1 and 2 with factor condition and block (1 3 2), teting the amplitude difference between the abtract feature MMN (125±185 m) and the ERAN (170± 230 m), revealed an interaction between the two factor (F(1,26) ˆ 12.43, p, 0.002). In the frequency MMN block, deviant tone elicited a frequency MMN at both the third and the fth poition, with a latency of around 100 m, with clear polarity inverion at matoidal ite, and with virtually the ame amplitude at the third and fth poition (Fig. 2, right, and Fig. 3). An ANOVA for the time window from 90 to 150 m revealed an effect of condition (F(1,26) ˆ 62.27, p, 0.0001), with no interaction between factor condition and poition (no interaction wa yielded between factor condition and hemiphere, although a light right hemipheric preponderance i viible in the potential map). 4 3 2 1 0 3rd 5th 3rd 5th 3rd 5th abtract feature ERAN to chord frequency MMN devation MMN Fig. 3. Amplitude of abtract-feature MMN (left), ERAN (middle) and MMN (right), eparately for poition 3 (white bar) and 5 (gray bar). Latency of the abtract feature MMN wa 160 m, of the ERAN 200 m, and of the frequency MMN 100 m. DISCUSSION The preent tudy invetigated the pre-attentively activated neural mechanim of an auditory deviance detection by comparing the ERP repone to an abtract ound change, a change of harmonic appropriatene of chord, and a pitch change. Only the ERAN (elicited by the harmonically incongruou chord in the econd block) differed igni cantly in amplitude between third and fth poition of a equence. Neither abtract feature MMN nor frequency MMN (elicited in a non-muical context) howed an amplitude modulation between thee poition. Since only the in-key chord of the econd block built up a muical context toward the end of each equence, thi nding indicate that the amplitude of the ERAN i peci cally correlated with the degree of harmonic incongruity induced by a preceding muical context. The preent reult thu demontrate that the ERAN re ect proceing of auditory information that refer to a complex rule-baed ytem, namely rule inherent in the major±minor tonal ytem which are far more complex than thoe known to elicit a phyical or an abtract feature MMN. Notably, the ERAN elicited at the fth poition had a longer latency, and a larger amplitude than the abtract feature MMN. The longer latency could eaily be explained within the claical MMN framework if one aume that the harmonic incongruitie are more complex, and thu more dif cult to differentiate than the deviant tone-pair: The MMN i known to have a longer latency when timuli are more dif cult to differentiate [8,18]. Then, however, the ERAN hould alo be maller in amplitude than the MMN [18]. Thi wa not the cae: on the contrary, the ERAN wa ditinctly larger than the abtract feature MMN, indicating that the ERAN doe not re ect the ame cognitive procee that underlie the MMN. The preent ERP thu provide evidence for a differentiation of cognitive procee underlying a pre-attentive proceing of auditory information in (a) enory memory procee on the one ide, and (b) relatively higher cognitive proceing of complex rule-baed information on the other. The fact that the ERAN peci cally correlate with the proceing of complex rule of major±minor tonal muic juti e the particular term ERAN for the effect oberved. It i important to note that MMN, ERAN, and ELAN all belong to a familiy of peri-ylvian (ee alo below) negativitie which re ect the proceing of irregularitie of auditory input. The preent tudy upport the notion that there are, however, coniderable difference with repect to the cognitive procee and the neuronal tructure mediating the proceing of a phyical irregularity like frequency on the one hand, and a language or muic yntactic violation on the other. The fact that an MMN can alo be elicited by abtract feature might indicate that all component (phyical MMN, abtract feature MMN, ERAN and ELAN) re ect tage on a continuum from rather imple (phyical) to fairly complex (yntactic) auditory feature proceing; thi conideration might ugget an expanion of the claical MMN framework. Notably, thi conideration i upported by functional neuroanatomical nding, which indicate that the more imple feature eem primarily to be generated in (or in the cloe vicinity of) primary auditory cortical area, with relatively mall contribution from the frontal area [19±22]. In contrat, the proceing 1388 Vol 12 No 7 25 May 2001

DIFFERENTIATING ERAN AND MMN: AN ERP STUDY of feature which refer to a complex rule ytem (a re ected in ERAN and ELAN) appear to involve more frontal, and le primary auditory tructure [6,23] (a recent tudy from Koelch et al. [6] revealed that the ERAN i mainly generated in the inferior part of BA 44 bilaterally). Finally, the ERAN wa (a the MMN) elicited under a condition in which participant were intructed to ignore the chord, upporting the hypothei that the ERAN re ect, like the MMN [8], pre-attentive neural procee [5]. Taken together, the preent reult thu upport the hypothei of a high adaptability and exibility of preattentive procee in the human brain [4]. CONCLUSION The preent reult demontrate that the ERAN re ect cognitive operation connected to the proceing of complex rule-baed muical information, in contrat to the MMN, which i known to re ect mainly enory memory operation. Thi nding i important for everal reaon. Firt, it indicate that partly different neuronal procee underlie the generation of ERAN and MMN (although both component hare everal feature). Since both MMN and ERAN can be elicited pre-attentively, the preent data provide evidence for a differentiation of fat and preattentive neural mechanim underlying auditory deviance detection in the human brain. Second, although the ERAN doe not eaily t into the claical MMN framework, both component t into one concept if one conider that both MMN and ERAN belong to a family of periylvian negativite that mediate the proceing of irregularitie of auditory input. With thi repect, the preent reult upport the hypothei of a trong adaptability and exibility of fat and automatic cognitive procee in the human brain, probably indicating that the claical MMN framework might be expanded (at leat with repect to the proceing of major±minor tonal muic) to the proceing of complex, or yntactic, rule. Third, reult upport the hypothei that proceing of muical yntax a re ected in the ERAN i proceed pre-attentively [5]. REFERENCES 1. Tervaniemi M, Medvedev SV, Alho K et al. Hum Br Mapp 10, 74±79 (2000). 2. Alain C, Achim A and Wood DL. Pychophyiology 36, 737±744 (1999). 3. Patel AD, Gibon E, Ratner J et al. J Cogn Neuroci 10, 717±733 (1998). 4. Koelch S, Gunter T, Friederici AD et al. J Cogn Neuroci 12, 520±541 (2000). 5. Koelch S, SchroÈger E and Gunter T. ubmitted (2001). 6. Koelch S, Mae B and Fiederici AD. Neuroimage 11, 56 (2000). 7. Hahne A and Friederici AD. J Cogn Neuroci 11, 194±205 (1999). 8. NaÈaÈtaÈnen R. Attention and Brain Function. Hilldale, NJ: Erlbaum, 1992. 9. SchroÈger E. Behav Re Method Intr Comp 30, 131±145 (1998). 10. Krumhanl C and Keler E. Pych Rev 89, 334±368 (1982). 11. Bharucha J and Krumhanl C. Cognition 13, 63±102 (1983). 12. Swain J. Muical Language. UK: Norton, 1997. 13. Saarinen J, Paavilainen P, SchroÈger E et al. Neuroreport 3, 1149±1151 (1992). 14. Tervaniemi M, Maury S and NaÈaÈtaÈnen R. Neuroreport 5, 844±846 (1994). 15. Paavilainen P, Saarinen J, Tervaniemi M et al. Pychophyiology 9, 243±249 (1995). 16. Paavilainen P, Jaramillo M and NaÈaÈtaÈnen R. Pychophyiology 35, 483±487 (1998). 17. Hindemith P. Unterweiung im Tonatz, 1. Theoreticher Teil. Mainz: Schott; 1940. 18. Tiitinen H, May P, Reinkainen K and NaÈaÈtaÈnen R. Nature 372, 90±92 (1994). 19. Alho K. Ear Hear 16, 38±51 (1995). 20. Giard M, Perrin F and Pernier J. Pychophyiology 27, 627±640 (1990). 21. Alain C, Wood DL and Knight RT. Brain Re 812, 23±37 (1998). 22. Opitz B, Mecklinger A, von Cramon DY et al. Pychophyiology 36, 142±147 (1999). 23. Friederici AD, Wang Y, Herrmann C et al. Hum Br Map 11, 1±11 (2000). Acknowledgement: The work wa upported by the Leibniz Science Prize awarded to A.D. Friederici by the German Reearch Foundation. Full color map, ound example of the timulation, and abtract with gure of the ubmitted article are available at www.tefan-koelch.de Vol 12 No 7 25 May 2001 1389