Neural Mechanisms of Involuntary Attentionto Acoustic Novelty and Change

Carles EsceraUniversity of Barcelona Catalonia

Kimmo AlhoUniversity of Helsinki and University of Barcelona Catalonia

Istvaacuten WinklerHungarian Academy of Sciences and University of Helsinki

Risto NaumlaumltaumlnenUniversity of Helsinki

Abstract

n Behavioral and event-related brain potential (ERP) measureswere used to elucidate the neural mechanisms of involuntaryengagement of attention by novelty and change in the acousticenvironment The behavioral measures consisted of the reac-tion time (RT) and performance accuracy (hit rate) in a forced-choice visual RT task where subjects were to discriminatebetween odd and even numbers Each visual stimulus waspreceded by an irrelevant auditory stimulus which was ran-domly either a ldquostandardrdquo tone (80) a slightly higher ldquodeviantrdquotone (10) or a natural ldquonovelrdquo sound (10) Novel soundsprolonged the RT to successive visual stimuli by 17 msec ascompared with the RT to visual stimuli that followed standardtones Deviant tones in turn decreased the hit rate but did notsignicantly affect the RT In the ERPs to deviant tones themismatch negativity (MMN) peaking at 150 msec and a second

negativity peaking at 400 msec could be observed Novelsounds elicited an enhanced N1 with a probable overlap bythe MMN and a large positive P3a response with two differentsubcomponents an early centrally dominant P3a peaking at230 msec and a late P3a peaking at 315 msec with a right-fron-tal scalp maximum The present results suggest the involve-ment of two different neural mechanisms in triggeringinvoluntary attention to acoustic novelty and change a tran-sient-detector mechanism activated by novel sounds andreected in the N1 and a stimulus-change detector mechanismactivated by deviant tones and novel sounds and reected inthe MMN The observed differential distracting effects byslightly deviant tones and widely deviant novel sounds supportthe notion of two separate mechanisms of involuntary atten-tion n

INTRODUCTION

It is a common experience that even during intensivetask performance our attention can be involuntarily en-gaged by acoustic changes occurring unexpectedly inthe environment The present study used a combinationof behavioral and event-related brain potential (ERP)measures to elucidate the neural mechanisms involvedin such involuntary attention Irrelevant novel soundsand slight sound changes were used to cause involuntaryattention to auditory stimuli in a forced-choice visualdiscrimination task In this task each visual stimulus waspreceded by a task-irrelevant auditory stimulus that wasa repetitive tone occasionally replaced by a slightlyhigher (ldquodeviantrdquo) tone or by a natural ldquonovelrdquo sound Itwas predicted that these unexpected acoustic eventswould prolong the reaction time (RT) to visual stimuli

copy 1998 Massachusetts Institute of Technology Journal of Cognitive Neuroscience 105 pp 590ndash604

and decrease the performance accuracy (hit rate) byengaging the subjectrsquos attention ERPs elicited by theauditory-visual stimulus pairs were simultaneously re-corded in order to reveal neural events underlying theexpected behavioral effects

Behavioral evidence of attention switching to irrele-vant stimuli was rst reported by Cherry (1953) Hissubjects noticed changes in the voice delivering an un-attended speech message to one ear while they wereattending to the speech message presented to the oppo-site ear Further evidence was provided by RTs to probestimuli in a secondary task (Dawson 1990 Posner 1978)These RTs were prolonged when target stimuli werepreceded by unexpected irrelevant changes in the re-petitive stimuli delivered in the primary task indicatingaccording to the authors that processing resources wereengaged by these stimulus changes occurring in the

primary task (Dawson Fillion amp Schell 1989 FillionDawson Schell amp Hazlett 1991 Siddle 1991 WoodwardBrown March amp Dawson 1991)

Involuntary attention shifts were originally explainedby the orienting-reex (OR) theory (Sokolov 1963) pro-posing that a neuronal model is built from the repetitivefeatures of the external environment inhibiting the ORto identical but not to different stimuli Further Oumlhman(1979 1992) proposed that a ldquocallrdquo for reallocation ofcentral processing resources is issued by preattentivemechanisms detecting signicant changes in incomingstimuli

Attention switching can also be elicited by stimulusonsets (regardless of repetitionchange) in particular ifthey appear after long ldquosilentrdquo intervals and by offsets ofcontinuous stimulation (Folk Remington amp Johnston1992 Gati amp Ben-Shakhar 1990 Hikosaka Miyauchi ampShimojo 1993 Oumlhman 1979 Theeuwes 1994 Yantis1993 Yantis amp Hillstrom 1994) According to Naumlaumltaumlnenrsquos(1990 1992) model stimulus onsets and offsets activatea transient-detector system reected in the auditory mo-dality in the supratemporal and nonspecic componentsof the N1 wave peaking at about 100 msec from stimu-lus onset (Naumlaumltaumlnen amp Picton 1987) The supratemporalN1 component is generated by bilateral dipoles locatedin the auditory cortices on the supratemporal plane(Giard et al 1994 Vaughan amp Ritter 1970) and its am-plitude increases with an increasing interval from theprevious sound (Davis Mast Yoshie amp Zerlin 1966 HariKaila Katila Tuomisto amp Varpula 1982 Maumlkelauml Hari ampLeinonen 1988 Ritter Vaughan amp Costa 1968) Thiscomponent is sensitive in particular to the transientaspects of stimulation (Davis amp Zerlin 1966 McMillan1973 Pfefferbaum Buchsbaum amp Gips 1971) whichsuggests that the neuronal process generating the su-pratemporal N1 triggers an attention-capturing signal forconscious perception of the stimulus (Naumlaumltaumlnen 19901992) The nonspecic N1 component is even moresensitive to interstimulus interval manipulations than thesupratemporal component (Naumlaumltaumlnen amp Picton 1987)and in contrast to the supratemporal N1 component itcan be elicited also by stimuli of other modalities (Leh-tonen 1973 Velasco amp Velasco 1986 Velasco Velascoamp Olvera 1985) The function of the nonspecic N1neural generators probably is to trigger a transientarousal burst facilitating sensory and motor responses tothe eliciting stimulus (Naumlaumltaumlnen amp Picton 1987) How-ever Giard et al (1994) described quite recently a frontalsubcomponent of the N1 wave This subcomponentmight be a better candidate for the attention-switchingfunction originally attributed to the supratemporal-N1generator

The stimulus-change detector mechanism postulatedby the OR theory is represented in Naumlaumltaumlnenrsquos (19901992) model as the neuronal process generating themismatch negativity (MMN) The MMN elicited by anydiscernible change in the physical features of a repeti-

tive even unattended sound (eg Naumlaumltaumlnen 1992) isgenerated mainly in the supratemporal auditory cortex(for an overview see Alho 1995) as revealed by ERP-source modeling (Giard et al 1995 Giard Perrin Pernieramp Bouchet 1990 Scherg Vajsar amp Picton 1989) neuro-magnetic data (Hari et al 1984) and intracranial record-ings in animals (Cseacutepe Karmos amp Molnaacuter 1987 JavittSchroeder Steinschneider Arezzo amp Vaughan 1992) aswell as in humans (Halgren Baudena Clarke Heit et al1995 Kropotov et al 1995) According to Naumlaumltaumlnenrsquos(1990 1992) model the physical features of auditorystimuli are fully analyzed and encoded into neural tracesof auditory sensory memory the MMN being automat-ically elicited when the auditory input does not matchwith the neuronal trace formed by the repetitive stan-dard stimulus (Naumlaumltaumlnen Paavilainen Alho Reinikainenamp Sams 1989) while the trace still is in an ldquoactiverdquo state(Cowan Winkler Teder amp Naumlaumltaumlnen 1993) Further themodel proposes that the MMN-generating process is in-volved in triggering a signal for attention switching afterthe automatic preperceptual detection of auditorychange (Naumlaumltaumlnen 1990)

Converging evidence supports the notion that theprocess generating the MMN may be associated withinvoluntary attention switching First the MMN is ofpreattentive nature because it does not depend onwhether the subject is engaged in an easy or a difcultvisual task (Alho Woods Algazi amp Naumlaumltaumlnen 1992 Dun-can amp Kaye 1987)1 Second in addition to its supratem-poral generators MMN appears to get a furthercontribution from the frontal lobe activity (Giard et al1990 Molnaacuter Skinner Cseacutepe Winkler amp Karmos 1995)a critical structure for controlling both voluntary andinvoluntary attention (Fuster 1989) Third the MMNtends to be accompanied especially in the beginning ofthe stimulus sequence by autonomous nervous system(ANS) responses even when stimulus change is of verysmall magnitude (Lyytinen Blomberg amp Naumlaumltaumlnen 1992)These autonomic responses indicate apparently atten-tion switching triggered by the automatic change-detec-tion process reected by the MMN

The strongest evidence supporting the causal relation-ship between the neural process generating the MMNand involuntary attention switching was recently ob-tained by Schroumlger (1996) He found that MMN-elicitingdeviant tones occurring in a repetitive sequence of stan-dard tones presented to the unattended ear prolongedthe RT and decreased the performance accuracy to sub-sequent target tones delivered to the other ear Thiseffect was observed only for irrelevant tones precedingtarget stimuli by a short interval of 200 msec but not forlonger intervals of 560 msec and was strengthened withincreasing frequency difference between the deviantand standard tones Both of these ndings support thefunctional role of the MMN-generating process in trigger-ing this attention switch

Another ERP component that has been associated

Escera et al 591

with orienting of attention is the P3a This positive com-ponent is elicited by irrelevant rare tones and novelsounds in a sequence of repetitive standard tones andcan be distinguished from the P300 or P3b componentto target stimuli by its shorter latency and its morefrontal scalp distribution (Courchesne Hillyard amp Galam-bos 1975 Ford Roth amp Kopell 1976 Squires Squires ampHillyard 1975) The association of the P3a-generatingprocess with the orienting response is supported by itselicitation by widely deviant or novel sounds and also bythe frontal-lobe and hippocampal contributions to itsgeneration (Knight 1984 1996) these brain regions be-ing involved also in the orienting response (Fuster 1989Sokolov 1975) The role of the P3a generator process ininvoluntary engagement of attention is further sup-ported by studies showing delayed RTs to target stimulifollowing irrelevant novel sounds (Grillon CourchesneAmeli Geyer amp Braff 1990 Woods 1992) When novelsounds occurred in the attended sequences containingthe target stimuli the P3a elicited was larger in ampli-tude as compared with the unattended-sequence P3a tonovel sounds and the RT to consecutive targets wasdelayed (Woods 1992) This attentional modulation sug-gests that the P3a reects an actual reorientation ofattention more closely than the MMN or the differentsubcomponents of the N1 which would rather be asso-ciated with the involuntary call for attention

The purpose of the present experiment was to inves-tigate the neural mechanisms of involuntary attention toacoustic novelty and change by analyzing the ERP con-comitants elicited by these irrelevant auditory eventsduring visual performance In the main experimentalcondition (Figure 1) the Auditory-Visual condition (AV)a response was required to each visual stimulus of anauditory-visual stimulus pair Subjects were instructed toignore the auditory stimuli which were a repetitiveldquostandardrdquo tone randomly replaced by a slightly higherldquodeviantrdquo tone (p = 01) or by a natural ldquonovelrdquo sound(p = 01) and to press a response button to even num-bers and another to odd numbers Two additional condi-tions were used as controls In the Visual-alonecondition (Va) the auditory stimuli were omitted andsubjects performed the forced-choice visual RT task Inthe Auditory-alone (Aa) condition the visual stimuliwere omitted and subjects were instructed to concen-trate on reading a book

RESULTS

Performance

In the Va condition the mean hit rate was 915 (SEM plusmn17) and the mean RT was 461 plusmn 11 msec Subjectsmissed the target on an average 45 plusmn 16 of thetrials and pressed the wrong button in 40 plusmn 08 ofthe trials

Performance data in the Va and AV conditions areshown in Figure 2 In the AV condition subjects tendedto respond faster to visual stimuli preceded by a standardtone than to those preceded by no sound (Va condition)(444 plusmn 9 and 461 plusmn 11 msec respectively) although thiseffect did not quite reach statistical signicance (F(19) = 38 p lt 009) The hit error and miss rates weresimilar in the Va condition and in the AV condition aftera standard tone In the AV condition the occurrence ofa deviant tone before the visual stimulus caused a sig-nicant hit rate decrease of 23 (F(1 9) = 650 p lt004) which was probably due to an increased numberof wrong responses to target stimuli (F(1 9) = 478 p lt006 see Figure 2) the number of missing responsesbeing identical after standard and deviant tones A novelsound occurring before a visual stimulus caused an av-erage increase of 17 msec in the RT to visual stimulicompared with the RT to these stimuli that were pre-ceded by standard tones (F(1 9) = 2219 p lt 0002) TheRT to visual stimuli after novel sounds was also sig-nicantly longer than that to visual stimuli followingdeviant tones (F(19) = 1107 p lt 0009)

Figure 1 Experimental conditions In the main condition [Auditory-Visual (AV) top] pairs of stimuli consisting of an auditory stimulusfollowed at 300 msec (onset-to-onset) by a visual stimulus (an arabicnumber between 1 and 8) were presented with an interpair intervalof 1200 msec The auditory stimulus was either a standard tone(80) a deviant tone (10) or a novel sound (10) Subjects wereinstructed to press one response button to odd (50) and the otherto even (50) numbers and to ignore auditory stimulation In theVisual-alone (Va) condition the auditory stimuli were omitted In theAuditory-alone (Aa) condition the visual stimuli were omitted thesubject being instructed to read a self-selected book and to ignorethe auditory stimuli

592 Journal of Cognitive Neuroscience Volume 10 Number 5

Auditory ERPs

In the Aa condition standard tones elicited N1 (meanpeak amplitude - 20 m V at Cz with a mean peak latencyof 93 msec) and P2 deections (34 m V 158 msec at Cz)which were largest over the fronto-central scalp loca-tions (Figure 3a) Deviant tones elicited N1 and P2 wavesat latencies similar to those in response to standardtones and the mismatch negativity (MMN) The MMN

was largest over the frontal electrodes (mean peak am-plitude - 22 m V mean peak latency 153 msec at Fz) andinverted in polarity at the mastoid electrodes (Figure 3b)being signicantly different from zero at F3 Fz and F4in the intervals 100 to 150 msec (F(1 9) = 1031 p lt002) and 150 to 200 msec (F(1 9) = 563 p lt 005) TheMMN was followed by a smaller fronto-centrally distrib-uted negative wave peaking at around 400 msec (Figure3b) Mean amplitude comparisons between ERPs to de-viant and standard tones at Fz and Cz revealed that thissecond negativity was signicant in the 350- to 400-msec(F(1 9) = 756 p lt 003) and in the 400- to 450-msec(F(1 9) = 873 p lt 002) intervals In the AV conditionthe MMN was similar to that obtained in the Aa condition(Figure 4)

In the Aa condition novel sounds elicited an ERPcharacterized by a prominent N1 deection (mean peakamplitude at Cz - 56 m V peak latency 96 msec) and bya long-duration broadly distributed P3a wave (Figure 3as seen in Figure 3b even the deviant tones elicitedsome P3a) The novel-sound N1 was larger than thestandard-tone N1 (F(1 9) = 1099 p lt 001) and wasprobably composed of overlapping N1 and MMN com-ponents The difference between the standard-tone andnovel-sound ERPs extended beyond the N1 latencyrange there being a signicant difference at Fz betweenthe standard and novel ERP amplitudes measured asmean voltages over 120 to 160 msec latency window(F(1 9) = 523 p lt 005 Figure 5)

The present P3a to novel sounds had a double peakover the frontal scalp locations (Figure 3b) suggestingtwo different subcomponents The early part of the P3a(peak latency 230 msec in the grand-mean ERP at Fz)was largest over the central scalp areas appearing withinverted polarity at lateral and posterior scalp sites (Fig-ures 3b and 5) The late part of the P3a (peak latency315 msec in the grand-mean ERP at Fz) in turn waslargest over frontal areas and did not show polarityinversion at any recording site (Figures 3b and 5) Thedifferent scalp distributions of the two P3a subcompo-nents in the Aa condition are shown in Figure 6 (top)

The component structure of the P3a deection wasevaluated by means of two different analyses of variance(ANOVA) for normalized ERP amplitudes (McCarthy ampWood 1985) First an ANOVA was performed to prove apossible scalp-distribution difference between the P2 tostandard tones and the early P3a to novel sounds Thenanother ANOVA was performed to conrm the differentscalp distributions of the two subcomponents of the P3aThe P2 amplitude was identied as the mean voltagebetween 150 and 200 msec from stimulus onset in thestandard-tone ERP The early P3a was measured as themean voltage between 175 and 275 msec in the differ-ence wave obtained by subtracting the ERP to the stan-dard tone from that to the novel sound The late P3a wasmeasured as the mean voltage between 275 and 375msec of the same difference wave A three-factor design

Figure 2 Mean RT (top panel) hit rate (middle panel) and errorrate (bottom panel) in the Visual-alone (Va) condition and in theAuditory-Visual (AV) condition to the visual stimulus occurring afterthe standard tone (std) the deviant tone (dev) or the novel sound(nov) The bars indicate the standard error of mean

Escera et al 593

was used for both analyses Wave (P2 versus early P3afor the rst analysis and early P3a versus late P3a for thesecond analysis) acute Frontality (F7 F3 Fz F4 F8 versus T3C3 Cz C4 T4 versus T5 P3 Pz P4 T6) acute Laterality (F7T3 T5 versus F3 C3 P3 versus Fz Cz Pz versus F4 C4P4 versus F8 T4 T6) The results of the rst analysisrevealed a signicant Wave acute Frontality interaction (F(218) = 509 p lt 003 e = 07547) due to the early P3abeing posterior to the P2 and a signicant Wave acuteLaterality interaction (F(4 36) = 406 p lt 003 e =05469) due to the more predominantly right-hemi-spheric scalp distribution of the P2 than that of the earlyP3a The results of the second analysis revealed that thelate P3a was anterior in distribution to the early P3a(signicant Wave Frontality interaction F(2 18) = 939p lt 0006 e = 07187) and further that the late P3a wasmore preponderant over the right hemisphere than the

early P3a (signicant Wave acute Laterality interactionF(4 36) = 1207 p lt 00002 e = 05740) The secondanalysis also revealed a signicant third-order interaction(Wave acute Frontality acute Laterality F(8 72) = 898 p lt00004 e = 03441) that was due to the predominantlyright-frontal distribution of the late phase of the P3adeection (Figure 6 top)

In the AV condition novel sounds elicited ERPs similarto those elicited by them in the Aa condition (large N1and P3a deections) although the early peak of the P3aseemed to be partly abolished at frontal electrodes (Fig-ure 7) However difference waves (ERP to novel soundsminus ERP to standard tones) revealed that the earlyphase of the P3a was similar in the Aa and AV conditionswhereas the late P3a was enhanced in the AV condition(Figure 7) The statistical scalp-distribution analyses ofthe two P3a phases across the conditions [ANOVA with

Figure 3a Grand-mean ERPsaveraged across subjects tostandard tones (600 Hz) devi-ant tones (700 Hz) and novelsounds in the Auditory-alone(Aa) condition

594 Journal of Cognitive Neuroscience Volume 10 Number 5

four factors Condition (Aa versus AV) acute Wave (early P3aversus late P3a) acute Frontality acute Laterality] conrmed thescalp-distribution difference between the two P3aphases (Wave acute Frontality interaction F(2 18) = 996p lt 0003 e = 08361 Wave acute Laterality F(4 36) = 688p lt 002 e = 03977 Wave acute Frontality acute LateralityF(872) = 981 p lt 00001 acute = 04381) and also revealeda signicant increase of the late P3a amplitude over theright hemisphere in the AV condition (Wave acute Condi-tion acute Laterality F(4 36) = 384 p lt 004 e = 05367)

Visual ERPs

Visual stimuli in the Va condition and those after stan-dard tones in the AV condition elicited almost identicalERPs characterized by P1 N1 P2 N2 and P3 deections(Figure 8) ANOVA for these visual ERP components with

Condition (after standard in AV versus Va) as a functionrevealed no signicant effects of standard tones on visualERPs

Figure 9 shows grand-average ERPs to auditory-visualstimulus pairs in the AV condition The AV stimulus pairselicited an auditory ERP lasting up to 400 msec post-stimulus for novel sounds and up to 300 msec post-stimulus for deviant tones followed by a visual ERPsimilar to that obtained in the Va condition All threetypes of AV stimulus pairs (standard-visual deviant-visualnovel-visual) were also characterized by a long-lastingfronto-polar negativity that was larger to visual stimulifollowing novel sounds than to those following standardor deviant tones (Figure 9) ANOVA for mean amplitudesat Fp1 and Fp2 in consecutive 100-msec latency win-dows yielded signicant effects for the novel-standardcomparison (F(1 9) ranging from 779 to 1196 p lt 003

Figure 3b Grand-average dif-ference waves obtained in theAa condition by subtractingthe ERP to standard tonesfrom that to deviant tones(continuous line) and tonovel sounds (dashed line)

Escera et al 595

in all cases) at latency windows 200 to 300 msec 300to 400 msec and 400 to 500 msec from visual-stimulusonset

Novel sounds and deviant tones enhanced the frontalN2 deection elicited by the subsequent visual stimulusin comparison with the visual N2 elicited after standardtones (Figure 9) This was revealed by two-way ANOVAsusing the Electrode (F3 Fz F4) and the auditory Stimulus(standard versus either deviant or novel) as factors Themean amplitude in the 200- to 300-msec range fromvisual-stimulus onset at Fz was - 19 m V after standardtones - 33 m V after deviant tones (F(1 9) = 641 p lt004) and - 36 m V after novel sounds (F(1 9) = 899 plt 002) The mean amplitude in the 300- to 400-msecinterval was also signicantly larger for visual stimulifollowing deviant tones or novel sounds than for thosefollowing standard tones F(1 9) = 815 p lt 002 for

deviant versus standard F(1 9) = 795 p lt 001 for novelversus standard

Preceding novel sounds and deviant tones also en-hanced the terminal slope of the visual P3b (Figure 9)this was revealed by the signicantly larger mean ampli-tude of the 500- to 600-msec latency window (at P3 Pzand P4) in the ERP to visual stimuli following devianttones than in the ERP to visual stimuli following standardtones (F(19) = 1556 p lt 004) The mean amplitude inthe 400- to 500-msec (F(1 9) = 694 p lt 003) and 500-to 600-msec (F(1 9) = 1199 p lt 0008) intervals was

Figure 4 The MMN was similar in the Auditory-alone (Aa) andAuditory-Visual (AV) conditions as revealed by nonsignicant differ-ences of the mean amplitude of ve consecutive 50-msec intervalsstarting at MMN onset (50 msec) [two-way ANOVAs Condition (Aaversus AV) acute Electrode (F3 Fz F4 C3 Cz C4)] Top Grand-averageERPs to standard and deviant tones at Fz in the two conditionsBottom Deviant-standard difference waves

Figure 5 The component structure of the ERPs to deviant tonesand novel sounds as revealed by the deviant-standard and novel-stan-dard difference waves at Fz and right-mastoid (RM) recordings Devi-ant tones elicited the MMN with a peak latency of 150 msec at Fzand with an inverted polarity at RM (top) Novel sounds elicited alarge N1 deection at Fz that was presumably composed of contribu-tions from two different processes for which the signicant differ-ences between the standard-tone and novel-sound ERPs extendedbeyond the N1 latency range (bottom) The novel sounds also elic-ited a P3a wave with two peaks at Fz suggesting two different sub-components contributing to the recorded P3a an early P3a (peaklatency 230 msec) which inverted polarity at RM and other poste-rior and lateral sites and a late P3a (peak latency 315 msec) whichinverted polarity at no recording site

596 Journal of Cognitive Neuroscience Volume 10 Number 5

also signicantly larger for visual stimuli following novelsounds than after those following the standard tones

DISCUSSION

Engagement of Attention During VisualPerformance

The present results provide evidence that attention dur-ing visual performance can be engaged by shortly pre-ceding unexpected novel sounds or small changes in thetask-irrelevant acoustic environment These results alsorevealed that this phenomenon depends on the intrinsicnature of the eliciting acoustic events In addition the

observed effects on visual performance were associatedwith different patterns of neural events activated bydeviant tones and novel sounds as reected by thecomponent structure of the auditory ERPs elicited

As expected novel sounds prolonged the RT to suc-cessive visual stimuli This result is in agreement with theextensive literature reporting delayed RTs to target stim-uli caused by preceding irrelevant sound changes ornovels sounds (Alho et al 1992 Cherry 1953 Dawsonet al 1989 Fillion et al 1991 Grillon et al 1990Schroumlger 1996 Woods 1992 Woodward et al 1991) andit provides experimental support for the everyday expe-rience that attention can be engaged by irrelevant acous-tic events

Figure 6 The scalp distribu-tion of the P3a to novelsounds Left early P3a rightlate P3a top Auditory-alone(Aa) condition bottom Audi-tory-Visual (AV) condition Thecentral plot shows the novel-standard difference waves atFz in the Aa (thin) and AV(thick) conditions as well asthe 100-msec latency win-dows used to measure thetwo phases of the P3a (e =early P3a l = late P3a)

Escera et al 597

A surprising nding of this study was however thatthe slightly deviant tones degraded the visual perfor-mance accuracy by decreasing the rate of hit responseswhereas the widely deviant novel sounds did not It wasexpected that the amount of distraction caused by devi-ant tones and novel sounds would be proportional tothe magnitude of their deviance from the standard toneand indeed such an effect was found for the RT (17-msecincrease after novel sounds) conrming other recentndings (Jaumlaumlskelaumlinen et al 1996 Schroumlger 1996) How-ever this RT effect was not paralleled by a similar effecton the hit rate Indeed deviant tones were associatedwith a lower hit rate whereas the hit rate after novelsounds was similar to that after standard tones Similarresults were obtained in a related study by Jaumlaumlskelaumlinenet al (1996) and recently replicated for deviant tones(Alho Escera Diacuteaz Yago amp Serra 1997) These ndingssuggest that the distracting effects of an irrelevant acous-tic event on visual performance may depend on thenature of the eliciting sound That is obtrusive novelsounds appear to cause a stronger effect on the RT tosuccessive visual target stimuli than minor changes inthe acoustic stream but in contrast these minor changes

seem to cause a signicant decrement in the number ofhit responses to visual targets not observed after thenovel sounds

The differential effects of distracting novel and deviantsounds on visual performance may be explained by trac-ing these behavioral effects back to the associated neuralevents as reected by ERPs Novel sounds enhanced theN1 amplitude and elicited the MMN as suggested bysignicant differences between the standard-tone andnovel-sound ERP that extended over the N1 latencyrange The present MMN to novel sounds peaked andterminated (returned to baseline) earlier than the MMNto deviant tones in concordance with previous studiesshowing that with larger deviances MMN becomes ear-lier (Tiitinen May Reinikainen amp Naumlaumltaumlnen 1994) over-lapping the N1 peak for wider deviances (Tiitinen et al1994 Scherg et al 1989) The present MMN to novelsounds is also supported by recent magnetoencepha-lograph (MEG) data showing that novel sounds activatedan MMN source in the human auditory cortex (Alhoet al 1998) Thus for novel sounds the attention-switch-ing signal was probably triggered by a combined re-sponse of the transient-detector system reected in thesupratemporal N1 (Naumlaumltaumlnen 1990 1992 Naumlaumltaumlnen ampPicton 1987) or some other N1 component such as thefrontal N1 reported by Giard et al (1994) and the stimu-lus-change detector system reected in the MMN(Naumlaumltaumlnen 1990 1992) This combined signal for atten-tion capture started with the N1 generation and wasoverlapped by the MMN process resulting in an effectiveattentional reorientation as reected by the subsequentlarge P3a wave and the clearly delayed RT to the follow-ing visual targets

Figure 7 Grand-average ERPs to standard tones and novel soundsin the Auditory-alone (Aa left column) and Auditory-Visual (AV rightcolumn) conditions The novel-standard difference waves are shownin the middle column The late P3a was larger in the AV conditionthan in the Aa condition whereas the early P3a was of similar ampli-tude in the two conditions

Figure 8 Grand-average visual ERPs at Pz O1 and O2 in the Visualalone (Va) condition and in the Auditory-Visual (AV) condition afterstandard tones

598 Journal of Cognitive Neuroscience Volume 10 Number 5

Deviant tones in turn elicited a distinct MMN whichwas followed by a small P3a wave and by a secondnegative peak (Figure 5) Thus the neural mechanismresponsible for triggering the attention-switching signalwas based for deviant tones on the stimulus-changedetector system reected in the MMN (Naumlaumltaumlnen 19901992 Schroumlger 1996) This attention capture signal asso-ciated with the MMN was sporadically followed by an

orienting of attention as suggested by the small (aver-aged) P3a and also by the 23 hit rate decrease to visualtargets that followed deviant tones

Although the subjects were instructed to ignore theauditory stimuli in the AV condition the RT to visualtargets following standard tones was slightly (insig-nicantly) shorter than the RT in the Va condition withno auditory stimuli This suggests that the auditory stim-

Figure 9 Grand-averageERPs in the Auditory-Visual(AV) condition to standard-visual (thin line) deviant-visual (thick line) andnovel-visual (dashed line)stimulus pairs Only the mostrelevant electrodes are shownThe epoch starts 100 msec be-fore auditory-stimulus onsetwhich was used as the base-line period to avoid the largeamplitude variability due tothe auditory response in the100 msec preceding the vis-ual stimulus Novel sounds en-hanced the fronto-polarnegativity and the frontal N2to visual stimuli was en-hanced after novel soundsand deviant tones which alsoaffected the offset slope ofP3b Notice that whereas thetraces for standard-visual devi-ant-visual and novel-visualstimulus pairs perfectly matcheach other during the onsetslope of the P3b they start todiffer from the P3b peak forthe novel-visual stimulus pairand about 100 msec later forthe deviant-visual stimuluspair (see P3 Pz P4)

Escera et al 599

uli were used as warning signals for the following visualtargets and thus attention may have been bimodal in theAV condition However because the occurrence of an aud-itory stimulus change or a novel sound was task-irrelevantand unpredictable to the subjects attention was involun-tarily engaged by these events In a more controlledexperiment attention was more effectively withdrawnfrom the auditory stimuli and their occurrence was madenoninformative by the simultaneous presentation of avisual cue stimulus informing about the successive ap-pearance of a visual target (Alho et al 1997) Like in thepresent study a decreased hit rate for visual stimuli thatfollowed MMN- and P3a-eliciting deviant tones was foundThis conrms the involuntary nature of the attention-switching mechanism indicated by the MMN P3a anddeterioration of the visual-task performance

Component Structure of the P3a to Novel Sounds

In both the Aa and AV conditions novel sounds eliciteda large long-lasting and broadly distributed P3a wavewith a double peak at Fz (in the Aa condition) suggestinga contribution of two different subcomponents Theearly part of this P3a had a peak latency of 230 msecand showed a centrally dominant scalp distribution thatwas similar in the Aa and AV conditions (Figure 6) Theearly P3a inverted in polarity at posterior and lateralelectrodes (Figures 3b and 5) suggesting bilateral gener-ators located in the vicinity of the temporal and parietallobes (cf Scherg amp von Cramon 1986) The late part ofthe P3a in turn had a peak latency of 315 msec and didnot invert in polarity at posterior electrodes and its scalpdistribution was centered over the right frontal areas Incontrast to the early P3a the late P3a was enhanced inthe AV condition particularly over the right hemisphere(Figure 6) when subjects apparently used auditory stim-uli as warning signals to prepare for visual targets

The component structure of the present P3a and thescalp distribution of the two subcomponents is in agree-ment with previous results in patients with brain lesionssuggesting a critical role of the temporal-parietal junc-tion in the generation of the auditory (Halgren BaudenaClarke Heit Marinkovic et al 1995 Halgren BaudenaClarke Heit Lieacutegeois et al 1995 Knight Scabini Woodsamp Clayworth 1989) somatosensory (Yamaguchi ampKnight 1991 1992) and visual P3a (Knight 1991 1997)It is also compatible with results suggesting a frontalcontribution to P3a such as those obtained in patientswith frontal lesions (Knight 1984) in topographicalanalysis (Friedman amp Simpson 1994 Friedman Simpsonamp Hamberger 1993) in dipole-source modeling (Meck-linger amp Ullsperger 1995) and in intracerebral record-ings in humans (Baudena Halgren Heit amp Clarke 1995)

The late P3a amplitude in the AV condition was en-hanced with regard to its amplitude in the Aa conditionThis amplitude enhancement of the late P3a reects an

attentional monitoring of auditory stimuli in the AV con-dition these stimuli being used as warning signals for thefollowing visual targets The attentional modulation ofthe P3a amplitude is in agreement with that reported inprevious studies (Holdstock amp Rugg 1995 Woods 1992)and suggests that the late P3a may more closely reectan actual orienting of attention than the involuntary N1and MMN processes triggering this attentional orientingAlso several other ndings support the association of thelate phase of the P3a with the actual reorientation ofattention or orienting response First the right-frontaldominance in the late P3a scalp distribution observed inthe present experiment agrees with ndings suggestingthe involvement of right-frontal areas in the reorienta-tion of attention (Fuster 1989 Mesulam 1981) Secondit has been shown that the P3a amplitude is attenuatedwith the repetition of the eliciting novel event(Courchesne 1978 Friedman amp Simpson 1994 Knight1984) This habituating behavior is a characteristic fea-ture of the orienting response (Oumlhman 1979 1992 Sok-olov 1963)

The early P3a was insensitive to attentional manipula-tions as it was to similar amplitude and scalp distributionin the Aa and AV conditions Its peak latency was rathershort (230 msec) and its scalp distribution suggestsneural generators probably located in the temporal-parietal cortex (Knight 1991 1997 Knight et al 1989Yamaguchi amp Knight 1991) Thus the early P3a mightreect a neural process other than attentional reorienta-tion such as the violation of a polisensorial model of theexternal world maintained in the temporal-parietal asso-ciation cortex (Yamaguchi amp Knight 1991)

Conclusions

The results reported in the present paper support thenotion of two different brain mechanisms of involuntaryattention differentially activated by different acousticevents The ERP component structure to novel soundssuggests that unexpected novelty in the acoustic envi-ronment involuntary captures attention by simultane-ously activating two different neural mechanisms atransient-detector mechanism associated with the su-pratemporal N1 component of the auditory ERPs and achange-detector mechanism reected in the MMN Pre-sumably the attention-switching signal generated by thecombined activation of these two mechanisms resultedin an effective engagement of attention as indicated bythe large P3a component elicited by the novel soundsand the increased RT to subsequent visual targets Devi-ant tones in turn resulted in different behavioral effectsand ERP component structure which was characterizedby the MMN This supports the notion that small changesin the acoustic environment capture attention involun-tarily by activating the stimulus-change detector mecha-nism reected in the MMN

600 Journal of Cognitive Neuroscience Volume 10 Number 5

METHODS

Subjects Stimuli and Procedure

Ten right-handed paid students (21 to 40 years 3 males)with normal hearing and normal or corrected-to-normalvisual acuity participated in the experiment Two furthersubjects were discarded due to their inability to reachthe required 90 hit rate in a practice session

Subjects were presented with four blocks of 400stimulus pairs (AV condition) in a visual forced-choiceRT task Each pair consisted of an auditory stimulusfollowed after 300 msec (onset-to-onset) by a visualstimulus The interpair interval (onset-to-onset) was 12sec (Figure 1) The auditory stimuli were either standardor deviant tones and novel sounds presented in randomorder with probabilities of 08 01 and 01 respectivelyThe standard and deviant tones were sinusoidal tonebursts of 200-msec duration including 10-msec rise andfall times presented binaurally at 75 dB SPL throughheadphones with respective frequencies of 600 and 700Hz Sixty different environmental sounds such as thoseproduced by a drill hammer rain door telephone ring-ing etc were used as novel sounds They were digitallyrecorded then low-pass ltered at 10000 Hz and editedto have a duration of 200 msec including rise and falltimes of 10 msec and an intensity maximum of 70 to 80dB SPL Each different novel sound occurred only oncewithin a stimulus block and was presented twice orthree times during the whole experiment Both devianttones and novel sounds were always preceded by at leastone AV pair containing a standard tone The visual stimuliwere the digits 1 to 8 presented one at a time in randomorder on a computer screen for 200 msec They sub-tended a vertical angle of 17deg and a horizontal angle of11deg (30 mm acute 20 mm 100 cm from the subjectrsquos eyes)

Subjects were comfortably seated in a reclining chairin a dimly lit electrically and acoustically shielded roomThey were instructed to focus on the small xation crossappearing in the middle of the screen and to press oneresponse button with the right index nger for oddnumbers and another button with the right middlenger for even numbers Subjects were also instructedto ignore the auditory stimulation Both speed and accu-racy were emphasized for the primary visual task

In two further conditions visual stimuli were omitted(Va condition Figure 1) or auditory stimuli were omitted(Aa condition Figure 1) In the Va condition subjectsreceived three blocks of 200 visual stimulus each theinstruction being identical to that in the AV condition Inthe Aa condition subjects were instructed to concen-trate on reading a self-selected book and to ignore theauditory stimulation

The Aa and AV blocks were presented in an alternatingorder with half of the subjects starting with the AVcondition The Va blocks were presented one at thebeginning one at the middle and one at the end of the

experiment Before the experimental blocks subjectsparticipated in a practice session consisting of two Vablocks They were required to reach a 90 hit rate beforethe experiment could start Two out of 12 subjects failedto reach this level even after several additional blocksand were nally discarded

EEG Recording and Averaging

The electroencephalogram (EEG) (bandpass 0ndash100 Hz)was continuously digitized at a rate of 500 Hz by aSynAmps amplier (NeuroScan Inc) from 19 scalp elec-trodes positioned according to the 10ndash20 system Fp1Fp2 F7 F3 Fz F4 F8 T3 C3 Cz C4 T4 T5 P3 Pz P4 T6O1 and O2 One additional electrode was placed at theright mastoid (RM) and another at the left mastoid (LM)The horizontal EOG was recorded with electrodes at-tached to the left and right canthi The vertical elec-trooculogram (EOG) was assessed using recordings fromFp1 and Fp2 electrodes The common reference elec-trode was placed on the tip of the nose

ERPs were averaged off-line for each auditory stimulusclass for an epoch of 1300 msec including a preauditorystimulus period of 100 msec in the AV and Aa conditionsIn the Va condition an ERP was also obtained for anepoch of 1300 msec including a previsual stimulus pe-riod of 400 msec Epochs in which the EEG or EOGexceeded plusmn100 m V as well as the ve rst epochs of eachblock were automatically excluded from averaging Inthe Aa and AV conditions standard-tone trials immedi-ately following deviant-tone or novel-sound trials werealso excluded from the averages Individual ERPs wereband-pass ltered between 001 and 30 Hz

Data Analysis

In the AV and Va conditions a correct button presswithin 800 msec after visual-stimulus onset was regardedas a hit the mean RT being computed only for the hittrials An incorrect button press during this period wasclassied as an error and trials with no response withinthis time window were considered misses Hits errorsmisses and RTs were computed across odd and evennumbers

Auditory ERP amplitudes (N1 P2 MMN and P3a) werereferred to the mean amplitude in the 100-msec baselineperiod preceding the auditory stimulus in both the Aaand AV conditions For auditory ERP analysis each indi-vidual waveform in the Aa and AV conditions was linearlydetrended over an epoch of 600 msec (including base-line) to remove any possible asymmetrical contributionto EEG caused by slow EOG movements due to readingin the Aa condition The effects of the standard tones onthe visual ERPs were analyzed at O1 and O2 for peaksas follows (identied in the individual waveforms withinthe specied latency ranges) P1 75 to 140 msec N1 100

Escera et al 601

to 200 msec P2 180 to 280 msec N2 225 to 350 msecThe visual P3b was identied at Pz between 300 and 500msec from visual-stimulus onset For this analysis visualERP amplitudes were measured against the mean ampli-tude of the 100-msec baseline preceding visual-stimulusonset The fronto-polar negativity was analyzed at Fp1and Fp2 the visual frontal N2 at F3 Fz and F4 and thevisual P3b at P3 Pz and P4 The amplitude measurementsfor visual ERP components in the AV condition werereferred to the 100-msec preauditory stimulus baselinerather than to the 100-msec previsual stimulus baselinewhich was differentially affected by standard tones de-viant tones and novel sounds (see Figure 9)

The ERP and performance data were statistically ana-lyzed by ANOVA with repeated measures Where appro-priate nominal degrees of freedom and epsilon valuesafter the Greenhouse-Geisser correction are reportedScalp-distribution analyses were performed after normal-izing ERP amplitudes to prevent amplitude differencesbetween different components from washing out thegenuine scalp-distribution differences This normalizationwas done by dividing the amplitude at each electrodeby the sum of the squared amplitudes at all electrodes(McCarthy amp Wood 1985)

Acknowledgments

This research was supported by BIOMED-2 contract BMH4-CT96-0819-COBRAIN of the European Union the SpanishMinistry of Science and Education (DGICYT PB93-0802 UE96-0038) the Academy of Finland the scientic exchange programbetween the University of Barcelona and the University ofHelsinki and the National Science Fund of Hungary (OTKAT022681) The authors would like to thank Teemu Rinne andTeija Kujala for their help in making the gures

Reprint requests should be sent to Carles Escera Neurody-namics Laboratory Department of Psychiatry and Clinical Psy-chobiology University of Barcelona P Vall drsquoHebron 171 08035Barcelona Catalonia-Spain or via e-mail cescerapsiubes

Note

1 The MMN amplitude in the unattended channel can behowever attenuated under some conditions by strongly focus-ing attention on other sources of auditory stimulation(Naumlaumltaumlnen 1991 Naumlaumltaumlnen Paavilainen Tiitinen Jiang amp Alho1993 Oades amp Dittmannbalcar 1995 Trejo Ryan-Jones amp Kra-mer 1995 Woldorff Hackley amp Hillyard 1991)

REFERENCES

Alho K (1995) Cerebral generators of mismatch negativity(MMN) and its magnetic counterpart (MMNn) elicited bysound changes Ear and Hearing 16 38ndash51

Alho K Escera C Diacuteaz R Yago E amp Serra J M (1997)Effects of involuntary auditory attention on visual taskperformance and brain activity NeuroReport 8 3233ndash3237

Alho K Winkler I Escera C Huotilainen M Virtanen JJaumlaumlskelaumlinen I Pekkonen E amp Ilmoniemi R (1998) Proc-

essing of novel sounds and frequency changes in the hu-man auditory cortex Magnetoencephalographic record-ings Psychophysiology 35 211ndash224

Alho K Woods D L Algazi A amp Naumlaumltaumlnen R (1992) Inter-modal selective attention II Effects of attentional load onprocessing of auditory and visual stimuli in central spaceElectroencephalography and Clinical Neurophysiology82 356ndash368

Baudena P Halgren E Heit G amp Clarke J M (1995) Intra-cerebral potentials to rare target and distractor auditoryand visual stimuli III Frontal cortex Electroencephalogra-phy and Clinical Neurophysiology 94 251ndash264

Cherry E C (1953) Some experiments on the recognitionof speech with one and with two ears Journal of theAcoustical Society of America 29 975ndash979

Courchesne E (1978) Changes in P3 waves with event repe-tition Long-term effects on scalp distribution and ampli-tudes Electroencephalography and ClinicalNeurophysiology 45 468ndash482

Courchesne E Hillyard S A amp Galambos R (1975) Stimu-lus novelty task relevance and the visual evoked potentialsin man Electroencephalography and Clinical Neurophysi-ology 39 131ndash143

Cowan N Winkler I Teder W amp Naumlaumltaumlnen R (1993) Mem-ory prerequisites of the mismatch negativity in the audi-tory event-related potential (ERP) Journal ofExperimental Psychology Learning Memory and Cogni-tion 19 909ndash921

Cseacutepe V Karmos G amp Molnaacuter M (1987) Evoked potentialcorrelates of stimulus deviance during wakefulness andsleep in catmdashanimal model of mismatch negativity Elec-troencephalography and Clinical Neurophysiology 66571ndash578

Davis H Mast T Yoshie N amp Zerlin S (1966) The slow re-sponse of the human cortex to auditory stimulus Recov-ery process Electroencephalography and ClinicalNeurophysiology 21 105ndash113

Davis H amp Zerlin S (1966) Acoustic relations of the humanvertex potentials Journal of the Acoustical Society ofAmerica 39 109ndash116

Dawson M E (1990) Psychophysiology at the interface ofclinical science cognitive science and neuroscience Psy-chophysiology 27 243ndash255

Dawson M E Fillion D L amp Schell A M (1989) Is elicita-tion of the autonomic orienting response associated withallocation of processing resources Psychophysiology 26560ndash572

Duncan C C amp Kaye W H (1987) Effects of chlonidine onevent-related potential measures of information process-ing In R Johnson J W Rohrbaugh amp R Parasuraman(Eds) Current trends in event-related potentialresearch (EEG supplement 40) (pp 527ndash531) Amster-dam Elsevier

Fillion D J Dawson M E Schell A M amp Hazlett E A(1991) The relationship between skin conductance orient-ing and the allocation of processing resources Psycho-physiology 28 410ndash424

Folk C L Remington R W amp Johnston J C (1992) Involun-tary covert orienting is contingent on attentional controlsettings Journal of Experimental Psychology Human Per-ception and Performance 18 1030ndash1044

Ford J M Roth W T amp Kopell B S (1976) Auditory event-related potentials to unpredictable shifts in pitch Psycho-physiology 13 32ndash39

Friedman D amp Simpson G (1994) ERP amplitude and scalpdistribution to target and novel events Effects of temporalorder in young middle-aged and older adults CognitiveBrain Research 2 49ndash63

602 Journal of Cognitive Neuroscience Volume 10 Number 5

primary task (Dawson Fillion amp Schell 1989 FillionDawson Schell amp Hazlett 1991 Siddle 1991 WoodwardBrown March amp Dawson 1991)

Involuntary attention shifts were originally explainedby the orienting-reex (OR) theory (Sokolov 1963) pro-posing that a neuronal model is built from the repetitivefeatures of the external environment inhibiting the ORto identical but not to different stimuli Further Oumlhman(1979 1992) proposed that a ldquocallrdquo for reallocation ofcentral processing resources is issued by preattentivemechanisms detecting signicant changes in incomingstimuli

Attention switching can also be elicited by stimulusonsets (regardless of repetitionchange) in particular ifthey appear after long ldquosilentrdquo intervals and by offsets ofcontinuous stimulation (Folk Remington amp Johnston1992 Gati amp Ben-Shakhar 1990 Hikosaka Miyauchi ampShimojo 1993 Oumlhman 1979 Theeuwes 1994 Yantis1993 Yantis amp Hillstrom 1994) According to Naumlaumltaumlnenrsquos(1990 1992) model stimulus onsets and offsets activatea transient-detector system reected in the auditory mo-dality in the supratemporal and nonspecic componentsof the N1 wave peaking at about 100 msec from stimu-lus onset (Naumlaumltaumlnen amp Picton 1987) The supratemporalN1 component is generated by bilateral dipoles locatedin the auditory cortices on the supratemporal plane(Giard et al 1994 Vaughan amp Ritter 1970) and its am-plitude increases with an increasing interval from theprevious sound (Davis Mast Yoshie amp Zerlin 1966 HariKaila Katila Tuomisto amp Varpula 1982 Maumlkelauml Hari ampLeinonen 1988 Ritter Vaughan amp Costa 1968) Thiscomponent is sensitive in particular to the transientaspects of stimulation (Davis amp Zerlin 1966 McMillan1973 Pfefferbaum Buchsbaum amp Gips 1971) whichsuggests that the neuronal process generating the su-pratemporal N1 triggers an attention-capturing signal forconscious perception of the stimulus (Naumlaumltaumlnen 19901992) The nonspecic N1 component is even moresensitive to interstimulus interval manipulations than thesupratemporal component (Naumlaumltaumlnen amp Picton 1987)and in contrast to the supratemporal N1 component itcan be elicited also by stimuli of other modalities (Leh-tonen 1973 Velasco amp Velasco 1986 Velasco Velascoamp Olvera 1985) The function of the nonspecic N1neural generators probably is to trigger a transientarousal burst facilitating sensory and motor responses tothe eliciting stimulus (Naumlaumltaumlnen amp Picton 1987) How-ever Giard et al (1994) described quite recently a frontalsubcomponent of the N1 wave This subcomponentmight be a better candidate for the attention-switchingfunction originally attributed to the supratemporal-N1generator

The stimulus-change detector mechanism postulatedby the OR theory is represented in Naumlaumltaumlnenrsquos (19901992) model as the neuronal process generating themismatch negativity (MMN) The MMN elicited by anydiscernible change in the physical features of a repeti-

tive even unattended sound (eg Naumlaumltaumlnen 1992) isgenerated mainly in the supratemporal auditory cortex(for an overview see Alho 1995) as revealed by ERP-source modeling (Giard et al 1995 Giard Perrin Pernieramp Bouchet 1990 Scherg Vajsar amp Picton 1989) neuro-magnetic data (Hari et al 1984) and intracranial record-ings in animals (Cseacutepe Karmos amp Molnaacuter 1987 JavittSchroeder Steinschneider Arezzo amp Vaughan 1992) aswell as in humans (Halgren Baudena Clarke Heit et al1995 Kropotov et al 1995) According to Naumlaumltaumlnenrsquos(1990 1992) model the physical features of auditorystimuli are fully analyzed and encoded into neural tracesof auditory sensory memory the MMN being automat-ically elicited when the auditory input does not matchwith the neuronal trace formed by the repetitive stan-dard stimulus (Naumlaumltaumlnen Paavilainen Alho Reinikainenamp Sams 1989) while the trace still is in an ldquoactiverdquo state(Cowan Winkler Teder amp Naumlaumltaumlnen 1993) Further themodel proposes that the MMN-generating process is in-volved in triggering a signal for attention switching afterthe automatic preperceptual detection of auditorychange (Naumlaumltaumlnen 1990)

Converging evidence supports the notion that theprocess generating the MMN may be associated withinvoluntary attention switching First the MMN is ofpreattentive nature because it does not depend onwhether the subject is engaged in an easy or a difcultvisual task (Alho Woods Algazi amp Naumlaumltaumlnen 1992 Dun-can amp Kaye 1987)1 Second in addition to its supratem-poral generators MMN appears to get a furthercontribution from the frontal lobe activity (Giard et al1990 Molnaacuter Skinner Cseacutepe Winkler amp Karmos 1995)a critical structure for controlling both voluntary andinvoluntary attention (Fuster 1989) Third the MMNtends to be accompanied especially in the beginning ofthe stimulus sequence by autonomous nervous system(ANS) responses even when stimulus change is of verysmall magnitude (Lyytinen Blomberg amp Naumlaumltaumlnen 1992)These autonomic responses indicate apparently atten-tion switching triggered by the automatic change-detec-tion process reected by the MMN

The strongest evidence supporting the causal relation-ship between the neural process generating the MMNand involuntary attention switching was recently ob-tained by Schroumlger (1996) He found that MMN-elicitingdeviant tones occurring in a repetitive sequence of stan-dard tones presented to the unattended ear prolongedthe RT and decreased the performance accuracy to sub-sequent target tones delivered to the other ear Thiseffect was observed only for irrelevant tones precedingtarget stimuli by a short interval of 200 msec but not forlonger intervals of 560 msec and was strengthened withincreasing frequency difference between the deviantand standard tones Both of these ndings support thefunctional role of the MMN-generating process in trigger-ing this attention switch

Another ERP component that has been associated

Escera et al 591

with orienting of attention is the P3a This positive com-ponent is elicited by irrelevant rare tones and novelsounds in a sequence of repetitive standard tones andcan be distinguished from the P300 or P3b componentto target stimuli by its shorter latency and its morefrontal scalp distribution (Courchesne Hillyard amp Galam-bos 1975 Ford Roth amp Kopell 1976 Squires Squires ampHillyard 1975) The association of the P3a-generatingprocess with the orienting response is supported by itselicitation by widely deviant or novel sounds and also bythe frontal-lobe and hippocampal contributions to itsgeneration (Knight 1984 1996) these brain regions be-ing involved also in the orienting response (Fuster 1989Sokolov 1975) The role of the P3a generator process ininvoluntary engagement of attention is further sup-ported by studies showing delayed RTs to target stimulifollowing irrelevant novel sounds (Grillon CourchesneAmeli Geyer amp Braff 1990 Woods 1992) When novelsounds occurred in the attended sequences containingthe target stimuli the P3a elicited was larger in ampli-tude as compared with the unattended-sequence P3a tonovel sounds and the RT to consecutive targets wasdelayed (Woods 1992) This attentional modulation sug-gests that the P3a reects an actual reorientation ofattention more closely than the MMN or the differentsubcomponents of the N1 which would rather be asso-ciated with the involuntary call for attention

The purpose of the present experiment was to inves-tigate the neural mechanisms of involuntary attention toacoustic novelty and change by analyzing the ERP con-comitants elicited by these irrelevant auditory eventsduring visual performance In the main experimentalcondition (Figure 1) the Auditory-Visual condition (AV)a response was required to each visual stimulus of anauditory-visual stimulus pair Subjects were instructed toignore the auditory stimuli which were a repetitiveldquostandardrdquo tone randomly replaced by a slightly higherldquodeviantrdquo tone (p = 01) or by a natural ldquonovelrdquo sound(p = 01) and to press a response button to even num-bers and another to odd numbers Two additional condi-tions were used as controls In the Visual-alonecondition (Va) the auditory stimuli were omitted andsubjects performed the forced-choice visual RT task Inthe Auditory-alone (Aa) condition the visual stimuliwere omitted and subjects were instructed to concen-trate on reading a book

RESULTS

Performance

In the Va condition the mean hit rate was 915 (SEM plusmn17) and the mean RT was 461 plusmn 11 msec Subjectsmissed the target on an average 45 plusmn 16 of thetrials and pressed the wrong button in 40 plusmn 08 ofthe trials

Performance data in the Va and AV conditions areshown in Figure 2 In the AV condition subjects tendedto respond faster to visual stimuli preceded by a standardtone than to those preceded by no sound (Va condition)(444 plusmn 9 and 461 plusmn 11 msec respectively) although thiseffect did not quite reach statistical signicance (F(19) = 38 p lt 009) The hit error and miss rates weresimilar in the Va condition and in the AV condition aftera standard tone In the AV condition the occurrence ofa deviant tone before the visual stimulus caused a sig-nicant hit rate decrease of 23 (F(1 9) = 650 p lt004) which was probably due to an increased numberof wrong responses to target stimuli (F(1 9) = 478 p lt006 see Figure 2) the number of missing responsesbeing identical after standard and deviant tones A novelsound occurring before a visual stimulus caused an av-erage increase of 17 msec in the RT to visual stimulicompared with the RT to these stimuli that were pre-ceded by standard tones (F(1 9) = 2219 p lt 0002) TheRT to visual stimuli after novel sounds was also sig-nicantly longer than that to visual stimuli followingdeviant tones (F(19) = 1107 p lt 0009)

Figure 1 Experimental conditions In the main condition [Auditory-Visual (AV) top] pairs of stimuli consisting of an auditory stimulusfollowed at 300 msec (onset-to-onset) by a visual stimulus (an arabicnumber between 1 and 8) were presented with an interpair intervalof 1200 msec The auditory stimulus was either a standard tone(80) a deviant tone (10) or a novel sound (10) Subjects wereinstructed to press one response button to odd (50) and the otherto even (50) numbers and to ignore auditory stimulation In theVisual-alone (Va) condition the auditory stimuli were omitted In theAuditory-alone (Aa) condition the visual stimuli were omitted thesubject being instructed to read a self-selected book and to ignorethe auditory stimuli

592 Journal of Cognitive Neuroscience Volume 10 Number 5

Auditory ERPs

In the Aa condition standard tones elicited N1 (meanpeak amplitude - 20 m V at Cz with a mean peak latencyof 93 msec) and P2 deections (34 m V 158 msec at Cz)which were largest over the fronto-central scalp loca-tions (Figure 3a) Deviant tones elicited N1 and P2 wavesat latencies similar to those in response to standardtones and the mismatch negativity (MMN) The MMN

was largest over the frontal electrodes (mean peak am-plitude - 22 m V mean peak latency 153 msec at Fz) andinverted in polarity at the mastoid electrodes (Figure 3b)being signicantly different from zero at F3 Fz and F4in the intervals 100 to 150 msec (F(1 9) = 1031 p lt002) and 150 to 200 msec (F(1 9) = 563 p lt 005) TheMMN was followed by a smaller fronto-centrally distrib-uted negative wave peaking at around 400 msec (Figure3b) Mean amplitude comparisons between ERPs to de-viant and standard tones at Fz and Cz revealed that thissecond negativity was signicant in the 350- to 400-msec(F(1 9) = 756 p lt 003) and in the 400- to 450-msec(F(1 9) = 873 p lt 002) intervals In the AV conditionthe MMN was similar to that obtained in the Aa condition(Figure 4)

In the Aa condition novel sounds elicited an ERPcharacterized by a prominent N1 deection (mean peakamplitude at Cz - 56 m V peak latency 96 msec) and bya long-duration broadly distributed P3a wave (Figure 3as seen in Figure 3b even the deviant tones elicitedsome P3a) The novel-sound N1 was larger than thestandard-tone N1 (F(1 9) = 1099 p lt 001) and wasprobably composed of overlapping N1 and MMN com-ponents The difference between the standard-tone andnovel-sound ERPs extended beyond the N1 latencyrange there being a signicant difference at Fz betweenthe standard and novel ERP amplitudes measured asmean voltages over 120 to 160 msec latency window(F(1 9) = 523 p lt 005 Figure 5)

The present P3a to novel sounds had a double peakover the frontal scalp locations (Figure 3b) suggestingtwo different subcomponents The early part of the P3a(peak latency 230 msec in the grand-mean ERP at Fz)was largest over the central scalp areas appearing withinverted polarity at lateral and posterior scalp sites (Fig-ures 3b and 5) The late part of the P3a (peak latency315 msec in the grand-mean ERP at Fz) in turn waslargest over frontal areas and did not show polarityinversion at any recording site (Figures 3b and 5) Thedifferent scalp distributions of the two P3a subcompo-nents in the Aa condition are shown in Figure 6 (top)

The component structure of the P3a deection wasevaluated by means of two different analyses of variance(ANOVA) for normalized ERP amplitudes (McCarthy ampWood 1985) First an ANOVA was performed to prove apossible scalp-distribution difference between the P2 tostandard tones and the early P3a to novel sounds Thenanother ANOVA was performed to conrm the differentscalp distributions of the two subcomponents of the P3aThe P2 amplitude was identied as the mean voltagebetween 150 and 200 msec from stimulus onset in thestandard-tone ERP The early P3a was measured as themean voltage between 175 and 275 msec in the differ-ence wave obtained by subtracting the ERP to the stan-dard tone from that to the novel sound The late P3a wasmeasured as the mean voltage between 275 and 375msec of the same difference wave A three-factor design

Figure 2 Mean RT (top panel) hit rate (middle panel) and errorrate (bottom panel) in the Visual-alone (Va) condition and in theAuditory-Visual (AV) condition to the visual stimulus occurring afterthe standard tone (std) the deviant tone (dev) or the novel sound(nov) The bars indicate the standard error of mean

Escera et al 593

was used for both analyses Wave (P2 versus early P3afor the rst analysis and early P3a versus late P3a for thesecond analysis) acute Frontality (F7 F3 Fz F4 F8 versus T3C3 Cz C4 T4 versus T5 P3 Pz P4 T6) acute Laterality (F7T3 T5 versus F3 C3 P3 versus Fz Cz Pz versus F4 C4P4 versus F8 T4 T6) The results of the rst analysisrevealed a signicant Wave acute Frontality interaction (F(218) = 509 p lt 003 e = 07547) due to the early P3abeing posterior to the P2 and a signicant Wave acuteLaterality interaction (F(4 36) = 406 p lt 003 e =05469) due to the more predominantly right-hemi-spheric scalp distribution of the P2 than that of the earlyP3a The results of the second analysis revealed that thelate P3a was anterior in distribution to the early P3a(signicant Wave Frontality interaction F(2 18) = 939p lt 0006 e = 07187) and further that the late P3a wasmore preponderant over the right hemisphere than the

early P3a (signicant Wave acute Laterality interactionF(4 36) = 1207 p lt 00002 e = 05740) The secondanalysis also revealed a signicant third-order interaction(Wave acute Frontality acute Laterality F(8 72) = 898 p lt00004 e = 03441) that was due to the predominantlyright-frontal distribution of the late phase of the P3adeection (Figure 6 top)

In the AV condition novel sounds elicited ERPs similarto those elicited by them in the Aa condition (large N1and P3a deections) although the early peak of the P3aseemed to be partly abolished at frontal electrodes (Fig-ure 7) However difference waves (ERP to novel soundsminus ERP to standard tones) revealed that the earlyphase of the P3a was similar in the Aa and AV conditionswhereas the late P3a was enhanced in the AV condition(Figure 7) The statistical scalp-distribution analyses ofthe two P3a phases across the conditions [ANOVA with

Figure 3a Grand-mean ERPsaveraged across subjects tostandard tones (600 Hz) devi-ant tones (700 Hz) and novelsounds in the Auditory-alone(Aa) condition

594 Journal of Cognitive Neuroscience Volume 10 Number 5

four factors Condition (Aa versus AV) acute Wave (early P3aversus late P3a) acute Frontality acute Laterality] conrmed thescalp-distribution difference between the two P3aphases (Wave acute Frontality interaction F(2 18) = 996p lt 0003 e = 08361 Wave acute Laterality F(4 36) = 688p lt 002 e = 03977 Wave acute Frontality acute LateralityF(872) = 981 p lt 00001 acute = 04381) and also revealeda signicant increase of the late P3a amplitude over theright hemisphere in the AV condition (Wave acute Condi-tion acute Laterality F(4 36) = 384 p lt 004 e = 05367)

Visual ERPs

Visual stimuli in the Va condition and those after stan-dard tones in the AV condition elicited almost identicalERPs characterized by P1 N1 P2 N2 and P3 deections(Figure 8) ANOVA for these visual ERP components with

Condition (after standard in AV versus Va) as a functionrevealed no signicant effects of standard tones on visualERPs

Figure 9 shows grand-average ERPs to auditory-visualstimulus pairs in the AV condition The AV stimulus pairselicited an auditory ERP lasting up to 400 msec post-stimulus for novel sounds and up to 300 msec post-stimulus for deviant tones followed by a visual ERPsimilar to that obtained in the Va condition All threetypes of AV stimulus pairs (standard-visual deviant-visualnovel-visual) were also characterized by a long-lastingfronto-polar negativity that was larger to visual stimulifollowing novel sounds than to those following standardor deviant tones (Figure 9) ANOVA for mean amplitudesat Fp1 and Fp2 in consecutive 100-msec latency win-dows yielded signicant effects for the novel-standardcomparison (F(1 9) ranging from 779 to 1196 p lt 003

Figure 3b Grand-average dif-ference waves obtained in theAa condition by subtractingthe ERP to standard tonesfrom that to deviant tones(continuous line) and tonovel sounds (dashed line)

Escera et al 595

in all cases) at latency windows 200 to 300 msec 300to 400 msec and 400 to 500 msec from visual-stimulusonset

Novel sounds and deviant tones enhanced the frontalN2 deection elicited by the subsequent visual stimulusin comparison with the visual N2 elicited after standardtones (Figure 9) This was revealed by two-way ANOVAsusing the Electrode (F3 Fz F4) and the auditory Stimulus(standard versus either deviant or novel) as factors Themean amplitude in the 200- to 300-msec range fromvisual-stimulus onset at Fz was - 19 m V after standardtones - 33 m V after deviant tones (F(1 9) = 641 p lt004) and - 36 m V after novel sounds (F(1 9) = 899 plt 002) The mean amplitude in the 300- to 400-msecinterval was also signicantly larger for visual stimulifollowing deviant tones or novel sounds than for thosefollowing standard tones F(1 9) = 815 p lt 002 for

deviant versus standard F(1 9) = 795 p lt 001 for novelversus standard

Preceding novel sounds and deviant tones also en-hanced the terminal slope of the visual P3b (Figure 9)this was revealed by the signicantly larger mean ampli-tude of the 500- to 600-msec latency window (at P3 Pzand P4) in the ERP to visual stimuli following devianttones than in the ERP to visual stimuli following standardtones (F(19) = 1556 p lt 004) The mean amplitude inthe 400- to 500-msec (F(1 9) = 694 p lt 003) and 500-to 600-msec (F(1 9) = 1199 p lt 0008) intervals was

Figure 4 The MMN was similar in the Auditory-alone (Aa) andAuditory-Visual (AV) conditions as revealed by nonsignicant differ-ences of the mean amplitude of ve consecutive 50-msec intervalsstarting at MMN onset (50 msec) [two-way ANOVAs Condition (Aaversus AV) acute Electrode (F3 Fz F4 C3 Cz C4)] Top Grand-averageERPs to standard and deviant tones at Fz in the two conditionsBottom Deviant-standard difference waves

Figure 5 The component structure of the ERPs to deviant tonesand novel sounds as revealed by the deviant-standard and novel-stan-dard difference waves at Fz and right-mastoid (RM) recordings Devi-ant tones elicited the MMN with a peak latency of 150 msec at Fzand with an inverted polarity at RM (top) Novel sounds elicited alarge N1 deection at Fz that was presumably composed of contribu-tions from two different processes for which the signicant differ-ences between the standard-tone and novel-sound ERPs extendedbeyond the N1 latency range (bottom) The novel sounds also elic-ited a P3a wave with two peaks at Fz suggesting two different sub-components contributing to the recorded P3a an early P3a (peaklatency 230 msec) which inverted polarity at RM and other poste-rior and lateral sites and a late P3a (peak latency 315 msec) whichinverted polarity at no recording site

596 Journal of Cognitive Neuroscience Volume 10 Number 5

also signicantly larger for visual stimuli following novelsounds than after those following the standard tones

DISCUSSION

Engagement of Attention During VisualPerformance

The present results provide evidence that attention dur-ing visual performance can be engaged by shortly pre-ceding unexpected novel sounds or small changes in thetask-irrelevant acoustic environment These results alsorevealed that this phenomenon depends on the intrinsicnature of the eliciting acoustic events In addition the

observed effects on visual performance were associatedwith different patterns of neural events activated bydeviant tones and novel sounds as reected by thecomponent structure of the auditory ERPs elicited

As expected novel sounds prolonged the RT to suc-cessive visual stimuli This result is in agreement with theextensive literature reporting delayed RTs to target stim-uli caused by preceding irrelevant sound changes ornovels sounds (Alho et al 1992 Cherry 1953 Dawsonet al 1989 Fillion et al 1991 Grillon et al 1990Schroumlger 1996 Woods 1992 Woodward et al 1991) andit provides experimental support for the everyday expe-rience that attention can be engaged by irrelevant acous-tic events

Figure 6 The scalp distribu-tion of the P3a to novelsounds Left early P3a rightlate P3a top Auditory-alone(Aa) condition bottom Audi-tory-Visual (AV) condition Thecentral plot shows the novel-standard difference waves atFz in the Aa (thin) and AV(thick) conditions as well asthe 100-msec latency win-dows used to measure thetwo phases of the P3a (e =early P3a l = late P3a)

Escera et al 597

A surprising nding of this study was however thatthe slightly deviant tones degraded the visual perfor-mance accuracy by decreasing the rate of hit responseswhereas the widely deviant novel sounds did not It wasexpected that the amount of distraction caused by devi-ant tones and novel sounds would be proportional tothe magnitude of their deviance from the standard toneand indeed such an effect was found for the RT (17-msecincrease after novel sounds) conrming other recentndings (Jaumlaumlskelaumlinen et al 1996 Schroumlger 1996) How-ever this RT effect was not paralleled by a similar effecton the hit rate Indeed deviant tones were associatedwith a lower hit rate whereas the hit rate after novelsounds was similar to that after standard tones Similarresults were obtained in a related study by Jaumlaumlskelaumlinenet al (1996) and recently replicated for deviant tones(Alho Escera Diacuteaz Yago amp Serra 1997) These ndingssuggest that the distracting effects of an irrelevant acous-tic event on visual performance may depend on thenature of the eliciting sound That is obtrusive novelsounds appear to cause a stronger effect on the RT tosuccessive visual target stimuli than minor changes inthe acoustic stream but in contrast these minor changes

seem to cause a signicant decrement in the number ofhit responses to visual targets not observed after thenovel sounds

The differential effects of distracting novel and deviantsounds on visual performance may be explained by trac-ing these behavioral effects back to the associated neuralevents as reected by ERPs Novel sounds enhanced theN1 amplitude and elicited the MMN as suggested bysignicant differences between the standard-tone andnovel-sound ERP that extended over the N1 latencyrange The present MMN to novel sounds peaked andterminated (returned to baseline) earlier than the MMNto deviant tones in concordance with previous studiesshowing that with larger deviances MMN becomes ear-lier (Tiitinen May Reinikainen amp Naumlaumltaumlnen 1994) over-lapping the N1 peak for wider deviances (Tiitinen et al1994 Scherg et al 1989) The present MMN to novelsounds is also supported by recent magnetoencepha-lograph (MEG) data showing that novel sounds activatedan MMN source in the human auditory cortex (Alhoet al 1998) Thus for novel sounds the attention-switch-ing signal was probably triggered by a combined re-sponse of the transient-detector system reected in thesupratemporal N1 (Naumlaumltaumlnen 1990 1992 Naumlaumltaumlnen ampPicton 1987) or some other N1 component such as thefrontal N1 reported by Giard et al (1994) and the stimu-lus-change detector system reected in the MMN(Naumlaumltaumlnen 1990 1992) This combined signal for atten-tion capture started with the N1 generation and wasoverlapped by the MMN process resulting in an effectiveattentional reorientation as reected by the subsequentlarge P3a wave and the clearly delayed RT to the follow-ing visual targets

Figure 7 Grand-average ERPs to standard tones and novel soundsin the Auditory-alone (Aa left column) and Auditory-Visual (AV rightcolumn) conditions The novel-standard difference waves are shownin the middle column The late P3a was larger in the AV conditionthan in the Aa condition whereas the early P3a was of similar ampli-tude in the two conditions

Figure 8 Grand-average visual ERPs at Pz O1 and O2 in the Visualalone (Va) condition and in the Auditory-Visual (AV) condition afterstandard tones

598 Journal of Cognitive Neuroscience Volume 10 Number 5

Deviant tones in turn elicited a distinct MMN whichwas followed by a small P3a wave and by a secondnegative peak (Figure 5) Thus the neural mechanismresponsible for triggering the attention-switching signalwas based for deviant tones on the stimulus-changedetector system reected in the MMN (Naumlaumltaumlnen 19901992 Schroumlger 1996) This attention capture signal asso-ciated with the MMN was sporadically followed by an

orienting of attention as suggested by the small (aver-aged) P3a and also by the 23 hit rate decrease to visualtargets that followed deviant tones

Although the subjects were instructed to ignore theauditory stimuli in the AV condition the RT to visualtargets following standard tones was slightly (insig-nicantly) shorter than the RT in the Va condition withno auditory stimuli This suggests that the auditory stim-

Figure 9 Grand-averageERPs in the Auditory-Visual(AV) condition to standard-visual (thin line) deviant-visual (thick line) andnovel-visual (dashed line)stimulus pairs Only the mostrelevant electrodes are shownThe epoch starts 100 msec be-fore auditory-stimulus onsetwhich was used as the base-line period to avoid the largeamplitude variability due tothe auditory response in the100 msec preceding the vis-ual stimulus Novel sounds en-hanced the fronto-polarnegativity and the frontal N2to visual stimuli was en-hanced after novel soundsand deviant tones which alsoaffected the offset slope ofP3b Notice that whereas thetraces for standard-visual devi-ant-visual and novel-visualstimulus pairs perfectly matcheach other during the onsetslope of the P3b they start todiffer from the P3b peak forthe novel-visual stimulus pairand about 100 msec later forthe deviant-visual stimuluspair (see P3 Pz P4)

Escera et al 599

uli were used as warning signals for the following visualtargets and thus attention may have been bimodal in theAV condition However because the occurrence of an aud-itory stimulus change or a novel sound was task-irrelevantand unpredictable to the subjects attention was involun-tarily engaged by these events In a more controlledexperiment attention was more effectively withdrawnfrom the auditory stimuli and their occurrence was madenoninformative by the simultaneous presentation of avisual cue stimulus informing about the successive ap-pearance of a visual target (Alho et al 1997) Like in thepresent study a decreased hit rate for visual stimuli thatfollowed MMN- and P3a-eliciting deviant tones was foundThis conrms the involuntary nature of the attention-switching mechanism indicated by the MMN P3a anddeterioration of the visual-task performance

Component Structure of the P3a to Novel Sounds

In both the Aa and AV conditions novel sounds eliciteda large long-lasting and broadly distributed P3a wavewith a double peak at Fz (in the Aa condition) suggestinga contribution of two different subcomponents Theearly part of this P3a had a peak latency of 230 msecand showed a centrally dominant scalp distribution thatwas similar in the Aa and AV conditions (Figure 6) Theearly P3a inverted in polarity at posterior and lateralelectrodes (Figures 3b and 5) suggesting bilateral gener-ators located in the vicinity of the temporal and parietallobes (cf Scherg amp von Cramon 1986) The late part ofthe P3a in turn had a peak latency of 315 msec and didnot invert in polarity at posterior electrodes and its scalpdistribution was centered over the right frontal areas Incontrast to the early P3a the late P3a was enhanced inthe AV condition particularly over the right hemisphere(Figure 6) when subjects apparently used auditory stim-uli as warning signals to prepare for visual targets

The component structure of the present P3a and thescalp distribution of the two subcomponents is in agree-ment with previous results in patients with brain lesionssuggesting a critical role of the temporal-parietal junc-tion in the generation of the auditory (Halgren BaudenaClarke Heit Marinkovic et al 1995 Halgren BaudenaClarke Heit Lieacutegeois et al 1995 Knight Scabini Woodsamp Clayworth 1989) somatosensory (Yamaguchi ampKnight 1991 1992) and visual P3a (Knight 1991 1997)It is also compatible with results suggesting a frontalcontribution to P3a such as those obtained in patientswith frontal lesions (Knight 1984) in topographicalanalysis (Friedman amp Simpson 1994 Friedman Simpsonamp Hamberger 1993) in dipole-source modeling (Meck-linger amp Ullsperger 1995) and in intracerebral record-ings in humans (Baudena Halgren Heit amp Clarke 1995)

The late P3a amplitude in the AV condition was en-hanced with regard to its amplitude in the Aa conditionThis amplitude enhancement of the late P3a reects an

attentional monitoring of auditory stimuli in the AV con-dition these stimuli being used as warning signals for thefollowing visual targets The attentional modulation ofthe P3a amplitude is in agreement with that reported inprevious studies (Holdstock amp Rugg 1995 Woods 1992)and suggests that the late P3a may more closely reectan actual orienting of attention than the involuntary N1and MMN processes triggering this attentional orientingAlso several other ndings support the association of thelate phase of the P3a with the actual reorientation ofattention or orienting response First the right-frontaldominance in the late P3a scalp distribution observed inthe present experiment agrees with ndings suggestingthe involvement of right-frontal areas in the reorienta-tion of attention (Fuster 1989 Mesulam 1981) Secondit has been shown that the P3a amplitude is attenuatedwith the repetition of the eliciting novel event(Courchesne 1978 Friedman amp Simpson 1994 Knight1984) This habituating behavior is a characteristic fea-ture of the orienting response (Oumlhman 1979 1992 Sok-olov 1963)

The early P3a was insensitive to attentional manipula-tions as it was to similar amplitude and scalp distributionin the Aa and AV conditions Its peak latency was rathershort (230 msec) and its scalp distribution suggestsneural generators probably located in the temporal-parietal cortex (Knight 1991 1997 Knight et al 1989Yamaguchi amp Knight 1991) Thus the early P3a mightreect a neural process other than attentional reorienta-tion such as the violation of a polisensorial model of theexternal world maintained in the temporal-parietal asso-ciation cortex (Yamaguchi amp Knight 1991)

Conclusions

The results reported in the present paper support thenotion of two different brain mechanisms of involuntaryattention differentially activated by different acousticevents The ERP component structure to novel soundssuggests that unexpected novelty in the acoustic envi-ronment involuntary captures attention by simultane-ously activating two different neural mechanisms atransient-detector mechanism associated with the su-pratemporal N1 component of the auditory ERPs and achange-detector mechanism reected in the MMN Pre-sumably the attention-switching signal generated by thecombined activation of these two mechanisms resultedin an effective engagement of attention as indicated bythe large P3a component elicited by the novel soundsand the increased RT to subsequent visual targets Devi-ant tones in turn resulted in different behavioral effectsand ERP component structure which was characterizedby the MMN This supports the notion that small changesin the acoustic environment capture attention involun-tarily by activating the stimulus-change detector mecha-nism reected in the MMN

600 Journal of Cognitive Neuroscience Volume 10 Number 5

METHODS

Subjects Stimuli and Procedure

Ten right-handed paid students (21 to 40 years 3 males)with normal hearing and normal or corrected-to-normalvisual acuity participated in the experiment Two furthersubjects were discarded due to their inability to reachthe required 90 hit rate in a practice session

Subjects were presented with four blocks of 400stimulus pairs (AV condition) in a visual forced-choiceRT task Each pair consisted of an auditory stimulusfollowed after 300 msec (onset-to-onset) by a visualstimulus The interpair interval (onset-to-onset) was 12sec (Figure 1) The auditory stimuli were either standardor deviant tones and novel sounds presented in randomorder with probabilities of 08 01 and 01 respectivelyThe standard and deviant tones were sinusoidal tonebursts of 200-msec duration including 10-msec rise andfall times presented binaurally at 75 dB SPL throughheadphones with respective frequencies of 600 and 700Hz Sixty different environmental sounds such as thoseproduced by a drill hammer rain door telephone ring-ing etc were used as novel sounds They were digitallyrecorded then low-pass ltered at 10000 Hz and editedto have a duration of 200 msec including rise and falltimes of 10 msec and an intensity maximum of 70 to 80dB SPL Each different novel sound occurred only oncewithin a stimulus block and was presented twice orthree times during the whole experiment Both devianttones and novel sounds were always preceded by at leastone AV pair containing a standard tone The visual stimuliwere the digits 1 to 8 presented one at a time in randomorder on a computer screen for 200 msec They sub-tended a vertical angle of 17deg and a horizontal angle of11deg (30 mm acute 20 mm 100 cm from the subjectrsquos eyes)

Subjects were comfortably seated in a reclining chairin a dimly lit electrically and acoustically shielded roomThey were instructed to focus on the small xation crossappearing in the middle of the screen and to press oneresponse button with the right index nger for oddnumbers and another button with the right middlenger for even numbers Subjects were also instructedto ignore the auditory stimulation Both speed and accu-racy were emphasized for the primary visual task

In two further conditions visual stimuli were omitted(Va condition Figure 1) or auditory stimuli were omitted(Aa condition Figure 1) In the Va condition subjectsreceived three blocks of 200 visual stimulus each theinstruction being identical to that in the AV condition Inthe Aa condition subjects were instructed to concen-trate on reading a self-selected book and to ignore theauditory stimulation

The Aa and AV blocks were presented in an alternatingorder with half of the subjects starting with the AVcondition The Va blocks were presented one at thebeginning one at the middle and one at the end of the

experiment Before the experimental blocks subjectsparticipated in a practice session consisting of two Vablocks They were required to reach a 90 hit rate beforethe experiment could start Two out of 12 subjects failedto reach this level even after several additional blocksand were nally discarded

EEG Recording and Averaging

The electroencephalogram (EEG) (bandpass 0ndash100 Hz)was continuously digitized at a rate of 500 Hz by aSynAmps amplier (NeuroScan Inc) from 19 scalp elec-trodes positioned according to the 10ndash20 system Fp1Fp2 F7 F3 Fz F4 F8 T3 C3 Cz C4 T4 T5 P3 Pz P4 T6O1 and O2 One additional electrode was placed at theright mastoid (RM) and another at the left mastoid (LM)The horizontal EOG was recorded with electrodes at-tached to the left and right canthi The vertical elec-trooculogram (EOG) was assessed using recordings fromFp1 and Fp2 electrodes The common reference elec-trode was placed on the tip of the nose

ERPs were averaged off-line for each auditory stimulusclass for an epoch of 1300 msec including a preauditorystimulus period of 100 msec in the AV and Aa conditionsIn the Va condition an ERP was also obtained for anepoch of 1300 msec including a previsual stimulus pe-riod of 400 msec Epochs in which the EEG or EOGexceeded plusmn100 m V as well as the ve rst epochs of eachblock were automatically excluded from averaging Inthe Aa and AV conditions standard-tone trials immedi-ately following deviant-tone or novel-sound trials werealso excluded from the averages Individual ERPs wereband-pass ltered between 001 and 30 Hz

Data Analysis

In the AV and Va conditions a correct button presswithin 800 msec after visual-stimulus onset was regardedas a hit the mean RT being computed only for the hittrials An incorrect button press during this period wasclassied as an error and trials with no response withinthis time window were considered misses Hits errorsmisses and RTs were computed across odd and evennumbers

Auditory ERP amplitudes (N1 P2 MMN and P3a) werereferred to the mean amplitude in the 100-msec baselineperiod preceding the auditory stimulus in both the Aaand AV conditions For auditory ERP analysis each indi-vidual waveform in the Aa and AV conditions was linearlydetrended over an epoch of 600 msec (including base-line) to remove any possible asymmetrical contributionto EEG caused by slow EOG movements due to readingin the Aa condition The effects of the standard tones onthe visual ERPs were analyzed at O1 and O2 for peaksas follows (identied in the individual waveforms withinthe specied latency ranges) P1 75 to 140 msec N1 100

Escera et al 601

to 200 msec P2 180 to 280 msec N2 225 to 350 msecThe visual P3b was identied at Pz between 300 and 500msec from visual-stimulus onset For this analysis visualERP amplitudes were measured against the mean ampli-tude of the 100-msec baseline preceding visual-stimulusonset The fronto-polar negativity was analyzed at Fp1and Fp2 the visual frontal N2 at F3 Fz and F4 and thevisual P3b at P3 Pz and P4 The amplitude measurementsfor visual ERP components in the AV condition werereferred to the 100-msec preauditory stimulus baselinerather than to the 100-msec previsual stimulus baselinewhich was differentially affected by standard tones de-viant tones and novel sounds (see Figure 9)

The ERP and performance data were statistically ana-lyzed by ANOVA with repeated measures Where appro-priate nominal degrees of freedom and epsilon valuesafter the Greenhouse-Geisser correction are reportedScalp-distribution analyses were performed after normal-izing ERP amplitudes to prevent amplitude differencesbetween different components from washing out thegenuine scalp-distribution differences This normalizationwas done by dividing the amplitude at each electrodeby the sum of the squared amplitudes at all electrodes(McCarthy amp Wood 1985)

Acknowledgments

This research was supported by BIOMED-2 contract BMH4-CT96-0819-COBRAIN of the European Union the SpanishMinistry of Science and Education (DGICYT PB93-0802 UE96-0038) the Academy of Finland the scientic exchange programbetween the University of Barcelona and the University ofHelsinki and the National Science Fund of Hungary (OTKAT022681) The authors would like to thank Teemu Rinne andTeija Kujala for their help in making the gures

Reprint requests should be sent to Carles Escera Neurody-namics Laboratory Department of Psychiatry and Clinical Psy-chobiology University of Barcelona P Vall drsquoHebron 171 08035Barcelona Catalonia-Spain or via e-mail cescerapsiubes

Note

1 The MMN amplitude in the unattended channel can behowever attenuated under some conditions by strongly focus-ing attention on other sources of auditory stimulation(Naumlaumltaumlnen 1991 Naumlaumltaumlnen Paavilainen Tiitinen Jiang amp Alho1993 Oades amp Dittmannbalcar 1995 Trejo Ryan-Jones amp Kra-mer 1995 Woldorff Hackley amp Hillyard 1991)

REFERENCES

Alho K (1995) Cerebral generators of mismatch negativity(MMN) and its magnetic counterpart (MMNn) elicited bysound changes Ear and Hearing 16 38ndash51

Alho K Escera C Diacuteaz R Yago E amp Serra J M (1997)Effects of involuntary auditory attention on visual taskperformance and brain activity NeuroReport 8 3233ndash3237

Alho K Winkler I Escera C Huotilainen M Virtanen JJaumlaumlskelaumlinen I Pekkonen E amp Ilmoniemi R (1998) Proc-

essing of novel sounds and frequency changes in the hu-man auditory cortex Magnetoencephalographic record-ings Psychophysiology 35 211ndash224

Alho K Woods D L Algazi A amp Naumlaumltaumlnen R (1992) Inter-modal selective attention II Effects of attentional load onprocessing of auditory and visual stimuli in central spaceElectroencephalography and Clinical Neurophysiology82 356ndash368

Baudena P Halgren E Heit G amp Clarke J M (1995) Intra-cerebral potentials to rare target and distractor auditoryand visual stimuli III Frontal cortex Electroencephalogra-phy and Clinical Neurophysiology 94 251ndash264

Cherry E C (1953) Some experiments on the recognitionof speech with one and with two ears Journal of theAcoustical Society of America 29 975ndash979

Courchesne E (1978) Changes in P3 waves with event repe-tition Long-term effects on scalp distribution and ampli-tudes Electroencephalography and ClinicalNeurophysiology 45 468ndash482

Courchesne E Hillyard S A amp Galambos R (1975) Stimu-lus novelty task relevance and the visual evoked potentialsin man Electroencephalography and Clinical Neurophysi-ology 39 131ndash143

Cowan N Winkler I Teder W amp Naumlaumltaumlnen R (1993) Mem-ory prerequisites of the mismatch negativity in the audi-tory event-related potential (ERP) Journal ofExperimental Psychology Learning Memory and Cogni-tion 19 909ndash921

Cseacutepe V Karmos G amp Molnaacuter M (1987) Evoked potentialcorrelates of stimulus deviance during wakefulness andsleep in catmdashanimal model of mismatch negativity Elec-troencephalography and Clinical Neurophysiology 66571ndash578

Davis H Mast T Yoshie N amp Zerlin S (1966) The slow re-sponse of the human cortex to auditory stimulus Recov-ery process Electroencephalography and ClinicalNeurophysiology 21 105ndash113

Davis H amp Zerlin S (1966) Acoustic relations of the humanvertex potentials Journal of the Acoustical Society ofAmerica 39 109ndash116

Dawson M E (1990) Psychophysiology at the interface ofclinical science cognitive science and neuroscience Psy-chophysiology 27 243ndash255

Dawson M E Fillion D L amp Schell A M (1989) Is elicita-tion of the autonomic orienting response associated withallocation of processing resources Psychophysiology 26560ndash572

Duncan C C amp Kaye W H (1987) Effects of chlonidine onevent-related potential measures of information process-ing In R Johnson J W Rohrbaugh amp R Parasuraman(Eds) Current trends in event-related potentialresearch (EEG supplement 40) (pp 527ndash531) Amster-dam Elsevier

Fillion D J Dawson M E Schell A M amp Hazlett E A(1991) The relationship between skin conductance orient-ing and the allocation of processing resources Psycho-physiology 28 410ndash424

Folk C L Remington R W amp Johnston J C (1992) Involun-tary covert orienting is contingent on attentional controlsettings Journal of Experimental Psychology Human Per-ception and Performance 18 1030ndash1044

Ford J M Roth W T amp Kopell B S (1976) Auditory event-related potentials to unpredictable shifts in pitch Psycho-physiology 13 32ndash39

Friedman D amp Simpson G (1994) ERP amplitude and scalpdistribution to target and novel events Effects of temporalorder in young middle-aged and older adults CognitiveBrain Research 2 49ndash63

602 Journal of Cognitive Neuroscience Volume 10 Number 5

with orienting of attention is the P3a This positive com-ponent is elicited by irrelevant rare tones and novelsounds in a sequence of repetitive standard tones andcan be distinguished from the P300 or P3b componentto target stimuli by its shorter latency and its morefrontal scalp distribution (Courchesne Hillyard amp Galam-bos 1975 Ford Roth amp Kopell 1976 Squires Squires ampHillyard 1975) The association of the P3a-generatingprocess with the orienting response is supported by itselicitation by widely deviant or novel sounds and also bythe frontal-lobe and hippocampal contributions to itsgeneration (Knight 1984 1996) these brain regions be-ing involved also in the orienting response (Fuster 1989Sokolov 1975) The role of the P3a generator process ininvoluntary engagement of attention is further sup-ported by studies showing delayed RTs to target stimulifollowing irrelevant novel sounds (Grillon CourchesneAmeli Geyer amp Braff 1990 Woods 1992) When novelsounds occurred in the attended sequences containingthe target stimuli the P3a elicited was larger in ampli-tude as compared with the unattended-sequence P3a tonovel sounds and the RT to consecutive targets wasdelayed (Woods 1992) This attentional modulation sug-gests that the P3a reects an actual reorientation ofattention more closely than the MMN or the differentsubcomponents of the N1 which would rather be asso-ciated with the involuntary call for attention

The purpose of the present experiment was to inves-tigate the neural mechanisms of involuntary attention toacoustic novelty and change by analyzing the ERP con-comitants elicited by these irrelevant auditory eventsduring visual performance In the main experimentalcondition (Figure 1) the Auditory-Visual condition (AV)a response was required to each visual stimulus of anauditory-visual stimulus pair Subjects were instructed toignore the auditory stimuli which were a repetitiveldquostandardrdquo tone randomly replaced by a slightly higherldquodeviantrdquo tone (p = 01) or by a natural ldquonovelrdquo sound(p = 01) and to press a response button to even num-bers and another to odd numbers Two additional condi-tions were used as controls In the Visual-alonecondition (Va) the auditory stimuli were omitted andsubjects performed the forced-choice visual RT task Inthe Auditory-alone (Aa) condition the visual stimuliwere omitted and subjects were instructed to concen-trate on reading a book

RESULTS

Performance

In the Va condition the mean hit rate was 915 (SEM plusmn17) and the mean RT was 461 plusmn 11 msec Subjectsmissed the target on an average 45 plusmn 16 of thetrials and pressed the wrong button in 40 plusmn 08 ofthe trials

Performance data in the Va and AV conditions areshown in Figure 2 In the AV condition subjects tendedto respond faster to visual stimuli preceded by a standardtone than to those preceded by no sound (Va condition)(444 plusmn 9 and 461 plusmn 11 msec respectively) although thiseffect did not quite reach statistical signicance (F(19) = 38 p lt 009) The hit error and miss rates weresimilar in the Va condition and in the AV condition aftera standard tone In the AV condition the occurrence ofa deviant tone before the visual stimulus caused a sig-nicant hit rate decrease of 23 (F(1 9) = 650 p lt004) which was probably due to an increased numberof wrong responses to target stimuli (F(1 9) = 478 p lt006 see Figure 2) the number of missing responsesbeing identical after standard and deviant tones A novelsound occurring before a visual stimulus caused an av-erage increase of 17 msec in the RT to visual stimulicompared with the RT to these stimuli that were pre-ceded by standard tones (F(1 9) = 2219 p lt 0002) TheRT to visual stimuli after novel sounds was also sig-nicantly longer than that to visual stimuli followingdeviant tones (F(19) = 1107 p lt 0009)

Figure 1 Experimental conditions In the main condition [Auditory-Visual (AV) top] pairs of stimuli consisting of an auditory stimulusfollowed at 300 msec (onset-to-onset) by a visual stimulus (an arabicnumber between 1 and 8) were presented with an interpair intervalof 1200 msec The auditory stimulus was either a standard tone(80) a deviant tone (10) or a novel sound (10) Subjects wereinstructed to press one response button to odd (50) and the otherto even (50) numbers and to ignore auditory stimulation In theVisual-alone (Va) condition the auditory stimuli were omitted In theAuditory-alone (Aa) condition the visual stimuli were omitted thesubject being instructed to read a self-selected book and to ignorethe auditory stimuli

592 Journal of Cognitive Neuroscience Volume 10 Number 5

Auditory ERPs

In the Aa condition standard tones elicited N1 (meanpeak amplitude - 20 m V at Cz with a mean peak latencyof 93 msec) and P2 deections (34 m V 158 msec at Cz)which were largest over the fronto-central scalp loca-tions (Figure 3a) Deviant tones elicited N1 and P2 wavesat latencies similar to those in response to standardtones and the mismatch negativity (MMN) The MMN

was largest over the frontal electrodes (mean peak am-plitude - 22 m V mean peak latency 153 msec at Fz) andinverted in polarity at the mastoid electrodes (Figure 3b)being signicantly different from zero at F3 Fz and F4in the intervals 100 to 150 msec (F(1 9) = 1031 p lt002) and 150 to 200 msec (F(1 9) = 563 p lt 005) TheMMN was followed by a smaller fronto-centrally distrib-uted negative wave peaking at around 400 msec (Figure3b) Mean amplitude comparisons between ERPs to de-viant and standard tones at Fz and Cz revealed that thissecond negativity was signicant in the 350- to 400-msec(F(1 9) = 756 p lt 003) and in the 400- to 450-msec(F(1 9) = 873 p lt 002) intervals In the AV conditionthe MMN was similar to that obtained in the Aa condition(Figure 4)

In the Aa condition novel sounds elicited an ERPcharacterized by a prominent N1 deection (mean peakamplitude at Cz - 56 m V peak latency 96 msec) and bya long-duration broadly distributed P3a wave (Figure 3as seen in Figure 3b even the deviant tones elicitedsome P3a) The novel-sound N1 was larger than thestandard-tone N1 (F(1 9) = 1099 p lt 001) and wasprobably composed of overlapping N1 and MMN com-ponents The difference between the standard-tone andnovel-sound ERPs extended beyond the N1 latencyrange there being a signicant difference at Fz betweenthe standard and novel ERP amplitudes measured asmean voltages over 120 to 160 msec latency window(F(1 9) = 523 p lt 005 Figure 5)

The present P3a to novel sounds had a double peakover the frontal scalp locations (Figure 3b) suggestingtwo different subcomponents The early part of the P3a(peak latency 230 msec in the grand-mean ERP at Fz)was largest over the central scalp areas appearing withinverted polarity at lateral and posterior scalp sites (Fig-ures 3b and 5) The late part of the P3a (peak latency315 msec in the grand-mean ERP at Fz) in turn waslargest over frontal areas and did not show polarityinversion at any recording site (Figures 3b and 5) Thedifferent scalp distributions of the two P3a subcompo-nents in the Aa condition are shown in Figure 6 (top)

The component structure of the P3a deection wasevaluated by means of two different analyses of variance(ANOVA) for normalized ERP amplitudes (McCarthy ampWood 1985) First an ANOVA was performed to prove apossible scalp-distribution difference between the P2 tostandard tones and the early P3a to novel sounds Thenanother ANOVA was performed to conrm the differentscalp distributions of the two subcomponents of the P3aThe P2 amplitude was identied as the mean voltagebetween 150 and 200 msec from stimulus onset in thestandard-tone ERP The early P3a was measured as themean voltage between 175 and 275 msec in the differ-ence wave obtained by subtracting the ERP to the stan-dard tone from that to the novel sound The late P3a wasmeasured as the mean voltage between 275 and 375msec of the same difference wave A three-factor design

Figure 2 Mean RT (top panel) hit rate (middle panel) and errorrate (bottom panel) in the Visual-alone (Va) condition and in theAuditory-Visual (AV) condition to the visual stimulus occurring afterthe standard tone (std) the deviant tone (dev) or the novel sound(nov) The bars indicate the standard error of mean

Escera et al 593

was used for both analyses Wave (P2 versus early P3afor the rst analysis and early P3a versus late P3a for thesecond analysis) acute Frontality (F7 F3 Fz F4 F8 versus T3C3 Cz C4 T4 versus T5 P3 Pz P4 T6) acute Laterality (F7T3 T5 versus F3 C3 P3 versus Fz Cz Pz versus F4 C4P4 versus F8 T4 T6) The results of the rst analysisrevealed a signicant Wave acute Frontality interaction (F(218) = 509 p lt 003 e = 07547) due to the early P3abeing posterior to the P2 and a signicant Wave acuteLaterality interaction (F(4 36) = 406 p lt 003 e =05469) due to the more predominantly right-hemi-spheric scalp distribution of the P2 than that of the earlyP3a The results of the second analysis revealed that thelate P3a was anterior in distribution to the early P3a(signicant Wave Frontality interaction F(2 18) = 939p lt 0006 e = 07187) and further that the late P3a wasmore preponderant over the right hemisphere than the

early P3a (signicant Wave acute Laterality interactionF(4 36) = 1207 p lt 00002 e = 05740) The secondanalysis also revealed a signicant third-order interaction(Wave acute Frontality acute Laterality F(8 72) = 898 p lt00004 e = 03441) that was due to the predominantlyright-frontal distribution of the late phase of the P3adeection (Figure 6 top)

In the AV condition novel sounds elicited ERPs similarto those elicited by them in the Aa condition (large N1and P3a deections) although the early peak of the P3aseemed to be partly abolished at frontal electrodes (Fig-ure 7) However difference waves (ERP to novel soundsminus ERP to standard tones) revealed that the earlyphase of the P3a was similar in the Aa and AV conditionswhereas the late P3a was enhanced in the AV condition(Figure 7) The statistical scalp-distribution analyses ofthe two P3a phases across the conditions [ANOVA with

Figure 3a Grand-mean ERPsaveraged across subjects tostandard tones (600 Hz) devi-ant tones (700 Hz) and novelsounds in the Auditory-alone(Aa) condition

594 Journal of Cognitive Neuroscience Volume 10 Number 5

four factors Condition (Aa versus AV) acute Wave (early P3aversus late P3a) acute Frontality acute Laterality] conrmed thescalp-distribution difference between the two P3aphases (Wave acute Frontality interaction F(2 18) = 996p lt 0003 e = 08361 Wave acute Laterality F(4 36) = 688p lt 002 e = 03977 Wave acute Frontality acute LateralityF(872) = 981 p lt 00001 acute = 04381) and also revealeda signicant increase of the late P3a amplitude over theright hemisphere in the AV condition (Wave acute Condi-tion acute Laterality F(4 36) = 384 p lt 004 e = 05367)

Visual ERPs

Visual stimuli in the Va condition and those after stan-dard tones in the AV condition elicited almost identicalERPs characterized by P1 N1 P2 N2 and P3 deections(Figure 8) ANOVA for these visual ERP components with

Condition (after standard in AV versus Va) as a functionrevealed no signicant effects of standard tones on visualERPs

Figure 9 shows grand-average ERPs to auditory-visualstimulus pairs in the AV condition The AV stimulus pairselicited an auditory ERP lasting up to 400 msec post-stimulus for novel sounds and up to 300 msec post-stimulus for deviant tones followed by a visual ERPsimilar to that obtained in the Va condition All threetypes of AV stimulus pairs (standard-visual deviant-visualnovel-visual) were also characterized by a long-lastingfronto-polar negativity that was larger to visual stimulifollowing novel sounds than to those following standardor deviant tones (Figure 9) ANOVA for mean amplitudesat Fp1 and Fp2 in consecutive 100-msec latency win-dows yielded signicant effects for the novel-standardcomparison (F(1 9) ranging from 779 to 1196 p lt 003

Figure 3b Grand-average dif-ference waves obtained in theAa condition by subtractingthe ERP to standard tonesfrom that to deviant tones(continuous line) and tonovel sounds (dashed line)

Escera et al 595

in all cases) at latency windows 200 to 300 msec 300to 400 msec and 400 to 500 msec from visual-stimulusonset

Novel sounds and deviant tones enhanced the frontalN2 deection elicited by the subsequent visual stimulusin comparison with the visual N2 elicited after standardtones (Figure 9) This was revealed by two-way ANOVAsusing the Electrode (F3 Fz F4) and the auditory Stimulus(standard versus either deviant or novel) as factors Themean amplitude in the 200- to 300-msec range fromvisual-stimulus onset at Fz was - 19 m V after standardtones - 33 m V after deviant tones (F(1 9) = 641 p lt004) and - 36 m V after novel sounds (F(1 9) = 899 plt 002) The mean amplitude in the 300- to 400-msecinterval was also signicantly larger for visual stimulifollowing deviant tones or novel sounds than for thosefollowing standard tones F(1 9) = 815 p lt 002 for

deviant versus standard F(1 9) = 795 p lt 001 for novelversus standard

Preceding novel sounds and deviant tones also en-hanced the terminal slope of the visual P3b (Figure 9)this was revealed by the signicantly larger mean ampli-tude of the 500- to 600-msec latency window (at P3 Pzand P4) in the ERP to visual stimuli following devianttones than in the ERP to visual stimuli following standardtones (F(19) = 1556 p lt 004) The mean amplitude inthe 400- to 500-msec (F(1 9) = 694 p lt 003) and 500-to 600-msec (F(1 9) = 1199 p lt 0008) intervals was

Figure 4 The MMN was similar in the Auditory-alone (Aa) andAuditory-Visual (AV) conditions as revealed by nonsignicant differ-ences of the mean amplitude of ve consecutive 50-msec intervalsstarting at MMN onset (50 msec) [two-way ANOVAs Condition (Aaversus AV) acute Electrode (F3 Fz F4 C3 Cz C4)] Top Grand-averageERPs to standard and deviant tones at Fz in the two conditionsBottom Deviant-standard difference waves

Figure 5 The component structure of the ERPs to deviant tonesand novel sounds as revealed by the deviant-standard and novel-stan-dard difference waves at Fz and right-mastoid (RM) recordings Devi-ant tones elicited the MMN with a peak latency of 150 msec at Fzand with an inverted polarity at RM (top) Novel sounds elicited alarge N1 deection at Fz that was presumably composed of contribu-tions from two different processes for which the signicant differ-ences between the standard-tone and novel-sound ERPs extendedbeyond the N1 latency range (bottom) The novel sounds also elic-ited a P3a wave with two peaks at Fz suggesting two different sub-components contributing to the recorded P3a an early P3a (peaklatency 230 msec) which inverted polarity at RM and other poste-rior and lateral sites and a late P3a (peak latency 315 msec) whichinverted polarity at no recording site

596 Journal of Cognitive Neuroscience Volume 10 Number 5

also signicantly larger for visual stimuli following novelsounds than after those following the standard tones

DISCUSSION

Engagement of Attention During VisualPerformance

The present results provide evidence that attention dur-ing visual performance can be engaged by shortly pre-ceding unexpected novel sounds or small changes in thetask-irrelevant acoustic environment These results alsorevealed that this phenomenon depends on the intrinsicnature of the eliciting acoustic events In addition the

observed effects on visual performance were associatedwith different patterns of neural events activated bydeviant tones and novel sounds as reected by thecomponent structure of the auditory ERPs elicited

As expected novel sounds prolonged the RT to suc-cessive visual stimuli This result is in agreement with theextensive literature reporting delayed RTs to target stim-uli caused by preceding irrelevant sound changes ornovels sounds (Alho et al 1992 Cherry 1953 Dawsonet al 1989 Fillion et al 1991 Grillon et al 1990Schroumlger 1996 Woods 1992 Woodward et al 1991) andit provides experimental support for the everyday expe-rience that attention can be engaged by irrelevant acous-tic events

Figure 6 The scalp distribu-tion of the P3a to novelsounds Left early P3a rightlate P3a top Auditory-alone(Aa) condition bottom Audi-tory-Visual (AV) condition Thecentral plot shows the novel-standard difference waves atFz in the Aa (thin) and AV(thick) conditions as well asthe 100-msec latency win-dows used to measure thetwo phases of the P3a (e =early P3a l = late P3a)

Escera et al 597

A surprising nding of this study was however thatthe slightly deviant tones degraded the visual perfor-mance accuracy by decreasing the rate of hit responseswhereas the widely deviant novel sounds did not It wasexpected that the amount of distraction caused by devi-ant tones and novel sounds would be proportional tothe magnitude of their deviance from the standard toneand indeed such an effect was found for the RT (17-msecincrease after novel sounds) conrming other recentndings (Jaumlaumlskelaumlinen et al 1996 Schroumlger 1996) How-ever this RT effect was not paralleled by a similar effecton the hit rate Indeed deviant tones were associatedwith a lower hit rate whereas the hit rate after novelsounds was similar to that after standard tones Similarresults were obtained in a related study by Jaumlaumlskelaumlinenet al (1996) and recently replicated for deviant tones(Alho Escera Diacuteaz Yago amp Serra 1997) These ndingssuggest that the distracting effects of an irrelevant acous-tic event on visual performance may depend on thenature of the eliciting sound That is obtrusive novelsounds appear to cause a stronger effect on the RT tosuccessive visual target stimuli than minor changes inthe acoustic stream but in contrast these minor changes

seem to cause a signicant decrement in the number ofhit responses to visual targets not observed after thenovel sounds

The differential effects of distracting novel and deviantsounds on visual performance may be explained by trac-ing these behavioral effects back to the associated neuralevents as reected by ERPs Novel sounds enhanced theN1 amplitude and elicited the MMN as suggested bysignicant differences between the standard-tone andnovel-sound ERP that extended over the N1 latencyrange The present MMN to novel sounds peaked andterminated (returned to baseline) earlier than the MMNto deviant tones in concordance with previous studiesshowing that with larger deviances MMN becomes ear-lier (Tiitinen May Reinikainen amp Naumlaumltaumlnen 1994) over-lapping the N1 peak for wider deviances (Tiitinen et al1994 Scherg et al 1989) The present MMN to novelsounds is also supported by recent magnetoencepha-lograph (MEG) data showing that novel sounds activatedan MMN source in the human auditory cortex (Alhoet al 1998) Thus for novel sounds the attention-switch-ing signal was probably triggered by a combined re-sponse of the transient-detector system reected in thesupratemporal N1 (Naumlaumltaumlnen 1990 1992 Naumlaumltaumlnen ampPicton 1987) or some other N1 component such as thefrontal N1 reported by Giard et al (1994) and the stimu-lus-change detector system reected in the MMN(Naumlaumltaumlnen 1990 1992) This combined signal for atten-tion capture started with the N1 generation and wasoverlapped by the MMN process resulting in an effectiveattentional reorientation as reected by the subsequentlarge P3a wave and the clearly delayed RT to the follow-ing visual targets

Figure 7 Grand-average ERPs to standard tones and novel soundsin the Auditory-alone (Aa left column) and Auditory-Visual (AV rightcolumn) conditions The novel-standard difference waves are shownin the middle column The late P3a was larger in the AV conditionthan in the Aa condition whereas the early P3a was of similar ampli-tude in the two conditions

Figure 8 Grand-average visual ERPs at Pz O1 and O2 in the Visualalone (Va) condition and in the Auditory-Visual (AV) condition afterstandard tones

598 Journal of Cognitive Neuroscience Volume 10 Number 5

Deviant tones in turn elicited a distinct MMN whichwas followed by a small P3a wave and by a secondnegative peak (Figure 5) Thus the neural mechanismresponsible for triggering the attention-switching signalwas based for deviant tones on the stimulus-changedetector system reected in the MMN (Naumlaumltaumlnen 19901992 Schroumlger 1996) This attention capture signal asso-ciated with the MMN was sporadically followed by an

orienting of attention as suggested by the small (aver-aged) P3a and also by the 23 hit rate decrease to visualtargets that followed deviant tones

Although the subjects were instructed to ignore theauditory stimuli in the AV condition the RT to visualtargets following standard tones was slightly (insig-nicantly) shorter than the RT in the Va condition withno auditory stimuli This suggests that the auditory stim-

Figure 9 Grand-averageERPs in the Auditory-Visual(AV) condition to standard-visual (thin line) deviant-visual (thick line) andnovel-visual (dashed line)stimulus pairs Only the mostrelevant electrodes are shownThe epoch starts 100 msec be-fore auditory-stimulus onsetwhich was used as the base-line period to avoid the largeamplitude variability due tothe auditory response in the100 msec preceding the vis-ual stimulus Novel sounds en-hanced the fronto-polarnegativity and the frontal N2to visual stimuli was en-hanced after novel soundsand deviant tones which alsoaffected the offset slope ofP3b Notice that whereas thetraces for standard-visual devi-ant-visual and novel-visualstimulus pairs perfectly matcheach other during the onsetslope of the P3b they start todiffer from the P3b peak forthe novel-visual stimulus pairand about 100 msec later forthe deviant-visual stimuluspair (see P3 Pz P4)

Escera et al 599

uli were used as warning signals for the following visualtargets and thus attention may have been bimodal in theAV condition However because the occurrence of an aud-itory stimulus change or a novel sound was task-irrelevantand unpredictable to the subjects attention was involun-tarily engaged by these events In a more controlledexperiment attention was more effectively withdrawnfrom the auditory stimuli and their occurrence was madenoninformative by the simultaneous presentation of avisual cue stimulus informing about the successive ap-pearance of a visual target (Alho et al 1997) Like in thepresent study a decreased hit rate for visual stimuli thatfollowed MMN- and P3a-eliciting deviant tones was foundThis conrms the involuntary nature of the attention-switching mechanism indicated by the MMN P3a anddeterioration of the visual-task performance

Component Structure of the P3a to Novel Sounds

In both the Aa and AV conditions novel sounds eliciteda large long-lasting and broadly distributed P3a wavewith a double peak at Fz (in the Aa condition) suggestinga contribution of two different subcomponents Theearly part of this P3a had a peak latency of 230 msecand showed a centrally dominant scalp distribution thatwas similar in the Aa and AV conditions (Figure 6) Theearly P3a inverted in polarity at posterior and lateralelectrodes (Figures 3b and 5) suggesting bilateral gener-ators located in the vicinity of the temporal and parietallobes (cf Scherg amp von Cramon 1986) The late part ofthe P3a in turn had a peak latency of 315 msec and didnot invert in polarity at posterior electrodes and its scalpdistribution was centered over the right frontal areas Incontrast to the early P3a the late P3a was enhanced inthe AV condition particularly over the right hemisphere(Figure 6) when subjects apparently used auditory stim-uli as warning signals to prepare for visual targets

The component structure of the present P3a and thescalp distribution of the two subcomponents is in agree-ment with previous results in patients with brain lesionssuggesting a critical role of the temporal-parietal junc-tion in the generation of the auditory (Halgren BaudenaClarke Heit Marinkovic et al 1995 Halgren BaudenaClarke Heit Lieacutegeois et al 1995 Knight Scabini Woodsamp Clayworth 1989) somatosensory (Yamaguchi ampKnight 1991 1992) and visual P3a (Knight 1991 1997)It is also compatible with results suggesting a frontalcontribution to P3a such as those obtained in patientswith frontal lesions (Knight 1984) in topographicalanalysis (Friedman amp Simpson 1994 Friedman Simpsonamp Hamberger 1993) in dipole-source modeling (Meck-linger amp Ullsperger 1995) and in intracerebral record-ings in humans (Baudena Halgren Heit amp Clarke 1995)

The late P3a amplitude in the AV condition was en-hanced with regard to its amplitude in the Aa conditionThis amplitude enhancement of the late P3a reects an

attentional monitoring of auditory stimuli in the AV con-dition these stimuli being used as warning signals for thefollowing visual targets The attentional modulation ofthe P3a amplitude is in agreement with that reported inprevious studies (Holdstock amp Rugg 1995 Woods 1992)and suggests that the late P3a may more closely reectan actual orienting of attention than the involuntary N1and MMN processes triggering this attentional orientingAlso several other ndings support the association of thelate phase of the P3a with the actual reorientation ofattention or orienting response First the right-frontaldominance in the late P3a scalp distribution observed inthe present experiment agrees with ndings suggestingthe involvement of right-frontal areas in the reorienta-tion of attention (Fuster 1989 Mesulam 1981) Secondit has been shown that the P3a amplitude is attenuatedwith the repetition of the eliciting novel event(Courchesne 1978 Friedman amp Simpson 1994 Knight1984) This habituating behavior is a characteristic fea-ture of the orienting response (Oumlhman 1979 1992 Sok-olov 1963)

The early P3a was insensitive to attentional manipula-tions as it was to similar amplitude and scalp distributionin the Aa and AV conditions Its peak latency was rathershort (230 msec) and its scalp distribution suggestsneural generators probably located in the temporal-parietal cortex (Knight 1991 1997 Knight et al 1989Yamaguchi amp Knight 1991) Thus the early P3a mightreect a neural process other than attentional reorienta-tion such as the violation of a polisensorial model of theexternal world maintained in the temporal-parietal asso-ciation cortex (Yamaguchi amp Knight 1991)

Conclusions

The results reported in the present paper support thenotion of two different brain mechanisms of involuntaryattention differentially activated by different acousticevents The ERP component structure to novel soundssuggests that unexpected novelty in the acoustic envi-ronment involuntary captures attention by simultane-ously activating two different neural mechanisms atransient-detector mechanism associated with the su-pratemporal N1 component of the auditory ERPs and achange-detector mechanism reected in the MMN Pre-sumably the attention-switching signal generated by thecombined activation of these two mechanisms resultedin an effective engagement of attention as indicated bythe large P3a component elicited by the novel soundsand the increased RT to subsequent visual targets Devi-ant tones in turn resulted in different behavioral effectsand ERP component structure which was characterizedby the MMN This supports the notion that small changesin the acoustic environment capture attention involun-tarily by activating the stimulus-change detector mecha-nism reected in the MMN

600 Journal of Cognitive Neuroscience Volume 10 Number 5

METHODS

Subjects Stimuli and Procedure

Ten right-handed paid students (21 to 40 years 3 males)with normal hearing and normal or corrected-to-normalvisual acuity participated in the experiment Two furthersubjects were discarded due to their inability to reachthe required 90 hit rate in a practice session

Subjects were presented with four blocks of 400stimulus pairs (AV condition) in a visual forced-choiceRT task Each pair consisted of an auditory stimulusfollowed after 300 msec (onset-to-onset) by a visualstimulus The interpair interval (onset-to-onset) was 12sec (Figure 1) The auditory stimuli were either standardor deviant tones and novel sounds presented in randomorder with probabilities of 08 01 and 01 respectivelyThe standard and deviant tones were sinusoidal tonebursts of 200-msec duration including 10-msec rise andfall times presented binaurally at 75 dB SPL throughheadphones with respective frequencies of 600 and 700Hz Sixty different environmental sounds such as thoseproduced by a drill hammer rain door telephone ring-ing etc were used as novel sounds They were digitallyrecorded then low-pass ltered at 10000 Hz and editedto have a duration of 200 msec including rise and falltimes of 10 msec and an intensity maximum of 70 to 80dB SPL Each different novel sound occurred only oncewithin a stimulus block and was presented twice orthree times during the whole experiment Both devianttones and novel sounds were always preceded by at leastone AV pair containing a standard tone The visual stimuliwere the digits 1 to 8 presented one at a time in randomorder on a computer screen for 200 msec They sub-tended a vertical angle of 17deg and a horizontal angle of11deg (30 mm acute 20 mm 100 cm from the subjectrsquos eyes)

Subjects were comfortably seated in a reclining chairin a dimly lit electrically and acoustically shielded roomThey were instructed to focus on the small xation crossappearing in the middle of the screen and to press oneresponse button with the right index nger for oddnumbers and another button with the right middlenger for even numbers Subjects were also instructedto ignore the auditory stimulation Both speed and accu-racy were emphasized for the primary visual task

In two further conditions visual stimuli were omitted(Va condition Figure 1) or auditory stimuli were omitted(Aa condition Figure 1) In the Va condition subjectsreceived three blocks of 200 visual stimulus each theinstruction being identical to that in the AV condition Inthe Aa condition subjects were instructed to concen-trate on reading a self-selected book and to ignore theauditory stimulation

The Aa and AV blocks were presented in an alternatingorder with half of the subjects starting with the AVcondition The Va blocks were presented one at thebeginning one at the middle and one at the end of the

experiment Before the experimental blocks subjectsparticipated in a practice session consisting of two Vablocks They were required to reach a 90 hit rate beforethe experiment could start Two out of 12 subjects failedto reach this level even after several additional blocksand were nally discarded

EEG Recording and Averaging

The electroencephalogram (EEG) (bandpass 0ndash100 Hz)was continuously digitized at a rate of 500 Hz by aSynAmps amplier (NeuroScan Inc) from 19 scalp elec-trodes positioned according to the 10ndash20 system Fp1Fp2 F7 F3 Fz F4 F8 T3 C3 Cz C4 T4 T5 P3 Pz P4 T6O1 and O2 One additional electrode was placed at theright mastoid (RM) and another at the left mastoid (LM)The horizontal EOG was recorded with electrodes at-tached to the left and right canthi The vertical elec-trooculogram (EOG) was assessed using recordings fromFp1 and Fp2 electrodes The common reference elec-trode was placed on the tip of the nose

ERPs were averaged off-line for each auditory stimulusclass for an epoch of 1300 msec including a preauditorystimulus period of 100 msec in the AV and Aa conditionsIn the Va condition an ERP was also obtained for anepoch of 1300 msec including a previsual stimulus pe-riod of 400 msec Epochs in which the EEG or EOGexceeded plusmn100 m V as well as the ve rst epochs of eachblock were automatically excluded from averaging Inthe Aa and AV conditions standard-tone trials immedi-ately following deviant-tone or novel-sound trials werealso excluded from the averages Individual ERPs wereband-pass ltered between 001 and 30 Hz

Data Analysis

In the AV and Va conditions a correct button presswithin 800 msec after visual-stimulus onset was regardedas a hit the mean RT being computed only for the hittrials An incorrect button press during this period wasclassied as an error and trials with no response withinthis time window were considered misses Hits errorsmisses and RTs were computed across odd and evennumbers

Auditory ERP amplitudes (N1 P2 MMN and P3a) werereferred to the mean amplitude in the 100-msec baselineperiod preceding the auditory stimulus in both the Aaand AV conditions For auditory ERP analysis each indi-vidual waveform in the Aa and AV conditions was linearlydetrended over an epoch of 600 msec (including base-line) to remove any possible asymmetrical contributionto EEG caused by slow EOG movements due to readingin the Aa condition The effects of the standard tones onthe visual ERPs were analyzed at O1 and O2 for peaksas follows (identied in the individual waveforms withinthe specied latency ranges) P1 75 to 140 msec N1 100

Escera et al 601

to 200 msec P2 180 to 280 msec N2 225 to 350 msecThe visual P3b was identied at Pz between 300 and 500msec from visual-stimulus onset For this analysis visualERP amplitudes were measured against the mean ampli-tude of the 100-msec baseline preceding visual-stimulusonset The fronto-polar negativity was analyzed at Fp1and Fp2 the visual frontal N2 at F3 Fz and F4 and thevisual P3b at P3 Pz and P4 The amplitude measurementsfor visual ERP components in the AV condition werereferred to the 100-msec preauditory stimulus baselinerather than to the 100-msec previsual stimulus baselinewhich was differentially affected by standard tones de-viant tones and novel sounds (see Figure 9)

The ERP and performance data were statistically ana-lyzed by ANOVA with repeated measures Where appro-priate nominal degrees of freedom and epsilon valuesafter the Greenhouse-Geisser correction are reportedScalp-distribution analyses were performed after normal-izing ERP amplitudes to prevent amplitude differencesbetween different components from washing out thegenuine scalp-distribution differences This normalizationwas done by dividing the amplitude at each electrodeby the sum of the squared amplitudes at all electrodes(McCarthy amp Wood 1985)

Acknowledgments

This research was supported by BIOMED-2 contract BMH4-CT96-0819-COBRAIN of the European Union the SpanishMinistry of Science and Education (DGICYT PB93-0802 UE96-0038) the Academy of Finland the scientic exchange programbetween the University of Barcelona and the University ofHelsinki and the National Science Fund of Hungary (OTKAT022681) The authors would like to thank Teemu Rinne andTeija Kujala for their help in making the gures

Reprint requests should be sent to Carles Escera Neurody-namics Laboratory Department of Psychiatry and Clinical Psy-chobiology University of Barcelona P Vall drsquoHebron 171 08035Barcelona Catalonia-Spain or via e-mail cescerapsiubes

Note

1 The MMN amplitude in the unattended channel can behowever attenuated under some conditions by strongly focus-ing attention on other sources of auditory stimulation(Naumlaumltaumlnen 1991 Naumlaumltaumlnen Paavilainen Tiitinen Jiang amp Alho1993 Oades amp Dittmannbalcar 1995 Trejo Ryan-Jones amp Kra-mer 1995 Woldorff Hackley amp Hillyard 1991)

REFERENCES

Alho K (1995) Cerebral generators of mismatch negativity(MMN) and its magnetic counterpart (MMNn) elicited bysound changes Ear and Hearing 16 38ndash51

Alho K Escera C Diacuteaz R Yago E amp Serra J M (1997)Effects of involuntary auditory attention on visual taskperformance and brain activity NeuroReport 8 3233ndash3237

Alho K Winkler I Escera C Huotilainen M Virtanen JJaumlaumlskelaumlinen I Pekkonen E amp Ilmoniemi R (1998) Proc-

essing of novel sounds and frequency changes in the hu-man auditory cortex Magnetoencephalographic record-ings Psychophysiology 35 211ndash224

Alho K Woods D L Algazi A amp Naumlaumltaumlnen R (1992) Inter-modal selective attention II Effects of attentional load onprocessing of auditory and visual stimuli in central spaceElectroencephalography and Clinical Neurophysiology82 356ndash368

Baudena P Halgren E Heit G amp Clarke J M (1995) Intra-cerebral potentials to rare target and distractor auditoryand visual stimuli III Frontal cortex Electroencephalogra-phy and Clinical Neurophysiology 94 251ndash264

Cherry E C (1953) Some experiments on the recognitionof speech with one and with two ears Journal of theAcoustical Society of America 29 975ndash979

Courchesne E (1978) Changes in P3 waves with event repe-tition Long-term effects on scalp distribution and ampli-tudes Electroencephalography and ClinicalNeurophysiology 45 468ndash482

Courchesne E Hillyard S A amp Galambos R (1975) Stimu-lus novelty task relevance and the visual evoked potentialsin man Electroencephalography and Clinical Neurophysi-ology 39 131ndash143

Cowan N Winkler I Teder W amp Naumlaumltaumlnen R (1993) Mem-ory prerequisites of the mismatch negativity in the audi-tory event-related potential (ERP) Journal ofExperimental Psychology Learning Memory and Cogni-tion 19 909ndash921

Cseacutepe V Karmos G amp Molnaacuter M (1987) Evoked potentialcorrelates of stimulus deviance during wakefulness andsleep in catmdashanimal model of mismatch negativity Elec-troencephalography and Clinical Neurophysiology 66571ndash578

Davis H Mast T Yoshie N amp Zerlin S (1966) The slow re-sponse of the human cortex to auditory stimulus Recov-ery process Electroencephalography and ClinicalNeurophysiology 21 105ndash113

Davis H amp Zerlin S (1966) Acoustic relations of the humanvertex potentials Journal of the Acoustical Society ofAmerica 39 109ndash116

Dawson M E (1990) Psychophysiology at the interface ofclinical science cognitive science and neuroscience Psy-chophysiology 27 243ndash255

Dawson M E Fillion D L amp Schell A M (1989) Is elicita-tion of the autonomic orienting response associated withallocation of processing resources Psychophysiology 26560ndash572

Duncan C C amp Kaye W H (1987) Effects of chlonidine onevent-related potential measures of information process-ing In R Johnson J W Rohrbaugh amp R Parasuraman(Eds) Current trends in event-related potentialresearch (EEG supplement 40) (pp 527ndash531) Amster-dam Elsevier

Fillion D J Dawson M E Schell A M amp Hazlett E A(1991) The relationship between skin conductance orient-ing and the allocation of processing resources Psycho-physiology 28 410ndash424

Folk C L Remington R W amp Johnston J C (1992) Involun-tary covert orienting is contingent on attentional controlsettings Journal of Experimental Psychology Human Per-ception and Performance 18 1030ndash1044

Ford J M Roth W T amp Kopell B S (1976) Auditory event-related potentials to unpredictable shifts in pitch Psycho-physiology 13 32ndash39

Friedman D amp Simpson G (1994) ERP amplitude and scalpdistribution to target and novel events Effects of temporalorder in young middle-aged and older adults CognitiveBrain Research 2 49ndash63

602 Journal of Cognitive Neuroscience Volume 10 Number 5

Auditory ERPs

In the Aa condition standard tones elicited N1 (meanpeak amplitude - 20 m V at Cz with a mean peak latencyof 93 msec) and P2 deections (34 m V 158 msec at Cz)which were largest over the fronto-central scalp loca-tions (Figure 3a) Deviant tones elicited N1 and P2 wavesat latencies similar to those in response to standardtones and the mismatch negativity (MMN) The MMN

was largest over the frontal electrodes (mean peak am-plitude - 22 m V mean peak latency 153 msec at Fz) andinverted in polarity at the mastoid electrodes (Figure 3b)being signicantly different from zero at F3 Fz and F4in the intervals 100 to 150 msec (F(1 9) = 1031 p lt002) and 150 to 200 msec (F(1 9) = 563 p lt 005) TheMMN was followed by a smaller fronto-centrally distrib-uted negative wave peaking at around 400 msec (Figure3b) Mean amplitude comparisons between ERPs to de-viant and standard tones at Fz and Cz revealed that thissecond negativity was signicant in the 350- to 400-msec(F(1 9) = 756 p lt 003) and in the 400- to 450-msec(F(1 9) = 873 p lt 002) intervals In the AV conditionthe MMN was similar to that obtained in the Aa condition(Figure 4)

In the Aa condition novel sounds elicited an ERPcharacterized by a prominent N1 deection (mean peakamplitude at Cz - 56 m V peak latency 96 msec) and bya long-duration broadly distributed P3a wave (Figure 3as seen in Figure 3b even the deviant tones elicitedsome P3a) The novel-sound N1 was larger than thestandard-tone N1 (F(1 9) = 1099 p lt 001) and wasprobably composed of overlapping N1 and MMN com-ponents The difference between the standard-tone andnovel-sound ERPs extended beyond the N1 latencyrange there being a signicant difference at Fz betweenthe standard and novel ERP amplitudes measured asmean voltages over 120 to 160 msec latency window(F(1 9) = 523 p lt 005 Figure 5)

The present P3a to novel sounds had a double peakover the frontal scalp locations (Figure 3b) suggestingtwo different subcomponents The early part of the P3a(peak latency 230 msec in the grand-mean ERP at Fz)was largest over the central scalp areas appearing withinverted polarity at lateral and posterior scalp sites (Fig-ures 3b and 5) The late part of the P3a (peak latency315 msec in the grand-mean ERP at Fz) in turn waslargest over frontal areas and did not show polarityinversion at any recording site (Figures 3b and 5) Thedifferent scalp distributions of the two P3a subcompo-nents in the Aa condition are shown in Figure 6 (top)

The component structure of the P3a deection wasevaluated by means of two different analyses of variance(ANOVA) for normalized ERP amplitudes (McCarthy ampWood 1985) First an ANOVA was performed to prove apossible scalp-distribution difference between the P2 tostandard tones and the early P3a to novel sounds Thenanother ANOVA was performed to conrm the differentscalp distributions of the two subcomponents of the P3aThe P2 amplitude was identied as the mean voltagebetween 150 and 200 msec from stimulus onset in thestandard-tone ERP The early P3a was measured as themean voltage between 175 and 275 msec in the differ-ence wave obtained by subtracting the ERP to the stan-dard tone from that to the novel sound The late P3a wasmeasured as the mean voltage between 275 and 375msec of the same difference wave A three-factor design

Figure 2 Mean RT (top panel) hit rate (middle panel) and errorrate (bottom panel) in the Visual-alone (Va) condition and in theAuditory-Visual (AV) condition to the visual stimulus occurring afterthe standard tone (std) the deviant tone (dev) or the novel sound(nov) The bars indicate the standard error of mean

Escera et al 593

was used for both analyses Wave (P2 versus early P3afor the rst analysis and early P3a versus late P3a for thesecond analysis) acute Frontality (F7 F3 Fz F4 F8 versus T3C3 Cz C4 T4 versus T5 P3 Pz P4 T6) acute Laterality (F7T3 T5 versus F3 C3 P3 versus Fz Cz Pz versus F4 C4P4 versus F8 T4 T6) The results of the rst analysisrevealed a signicant Wave acute Frontality interaction (F(218) = 509 p lt 003 e = 07547) due to the early P3abeing posterior to the P2 and a signicant Wave acuteLaterality interaction (F(4 36) = 406 p lt 003 e =05469) due to the more predominantly right-hemi-spheric scalp distribution of the P2 than that of the earlyP3a The results of the second analysis revealed that thelate P3a was anterior in distribution to the early P3a(signicant Wave Frontality interaction F(2 18) = 939p lt 0006 e = 07187) and further that the late P3a wasmore preponderant over the right hemisphere than the

early P3a (signicant Wave acute Laterality interactionF(4 36) = 1207 p lt 00002 e = 05740) The secondanalysis also revealed a signicant third-order interaction(Wave acute Frontality acute Laterality F(8 72) = 898 p lt00004 e = 03441) that was due to the predominantlyright-frontal distribution of the late phase of the P3adeection (Figure 6 top)

In the AV condition novel sounds elicited ERPs similarto those elicited by them in the Aa condition (large N1and P3a deections) although the early peak of the P3aseemed to be partly abolished at frontal electrodes (Fig-ure 7) However difference waves (ERP to novel soundsminus ERP to standard tones) revealed that the earlyphase of the P3a was similar in the Aa and AV conditionswhereas the late P3a was enhanced in the AV condition(Figure 7) The statistical scalp-distribution analyses ofthe two P3a phases across the conditions [ANOVA with

Figure 3a Grand-mean ERPsaveraged across subjects tostandard tones (600 Hz) devi-ant tones (700 Hz) and novelsounds in the Auditory-alone(Aa) condition

594 Journal of Cognitive Neuroscience Volume 10 Number 5

four factors Condition (Aa versus AV) acute Wave (early P3aversus late P3a) acute Frontality acute Laterality] conrmed thescalp-distribution difference between the two P3aphases (Wave acute Frontality interaction F(2 18) = 996p lt 0003 e = 08361 Wave acute Laterality F(4 36) = 688p lt 002 e = 03977 Wave acute Frontality acute LateralityF(872) = 981 p lt 00001 acute = 04381) and also revealeda signicant increase of the late P3a amplitude over theright hemisphere in the AV condition (Wave acute Condi-tion acute Laterality F(4 36) = 384 p lt 004 e = 05367)

Visual ERPs

Visual stimuli in the Va condition and those after stan-dard tones in the AV condition elicited almost identicalERPs characterized by P1 N1 P2 N2 and P3 deections(Figure 8) ANOVA for these visual ERP components with

Condition (after standard in AV versus Va) as a functionrevealed no signicant effects of standard tones on visualERPs

Figure 9 shows grand-average ERPs to auditory-visualstimulus pairs in the AV condition The AV stimulus pairselicited an auditory ERP lasting up to 400 msec post-stimulus for novel sounds and up to 300 msec post-stimulus for deviant tones followed by a visual ERPsimilar to that obtained in the Va condition All threetypes of AV stimulus pairs (standard-visual deviant-visualnovel-visual) were also characterized by a long-lastingfronto-polar negativity that was larger to visual stimulifollowing novel sounds than to those following standardor deviant tones (Figure 9) ANOVA for mean amplitudesat Fp1 and Fp2 in consecutive 100-msec latency win-dows yielded signicant effects for the novel-standardcomparison (F(1 9) ranging from 779 to 1196 p lt 003

Figure 3b Grand-average dif-ference waves obtained in theAa condition by subtractingthe ERP to standard tonesfrom that to deviant tones(continuous line) and tonovel sounds (dashed line)

Escera et al 595

in all cases) at latency windows 200 to 300 msec 300to 400 msec and 400 to 500 msec from visual-stimulusonset

Novel sounds and deviant tones enhanced the frontalN2 deection elicited by the subsequent visual stimulusin comparison with the visual N2 elicited after standardtones (Figure 9) This was revealed by two-way ANOVAsusing the Electrode (F3 Fz F4) and the auditory Stimulus(standard versus either deviant or novel) as factors Themean amplitude in the 200- to 300-msec range fromvisual-stimulus onset at Fz was - 19 m V after standardtones - 33 m V after deviant tones (F(1 9) = 641 p lt004) and - 36 m V after novel sounds (F(1 9) = 899 plt 002) The mean amplitude in the 300- to 400-msecinterval was also signicantly larger for visual stimulifollowing deviant tones or novel sounds than for thosefollowing standard tones F(1 9) = 815 p lt 002 for

deviant versus standard F(1 9) = 795 p lt 001 for novelversus standard

Preceding novel sounds and deviant tones also en-hanced the terminal slope of the visual P3b (Figure 9)this was revealed by the signicantly larger mean ampli-tude of the 500- to 600-msec latency window (at P3 Pzand P4) in the ERP to visual stimuli following devianttones than in the ERP to visual stimuli following standardtones (F(19) = 1556 p lt 004) The mean amplitude inthe 400- to 500-msec (F(1 9) = 694 p lt 003) and 500-to 600-msec (F(1 9) = 1199 p lt 0008) intervals was

Figure 4 The MMN was similar in the Auditory-alone (Aa) andAuditory-Visual (AV) conditions as revealed by nonsignicant differ-ences of the mean amplitude of ve consecutive 50-msec intervalsstarting at MMN onset (50 msec) [two-way ANOVAs Condition (Aaversus AV) acute Electrode (F3 Fz F4 C3 Cz C4)] Top Grand-averageERPs to standard and deviant tones at Fz in the two conditionsBottom Deviant-standard difference waves

Figure 5 The component structure of the ERPs to deviant tonesand novel sounds as revealed by the deviant-standard and novel-stan-dard difference waves at Fz and right-mastoid (RM) recordings Devi-ant tones elicited the MMN with a peak latency of 150 msec at Fzand with an inverted polarity at RM (top) Novel sounds elicited alarge N1 deection at Fz that was presumably composed of contribu-tions from two different processes for which the signicant differ-ences between the standard-tone and novel-sound ERPs extendedbeyond the N1 latency range (bottom) The novel sounds also elic-ited a P3a wave with two peaks at Fz suggesting two different sub-components contributing to the recorded P3a an early P3a (peaklatency 230 msec) which inverted polarity at RM and other poste-rior and lateral sites and a late P3a (peak latency 315 msec) whichinverted polarity at no recording site

596 Journal of Cognitive Neuroscience Volume 10 Number 5

also signicantly larger for visual stimuli following novelsounds than after those following the standard tones

DISCUSSION

Engagement of Attention During VisualPerformance

The present results provide evidence that attention dur-ing visual performance can be engaged by shortly pre-ceding unexpected novel sounds or small changes in thetask-irrelevant acoustic environment These results alsorevealed that this phenomenon depends on the intrinsicnature of the eliciting acoustic events In addition the

observed effects on visual performance were associatedwith different patterns of neural events activated bydeviant tones and novel sounds as reected by thecomponent structure of the auditory ERPs elicited

As expected novel sounds prolonged the RT to suc-cessive visual stimuli This result is in agreement with theextensive literature reporting delayed RTs to target stim-uli caused by preceding irrelevant sound changes ornovels sounds (Alho et al 1992 Cherry 1953 Dawsonet al 1989 Fillion et al 1991 Grillon et al 1990Schroumlger 1996 Woods 1992 Woodward et al 1991) andit provides experimental support for the everyday expe-rience that attention can be engaged by irrelevant acous-tic events

Figure 6 The scalp distribu-tion of the P3a to novelsounds Left early P3a rightlate P3a top Auditory-alone(Aa) condition bottom Audi-tory-Visual (AV) condition Thecentral plot shows the novel-standard difference waves atFz in the Aa (thin) and AV(thick) conditions as well asthe 100-msec latency win-dows used to measure thetwo phases of the P3a (e =early P3a l = late P3a)

Escera et al 597

A surprising nding of this study was however thatthe slightly deviant tones degraded the visual perfor-mance accuracy by decreasing the rate of hit responseswhereas the widely deviant novel sounds did not It wasexpected that the amount of distraction caused by devi-ant tones and novel sounds would be proportional tothe magnitude of their deviance from the standard toneand indeed such an effect was found for the RT (17-msecincrease after novel sounds) conrming other recentndings (Jaumlaumlskelaumlinen et al 1996 Schroumlger 1996) How-ever this RT effect was not paralleled by a similar effecton the hit rate Indeed deviant tones were associatedwith a lower hit rate whereas the hit rate after novelsounds was similar to that after standard tones Similarresults were obtained in a related study by Jaumlaumlskelaumlinenet al (1996) and recently replicated for deviant tones(Alho Escera Diacuteaz Yago amp Serra 1997) These ndingssuggest that the distracting effects of an irrelevant acous-tic event on visual performance may depend on thenature of the eliciting sound That is obtrusive novelsounds appear to cause a stronger effect on the RT tosuccessive visual target stimuli than minor changes inthe acoustic stream but in contrast these minor changes

seem to cause a signicant decrement in the number ofhit responses to visual targets not observed after thenovel sounds

The differential effects of distracting novel and deviantsounds on visual performance may be explained by trac-ing these behavioral effects back to the associated neuralevents as reected by ERPs Novel sounds enhanced theN1 amplitude and elicited the MMN as suggested bysignicant differences between the standard-tone andnovel-sound ERP that extended over the N1 latencyrange The present MMN to novel sounds peaked andterminated (returned to baseline) earlier than the MMNto deviant tones in concordance with previous studiesshowing that with larger deviances MMN becomes ear-lier (Tiitinen May Reinikainen amp Naumlaumltaumlnen 1994) over-lapping the N1 peak for wider deviances (Tiitinen et al1994 Scherg et al 1989) The present MMN to novelsounds is also supported by recent magnetoencepha-lograph (MEG) data showing that novel sounds activatedan MMN source in the human auditory cortex (Alhoet al 1998) Thus for novel sounds the attention-switch-ing signal was probably triggered by a combined re-sponse of the transient-detector system reected in thesupratemporal N1 (Naumlaumltaumlnen 1990 1992 Naumlaumltaumlnen ampPicton 1987) or some other N1 component such as thefrontal N1 reported by Giard et al (1994) and the stimu-lus-change detector system reected in the MMN(Naumlaumltaumlnen 1990 1992) This combined signal for atten-tion capture started with the N1 generation and wasoverlapped by the MMN process resulting in an effectiveattentional reorientation as reected by the subsequentlarge P3a wave and the clearly delayed RT to the follow-ing visual targets

Figure 7 Grand-average ERPs to standard tones and novel soundsin the Auditory-alone (Aa left column) and Auditory-Visual (AV rightcolumn) conditions The novel-standard difference waves are shownin the middle column The late P3a was larger in the AV conditionthan in the Aa condition whereas the early P3a was of similar ampli-tude in the two conditions

Figure 8 Grand-average visual ERPs at Pz O1 and O2 in the Visualalone (Va) condition and in the Auditory-Visual (AV) condition afterstandard tones

598 Journal of Cognitive Neuroscience Volume 10 Number 5

Deviant tones in turn elicited a distinct MMN whichwas followed by a small P3a wave and by a secondnegative peak (Figure 5) Thus the neural mechanismresponsible for triggering the attention-switching signalwas based for deviant tones on the stimulus-changedetector system reected in the MMN (Naumlaumltaumlnen 19901992 Schroumlger 1996) This attention capture signal asso-ciated with the MMN was sporadically followed by an

orienting of attention as suggested by the small (aver-aged) P3a and also by the 23 hit rate decrease to visualtargets that followed deviant tones

Although the subjects were instructed to ignore theauditory stimuli in the AV condition the RT to visualtargets following standard tones was slightly (insig-nicantly) shorter than the RT in the Va condition withno auditory stimuli This suggests that the auditory stim-

Figure 9 Grand-averageERPs in the Auditory-Visual(AV) condition to standard-visual (thin line) deviant-visual (thick line) andnovel-visual (dashed line)stimulus pairs Only the mostrelevant electrodes are shownThe epoch starts 100 msec be-fore auditory-stimulus onsetwhich was used as the base-line period to avoid the largeamplitude variability due tothe auditory response in the100 msec preceding the vis-ual stimulus Novel sounds en-hanced the fronto-polarnegativity and the frontal N2to visual stimuli was en-hanced after novel soundsand deviant tones which alsoaffected the offset slope ofP3b Notice that whereas thetraces for standard-visual devi-ant-visual and novel-visualstimulus pairs perfectly matcheach other during the onsetslope of the P3b they start todiffer from the P3b peak forthe novel-visual stimulus pairand about 100 msec later forthe deviant-visual stimuluspair (see P3 Pz P4)

Escera et al 599

uli were used as warning signals for the following visualtargets and thus attention may have been bimodal in theAV condition However because the occurrence of an aud-itory stimulus change or a novel sound was task-irrelevantand unpredictable to the subjects attention was involun-tarily engaged by these events In a more controlledexperiment attention was more effectively withdrawnfrom the auditory stimuli and their occurrence was madenoninformative by the simultaneous presentation of avisual cue stimulus informing about the successive ap-pearance of a visual target (Alho et al 1997) Like in thepresent study a decreased hit rate for visual stimuli thatfollowed MMN- and P3a-eliciting deviant tones was foundThis conrms the involuntary nature of the attention-switching mechanism indicated by the MMN P3a anddeterioration of the visual-task performance

Component Structure of the P3a to Novel Sounds

In both the Aa and AV conditions novel sounds eliciteda large long-lasting and broadly distributed P3a wavewith a double peak at Fz (in the Aa condition) suggestinga contribution of two different subcomponents Theearly part of this P3a had a peak latency of 230 msecand showed a centrally dominant scalp distribution thatwas similar in the Aa and AV conditions (Figure 6) Theearly P3a inverted in polarity at posterior and lateralelectrodes (Figures 3b and 5) suggesting bilateral gener-ators located in the vicinity of the temporal and parietallobes (cf Scherg amp von Cramon 1986) The late part ofthe P3a in turn had a peak latency of 315 msec and didnot invert in polarity at posterior electrodes and its scalpdistribution was centered over the right frontal areas Incontrast to the early P3a the late P3a was enhanced inthe AV condition particularly over the right hemisphere(Figure 6) when subjects apparently used auditory stim-uli as warning signals to prepare for visual targets

The component structure of the present P3a and thescalp distribution of the two subcomponents is in agree-ment with previous results in patients with brain lesionssuggesting a critical role of the temporal-parietal junc-tion in the generation of the auditory (Halgren BaudenaClarke Heit Marinkovic et al 1995 Halgren BaudenaClarke Heit Lieacutegeois et al 1995 Knight Scabini Woodsamp Clayworth 1989) somatosensory (Yamaguchi ampKnight 1991 1992) and visual P3a (Knight 1991 1997)It is also compatible with results suggesting a frontalcontribution to P3a such as those obtained in patientswith frontal lesions (Knight 1984) in topographicalanalysis (Friedman amp Simpson 1994 Friedman Simpsonamp Hamberger 1993) in dipole-source modeling (Meck-linger amp Ullsperger 1995) and in intracerebral record-ings in humans (Baudena Halgren Heit amp Clarke 1995)

The late P3a amplitude in the AV condition was en-hanced with regard to its amplitude in the Aa conditionThis amplitude enhancement of the late P3a reects an

attentional monitoring of auditory stimuli in the AV con-dition these stimuli being used as warning signals for thefollowing visual targets The attentional modulation ofthe P3a amplitude is in agreement with that reported inprevious studies (Holdstock amp Rugg 1995 Woods 1992)and suggests that the late P3a may more closely reectan actual orienting of attention than the involuntary N1and MMN processes triggering this attentional orientingAlso several other ndings support the association of thelate phase of the P3a with the actual reorientation ofattention or orienting response First the right-frontaldominance in the late P3a scalp distribution observed inthe present experiment agrees with ndings suggestingthe involvement of right-frontal areas in the reorienta-tion of attention (Fuster 1989 Mesulam 1981) Secondit has been shown that the P3a amplitude is attenuatedwith the repetition of the eliciting novel event(Courchesne 1978 Friedman amp Simpson 1994 Knight1984) This habituating behavior is a characteristic fea-ture of the orienting response (Oumlhman 1979 1992 Sok-olov 1963)

The early P3a was insensitive to attentional manipula-tions as it was to similar amplitude and scalp distributionin the Aa and AV conditions Its peak latency was rathershort (230 msec) and its scalp distribution suggestsneural generators probably located in the temporal-parietal cortex (Knight 1991 1997 Knight et al 1989Yamaguchi amp Knight 1991) Thus the early P3a mightreect a neural process other than attentional reorienta-tion such as the violation of a polisensorial model of theexternal world maintained in the temporal-parietal asso-ciation cortex (Yamaguchi amp Knight 1991)

Conclusions

The results reported in the present paper support thenotion of two different brain mechanisms of involuntaryattention differentially activated by different acousticevents The ERP component structure to novel soundssuggests that unexpected novelty in the acoustic envi-ronment involuntary captures attention by simultane-ously activating two different neural mechanisms atransient-detector mechanism associated with the su-pratemporal N1 component of the auditory ERPs and achange-detector mechanism reected in the MMN Pre-sumably the attention-switching signal generated by thecombined activation of these two mechanisms resultedin an effective engagement of attention as indicated bythe large P3a component elicited by the novel soundsand the increased RT to subsequent visual targets Devi-ant tones in turn resulted in different behavioral effectsand ERP component structure which was characterizedby the MMN This supports the notion that small changesin the acoustic environment capture attention involun-tarily by activating the stimulus-change detector mecha-nism reected in the MMN

600 Journal of Cognitive Neuroscience Volume 10 Number 5

METHODS

Subjects Stimuli and Procedure

Ten right-handed paid students (21 to 40 years 3 males)with normal hearing and normal or corrected-to-normalvisual acuity participated in the experiment Two furthersubjects were discarded due to their inability to reachthe required 90 hit rate in a practice session

Subjects were presented with four blocks of 400stimulus pairs (AV condition) in a visual forced-choiceRT task Each pair consisted of an auditory stimulusfollowed after 300 msec (onset-to-onset) by a visualstimulus The interpair interval (onset-to-onset) was 12sec (Figure 1) The auditory stimuli were either standardor deviant tones and novel sounds presented in randomorder with probabilities of 08 01 and 01 respectivelyThe standard and deviant tones were sinusoidal tonebursts of 200-msec duration including 10-msec rise andfall times presented binaurally at 75 dB SPL throughheadphones with respective frequencies of 600 and 700Hz Sixty different environmental sounds such as thoseproduced by a drill hammer rain door telephone ring-ing etc were used as novel sounds They were digitallyrecorded then low-pass ltered at 10000 Hz and editedto have a duration of 200 msec including rise and falltimes of 10 msec and an intensity maximum of 70 to 80dB SPL Each different novel sound occurred only oncewithin a stimulus block and was presented twice orthree times during the whole experiment Both devianttones and novel sounds were always preceded by at leastone AV pair containing a standard tone The visual stimuliwere the digits 1 to 8 presented one at a time in randomorder on a computer screen for 200 msec They sub-tended a vertical angle of 17deg and a horizontal angle of11deg (30 mm acute 20 mm 100 cm from the subjectrsquos eyes)

Subjects were comfortably seated in a reclining chairin a dimly lit electrically and acoustically shielded roomThey were instructed to focus on the small xation crossappearing in the middle of the screen and to press oneresponse button with the right index nger for oddnumbers and another button with the right middlenger for even numbers Subjects were also instructedto ignore the auditory stimulation Both speed and accu-racy were emphasized for the primary visual task

In two further conditions visual stimuli were omitted(Va condition Figure 1) or auditory stimuli were omitted(Aa condition Figure 1) In the Va condition subjectsreceived three blocks of 200 visual stimulus each theinstruction being identical to that in the AV condition Inthe Aa condition subjects were instructed to concen-trate on reading a self-selected book and to ignore theauditory stimulation

The Aa and AV blocks were presented in an alternatingorder with half of the subjects starting with the AVcondition The Va blocks were presented one at thebeginning one at the middle and one at the end of the

experiment Before the experimental blocks subjectsparticipated in a practice session consisting of two Vablocks They were required to reach a 90 hit rate beforethe experiment could start Two out of 12 subjects failedto reach this level even after several additional blocksand were nally discarded

EEG Recording and Averaging

The electroencephalogram (EEG) (bandpass 0ndash100 Hz)was continuously digitized at a rate of 500 Hz by aSynAmps amplier (NeuroScan Inc) from 19 scalp elec-trodes positioned according to the 10ndash20 system Fp1Fp2 F7 F3 Fz F4 F8 T3 C3 Cz C4 T4 T5 P3 Pz P4 T6O1 and O2 One additional electrode was placed at theright mastoid (RM) and another at the left mastoid (LM)The horizontal EOG was recorded with electrodes at-tached to the left and right canthi The vertical elec-trooculogram (EOG) was assessed using recordings fromFp1 and Fp2 electrodes The common reference elec-trode was placed on the tip of the nose

ERPs were averaged off-line for each auditory stimulusclass for an epoch of 1300 msec including a preauditorystimulus period of 100 msec in the AV and Aa conditionsIn the Va condition an ERP was also obtained for anepoch of 1300 msec including a previsual stimulus pe-riod of 400 msec Epochs in which the EEG or EOGexceeded plusmn100 m V as well as the ve rst epochs of eachblock were automatically excluded from averaging Inthe Aa and AV conditions standard-tone trials immedi-ately following deviant-tone or novel-sound trials werealso excluded from the averages Individual ERPs wereband-pass ltered between 001 and 30 Hz

Data Analysis

In the AV and Va conditions a correct button presswithin 800 msec after visual-stimulus onset was regardedas a hit the mean RT being computed only for the hittrials An incorrect button press during this period wasclassied as an error and trials with no response withinthis time window were considered misses Hits errorsmisses and RTs were computed across odd and evennumbers

Auditory ERP amplitudes (N1 P2 MMN and P3a) werereferred to the mean amplitude in the 100-msec baselineperiod preceding the auditory stimulus in both the Aaand AV conditions For auditory ERP analysis each indi-vidual waveform in the Aa and AV conditions was linearlydetrended over an epoch of 600 msec (including base-line) to remove any possible asymmetrical contributionto EEG caused by slow EOG movements due to readingin the Aa condition The effects of the standard tones onthe visual ERPs were analyzed at O1 and O2 for peaksas follows (identied in the individual waveforms withinthe specied latency ranges) P1 75 to 140 msec N1 100

Escera et al 601

to 200 msec P2 180 to 280 msec N2 225 to 350 msecThe visual P3b was identied at Pz between 300 and 500msec from visual-stimulus onset For this analysis visualERP amplitudes were measured against the mean ampli-tude of the 100-msec baseline preceding visual-stimulusonset The fronto-polar negativity was analyzed at Fp1and Fp2 the visual frontal N2 at F3 Fz and F4 and thevisual P3b at P3 Pz and P4 The amplitude measurementsfor visual ERP components in the AV condition werereferred to the 100-msec preauditory stimulus baselinerather than to the 100-msec previsual stimulus baselinewhich was differentially affected by standard tones de-viant tones and novel sounds (see Figure 9)

The ERP and performance data were statistically ana-lyzed by ANOVA with repeated measures Where appro-priate nominal degrees of freedom and epsilon valuesafter the Greenhouse-Geisser correction are reportedScalp-distribution analyses were performed after normal-izing ERP amplitudes to prevent amplitude differencesbetween different components from washing out thegenuine scalp-distribution differences This normalizationwas done by dividing the amplitude at each electrodeby the sum of the squared amplitudes at all electrodes(McCarthy amp Wood 1985)

Acknowledgments

This research was supported by BIOMED-2 contract BMH4-CT96-0819-COBRAIN of the European Union the SpanishMinistry of Science and Education (DGICYT PB93-0802 UE96-0038) the Academy of Finland the scientic exchange programbetween the University of Barcelona and the University ofHelsinki and the National Science Fund of Hungary (OTKAT022681) The authors would like to thank Teemu Rinne andTeija Kujala for their help in making the gures

Reprint requests should be sent to Carles Escera Neurody-namics Laboratory Department of Psychiatry and Clinical Psy-chobiology University of Barcelona P Vall drsquoHebron 171 08035Barcelona Catalonia-Spain or via e-mail cescerapsiubes

Note

1 The MMN amplitude in the unattended channel can behowever attenuated under some conditions by strongly focus-ing attention on other sources of auditory stimulation(Naumlaumltaumlnen 1991 Naumlaumltaumlnen Paavilainen Tiitinen Jiang amp Alho1993 Oades amp Dittmannbalcar 1995 Trejo Ryan-Jones amp Kra-mer 1995 Woldorff Hackley amp Hillyard 1991)

REFERENCES

Alho K (1995) Cerebral generators of mismatch negativity(MMN) and its magnetic counterpart (MMNn) elicited bysound changes Ear and Hearing 16 38ndash51

Alho K Escera C Diacuteaz R Yago E amp Serra J M (1997)Effects of involuntary auditory attention on visual taskperformance and brain activity NeuroReport 8 3233ndash3237

Alho K Winkler I Escera C Huotilainen M Virtanen JJaumlaumlskelaumlinen I Pekkonen E amp Ilmoniemi R (1998) Proc-

essing of novel sounds and frequency changes in the hu-man auditory cortex Magnetoencephalographic record-ings Psychophysiology 35 211ndash224

Alho K Woods D L Algazi A amp Naumlaumltaumlnen R (1992) Inter-modal selective attention II Effects of attentional load onprocessing of auditory and visual stimuli in central spaceElectroencephalography and Clinical Neurophysiology82 356ndash368

Baudena P Halgren E Heit G amp Clarke J M (1995) Intra-cerebral potentials to rare target and distractor auditoryand visual stimuli III Frontal cortex Electroencephalogra-phy and Clinical Neurophysiology 94 251ndash264

Cherry E C (1953) Some experiments on the recognitionof speech with one and with two ears Journal of theAcoustical Society of America 29 975ndash979

Courchesne E (1978) Changes in P3 waves with event repe-tition Long-term effects on scalp distribution and ampli-tudes Electroencephalography and ClinicalNeurophysiology 45 468ndash482

Courchesne E Hillyard S A amp Galambos R (1975) Stimu-lus novelty task relevance and the visual evoked potentialsin man Electroencephalography and Clinical Neurophysi-ology 39 131ndash143

Cowan N Winkler I Teder W amp Naumlaumltaumlnen R (1993) Mem-ory prerequisites of the mismatch negativity in the audi-tory event-related potential (ERP) Journal ofExperimental Psychology Learning Memory and Cogni-tion 19 909ndash921

Cseacutepe V Karmos G amp Molnaacuter M (1987) Evoked potentialcorrelates of stimulus deviance during wakefulness andsleep in catmdashanimal model of mismatch negativity Elec-troencephalography and Clinical Neurophysiology 66571ndash578

Davis H Mast T Yoshie N amp Zerlin S (1966) The slow re-sponse of the human cortex to auditory stimulus Recov-ery process Electroencephalography and ClinicalNeurophysiology 21 105ndash113

Davis H amp Zerlin S (1966) Acoustic relations of the humanvertex potentials Journal of the Acoustical Society ofAmerica 39 109ndash116

Dawson M E (1990) Psychophysiology at the interface ofclinical science cognitive science and neuroscience Psy-chophysiology 27 243ndash255

Dawson M E Fillion D L amp Schell A M (1989) Is elicita-tion of the autonomic orienting response associated withallocation of processing resources Psychophysiology 26560ndash572

Duncan C C amp Kaye W H (1987) Effects of chlonidine onevent-related potential measures of information process-ing In R Johnson J W Rohrbaugh amp R Parasuraman(Eds) Current trends in event-related potentialresearch (EEG supplement 40) (pp 527ndash531) Amster-dam Elsevier

Fillion D J Dawson M E Schell A M amp Hazlett E A(1991) The relationship between skin conductance orient-ing and the allocation of processing resources Psycho-physiology 28 410ndash424

Folk C L Remington R W amp Johnston J C (1992) Involun-tary covert orienting is contingent on attentional controlsettings Journal of Experimental Psychology Human Per-ception and Performance 18 1030ndash1044

Ford J M Roth W T amp Kopell B S (1976) Auditory event-related potentials to unpredictable shifts in pitch Psycho-physiology 13 32ndash39

Friedman D amp Simpson G (1994) ERP amplitude and scalpdistribution to target and novel events Effects of temporalorder in young middle-aged and older adults CognitiveBrain Research 2 49ndash63

602 Journal of Cognitive Neuroscience Volume 10 Number 5

was used for both analyses Wave (P2 versus early P3afor the rst analysis and early P3a versus late P3a for thesecond analysis) acute Frontality (F7 F3 Fz F4 F8 versus T3C3 Cz C4 T4 versus T5 P3 Pz P4 T6) acute Laterality (F7T3 T5 versus F3 C3 P3 versus Fz Cz Pz versus F4 C4P4 versus F8 T4 T6) The results of the rst analysisrevealed a signicant Wave acute Frontality interaction (F(218) = 509 p lt 003 e = 07547) due to the early P3abeing posterior to the P2 and a signicant Wave acuteLaterality interaction (F(4 36) = 406 p lt 003 e =05469) due to the more predominantly right-hemi-spheric scalp distribution of the P2 than that of the earlyP3a The results of the second analysis revealed that thelate P3a was anterior in distribution to the early P3a(signicant Wave Frontality interaction F(2 18) = 939p lt 0006 e = 07187) and further that the late P3a wasmore preponderant over the right hemisphere than the

early P3a (signicant Wave acute Laterality interactionF(4 36) = 1207 p lt 00002 e = 05740) The secondanalysis also revealed a signicant third-order interaction(Wave acute Frontality acute Laterality F(8 72) = 898 p lt00004 e = 03441) that was due to the predominantlyright-frontal distribution of the late phase of the P3adeection (Figure 6 top)

In the AV condition novel sounds elicited ERPs similarto those elicited by them in the Aa condition (large N1and P3a deections) although the early peak of the P3aseemed to be partly abolished at frontal electrodes (Fig-ure 7) However difference waves (ERP to novel soundsminus ERP to standard tones) revealed that the earlyphase of the P3a was similar in the Aa and AV conditionswhereas the late P3a was enhanced in the AV condition(Figure 7) The statistical scalp-distribution analyses ofthe two P3a phases across the conditions [ANOVA with

Figure 3a Grand-mean ERPsaveraged across subjects tostandard tones (600 Hz) devi-ant tones (700 Hz) and novelsounds in the Auditory-alone(Aa) condition

594 Journal of Cognitive Neuroscience Volume 10 Number 5

four factors Condition (Aa versus AV) acute Wave (early P3aversus late P3a) acute Frontality acute Laterality] conrmed thescalp-distribution difference between the two P3aphases (Wave acute Frontality interaction F(2 18) = 996p lt 0003 e = 08361 Wave acute Laterality F(4 36) = 688p lt 002 e = 03977 Wave acute Frontality acute LateralityF(872) = 981 p lt 00001 acute = 04381) and also revealeda signicant increase of the late P3a amplitude over theright hemisphere in the AV condition (Wave acute Condi-tion acute Laterality F(4 36) = 384 p lt 004 e = 05367)

Visual ERPs

Visual stimuli in the Va condition and those after stan-dard tones in the AV condition elicited almost identicalERPs characterized by P1 N1 P2 N2 and P3 deections(Figure 8) ANOVA for these visual ERP components with

Condition (after standard in AV versus Va) as a functionrevealed no signicant effects of standard tones on visualERPs

Figure 9 shows grand-average ERPs to auditory-visualstimulus pairs in the AV condition The AV stimulus pairselicited an auditory ERP lasting up to 400 msec post-stimulus for novel sounds and up to 300 msec post-stimulus for deviant tones followed by a visual ERPsimilar to that obtained in the Va condition All threetypes of AV stimulus pairs (standard-visual deviant-visualnovel-visual) were also characterized by a long-lastingfronto-polar negativity that was larger to visual stimulifollowing novel sounds than to those following standardor deviant tones (Figure 9) ANOVA for mean amplitudesat Fp1 and Fp2 in consecutive 100-msec latency win-dows yielded signicant effects for the novel-standardcomparison (F(1 9) ranging from 779 to 1196 p lt 003

Figure 3b Grand-average dif-ference waves obtained in theAa condition by subtractingthe ERP to standard tonesfrom that to deviant tones(continuous line) and tonovel sounds (dashed line)

Escera et al 595

in all cases) at latency windows 200 to 300 msec 300to 400 msec and 400 to 500 msec from visual-stimulusonset

Novel sounds and deviant tones enhanced the frontalN2 deection elicited by the subsequent visual stimulusin comparison with the visual N2 elicited after standardtones (Figure 9) This was revealed by two-way ANOVAsusing the Electrode (F3 Fz F4) and the auditory Stimulus(standard versus either deviant or novel) as factors Themean amplitude in the 200- to 300-msec range fromvisual-stimulus onset at Fz was - 19 m V after standardtones - 33 m V after deviant tones (F(1 9) = 641 p lt004) and - 36 m V after novel sounds (F(1 9) = 899 plt 002) The mean amplitude in the 300- to 400-msecinterval was also signicantly larger for visual stimulifollowing deviant tones or novel sounds than for thosefollowing standard tones F(1 9) = 815 p lt 002 for

deviant versus standard F(1 9) = 795 p lt 001 for novelversus standard

Preceding novel sounds and deviant tones also en-hanced the terminal slope of the visual P3b (Figure 9)this was revealed by the signicantly larger mean ampli-tude of the 500- to 600-msec latency window (at P3 Pzand P4) in the ERP to visual stimuli following devianttones than in the ERP to visual stimuli following standardtones (F(19) = 1556 p lt 004) The mean amplitude inthe 400- to 500-msec (F(1 9) = 694 p lt 003) and 500-to 600-msec (F(1 9) = 1199 p lt 0008) intervals was

Figure 4 The MMN was similar in the Auditory-alone (Aa) andAuditory-Visual (AV) conditions as revealed by nonsignicant differ-ences of the mean amplitude of ve consecutive 50-msec intervalsstarting at MMN onset (50 msec) [two-way ANOVAs Condition (Aaversus AV) acute Electrode (F3 Fz F4 C3 Cz C4)] Top Grand-averageERPs to standard and deviant tones at Fz in the two conditionsBottom Deviant-standard difference waves

Figure 5 The component structure of the ERPs to deviant tonesand novel sounds as revealed by the deviant-standard and novel-stan-dard difference waves at Fz and right-mastoid (RM) recordings Devi-ant tones elicited the MMN with a peak latency of 150 msec at Fzand with an inverted polarity at RM (top) Novel sounds elicited alarge N1 deection at Fz that was presumably composed of contribu-tions from two different processes for which the signicant differ-ences between the standard-tone and novel-sound ERPs extendedbeyond the N1 latency range (bottom) The novel sounds also elic-ited a P3a wave with two peaks at Fz suggesting two different sub-components contributing to the recorded P3a an early P3a (peaklatency 230 msec) which inverted polarity at RM and other poste-rior and lateral sites and a late P3a (peak latency 315 msec) whichinverted polarity at no recording site

596 Journal of Cognitive Neuroscience Volume 10 Number 5

also signicantly larger for visual stimuli following novelsounds than after those following the standard tones

DISCUSSION

Engagement of Attention During VisualPerformance

The present results provide evidence that attention dur-ing visual performance can be engaged by shortly pre-ceding unexpected novel sounds or small changes in thetask-irrelevant acoustic environment These results alsorevealed that this phenomenon depends on the intrinsicnature of the eliciting acoustic events In addition the

observed effects on visual performance were associatedwith different patterns of neural events activated bydeviant tones and novel sounds as reected by thecomponent structure of the auditory ERPs elicited

As expected novel sounds prolonged the RT to suc-cessive visual stimuli This result is in agreement with theextensive literature reporting delayed RTs to target stim-uli caused by preceding irrelevant sound changes ornovels sounds (Alho et al 1992 Cherry 1953 Dawsonet al 1989 Fillion et al 1991 Grillon et al 1990Schroumlger 1996 Woods 1992 Woodward et al 1991) andit provides experimental support for the everyday expe-rience that attention can be engaged by irrelevant acous-tic events

Figure 6 The scalp distribu-tion of the P3a to novelsounds Left early P3a rightlate P3a top Auditory-alone(Aa) condition bottom Audi-tory-Visual (AV) condition Thecentral plot shows the novel-standard difference waves atFz in the Aa (thin) and AV(thick) conditions as well asthe 100-msec latency win-dows used to measure thetwo phases of the P3a (e =early P3a l = late P3a)

Escera et al 597

A surprising nding of this study was however thatthe slightly deviant tones degraded the visual perfor-mance accuracy by decreasing the rate of hit responseswhereas the widely deviant novel sounds did not It wasexpected that the amount of distraction caused by devi-ant tones and novel sounds would be proportional tothe magnitude of their deviance from the standard toneand indeed such an effect was found for the RT (17-msecincrease after novel sounds) conrming other recentndings (Jaumlaumlskelaumlinen et al 1996 Schroumlger 1996) How-ever this RT effect was not paralleled by a similar effecton the hit rate Indeed deviant tones were associatedwith a lower hit rate whereas the hit rate after novelsounds was similar to that after standard tones Similarresults were obtained in a related study by Jaumlaumlskelaumlinenet al (1996) and recently replicated for deviant tones(Alho Escera Diacuteaz Yago amp Serra 1997) These ndingssuggest that the distracting effects of an irrelevant acous-tic event on visual performance may depend on thenature of the eliciting sound That is obtrusive novelsounds appear to cause a stronger effect on the RT tosuccessive visual target stimuli than minor changes inthe acoustic stream but in contrast these minor changes

seem to cause a signicant decrement in the number ofhit responses to visual targets not observed after thenovel sounds

The differential effects of distracting novel and deviantsounds on visual performance may be explained by trac-ing these behavioral effects back to the associated neuralevents as reected by ERPs Novel sounds enhanced theN1 amplitude and elicited the MMN as suggested bysignicant differences between the standard-tone andnovel-sound ERP that extended over the N1 latencyrange The present MMN to novel sounds peaked andterminated (returned to baseline) earlier than the MMNto deviant tones in concordance with previous studiesshowing that with larger deviances MMN becomes ear-lier (Tiitinen May Reinikainen amp Naumlaumltaumlnen 1994) over-lapping the N1 peak for wider deviances (Tiitinen et al1994 Scherg et al 1989) The present MMN to novelsounds is also supported by recent magnetoencepha-lograph (MEG) data showing that novel sounds activatedan MMN source in the human auditory cortex (Alhoet al 1998) Thus for novel sounds the attention-switch-ing signal was probably triggered by a combined re-sponse of the transient-detector system reected in thesupratemporal N1 (Naumlaumltaumlnen 1990 1992 Naumlaumltaumlnen ampPicton 1987) or some other N1 component such as thefrontal N1 reported by Giard et al (1994) and the stimu-lus-change detector system reected in the MMN(Naumlaumltaumlnen 1990 1992) This combined signal for atten-tion capture started with the N1 generation and wasoverlapped by the MMN process resulting in an effectiveattentional reorientation as reected by the subsequentlarge P3a wave and the clearly delayed RT to the follow-ing visual targets

Figure 7 Grand-average ERPs to standard tones and novel soundsin the Auditory-alone (Aa left column) and Auditory-Visual (AV rightcolumn) conditions The novel-standard difference waves are shownin the middle column The late P3a was larger in the AV conditionthan in the Aa condition whereas the early P3a was of similar ampli-tude in the two conditions

Figure 8 Grand-average visual ERPs at Pz O1 and O2 in the Visualalone (Va) condition and in the Auditory-Visual (AV) condition afterstandard tones

598 Journal of Cognitive Neuroscience Volume 10 Number 5

Deviant tones in turn elicited a distinct MMN whichwas followed by a small P3a wave and by a secondnegative peak (Figure 5) Thus the neural mechanismresponsible for triggering the attention-switching signalwas based for deviant tones on the stimulus-changedetector system reected in the MMN (Naumlaumltaumlnen 19901992 Schroumlger 1996) This attention capture signal asso-ciated with the MMN was sporadically followed by an

orienting of attention as suggested by the small (aver-aged) P3a and also by the 23 hit rate decrease to visualtargets that followed deviant tones

Although the subjects were instructed to ignore theauditory stimuli in the AV condition the RT to visualtargets following standard tones was slightly (insig-nicantly) shorter than the RT in the Va condition withno auditory stimuli This suggests that the auditory stim-

Figure 9 Grand-averageERPs in the Auditory-Visual(AV) condition to standard-visual (thin line) deviant-visual (thick line) andnovel-visual (dashed line)stimulus pairs Only the mostrelevant electrodes are shownThe epoch starts 100 msec be-fore auditory-stimulus onsetwhich was used as the base-line period to avoid the largeamplitude variability due tothe auditory response in the100 msec preceding the vis-ual stimulus Novel sounds en-hanced the fronto-polarnegativity and the frontal N2to visual stimuli was en-hanced after novel soundsand deviant tones which alsoaffected the offset slope ofP3b Notice that whereas thetraces for standard-visual devi-ant-visual and novel-visualstimulus pairs perfectly matcheach other during the onsetslope of the P3b they start todiffer from the P3b peak forthe novel-visual stimulus pairand about 100 msec later forthe deviant-visual stimuluspair (see P3 Pz P4)

Escera et al 599

uli were used as warning signals for the following visualtargets and thus attention may have been bimodal in theAV condition However because the occurrence of an aud-itory stimulus change or a novel sound was task-irrelevantand unpredictable to the subjects attention was involun-tarily engaged by these events In a more controlledexperiment attention was more effectively withdrawnfrom the auditory stimuli and their occurrence was madenoninformative by the simultaneous presentation of avisual cue stimulus informing about the successive ap-pearance of a visual target (Alho et al 1997) Like in thepresent study a decreased hit rate for visual stimuli thatfollowed MMN- and P3a-eliciting deviant tones was foundThis conrms the involuntary nature of the attention-switching mechanism indicated by the MMN P3a anddeterioration of the visual-task performance

Component Structure of the P3a to Novel Sounds

In both the Aa and AV conditions novel sounds eliciteda large long-lasting and broadly distributed P3a wavewith a double peak at Fz (in the Aa condition) suggestinga contribution of two different subcomponents Theearly part of this P3a had a peak latency of 230 msecand showed a centrally dominant scalp distribution thatwas similar in the Aa and AV conditions (Figure 6) Theearly P3a inverted in polarity at posterior and lateralelectrodes (Figures 3b and 5) suggesting bilateral gener-ators located in the vicinity of the temporal and parietallobes (cf Scherg amp von Cramon 1986) The late part ofthe P3a in turn had a peak latency of 315 msec and didnot invert in polarity at posterior electrodes and its scalpdistribution was centered over the right frontal areas Incontrast to the early P3a the late P3a was enhanced inthe AV condition particularly over the right hemisphere(Figure 6) when subjects apparently used auditory stim-uli as warning signals to prepare for visual targets

The component structure of the present P3a and thescalp distribution of the two subcomponents is in agree-ment with previous results in patients with brain lesionssuggesting a critical role of the temporal-parietal junc-tion in the generation of the auditory (Halgren BaudenaClarke Heit Marinkovic et al 1995 Halgren BaudenaClarke Heit Lieacutegeois et al 1995 Knight Scabini Woodsamp Clayworth 1989) somatosensory (Yamaguchi ampKnight 1991 1992) and visual P3a (Knight 1991 1997)It is also compatible with results suggesting a frontalcontribution to P3a such as those obtained in patientswith frontal lesions (Knight 1984) in topographicalanalysis (Friedman amp Simpson 1994 Friedman Simpsonamp Hamberger 1993) in dipole-source modeling (Meck-linger amp Ullsperger 1995) and in intracerebral record-ings in humans (Baudena Halgren Heit amp Clarke 1995)

The late P3a amplitude in the AV condition was en-hanced with regard to its amplitude in the Aa conditionThis amplitude enhancement of the late P3a reects an

attentional monitoring of auditory stimuli in the AV con-dition these stimuli being used as warning signals for thefollowing visual targets The attentional modulation ofthe P3a amplitude is in agreement with that reported inprevious studies (Holdstock amp Rugg 1995 Woods 1992)and suggests that the late P3a may more closely reectan actual orienting of attention than the involuntary N1and MMN processes triggering this attentional orientingAlso several other ndings support the association of thelate phase of the P3a with the actual reorientation ofattention or orienting response First the right-frontaldominance in the late P3a scalp distribution observed inthe present experiment agrees with ndings suggestingthe involvement of right-frontal areas in the reorienta-tion of attention (Fuster 1989 Mesulam 1981) Secondit has been shown that the P3a amplitude is attenuatedwith the repetition of the eliciting novel event(Courchesne 1978 Friedman amp Simpson 1994 Knight1984) This habituating behavior is a characteristic fea-ture of the orienting response (Oumlhman 1979 1992 Sok-olov 1963)

The early P3a was insensitive to attentional manipula-tions as it was to similar amplitude and scalp distributionin the Aa and AV conditions Its peak latency was rathershort (230 msec) and its scalp distribution suggestsneural generators probably located in the temporal-parietal cortex (Knight 1991 1997 Knight et al 1989Yamaguchi amp Knight 1991) Thus the early P3a mightreect a neural process other than attentional reorienta-tion such as the violation of a polisensorial model of theexternal world maintained in the temporal-parietal asso-ciation cortex (Yamaguchi amp Knight 1991)

Conclusions

The results reported in the present paper support thenotion of two different brain mechanisms of involuntaryattention differentially activated by different acousticevents The ERP component structure to novel soundssuggests that unexpected novelty in the acoustic envi-ronment involuntary captures attention by simultane-ously activating two different neural mechanisms atransient-detector mechanism associated with the su-pratemporal N1 component of the auditory ERPs and achange-detector mechanism reected in the MMN Pre-sumably the attention-switching signal generated by thecombined activation of these two mechanisms resultedin an effective engagement of attention as indicated bythe large P3a component elicited by the novel soundsand the increased RT to subsequent visual targets Devi-ant tones in turn resulted in different behavioral effectsand ERP component structure which was characterizedby the MMN This supports the notion that small changesin the acoustic environment capture attention involun-tarily by activating the stimulus-change detector mecha-nism reected in the MMN

600 Journal of Cognitive Neuroscience Volume 10 Number 5

METHODS

Subjects Stimuli and Procedure

Ten right-handed paid students (21 to 40 years 3 males)with normal hearing and normal or corrected-to-normalvisual acuity participated in the experiment Two furthersubjects were discarded due to their inability to reachthe required 90 hit rate in a practice session

Subjects were presented with four blocks of 400stimulus pairs (AV condition) in a visual forced-choiceRT task Each pair consisted of an auditory stimulusfollowed after 300 msec (onset-to-onset) by a visualstimulus The interpair interval (onset-to-onset) was 12sec (Figure 1) The auditory stimuli were either standardor deviant tones and novel sounds presented in randomorder with probabilities of 08 01 and 01 respectivelyThe standard and deviant tones were sinusoidal tonebursts of 200-msec duration including 10-msec rise andfall times presented binaurally at 75 dB SPL throughheadphones with respective frequencies of 600 and 700Hz Sixty different environmental sounds such as thoseproduced by a drill hammer rain door telephone ring-ing etc were used as novel sounds They were digitallyrecorded then low-pass ltered at 10000 Hz and editedto have a duration of 200 msec including rise and falltimes of 10 msec and an intensity maximum of 70 to 80dB SPL Each different novel sound occurred only oncewithin a stimulus block and was presented twice orthree times during the whole experiment Both devianttones and novel sounds were always preceded by at leastone AV pair containing a standard tone The visual stimuliwere the digits 1 to 8 presented one at a time in randomorder on a computer screen for 200 msec They sub-tended a vertical angle of 17deg and a horizontal angle of11deg (30 mm acute 20 mm 100 cm from the subjectrsquos eyes)

Subjects were comfortably seated in a reclining chairin a dimly lit electrically and acoustically shielded roomThey were instructed to focus on the small xation crossappearing in the middle of the screen and to press oneresponse button with the right index nger for oddnumbers and another button with the right middlenger for even numbers Subjects were also instructedto ignore the auditory stimulation Both speed and accu-racy were emphasized for the primary visual task

In two further conditions visual stimuli were omitted(Va condition Figure 1) or auditory stimuli were omitted(Aa condition Figure 1) In the Va condition subjectsreceived three blocks of 200 visual stimulus each theinstruction being identical to that in the AV condition Inthe Aa condition subjects were instructed to concen-trate on reading a self-selected book and to ignore theauditory stimulation

The Aa and AV blocks were presented in an alternatingorder with half of the subjects starting with the AVcondition The Va blocks were presented one at thebeginning one at the middle and one at the end of the

experiment Before the experimental blocks subjectsparticipated in a practice session consisting of two Vablocks They were required to reach a 90 hit rate beforethe experiment could start Two out of 12 subjects failedto reach this level even after several additional blocksand were nally discarded

EEG Recording and Averaging

The electroencephalogram (EEG) (bandpass 0ndash100 Hz)was continuously digitized at a rate of 500 Hz by aSynAmps amplier (NeuroScan Inc) from 19 scalp elec-trodes positioned according to the 10ndash20 system Fp1Fp2 F7 F3 Fz F4 F8 T3 C3 Cz C4 T4 T5 P3 Pz P4 T6O1 and O2 One additional electrode was placed at theright mastoid (RM) and another at the left mastoid (LM)The horizontal EOG was recorded with electrodes at-tached to the left and right canthi The vertical elec-trooculogram (EOG) was assessed using recordings fromFp1 and Fp2 electrodes The common reference elec-trode was placed on the tip of the nose

ERPs were averaged off-line for each auditory stimulusclass for an epoch of 1300 msec including a preauditorystimulus period of 100 msec in the AV and Aa conditionsIn the Va condition an ERP was also obtained for anepoch of 1300 msec including a previsual stimulus pe-riod of 400 msec Epochs in which the EEG or EOGexceeded plusmn100 m V as well as the ve rst epochs of eachblock were automatically excluded from averaging Inthe Aa and AV conditions standard-tone trials immedi-ately following deviant-tone or novel-sound trials werealso excluded from the averages Individual ERPs wereband-pass ltered between 001 and 30 Hz

Data Analysis

In the AV and Va conditions a correct button presswithin 800 msec after visual-stimulus onset was regardedas a hit the mean RT being computed only for the hittrials An incorrect button press during this period wasclassied as an error and trials with no response withinthis time window were considered misses Hits errorsmisses and RTs were computed across odd and evennumbers

Auditory ERP amplitudes (N1 P2 MMN and P3a) werereferred to the mean amplitude in the 100-msec baselineperiod preceding the auditory stimulus in both the Aaand AV conditions For auditory ERP analysis each indi-vidual waveform in the Aa and AV conditions was linearlydetrended over an epoch of 600 msec (including base-line) to remove any possible asymmetrical contributionto EEG caused by slow EOG movements due to readingin the Aa condition The effects of the standard tones onthe visual ERPs were analyzed at O1 and O2 for peaksas follows (identied in the individual waveforms withinthe specied latency ranges) P1 75 to 140 msec N1 100

Escera et al 601

to 200 msec P2 180 to 280 msec N2 225 to 350 msecThe visual P3b was identied at Pz between 300 and 500msec from visual-stimulus onset For this analysis visualERP amplitudes were measured against the mean ampli-tude of the 100-msec baseline preceding visual-stimulusonset The fronto-polar negativity was analyzed at Fp1and Fp2 the visual frontal N2 at F3 Fz and F4 and thevisual P3b at P3 Pz and P4 The amplitude measurementsfor visual ERP components in the AV condition werereferred to the 100-msec preauditory stimulus baselinerather than to the 100-msec previsual stimulus baselinewhich was differentially affected by standard tones de-viant tones and novel sounds (see Figure 9)

The ERP and performance data were statistically ana-lyzed by ANOVA with repeated measures Where appro-priate nominal degrees of freedom and epsilon valuesafter the Greenhouse-Geisser correction are reportedScalp-distribution analyses were performed after normal-izing ERP amplitudes to prevent amplitude differencesbetween different components from washing out thegenuine scalp-distribution differences This normalizationwas done by dividing the amplitude at each electrodeby the sum of the squared amplitudes at all electrodes(McCarthy amp Wood 1985)

Acknowledgments

This research was supported by BIOMED-2 contract BMH4-CT96-0819-COBRAIN of the European Union the SpanishMinistry of Science and Education (DGICYT PB93-0802 UE96-0038) the Academy of Finland the scientic exchange programbetween the University of Barcelona and the University ofHelsinki and the National Science Fund of Hungary (OTKAT022681) The authors would like to thank Teemu Rinne andTeija Kujala for their help in making the gures

Reprint requests should be sent to Carles Escera Neurody-namics Laboratory Department of Psychiatry and Clinical Psy-chobiology University of Barcelona P Vall drsquoHebron 171 08035Barcelona Catalonia-Spain or via e-mail cescerapsiubes

Note

1 The MMN amplitude in the unattended channel can behowever attenuated under some conditions by strongly focus-ing attention on other sources of auditory stimulation(Naumlaumltaumlnen 1991 Naumlaumltaumlnen Paavilainen Tiitinen Jiang amp Alho1993 Oades amp Dittmannbalcar 1995 Trejo Ryan-Jones amp Kra-mer 1995 Woldorff Hackley amp Hillyard 1991)

REFERENCES

Alho K (1995) Cerebral generators of mismatch negativity(MMN) and its magnetic counterpart (MMNn) elicited bysound changes Ear and Hearing 16 38ndash51

Alho K Escera C Diacuteaz R Yago E amp Serra J M (1997)Effects of involuntary auditory attention on visual taskperformance and brain activity NeuroReport 8 3233ndash3237

Alho K Winkler I Escera C Huotilainen M Virtanen JJaumlaumlskelaumlinen I Pekkonen E amp Ilmoniemi R (1998) Proc-

essing of novel sounds and frequency changes in the hu-man auditory cortex Magnetoencephalographic record-ings Psychophysiology 35 211ndash224

Alho K Woods D L Algazi A amp Naumlaumltaumlnen R (1992) Inter-modal selective attention II Effects of attentional load onprocessing of auditory and visual stimuli in central spaceElectroencephalography and Clinical Neurophysiology82 356ndash368

Baudena P Halgren E Heit G amp Clarke J M (1995) Intra-cerebral potentials to rare target and distractor auditoryand visual stimuli III Frontal cortex Electroencephalogra-phy and Clinical Neurophysiology 94 251ndash264

Cherry E C (1953) Some experiments on the recognitionof speech with one and with two ears Journal of theAcoustical Society of America 29 975ndash979

Courchesne E (1978) Changes in P3 waves with event repe-tition Long-term effects on scalp distribution and ampli-tudes Electroencephalography and ClinicalNeurophysiology 45 468ndash482

Courchesne E Hillyard S A amp Galambos R (1975) Stimu-lus novelty task relevance and the visual evoked potentialsin man Electroencephalography and Clinical Neurophysi-ology 39 131ndash143

Cowan N Winkler I Teder W amp Naumlaumltaumlnen R (1993) Mem-ory prerequisites of the mismatch negativity in the audi-tory event-related potential (ERP) Journal ofExperimental Psychology Learning Memory and Cogni-tion 19 909ndash921

Cseacutepe V Karmos G amp Molnaacuter M (1987) Evoked potentialcorrelates of stimulus deviance during wakefulness andsleep in catmdashanimal model of mismatch negativity Elec-troencephalography and Clinical Neurophysiology 66571ndash578

Davis H Mast T Yoshie N amp Zerlin S (1966) The slow re-sponse of the human cortex to auditory stimulus Recov-ery process Electroencephalography and ClinicalNeurophysiology 21 105ndash113

Davis H amp Zerlin S (1966) Acoustic relations of the humanvertex potentials Journal of the Acoustical Society ofAmerica 39 109ndash116

Dawson M E (1990) Psychophysiology at the interface ofclinical science cognitive science and neuroscience Psy-chophysiology 27 243ndash255

Dawson M E Fillion D L amp Schell A M (1989) Is elicita-tion of the autonomic orienting response associated withallocation of processing resources Psychophysiology 26560ndash572

Duncan C C amp Kaye W H (1987) Effects of chlonidine onevent-related potential measures of information process-ing In R Johnson J W Rohrbaugh amp R Parasuraman(Eds) Current trends in event-related potentialresearch (EEG supplement 40) (pp 527ndash531) Amster-dam Elsevier

Fillion D J Dawson M E Schell A M amp Hazlett E A(1991) The relationship between skin conductance orient-ing and the allocation of processing resources Psycho-physiology 28 410ndash424

Folk C L Remington R W amp Johnston J C (1992) Involun-tary covert orienting is contingent on attentional controlsettings Journal of Experimental Psychology Human Per-ception and Performance 18 1030ndash1044

Ford J M Roth W T amp Kopell B S (1976) Auditory event-related potentials to unpredictable shifts in pitch Psycho-physiology 13 32ndash39

Friedman D amp Simpson G (1994) ERP amplitude and scalpdistribution to target and novel events Effects of temporalorder in young middle-aged and older adults CognitiveBrain Research 2 49ndash63

602 Journal of Cognitive Neuroscience Volume 10 Number 5

four factors Condition (Aa versus AV) acute Wave (early P3aversus late P3a) acute Frontality acute Laterality] conrmed thescalp-distribution difference between the two P3aphases (Wave acute Frontality interaction F(2 18) = 996p lt 0003 e = 08361 Wave acute Laterality F(4 36) = 688p lt 002 e = 03977 Wave acute Frontality acute LateralityF(872) = 981 p lt 00001 acute = 04381) and also revealeda signicant increase of the late P3a amplitude over theright hemisphere in the AV condition (Wave acute Condi-tion acute Laterality F(4 36) = 384 p lt 004 e = 05367)

Visual ERPs

Visual stimuli in the Va condition and those after stan-dard tones in the AV condition elicited almost identicalERPs characterized by P1 N1 P2 N2 and P3 deections(Figure 8) ANOVA for these visual ERP components with

Condition (after standard in AV versus Va) as a functionrevealed no signicant effects of standard tones on visualERPs

Figure 9 shows grand-average ERPs to auditory-visualstimulus pairs in the AV condition The AV stimulus pairselicited an auditory ERP lasting up to 400 msec post-stimulus for novel sounds and up to 300 msec post-stimulus for deviant tones followed by a visual ERPsimilar to that obtained in the Va condition All threetypes of AV stimulus pairs (standard-visual deviant-visualnovel-visual) were also characterized by a long-lastingfronto-polar negativity that was larger to visual stimulifollowing novel sounds than to those following standardor deviant tones (Figure 9) ANOVA for mean amplitudesat Fp1 and Fp2 in consecutive 100-msec latency win-dows yielded signicant effects for the novel-standardcomparison (F(1 9) ranging from 779 to 1196 p lt 003

Figure 3b Grand-average dif-ference waves obtained in theAa condition by subtractingthe ERP to standard tonesfrom that to deviant tones(continuous line) and tonovel sounds (dashed line)

Escera et al 595

in all cases) at latency windows 200 to 300 msec 300to 400 msec and 400 to 500 msec from visual-stimulusonset

Novel sounds and deviant tones enhanced the frontalN2 deection elicited by the subsequent visual stimulusin comparison with the visual N2 elicited after standardtones (Figure 9) This was revealed by two-way ANOVAsusing the Electrode (F3 Fz F4) and the auditory Stimulus(standard versus either deviant or novel) as factors Themean amplitude in the 200- to 300-msec range fromvisual-stimulus onset at Fz was - 19 m V after standardtones - 33 m V after deviant tones (F(1 9) = 641 p lt004) and - 36 m V after novel sounds (F(1 9) = 899 plt 002) The mean amplitude in the 300- to 400-msecinterval was also signicantly larger for visual stimulifollowing deviant tones or novel sounds than for thosefollowing standard tones F(1 9) = 815 p lt 002 for

deviant versus standard F(1 9) = 795 p lt 001 for novelversus standard

Preceding novel sounds and deviant tones also en-hanced the terminal slope of the visual P3b (Figure 9)this was revealed by the signicantly larger mean ampli-tude of the 500- to 600-msec latency window (at P3 Pzand P4) in the ERP to visual stimuli following devianttones than in the ERP to visual stimuli following standardtones (F(19) = 1556 p lt 004) The mean amplitude inthe 400- to 500-msec (F(1 9) = 694 p lt 003) and 500-to 600-msec (F(1 9) = 1199 p lt 0008) intervals was

Figure 4 The MMN was similar in the Auditory-alone (Aa) andAuditory-Visual (AV) conditions as revealed by nonsignicant differ-ences of the mean amplitude of ve consecutive 50-msec intervalsstarting at MMN onset (50 msec) [two-way ANOVAs Condition (Aaversus AV) acute Electrode (F3 Fz F4 C3 Cz C4)] Top Grand-averageERPs to standard and deviant tones at Fz in the two conditionsBottom Deviant-standard difference waves

Figure 5 The component structure of the ERPs to deviant tonesand novel sounds as revealed by the deviant-standard and novel-stan-dard difference waves at Fz and right-mastoid (RM) recordings Devi-ant tones elicited the MMN with a peak latency of 150 msec at Fzand with an inverted polarity at RM (top) Novel sounds elicited alarge N1 deection at Fz that was presumably composed of contribu-tions from two different processes for which the signicant differ-ences between the standard-tone and novel-sound ERPs extendedbeyond the N1 latency range (bottom) The novel sounds also elic-ited a P3a wave with two peaks at Fz suggesting two different sub-components contributing to the recorded P3a an early P3a (peaklatency 230 msec) which inverted polarity at RM and other poste-rior and lateral sites and a late P3a (peak latency 315 msec) whichinverted polarity at no recording site

596 Journal of Cognitive Neuroscience Volume 10 Number 5

also signicantly larger for visual stimuli following novelsounds than after those following the standard tones

DISCUSSION

Engagement of Attention During VisualPerformance

The present results provide evidence that attention dur-ing visual performance can be engaged by shortly pre-ceding unexpected novel sounds or small changes in thetask-irrelevant acoustic environment These results alsorevealed that this phenomenon depends on the intrinsicnature of the eliciting acoustic events In addition the

observed effects on visual performance were associatedwith different patterns of neural events activated bydeviant tones and novel sounds as reected by thecomponent structure of the auditory ERPs elicited

As expected novel sounds prolonged the RT to suc-cessive visual stimuli This result is in agreement with theextensive literature reporting delayed RTs to target stim-uli caused by preceding irrelevant sound changes ornovels sounds (Alho et al 1992 Cherry 1953 Dawsonet al 1989 Fillion et al 1991 Grillon et al 1990Schroumlger 1996 Woods 1992 Woodward et al 1991) andit provides experimental support for the everyday expe-rience that attention can be engaged by irrelevant acous-tic events

Figure 6 The scalp distribu-tion of the P3a to novelsounds Left early P3a rightlate P3a top Auditory-alone(Aa) condition bottom Audi-tory-Visual (AV) condition Thecentral plot shows the novel-standard difference waves atFz in the Aa (thin) and AV(thick) conditions as well asthe 100-msec latency win-dows used to measure thetwo phases of the P3a (e =early P3a l = late P3a)

Escera et al 597

A surprising nding of this study was however thatthe slightly deviant tones degraded the visual perfor-mance accuracy by decreasing the rate of hit responseswhereas the widely deviant novel sounds did not It wasexpected that the amount of distraction caused by devi-ant tones and novel sounds would be proportional tothe magnitude of their deviance from the standard toneand indeed such an effect was found for the RT (17-msecincrease after novel sounds) conrming other recentndings (Jaumlaumlskelaumlinen et al 1996 Schroumlger 1996) How-ever this RT effect was not paralleled by a similar effecton the hit rate Indeed deviant tones were associatedwith a lower hit rate whereas the hit rate after novelsounds was similar to that after standard tones Similarresults were obtained in a related study by Jaumlaumlskelaumlinenet al (1996) and recently replicated for deviant tones(Alho Escera Diacuteaz Yago amp Serra 1997) These ndingssuggest that the distracting effects of an irrelevant acous-tic event on visual performance may depend on thenature of the eliciting sound That is obtrusive novelsounds appear to cause a stronger effect on the RT tosuccessive visual target stimuli than minor changes inthe acoustic stream but in contrast these minor changes

seem to cause a signicant decrement in the number ofhit responses to visual targets not observed after thenovel sounds

The differential effects of distracting novel and deviantsounds on visual performance may be explained by trac-ing these behavioral effects back to the associated neuralevents as reected by ERPs Novel sounds enhanced theN1 amplitude and elicited the MMN as suggested bysignicant differences between the standard-tone andnovel-sound ERP that extended over the N1 latencyrange The present MMN to novel sounds peaked andterminated (returned to baseline) earlier than the MMNto deviant tones in concordance with previous studiesshowing that with larger deviances MMN becomes ear-lier (Tiitinen May Reinikainen amp Naumlaumltaumlnen 1994) over-lapping the N1 peak for wider deviances (Tiitinen et al1994 Scherg et al 1989) The present MMN to novelsounds is also supported by recent magnetoencepha-lograph (MEG) data showing that novel sounds activatedan MMN source in the human auditory cortex (Alhoet al 1998) Thus for novel sounds the attention-switch-ing signal was probably triggered by a combined re-sponse of the transient-detector system reected in thesupratemporal N1 (Naumlaumltaumlnen 1990 1992 Naumlaumltaumlnen ampPicton 1987) or some other N1 component such as thefrontal N1 reported by Giard et al (1994) and the stimu-lus-change detector system reected in the MMN(Naumlaumltaumlnen 1990 1992) This combined signal for atten-tion capture started with the N1 generation and wasoverlapped by the MMN process resulting in an effectiveattentional reorientation as reected by the subsequentlarge P3a wave and the clearly delayed RT to the follow-ing visual targets

Figure 7 Grand-average ERPs to standard tones and novel soundsin the Auditory-alone (Aa left column) and Auditory-Visual (AV rightcolumn) conditions The novel-standard difference waves are shownin the middle column The late P3a was larger in the AV conditionthan in the Aa condition whereas the early P3a was of similar ampli-tude in the two conditions

Figure 8 Grand-average visual ERPs at Pz O1 and O2 in the Visualalone (Va) condition and in the Auditory-Visual (AV) condition afterstandard tones

598 Journal of Cognitive Neuroscience Volume 10 Number 5

Deviant tones in turn elicited a distinct MMN whichwas followed by a small P3a wave and by a secondnegative peak (Figure 5) Thus the neural mechanismresponsible for triggering the attention-switching signalwas based for deviant tones on the stimulus-changedetector system reected in the MMN (Naumlaumltaumlnen 19901992 Schroumlger 1996) This attention capture signal asso-ciated with the MMN was sporadically followed by an

orienting of attention as suggested by the small (aver-aged) P3a and also by the 23 hit rate decrease to visualtargets that followed deviant tones

Although the subjects were instructed to ignore theauditory stimuli in the AV condition the RT to visualtargets following standard tones was slightly (insig-nicantly) shorter than the RT in the Va condition withno auditory stimuli This suggests that the auditory stim-

Figure 9 Grand-averageERPs in the Auditory-Visual(AV) condition to standard-visual (thin line) deviant-visual (thick line) andnovel-visual (dashed line)stimulus pairs Only the mostrelevant electrodes are shownThe epoch starts 100 msec be-fore auditory-stimulus onsetwhich was used as the base-line period to avoid the largeamplitude variability due tothe auditory response in the100 msec preceding the vis-ual stimulus Novel sounds en-hanced the fronto-polarnegativity and the frontal N2to visual stimuli was en-hanced after novel soundsand deviant tones which alsoaffected the offset slope ofP3b Notice that whereas thetraces for standard-visual devi-ant-visual and novel-visualstimulus pairs perfectly matcheach other during the onsetslope of the P3b they start todiffer from the P3b peak forthe novel-visual stimulus pairand about 100 msec later forthe deviant-visual stimuluspair (see P3 Pz P4)

Escera et al 599

uli were used as warning signals for the following visualtargets and thus attention may have been bimodal in theAV condition However because the occurrence of an aud-itory stimulus change or a novel sound was task-irrelevantand unpredictable to the subjects attention was involun-tarily engaged by these events In a more controlledexperiment attention was more effectively withdrawnfrom the auditory stimuli and their occurrence was madenoninformative by the simultaneous presentation of avisual cue stimulus informing about the successive ap-pearance of a visual target (Alho et al 1997) Like in thepresent study a decreased hit rate for visual stimuli thatfollowed MMN- and P3a-eliciting deviant tones was foundThis conrms the involuntary nature of the attention-switching mechanism indicated by the MMN P3a anddeterioration of the visual-task performance

Component Structure of the P3a to Novel Sounds

In both the Aa and AV conditions novel sounds eliciteda large long-lasting and broadly distributed P3a wavewith a double peak at Fz (in the Aa condition) suggestinga contribution of two different subcomponents Theearly part of this P3a had a peak latency of 230 msecand showed a centrally dominant scalp distribution thatwas similar in the Aa and AV conditions (Figure 6) Theearly P3a inverted in polarity at posterior and lateralelectrodes (Figures 3b and 5) suggesting bilateral gener-ators located in the vicinity of the temporal and parietallobes (cf Scherg amp von Cramon 1986) The late part ofthe P3a in turn had a peak latency of 315 msec and didnot invert in polarity at posterior electrodes and its scalpdistribution was centered over the right frontal areas Incontrast to the early P3a the late P3a was enhanced inthe AV condition particularly over the right hemisphere(Figure 6) when subjects apparently used auditory stim-uli as warning signals to prepare for visual targets

The component structure of the present P3a and thescalp distribution of the two subcomponents is in agree-ment with previous results in patients with brain lesionssuggesting a critical role of the temporal-parietal junc-tion in the generation of the auditory (Halgren BaudenaClarke Heit Marinkovic et al 1995 Halgren BaudenaClarke Heit Lieacutegeois et al 1995 Knight Scabini Woodsamp Clayworth 1989) somatosensory (Yamaguchi ampKnight 1991 1992) and visual P3a (Knight 1991 1997)It is also compatible with results suggesting a frontalcontribution to P3a such as those obtained in patientswith frontal lesions (Knight 1984) in topographicalanalysis (Friedman amp Simpson 1994 Friedman Simpsonamp Hamberger 1993) in dipole-source modeling (Meck-linger amp Ullsperger 1995) and in intracerebral record-ings in humans (Baudena Halgren Heit amp Clarke 1995)

The late P3a amplitude in the AV condition was en-hanced with regard to its amplitude in the Aa conditionThis amplitude enhancement of the late P3a reects an

attentional monitoring of auditory stimuli in the AV con-dition these stimuli being used as warning signals for thefollowing visual targets The attentional modulation ofthe P3a amplitude is in agreement with that reported inprevious studies (Holdstock amp Rugg 1995 Woods 1992)and suggests that the late P3a may more closely reectan actual orienting of attention than the involuntary N1and MMN processes triggering this attentional orientingAlso several other ndings support the association of thelate phase of the P3a with the actual reorientation ofattention or orienting response First the right-frontaldominance in the late P3a scalp distribution observed inthe present experiment agrees with ndings suggestingthe involvement of right-frontal areas in the reorienta-tion of attention (Fuster 1989 Mesulam 1981) Secondit has been shown that the P3a amplitude is attenuatedwith the repetition of the eliciting novel event(Courchesne 1978 Friedman amp Simpson 1994 Knight1984) This habituating behavior is a characteristic fea-ture of the orienting response (Oumlhman 1979 1992 Sok-olov 1963)

The early P3a was insensitive to attentional manipula-tions as it was to similar amplitude and scalp distributionin the Aa and AV conditions Its peak latency was rathershort (230 msec) and its scalp distribution suggestsneural generators probably located in the temporal-parietal cortex (Knight 1991 1997 Knight et al 1989Yamaguchi amp Knight 1991) Thus the early P3a mightreect a neural process other than attentional reorienta-tion such as the violation of a polisensorial model of theexternal world maintained in the temporal-parietal asso-ciation cortex (Yamaguchi amp Knight 1991)

Conclusions

The results reported in the present paper support thenotion of two different brain mechanisms of involuntaryattention differentially activated by different acousticevents The ERP component structure to novel soundssuggests that unexpected novelty in the acoustic envi-ronment involuntary captures attention by simultane-ously activating two different neural mechanisms atransient-detector mechanism associated with the su-pratemporal N1 component of the auditory ERPs and achange-detector mechanism reected in the MMN Pre-sumably the attention-switching signal generated by thecombined activation of these two mechanisms resultedin an effective engagement of attention as indicated bythe large P3a component elicited by the novel soundsand the increased RT to subsequent visual targets Devi-ant tones in turn resulted in different behavioral effectsand ERP component structure which was characterizedby the MMN This supports the notion that small changesin the acoustic environment capture attention involun-tarily by activating the stimulus-change detector mecha-nism reected in the MMN

600 Journal of Cognitive Neuroscience Volume 10 Number 5

METHODS

Subjects Stimuli and Procedure

Ten right-handed paid students (21 to 40 years 3 males)with normal hearing and normal or corrected-to-normalvisual acuity participated in the experiment Two furthersubjects were discarded due to their inability to reachthe required 90 hit rate in a practice session

Subjects were presented with four blocks of 400stimulus pairs (AV condition) in a visual forced-choiceRT task Each pair consisted of an auditory stimulusfollowed after 300 msec (onset-to-onset) by a visualstimulus The interpair interval (onset-to-onset) was 12sec (Figure 1) The auditory stimuli were either standardor deviant tones and novel sounds presented in randomorder with probabilities of 08 01 and 01 respectivelyThe standard and deviant tones were sinusoidal tonebursts of 200-msec duration including 10-msec rise andfall times presented binaurally at 75 dB SPL throughheadphones with respective frequencies of 600 and 700Hz Sixty different environmental sounds such as thoseproduced by a drill hammer rain door telephone ring-ing etc were used as novel sounds They were digitallyrecorded then low-pass ltered at 10000 Hz and editedto have a duration of 200 msec including rise and falltimes of 10 msec and an intensity maximum of 70 to 80dB SPL Each different novel sound occurred only oncewithin a stimulus block and was presented twice orthree times during the whole experiment Both devianttones and novel sounds were always preceded by at leastone AV pair containing a standard tone The visual stimuliwere the digits 1 to 8 presented one at a time in randomorder on a computer screen for 200 msec They sub-tended a vertical angle of 17deg and a horizontal angle of11deg (30 mm acute 20 mm 100 cm from the subjectrsquos eyes)

Subjects were comfortably seated in a reclining chairin a dimly lit electrically and acoustically shielded roomThey were instructed to focus on the small xation crossappearing in the middle of the screen and to press oneresponse button with the right index nger for oddnumbers and another button with the right middlenger for even numbers Subjects were also instructedto ignore the auditory stimulation Both speed and accu-racy were emphasized for the primary visual task

In two further conditions visual stimuli were omitted(Va condition Figure 1) or auditory stimuli were omitted(Aa condition Figure 1) In the Va condition subjectsreceived three blocks of 200 visual stimulus each theinstruction being identical to that in the AV condition Inthe Aa condition subjects were instructed to concen-trate on reading a self-selected book and to ignore theauditory stimulation

The Aa and AV blocks were presented in an alternatingorder with half of the subjects starting with the AVcondition The Va blocks were presented one at thebeginning one at the middle and one at the end of the

experiment Before the experimental blocks subjectsparticipated in a practice session consisting of two Vablocks They were required to reach a 90 hit rate beforethe experiment could start Two out of 12 subjects failedto reach this level even after several additional blocksand were nally discarded

EEG Recording and Averaging

The electroencephalogram (EEG) (bandpass 0ndash100 Hz)was continuously digitized at a rate of 500 Hz by aSynAmps amplier (NeuroScan Inc) from 19 scalp elec-trodes positioned according to the 10ndash20 system Fp1Fp2 F7 F3 Fz F4 F8 T3 C3 Cz C4 T4 T5 P3 Pz P4 T6O1 and O2 One additional electrode was placed at theright mastoid (RM) and another at the left mastoid (LM)The horizontal EOG was recorded with electrodes at-tached to the left and right canthi The vertical elec-trooculogram (EOG) was assessed using recordings fromFp1 and Fp2 electrodes The common reference elec-trode was placed on the tip of the nose

ERPs were averaged off-line for each auditory stimulusclass for an epoch of 1300 msec including a preauditorystimulus period of 100 msec in the AV and Aa conditionsIn the Va condition an ERP was also obtained for anepoch of 1300 msec including a previsual stimulus pe-riod of 400 msec Epochs in which the EEG or EOGexceeded plusmn100 m V as well as the ve rst epochs of eachblock were automatically excluded from averaging Inthe Aa and AV conditions standard-tone trials immedi-ately following deviant-tone or novel-sound trials werealso excluded from the averages Individual ERPs wereband-pass ltered between 001 and 30 Hz

Data Analysis

In the AV and Va conditions a correct button presswithin 800 msec after visual-stimulus onset was regardedas a hit the mean RT being computed only for the hittrials An incorrect button press during this period wasclassied as an error and trials with no response withinthis time window were considered misses Hits errorsmisses and RTs were computed across odd and evennumbers

Auditory ERP amplitudes (N1 P2 MMN and P3a) werereferred to the mean amplitude in the 100-msec baselineperiod preceding the auditory stimulus in both the Aaand AV conditions For auditory ERP analysis each indi-vidual waveform in the Aa and AV conditions was linearlydetrended over an epoch of 600 msec (including base-line) to remove any possible asymmetrical contributionto EEG caused by slow EOG movements due to readingin the Aa condition The effects of the standard tones onthe visual ERPs were analyzed at O1 and O2 for peaksas follows (identied in the individual waveforms withinthe specied latency ranges) P1 75 to 140 msec N1 100

Escera et al 601

to 200 msec P2 180 to 280 msec N2 225 to 350 msecThe visual P3b was identied at Pz between 300 and 500msec from visual-stimulus onset For this analysis visualERP amplitudes were measured against the mean ampli-tude of the 100-msec baseline preceding visual-stimulusonset The fronto-polar negativity was analyzed at Fp1and Fp2 the visual frontal N2 at F3 Fz and F4 and thevisual P3b at P3 Pz and P4 The amplitude measurementsfor visual ERP components in the AV condition werereferred to the 100-msec preauditory stimulus baselinerather than to the 100-msec previsual stimulus baselinewhich was differentially affected by standard tones de-viant tones and novel sounds (see Figure 9)

The ERP and performance data were statistically ana-lyzed by ANOVA with repeated measures Where appro-priate nominal degrees of freedom and epsilon valuesafter the Greenhouse-Geisser correction are reportedScalp-distribution analyses were performed after normal-izing ERP amplitudes to prevent amplitude differencesbetween different components from washing out thegenuine scalp-distribution differences This normalizationwas done by dividing the amplitude at each electrodeby the sum of the squared amplitudes at all electrodes(McCarthy amp Wood 1985)

Acknowledgments

This research was supported by BIOMED-2 contract BMH4-CT96-0819-COBRAIN of the European Union the SpanishMinistry of Science and Education (DGICYT PB93-0802 UE96-0038) the Academy of Finland the scientic exchange programbetween the University of Barcelona and the University ofHelsinki and the National Science Fund of Hungary (OTKAT022681) The authors would like to thank Teemu Rinne andTeija Kujala for their help in making the gures

Reprint requests should be sent to Carles Escera Neurody-namics Laboratory Department of Psychiatry and Clinical Psy-chobiology University of Barcelona P Vall drsquoHebron 171 08035Barcelona Catalonia-Spain or via e-mail cescerapsiubes

Note

1 The MMN amplitude in the unattended channel can behowever attenuated under some conditions by strongly focus-ing attention on other sources of auditory stimulation(Naumlaumltaumlnen 1991 Naumlaumltaumlnen Paavilainen Tiitinen Jiang amp Alho1993 Oades amp Dittmannbalcar 1995 Trejo Ryan-Jones amp Kra-mer 1995 Woldorff Hackley amp Hillyard 1991)

REFERENCES

Alho K (1995) Cerebral generators of mismatch negativity(MMN) and its magnetic counterpart (MMNn) elicited bysound changes Ear and Hearing 16 38ndash51

Alho K Escera C Diacuteaz R Yago E amp Serra J M (1997)Effects of involuntary auditory attention on visual taskperformance and brain activity NeuroReport 8 3233ndash3237

Alho K Winkler I Escera C Huotilainen M Virtanen JJaumlaumlskelaumlinen I Pekkonen E amp Ilmoniemi R (1998) Proc-

essing of novel sounds and frequency changes in the hu-man auditory cortex Magnetoencephalographic record-ings Psychophysiology 35 211ndash224

Alho K Woods D L Algazi A amp Naumlaumltaumlnen R (1992) Inter-modal selective attention II Effects of attentional load onprocessing of auditory and visual stimuli in central spaceElectroencephalography and Clinical Neurophysiology82 356ndash368

Baudena P Halgren E Heit G amp Clarke J M (1995) Intra-cerebral potentials to rare target and distractor auditoryand visual stimuli III Frontal cortex Electroencephalogra-phy and Clinical Neurophysiology 94 251ndash264

Cherry E C (1953) Some experiments on the recognitionof speech with one and with two ears Journal of theAcoustical Society of America 29 975ndash979

Courchesne E (1978) Changes in P3 waves with event repe-tition Long-term effects on scalp distribution and ampli-tudes Electroencephalography and ClinicalNeurophysiology 45 468ndash482

Courchesne E Hillyard S A amp Galambos R (1975) Stimu-lus novelty task relevance and the visual evoked potentialsin man Electroencephalography and Clinical Neurophysi-ology 39 131ndash143

Cowan N Winkler I Teder W amp Naumlaumltaumlnen R (1993) Mem-ory prerequisites of the mismatch negativity in the audi-tory event-related potential (ERP) Journal ofExperimental Psychology Learning Memory and Cogni-tion 19 909ndash921

Cseacutepe V Karmos G amp Molnaacuter M (1987) Evoked potentialcorrelates of stimulus deviance during wakefulness andsleep in catmdashanimal model of mismatch negativity Elec-troencephalography and Clinical Neurophysiology 66571ndash578

Davis H Mast T Yoshie N amp Zerlin S (1966) The slow re-sponse of the human cortex to auditory stimulus Recov-ery process Electroencephalography and ClinicalNeurophysiology 21 105ndash113

Davis H amp Zerlin S (1966) Acoustic relations of the humanvertex potentials Journal of the Acoustical Society ofAmerica 39 109ndash116

Dawson M E (1990) Psychophysiology at the interface ofclinical science cognitive science and neuroscience Psy-chophysiology 27 243ndash255

Dawson M E Fillion D L amp Schell A M (1989) Is elicita-tion of the autonomic orienting response associated withallocation of processing resources Psychophysiology 26560ndash572

Duncan C C amp Kaye W H (1987) Effects of chlonidine onevent-related potential measures of information process-ing In R Johnson J W Rohrbaugh amp R Parasuraman(Eds) Current trends in event-related potentialresearch (EEG supplement 40) (pp 527ndash531) Amster-dam Elsevier

Fillion D J Dawson M E Schell A M amp Hazlett E A(1991) The relationship between skin conductance orient-ing and the allocation of processing resources Psycho-physiology 28 410ndash424

Folk C L Remington R W amp Johnston J C (1992) Involun-tary covert orienting is contingent on attentional controlsettings Journal of Experimental Psychology Human Per-ception and Performance 18 1030ndash1044

Ford J M Roth W T amp Kopell B S (1976) Auditory event-related potentials to unpredictable shifts in pitch Psycho-physiology 13 32ndash39

Friedman D amp Simpson G (1994) ERP amplitude and scalpdistribution to target and novel events Effects of temporalorder in young middle-aged and older adults CognitiveBrain Research 2 49ndash63

602 Journal of Cognitive Neuroscience Volume 10 Number 5

in all cases) at latency windows 200 to 300 msec 300to 400 msec and 400 to 500 msec from visual-stimulusonset

Novel sounds and deviant tones enhanced the frontalN2 deection elicited by the subsequent visual stimulusin comparison with the visual N2 elicited after standardtones (Figure 9) This was revealed by two-way ANOVAsusing the Electrode (F3 Fz F4) and the auditory Stimulus(standard versus either deviant or novel) as factors Themean amplitude in the 200- to 300-msec range fromvisual-stimulus onset at Fz was - 19 m V after standardtones - 33 m V after deviant tones (F(1 9) = 641 p lt004) and - 36 m V after novel sounds (F(1 9) = 899 plt 002) The mean amplitude in the 300- to 400-msecinterval was also signicantly larger for visual stimulifollowing deviant tones or novel sounds than for thosefollowing standard tones F(1 9) = 815 p lt 002 for

deviant versus standard F(1 9) = 795 p lt 001 for novelversus standard

Preceding novel sounds and deviant tones also en-hanced the terminal slope of the visual P3b (Figure 9)this was revealed by the signicantly larger mean ampli-tude of the 500- to 600-msec latency window (at P3 Pzand P4) in the ERP to visual stimuli following devianttones than in the ERP to visual stimuli following standardtones (F(19) = 1556 p lt 004) The mean amplitude inthe 400- to 500-msec (F(1 9) = 694 p lt 003) and 500-to 600-msec (F(1 9) = 1199 p lt 0008) intervals was

Figure 4 The MMN was similar in the Auditory-alone (Aa) andAuditory-Visual (AV) conditions as revealed by nonsignicant differ-ences of the mean amplitude of ve consecutive 50-msec intervalsstarting at MMN onset (50 msec) [two-way ANOVAs Condition (Aaversus AV) acute Electrode (F3 Fz F4 C3 Cz C4)] Top Grand-averageERPs to standard and deviant tones at Fz in the two conditionsBottom Deviant-standard difference waves

Figure 5 The component structure of the ERPs to deviant tonesand novel sounds as revealed by the deviant-standard and novel-stan-dard difference waves at Fz and right-mastoid (RM) recordings Devi-ant tones elicited the MMN with a peak latency of 150 msec at Fzand with an inverted polarity at RM (top) Novel sounds elicited alarge N1 deection at Fz that was presumably composed of contribu-tions from two different processes for which the signicant differ-ences between the standard-tone and novel-sound ERPs extendedbeyond the N1 latency range (bottom) The novel sounds also elic-ited a P3a wave with two peaks at Fz suggesting two different sub-components contributing to the recorded P3a an early P3a (peaklatency 230 msec) which inverted polarity at RM and other poste-rior and lateral sites and a late P3a (peak latency 315 msec) whichinverted polarity at no recording site

596 Journal of Cognitive Neuroscience Volume 10 Number 5

also signicantly larger for visual stimuli following novelsounds than after those following the standard tones

DISCUSSION

Engagement of Attention During VisualPerformance

The present results provide evidence that attention dur-ing visual performance can be engaged by shortly pre-ceding unexpected novel sounds or small changes in thetask-irrelevant acoustic environment These results alsorevealed that this phenomenon depends on the intrinsicnature of the eliciting acoustic events In addition the

observed effects on visual performance were associatedwith different patterns of neural events activated bydeviant tones and novel sounds as reected by thecomponent structure of the auditory ERPs elicited

As expected novel sounds prolonged the RT to suc-cessive visual stimuli This result is in agreement with theextensive literature reporting delayed RTs to target stim-uli caused by preceding irrelevant sound changes ornovels sounds (Alho et al 1992 Cherry 1953 Dawsonet al 1989 Fillion et al 1991 Grillon et al 1990Schroumlger 1996 Woods 1992 Woodward et al 1991) andit provides experimental support for the everyday expe-rience that attention can be engaged by irrelevant acous-tic events

Figure 6 The scalp distribu-tion of the P3a to novelsounds Left early P3a rightlate P3a top Auditory-alone(Aa) condition bottom Audi-tory-Visual (AV) condition Thecentral plot shows the novel-standard difference waves atFz in the Aa (thin) and AV(thick) conditions as well asthe 100-msec latency win-dows used to measure thetwo phases of the P3a (e =early P3a l = late P3a)

Escera et al 597

A surprising nding of this study was however thatthe slightly deviant tones degraded the visual perfor-mance accuracy by decreasing the rate of hit responseswhereas the widely deviant novel sounds did not It wasexpected that the amount of distraction caused by devi-ant tones and novel sounds would be proportional tothe magnitude of their deviance from the standard toneand indeed such an effect was found for the RT (17-msecincrease after novel sounds) conrming other recentndings (Jaumlaumlskelaumlinen et al 1996 Schroumlger 1996) How-ever this RT effect was not paralleled by a similar effecton the hit rate Indeed deviant tones were associatedwith a lower hit rate whereas the hit rate after novelsounds was similar to that after standard tones Similarresults were obtained in a related study by Jaumlaumlskelaumlinenet al (1996) and recently replicated for deviant tones(Alho Escera Diacuteaz Yago amp Serra 1997) These ndingssuggest that the distracting effects of an irrelevant acous-tic event on visual performance may depend on thenature of the eliciting sound That is obtrusive novelsounds appear to cause a stronger effect on the RT tosuccessive visual target stimuli than minor changes inthe acoustic stream but in contrast these minor changes

seem to cause a signicant decrement in the number ofhit responses to visual targets not observed after thenovel sounds

The differential effects of distracting novel and deviantsounds on visual performance may be explained by trac-ing these behavioral effects back to the associated neuralevents as reected by ERPs Novel sounds enhanced theN1 amplitude and elicited the MMN as suggested bysignicant differences between the standard-tone andnovel-sound ERP that extended over the N1 latencyrange The present MMN to novel sounds peaked andterminated (returned to baseline) earlier than the MMNto deviant tones in concordance with previous studiesshowing that with larger deviances MMN becomes ear-lier (Tiitinen May Reinikainen amp Naumlaumltaumlnen 1994) over-lapping the N1 peak for wider deviances (Tiitinen et al1994 Scherg et al 1989) The present MMN to novelsounds is also supported by recent magnetoencepha-lograph (MEG) data showing that novel sounds activatedan MMN source in the human auditory cortex (Alhoet al 1998) Thus for novel sounds the attention-switch-ing signal was probably triggered by a combined re-sponse of the transient-detector system reected in thesupratemporal N1 (Naumlaumltaumlnen 1990 1992 Naumlaumltaumlnen ampPicton 1987) or some other N1 component such as thefrontal N1 reported by Giard et al (1994) and the stimu-lus-change detector system reected in the MMN(Naumlaumltaumlnen 1990 1992) This combined signal for atten-tion capture started with the N1 generation and wasoverlapped by the MMN process resulting in an effectiveattentional reorientation as reected by the subsequentlarge P3a wave and the clearly delayed RT to the follow-ing visual targets

Figure 7 Grand-average ERPs to standard tones and novel soundsin the Auditory-alone (Aa left column) and Auditory-Visual (AV rightcolumn) conditions The novel-standard difference waves are shownin the middle column The late P3a was larger in the AV conditionthan in the Aa condition whereas the early P3a was of similar ampli-tude in the two conditions

Figure 8 Grand-average visual ERPs at Pz O1 and O2 in the Visualalone (Va) condition and in the Auditory-Visual (AV) condition afterstandard tones

598 Journal of Cognitive Neuroscience Volume 10 Number 5

Deviant tones in turn elicited a distinct MMN whichwas followed by a small P3a wave and by a secondnegative peak (Figure 5) Thus the neural mechanismresponsible for triggering the attention-switching signalwas based for deviant tones on the stimulus-changedetector system reected in the MMN (Naumlaumltaumlnen 19901992 Schroumlger 1996) This attention capture signal asso-ciated with the MMN was sporadically followed by an

orienting of attention as suggested by the small (aver-aged) P3a and also by the 23 hit rate decrease to visualtargets that followed deviant tones

Although the subjects were instructed to ignore theauditory stimuli in the AV condition the RT to visualtargets following standard tones was slightly (insig-nicantly) shorter than the RT in the Va condition withno auditory stimuli This suggests that the auditory stim-

Figure 9 Grand-averageERPs in the Auditory-Visual(AV) condition to standard-visual (thin line) deviant-visual (thick line) andnovel-visual (dashed line)stimulus pairs Only the mostrelevant electrodes are shownThe epoch starts 100 msec be-fore auditory-stimulus onsetwhich was used as the base-line period to avoid the largeamplitude variability due tothe auditory response in the100 msec preceding the vis-ual stimulus Novel sounds en-hanced the fronto-polarnegativity and the frontal N2to visual stimuli was en-hanced after novel soundsand deviant tones which alsoaffected the offset slope ofP3b Notice that whereas thetraces for standard-visual devi-ant-visual and novel-visualstimulus pairs perfectly matcheach other during the onsetslope of the P3b they start todiffer from the P3b peak forthe novel-visual stimulus pairand about 100 msec later forthe deviant-visual stimuluspair (see P3 Pz P4)

Escera et al 599

uli were used as warning signals for the following visualtargets and thus attention may have been bimodal in theAV condition However because the occurrence of an aud-itory stimulus change or a novel sound was task-irrelevantand unpredictable to the subjects attention was involun-tarily engaged by these events In a more controlledexperiment attention was more effectively withdrawnfrom the auditory stimuli and their occurrence was madenoninformative by the simultaneous presentation of avisual cue stimulus informing about the successive ap-pearance of a visual target (Alho et al 1997) Like in thepresent study a decreased hit rate for visual stimuli thatfollowed MMN- and P3a-eliciting deviant tones was foundThis conrms the involuntary nature of the attention-switching mechanism indicated by the MMN P3a anddeterioration of the visual-task performance

Component Structure of the P3a to Novel Sounds

In both the Aa and AV conditions novel sounds eliciteda large long-lasting and broadly distributed P3a wavewith a double peak at Fz (in the Aa condition) suggestinga contribution of two different subcomponents Theearly part of this P3a had a peak latency of 230 msecand showed a centrally dominant scalp distribution thatwas similar in the Aa and AV conditions (Figure 6) Theearly P3a inverted in polarity at posterior and lateralelectrodes (Figures 3b and 5) suggesting bilateral gener-ators located in the vicinity of the temporal and parietallobes (cf Scherg amp von Cramon 1986) The late part ofthe P3a in turn had a peak latency of 315 msec and didnot invert in polarity at posterior electrodes and its scalpdistribution was centered over the right frontal areas Incontrast to the early P3a the late P3a was enhanced inthe AV condition particularly over the right hemisphere(Figure 6) when subjects apparently used auditory stim-uli as warning signals to prepare for visual targets

The component structure of the present P3a and thescalp distribution of the two subcomponents is in agree-ment with previous results in patients with brain lesionssuggesting a critical role of the temporal-parietal junc-tion in the generation of the auditory (Halgren BaudenaClarke Heit Marinkovic et al 1995 Halgren BaudenaClarke Heit Lieacutegeois et al 1995 Knight Scabini Woodsamp Clayworth 1989) somatosensory (Yamaguchi ampKnight 1991 1992) and visual P3a (Knight 1991 1997)It is also compatible with results suggesting a frontalcontribution to P3a such as those obtained in patientswith frontal lesions (Knight 1984) in topographicalanalysis (Friedman amp Simpson 1994 Friedman Simpsonamp Hamberger 1993) in dipole-source modeling (Meck-linger amp Ullsperger 1995) and in intracerebral record-ings in humans (Baudena Halgren Heit amp Clarke 1995)

The late P3a amplitude in the AV condition was en-hanced with regard to its amplitude in the Aa conditionThis amplitude enhancement of the late P3a reects an

attentional monitoring of auditory stimuli in the AV con-dition these stimuli being used as warning signals for thefollowing visual targets The attentional modulation ofthe P3a amplitude is in agreement with that reported inprevious studies (Holdstock amp Rugg 1995 Woods 1992)and suggests that the late P3a may more closely reectan actual orienting of attention than the involuntary N1and MMN processes triggering this attentional orientingAlso several other ndings support the association of thelate phase of the P3a with the actual reorientation ofattention or orienting response First the right-frontaldominance in the late P3a scalp distribution observed inthe present experiment agrees with ndings suggestingthe involvement of right-frontal areas in the reorienta-tion of attention (Fuster 1989 Mesulam 1981) Secondit has been shown that the P3a amplitude is attenuatedwith the repetition of the eliciting novel event(Courchesne 1978 Friedman amp Simpson 1994 Knight1984) This habituating behavior is a characteristic fea-ture of the orienting response (Oumlhman 1979 1992 Sok-olov 1963)

The early P3a was insensitive to attentional manipula-tions as it was to similar amplitude and scalp distributionin the Aa and AV conditions Its peak latency was rathershort (230 msec) and its scalp distribution suggestsneural generators probably located in the temporal-parietal cortex (Knight 1991 1997 Knight et al 1989Yamaguchi amp Knight 1991) Thus the early P3a mightreect a neural process other than attentional reorienta-tion such as the violation of a polisensorial model of theexternal world maintained in the temporal-parietal asso-ciation cortex (Yamaguchi amp Knight 1991)

Conclusions

The results reported in the present paper support thenotion of two different brain mechanisms of involuntaryattention differentially activated by different acousticevents The ERP component structure to novel soundssuggests that unexpected novelty in the acoustic envi-ronment involuntary captures attention by simultane-ously activating two different neural mechanisms atransient-detector mechanism associated with the su-pratemporal N1 component of the auditory ERPs and achange-detector mechanism reected in the MMN Pre-sumably the attention-switching signal generated by thecombined activation of these two mechanisms resultedin an effective engagement of attention as indicated bythe large P3a component elicited by the novel soundsand the increased RT to subsequent visual targets Devi-ant tones in turn resulted in different behavioral effectsand ERP component structure which was characterizedby the MMN This supports the notion that small changesin the acoustic environment capture attention involun-tarily by activating the stimulus-change detector mecha-nism reected in the MMN

600 Journal of Cognitive Neuroscience Volume 10 Number 5

METHODS

Subjects Stimuli and Procedure

Ten right-handed paid students (21 to 40 years 3 males)with normal hearing and normal or corrected-to-normalvisual acuity participated in the experiment Two furthersubjects were discarded due to their inability to reachthe required 90 hit rate in a practice session

Subjects were presented with four blocks of 400stimulus pairs (AV condition) in a visual forced-choiceRT task Each pair consisted of an auditory stimulusfollowed after 300 msec (onset-to-onset) by a visualstimulus The interpair interval (onset-to-onset) was 12sec (Figure 1) The auditory stimuli were either standardor deviant tones and novel sounds presented in randomorder with probabilities of 08 01 and 01 respectivelyThe standard and deviant tones were sinusoidal tonebursts of 200-msec duration including 10-msec rise andfall times presented binaurally at 75 dB SPL throughheadphones with respective frequencies of 600 and 700Hz Sixty different environmental sounds such as thoseproduced by a drill hammer rain door telephone ring-ing etc were used as novel sounds They were digitallyrecorded then low-pass ltered at 10000 Hz and editedto have a duration of 200 msec including rise and falltimes of 10 msec and an intensity maximum of 70 to 80dB SPL Each different novel sound occurred only oncewithin a stimulus block and was presented twice orthree times during the whole experiment Both devianttones and novel sounds were always preceded by at leastone AV pair containing a standard tone The visual stimuliwere the digits 1 to 8 presented one at a time in randomorder on a computer screen for 200 msec They sub-tended a vertical angle of 17deg and a horizontal angle of11deg (30 mm acute 20 mm 100 cm from the subjectrsquos eyes)

Subjects were comfortably seated in a reclining chairin a dimly lit electrically and acoustically shielded roomThey were instructed to focus on the small xation crossappearing in the middle of the screen and to press oneresponse button with the right index nger for oddnumbers and another button with the right middlenger for even numbers Subjects were also instructedto ignore the auditory stimulation Both speed and accu-racy were emphasized for the primary visual task

In two further conditions visual stimuli were omitted(Va condition Figure 1) or auditory stimuli were omitted(Aa condition Figure 1) In the Va condition subjectsreceived three blocks of 200 visual stimulus each theinstruction being identical to that in the AV condition Inthe Aa condition subjects were instructed to concen-trate on reading a self-selected book and to ignore theauditory stimulation

The Aa and AV blocks were presented in an alternatingorder with half of the subjects starting with the AVcondition The Va blocks were presented one at thebeginning one at the middle and one at the end of the

experiment Before the experimental blocks subjectsparticipated in a practice session consisting of two Vablocks They were required to reach a 90 hit rate beforethe experiment could start Two out of 12 subjects failedto reach this level even after several additional blocksand were nally discarded

EEG Recording and Averaging

The electroencephalogram (EEG) (bandpass 0ndash100 Hz)was continuously digitized at a rate of 500 Hz by aSynAmps amplier (NeuroScan Inc) from 19 scalp elec-trodes positioned according to the 10ndash20 system Fp1Fp2 F7 F3 Fz F4 F8 T3 C3 Cz C4 T4 T5 P3 Pz P4 T6O1 and O2 One additional electrode was placed at theright mastoid (RM) and another at the left mastoid (LM)The horizontal EOG was recorded with electrodes at-tached to the left and right canthi The vertical elec-trooculogram (EOG) was assessed using recordings fromFp1 and Fp2 electrodes The common reference elec-trode was placed on the tip of the nose

ERPs were averaged off-line for each auditory stimulusclass for an epoch of 1300 msec including a preauditorystimulus period of 100 msec in the AV and Aa conditionsIn the Va condition an ERP was also obtained for anepoch of 1300 msec including a previsual stimulus pe-riod of 400 msec Epochs in which the EEG or EOGexceeded plusmn100 m V as well as the ve rst epochs of eachblock were automatically excluded from averaging Inthe Aa and AV conditions standard-tone trials immedi-ately following deviant-tone or novel-sound trials werealso excluded from the averages Individual ERPs wereband-pass ltered between 001 and 30 Hz

Data Analysis

In the AV and Va conditions a correct button presswithin 800 msec after visual-stimulus onset was regardedas a hit the mean RT being computed only for the hittrials An incorrect button press during this period wasclassied as an error and trials with no response withinthis time window were considered misses Hits errorsmisses and RTs were computed across odd and evennumbers

Auditory ERP amplitudes (N1 P2 MMN and P3a) werereferred to the mean amplitude in the 100-msec baselineperiod preceding the auditory stimulus in both the Aaand AV conditions For auditory ERP analysis each indi-vidual waveform in the Aa and AV conditions was linearlydetrended over an epoch of 600 msec (including base-line) to remove any possible asymmetrical contributionto EEG caused by slow EOG movements due to readingin the Aa condition The effects of the standard tones onthe visual ERPs were analyzed at O1 and O2 for peaksas follows (identied in the individual waveforms withinthe specied latency ranges) P1 75 to 140 msec N1 100

Escera et al 601

to 200 msec P2 180 to 280 msec N2 225 to 350 msecThe visual P3b was identied at Pz between 300 and 500msec from visual-stimulus onset For this analysis visualERP amplitudes were measured against the mean ampli-tude of the 100-msec baseline preceding visual-stimulusonset The fronto-polar negativity was analyzed at Fp1and Fp2 the visual frontal N2 at F3 Fz and F4 and thevisual P3b at P3 Pz and P4 The amplitude measurementsfor visual ERP components in the AV condition werereferred to the 100-msec preauditory stimulus baselinerather than to the 100-msec previsual stimulus baselinewhich was differentially affected by standard tones de-viant tones and novel sounds (see Figure 9)

The ERP and performance data were statistically ana-lyzed by ANOVA with repeated measures Where appro-priate nominal degrees of freedom and epsilon valuesafter the Greenhouse-Geisser correction are reportedScalp-distribution analyses were performed after normal-izing ERP amplitudes to prevent amplitude differencesbetween different components from washing out thegenuine scalp-distribution differences This normalizationwas done by dividing the amplitude at each electrodeby the sum of the squared amplitudes at all electrodes(McCarthy amp Wood 1985)

Acknowledgments

This research was supported by BIOMED-2 contract BMH4-CT96-0819-COBRAIN of the European Union the SpanishMinistry of Science and Education (DGICYT PB93-0802 UE96-0038) the Academy of Finland the scientic exchange programbetween the University of Barcelona and the University ofHelsinki and the National Science Fund of Hungary (OTKAT022681) The authors would like to thank Teemu Rinne andTeija Kujala for their help in making the gures

Reprint requests should be sent to Carles Escera Neurody-namics Laboratory Department of Psychiatry and Clinical Psy-chobiology University of Barcelona P Vall drsquoHebron 171 08035Barcelona Catalonia-Spain or via e-mail cescerapsiubes

Note

1 The MMN amplitude in the unattended channel can behowever attenuated under some conditions by strongly focus-ing attention on other sources of auditory stimulation(Naumlaumltaumlnen 1991 Naumlaumltaumlnen Paavilainen Tiitinen Jiang amp Alho1993 Oades amp Dittmannbalcar 1995 Trejo Ryan-Jones amp Kra-mer 1995 Woldorff Hackley amp Hillyard 1991)

REFERENCES

Alho K (1995) Cerebral generators of mismatch negativity(MMN) and its magnetic counterpart (MMNn) elicited bysound changes Ear and Hearing 16 38ndash51

Alho K Escera C Diacuteaz R Yago E amp Serra J M (1997)Effects of involuntary auditory attention on visual taskperformance and brain activity NeuroReport 8 3233ndash3237

Alho K Winkler I Escera C Huotilainen M Virtanen JJaumlaumlskelaumlinen I Pekkonen E amp Ilmoniemi R (1998) Proc-

essing of novel sounds and frequency changes in the hu-man auditory cortex Magnetoencephalographic record-ings Psychophysiology 35 211ndash224

Alho K Woods D L Algazi A amp Naumlaumltaumlnen R (1992) Inter-modal selective attention II Effects of attentional load onprocessing of auditory and visual stimuli in central spaceElectroencephalography and Clinical Neurophysiology82 356ndash368

Baudena P Halgren E Heit G amp Clarke J M (1995) Intra-cerebral potentials to rare target and distractor auditoryand visual stimuli III Frontal cortex Electroencephalogra-phy and Clinical Neurophysiology 94 251ndash264

Cherry E C (1953) Some experiments on the recognitionof speech with one and with two ears Journal of theAcoustical Society of America 29 975ndash979

Courchesne E (1978) Changes in P3 waves with event repe-tition Long-term effects on scalp distribution and ampli-tudes Electroencephalography and ClinicalNeurophysiology 45 468ndash482

Courchesne E Hillyard S A amp Galambos R (1975) Stimu-lus novelty task relevance and the visual evoked potentialsin man Electroencephalography and Clinical Neurophysi-ology 39 131ndash143

Cowan N Winkler I Teder W amp Naumlaumltaumlnen R (1993) Mem-ory prerequisites of the mismatch negativity in the audi-tory event-related potential (ERP) Journal ofExperimental Psychology Learning Memory and Cogni-tion 19 909ndash921

Cseacutepe V Karmos G amp Molnaacuter M (1987) Evoked potentialcorrelates of stimulus deviance during wakefulness andsleep in catmdashanimal model of mismatch negativity Elec-troencephalography and Clinical Neurophysiology 66571ndash578

Davis H Mast T Yoshie N amp Zerlin S (1966) The slow re-sponse of the human cortex to auditory stimulus Recov-ery process Electroencephalography and ClinicalNeurophysiology 21 105ndash113

Davis H amp Zerlin S (1966) Acoustic relations of the humanvertex potentials Journal of the Acoustical Society ofAmerica 39 109ndash116

Dawson M E (1990) Psychophysiology at the interface ofclinical science cognitive science and neuroscience Psy-chophysiology 27 243ndash255

Dawson M E Fillion D L amp Schell A M (1989) Is elicita-tion of the autonomic orienting response associated withallocation of processing resources Psychophysiology 26560ndash572

Duncan C C amp Kaye W H (1987) Effects of chlonidine onevent-related potential measures of information process-ing In R Johnson J W Rohrbaugh amp R Parasuraman(Eds) Current trends in event-related potentialresearch (EEG supplement 40) (pp 527ndash531) Amster-dam Elsevier

Fillion D J Dawson M E Schell A M amp Hazlett E A(1991) The relationship between skin conductance orient-ing and the allocation of processing resources Psycho-physiology 28 410ndash424

Folk C L Remington R W amp Johnston J C (1992) Involun-tary covert orienting is contingent on attentional controlsettings Journal of Experimental Psychology Human Per-ception and Performance 18 1030ndash1044

Ford J M Roth W T amp Kopell B S (1976) Auditory event-related potentials to unpredictable shifts in pitch Psycho-physiology 13 32ndash39

Friedman D amp Simpson G (1994) ERP amplitude and scalpdistribution to target and novel events Effects of temporalorder in young middle-aged and older adults CognitiveBrain Research 2 49ndash63

602 Journal of Cognitive Neuroscience Volume 10 Number 5

also signicantly larger for visual stimuli following novelsounds than after those following the standard tones

DISCUSSION

Engagement of Attention During VisualPerformance

The present results provide evidence that attention dur-ing visual performance can be engaged by shortly pre-ceding unexpected novel sounds or small changes in thetask-irrelevant acoustic environment These results alsorevealed that this phenomenon depends on the intrinsicnature of the eliciting acoustic events In addition the

observed effects on visual performance were associatedwith different patterns of neural events activated bydeviant tones and novel sounds as reected by thecomponent structure of the auditory ERPs elicited

As expected novel sounds prolonged the RT to suc-cessive visual stimuli This result is in agreement with theextensive literature reporting delayed RTs to target stim-uli caused by preceding irrelevant sound changes ornovels sounds (Alho et al 1992 Cherry 1953 Dawsonet al 1989 Fillion et al 1991 Grillon et al 1990Schroumlger 1996 Woods 1992 Woodward et al 1991) andit provides experimental support for the everyday expe-rience that attention can be engaged by irrelevant acous-tic events

Figure 6 The scalp distribu-tion of the P3a to novelsounds Left early P3a rightlate P3a top Auditory-alone(Aa) condition bottom Audi-tory-Visual (AV) condition Thecentral plot shows the novel-standard difference waves atFz in the Aa (thin) and AV(thick) conditions as well asthe 100-msec latency win-dows used to measure thetwo phases of the P3a (e =early P3a l = late P3a)

Escera et al 597

A surprising nding of this study was however thatthe slightly deviant tones degraded the visual perfor-mance accuracy by decreasing the rate of hit responseswhereas the widely deviant novel sounds did not It wasexpected that the amount of distraction caused by devi-ant tones and novel sounds would be proportional tothe magnitude of their deviance from the standard toneand indeed such an effect was found for the RT (17-msecincrease after novel sounds) conrming other recentndings (Jaumlaumlskelaumlinen et al 1996 Schroumlger 1996) How-ever this RT effect was not paralleled by a similar effecton the hit rate Indeed deviant tones were associatedwith a lower hit rate whereas the hit rate after novelsounds was similar to that after standard tones Similarresults were obtained in a related study by Jaumlaumlskelaumlinenet al (1996) and recently replicated for deviant tones(Alho Escera Diacuteaz Yago amp Serra 1997) These ndingssuggest that the distracting effects of an irrelevant acous-tic event on visual performance may depend on thenature of the eliciting sound That is obtrusive novelsounds appear to cause a stronger effect on the RT tosuccessive visual target stimuli than minor changes inthe acoustic stream but in contrast these minor changes

seem to cause a signicant decrement in the number ofhit responses to visual targets not observed after thenovel sounds

The differential effects of distracting novel and deviantsounds on visual performance may be explained by trac-ing these behavioral effects back to the associated neuralevents as reected by ERPs Novel sounds enhanced theN1 amplitude and elicited the MMN as suggested bysignicant differences between the standard-tone andnovel-sound ERP that extended over the N1 latencyrange The present MMN to novel sounds peaked andterminated (returned to baseline) earlier than the MMNto deviant tones in concordance with previous studiesshowing that with larger deviances MMN becomes ear-lier (Tiitinen May Reinikainen amp Naumlaumltaumlnen 1994) over-lapping the N1 peak for wider deviances (Tiitinen et al1994 Scherg et al 1989) The present MMN to novelsounds is also supported by recent magnetoencepha-lograph (MEG) data showing that novel sounds activatedan MMN source in the human auditory cortex (Alhoet al 1998) Thus for novel sounds the attention-switch-ing signal was probably triggered by a combined re-sponse of the transient-detector system reected in thesupratemporal N1 (Naumlaumltaumlnen 1990 1992 Naumlaumltaumlnen ampPicton 1987) or some other N1 component such as thefrontal N1 reported by Giard et al (1994) and the stimu-lus-change detector system reected in the MMN(Naumlaumltaumlnen 1990 1992) This combined signal for atten-tion capture started with the N1 generation and wasoverlapped by the MMN process resulting in an effectiveattentional reorientation as reected by the subsequentlarge P3a wave and the clearly delayed RT to the follow-ing visual targets

Figure 7 Grand-average ERPs to standard tones and novel soundsin the Auditory-alone (Aa left column) and Auditory-Visual (AV rightcolumn) conditions The novel-standard difference waves are shownin the middle column The late P3a was larger in the AV conditionthan in the Aa condition whereas the early P3a was of similar ampli-tude in the two conditions

Figure 8 Grand-average visual ERPs at Pz O1 and O2 in the Visualalone (Va) condition and in the Auditory-Visual (AV) condition afterstandard tones

598 Journal of Cognitive Neuroscience Volume 10 Number 5

Deviant tones in turn elicited a distinct MMN whichwas followed by a small P3a wave and by a secondnegative peak (Figure 5) Thus the neural mechanismresponsible for triggering the attention-switching signalwas based for deviant tones on the stimulus-changedetector system reected in the MMN (Naumlaumltaumlnen 19901992 Schroumlger 1996) This attention capture signal asso-ciated with the MMN was sporadically followed by an

orienting of attention as suggested by the small (aver-aged) P3a and also by the 23 hit rate decrease to visualtargets that followed deviant tones

Although the subjects were instructed to ignore theauditory stimuli in the AV condition the RT to visualtargets following standard tones was slightly (insig-nicantly) shorter than the RT in the Va condition withno auditory stimuli This suggests that the auditory stim-

Figure 9 Grand-averageERPs in the Auditory-Visual(AV) condition to standard-visual (thin line) deviant-visual (thick line) andnovel-visual (dashed line)stimulus pairs Only the mostrelevant electrodes are shownThe epoch starts 100 msec be-fore auditory-stimulus onsetwhich was used as the base-line period to avoid the largeamplitude variability due tothe auditory response in the100 msec preceding the vis-ual stimulus Novel sounds en-hanced the fronto-polarnegativity and the frontal N2to visual stimuli was en-hanced after novel soundsand deviant tones which alsoaffected the offset slope ofP3b Notice that whereas thetraces for standard-visual devi-ant-visual and novel-visualstimulus pairs perfectly matcheach other during the onsetslope of the P3b they start todiffer from the P3b peak forthe novel-visual stimulus pairand about 100 msec later forthe deviant-visual stimuluspair (see P3 Pz P4)

Escera et al 599

uli were used as warning signals for the following visualtargets and thus attention may have been bimodal in theAV condition However because the occurrence of an aud-itory stimulus change or a novel sound was task-irrelevantand unpredictable to the subjects attention was involun-tarily engaged by these events In a more controlledexperiment attention was more effectively withdrawnfrom the auditory stimuli and their occurrence was madenoninformative by the simultaneous presentation of avisual cue stimulus informing about the successive ap-pearance of a visual target (Alho et al 1997) Like in thepresent study a decreased hit rate for visual stimuli thatfollowed MMN- and P3a-eliciting deviant tones was foundThis conrms the involuntary nature of the attention-switching mechanism indicated by the MMN P3a anddeterioration of the visual-task performance

Component Structure of the P3a to Novel Sounds

In both the Aa and AV conditions novel sounds eliciteda large long-lasting and broadly distributed P3a wavewith a double peak at Fz (in the Aa condition) suggestinga contribution of two different subcomponents Theearly part of this P3a had a peak latency of 230 msecand showed a centrally dominant scalp distribution thatwas similar in the Aa and AV conditions (Figure 6) Theearly P3a inverted in polarity at posterior and lateralelectrodes (Figures 3b and 5) suggesting bilateral gener-ators located in the vicinity of the temporal and parietallobes (cf Scherg amp von Cramon 1986) The late part ofthe P3a in turn had a peak latency of 315 msec and didnot invert in polarity at posterior electrodes and its scalpdistribution was centered over the right frontal areas Incontrast to the early P3a the late P3a was enhanced inthe AV condition particularly over the right hemisphere(Figure 6) when subjects apparently used auditory stim-uli as warning signals to prepare for visual targets

The component structure of the present P3a and thescalp distribution of the two subcomponents is in agree-ment with previous results in patients with brain lesionssuggesting a critical role of the temporal-parietal junc-tion in the generation of the auditory (Halgren BaudenaClarke Heit Marinkovic et al 1995 Halgren BaudenaClarke Heit Lieacutegeois et al 1995 Knight Scabini Woodsamp Clayworth 1989) somatosensory (Yamaguchi ampKnight 1991 1992) and visual P3a (Knight 1991 1997)It is also compatible with results suggesting a frontalcontribution to P3a such as those obtained in patientswith frontal lesions (Knight 1984) in topographicalanalysis (Friedman amp Simpson 1994 Friedman Simpsonamp Hamberger 1993) in dipole-source modeling (Meck-linger amp Ullsperger 1995) and in intracerebral record-ings in humans (Baudena Halgren Heit amp Clarke 1995)

The late P3a amplitude in the AV condition was en-hanced with regard to its amplitude in the Aa conditionThis amplitude enhancement of the late P3a reects an

attentional monitoring of auditory stimuli in the AV con-dition these stimuli being used as warning signals for thefollowing visual targets The attentional modulation ofthe P3a amplitude is in agreement with that reported inprevious studies (Holdstock amp Rugg 1995 Woods 1992)and suggests that the late P3a may more closely reectan actual orienting of attention than the involuntary N1and MMN processes triggering this attentional orientingAlso several other ndings support the association of thelate phase of the P3a with the actual reorientation ofattention or orienting response First the right-frontaldominance in the late P3a scalp distribution observed inthe present experiment agrees with ndings suggestingthe involvement of right-frontal areas in the reorienta-tion of attention (Fuster 1989 Mesulam 1981) Secondit has been shown that the P3a amplitude is attenuatedwith the repetition of the eliciting novel event(Courchesne 1978 Friedman amp Simpson 1994 Knight1984) This habituating behavior is a characteristic fea-ture of the orienting response (Oumlhman 1979 1992 Sok-olov 1963)

The early P3a was insensitive to attentional manipula-tions as it was to similar amplitude and scalp distributionin the Aa and AV conditions Its peak latency was rathershort (230 msec) and its scalp distribution suggestsneural generators probably located in the temporal-parietal cortex (Knight 1991 1997 Knight et al 1989Yamaguchi amp Knight 1991) Thus the early P3a mightreect a neural process other than attentional reorienta-tion such as the violation of a polisensorial model of theexternal world maintained in the temporal-parietal asso-ciation cortex (Yamaguchi amp Knight 1991)

Conclusions

The results reported in the present paper support thenotion of two different brain mechanisms of involuntaryattention differentially activated by different acousticevents The ERP component structure to novel soundssuggests that unexpected novelty in the acoustic envi-ronment involuntary captures attention by simultane-ously activating two different neural mechanisms atransient-detector mechanism associated with the su-pratemporal N1 component of the auditory ERPs and achange-detector mechanism reected in the MMN Pre-sumably the attention-switching signal generated by thecombined activation of these two mechanisms resultedin an effective engagement of attention as indicated bythe large P3a component elicited by the novel soundsand the increased RT to subsequent visual targets Devi-ant tones in turn resulted in different behavioral effectsand ERP component structure which was characterizedby the MMN This supports the notion that small changesin the acoustic environment capture attention involun-tarily by activating the stimulus-change detector mecha-nism reected in the MMN

600 Journal of Cognitive Neuroscience Volume 10 Number 5

METHODS

Subjects Stimuli and Procedure

Ten right-handed paid students (21 to 40 years 3 males)with normal hearing and normal or corrected-to-normalvisual acuity participated in the experiment Two furthersubjects were discarded due to their inability to reachthe required 90 hit rate in a practice session

Subjects were presented with four blocks of 400stimulus pairs (AV condition) in a visual forced-choiceRT task Each pair consisted of an auditory stimulusfollowed after 300 msec (onset-to-onset) by a visualstimulus The interpair interval (onset-to-onset) was 12sec (Figure 1) The auditory stimuli were either standardor deviant tones and novel sounds presented in randomorder with probabilities of 08 01 and 01 respectivelyThe standard and deviant tones were sinusoidal tonebursts of 200-msec duration including 10-msec rise andfall times presented binaurally at 75 dB SPL throughheadphones with respective frequencies of 600 and 700Hz Sixty different environmental sounds such as thoseproduced by a drill hammer rain door telephone ring-ing etc were used as novel sounds They were digitallyrecorded then low-pass ltered at 10000 Hz and editedto have a duration of 200 msec including rise and falltimes of 10 msec and an intensity maximum of 70 to 80dB SPL Each different novel sound occurred only oncewithin a stimulus block and was presented twice orthree times during the whole experiment Both devianttones and novel sounds were always preceded by at leastone AV pair containing a standard tone The visual stimuliwere the digits 1 to 8 presented one at a time in randomorder on a computer screen for 200 msec They sub-tended a vertical angle of 17deg and a horizontal angle of11deg (30 mm acute 20 mm 100 cm from the subjectrsquos eyes)

Subjects were comfortably seated in a reclining chairin a dimly lit electrically and acoustically shielded roomThey were instructed to focus on the small xation crossappearing in the middle of the screen and to press oneresponse button with the right index nger for oddnumbers and another button with the right middlenger for even numbers Subjects were also instructedto ignore the auditory stimulation Both speed and accu-racy were emphasized for the primary visual task

In two further conditions visual stimuli were omitted(Va condition Figure 1) or auditory stimuli were omitted(Aa condition Figure 1) In the Va condition subjectsreceived three blocks of 200 visual stimulus each theinstruction being identical to that in the AV condition Inthe Aa condition subjects were instructed to concen-trate on reading a self-selected book and to ignore theauditory stimulation

The Aa and AV blocks were presented in an alternatingorder with half of the subjects starting with the AVcondition The Va blocks were presented one at thebeginning one at the middle and one at the end of the

experiment Before the experimental blocks subjectsparticipated in a practice session consisting of two Vablocks They were required to reach a 90 hit rate beforethe experiment could start Two out of 12 subjects failedto reach this level even after several additional blocksand were nally discarded

EEG Recording and Averaging

The electroencephalogram (EEG) (bandpass 0ndash100 Hz)was continuously digitized at a rate of 500 Hz by aSynAmps amplier (NeuroScan Inc) from 19 scalp elec-trodes positioned according to the 10ndash20 system Fp1Fp2 F7 F3 Fz F4 F8 T3 C3 Cz C4 T4 T5 P3 Pz P4 T6O1 and O2 One additional electrode was placed at theright mastoid (RM) and another at the left mastoid (LM)The horizontal EOG was recorded with electrodes at-tached to the left and right canthi The vertical elec-trooculogram (EOG) was assessed using recordings fromFp1 and Fp2 electrodes The common reference elec-trode was placed on the tip of the nose

ERPs were averaged off-line for each auditory stimulusclass for an epoch of 1300 msec including a preauditorystimulus period of 100 msec in the AV and Aa conditionsIn the Va condition an ERP was also obtained for anepoch of 1300 msec including a previsual stimulus pe-riod of 400 msec Epochs in which the EEG or EOGexceeded plusmn100 m V as well as the ve rst epochs of eachblock were automatically excluded from averaging Inthe Aa and AV conditions standard-tone trials immedi-ately following deviant-tone or novel-sound trials werealso excluded from the averages Individual ERPs wereband-pass ltered between 001 and 30 Hz

Data Analysis

In the AV and Va conditions a correct button presswithin 800 msec after visual-stimulus onset was regardedas a hit the mean RT being computed only for the hittrials An incorrect button press during this period wasclassied as an error and trials with no response withinthis time window were considered misses Hits errorsmisses and RTs were computed across odd and evennumbers

Auditory ERP amplitudes (N1 P2 MMN and P3a) werereferred to the mean amplitude in the 100-msec baselineperiod preceding the auditory stimulus in both the Aaand AV conditions For auditory ERP analysis each indi-vidual waveform in the Aa and AV conditions was linearlydetrended over an epoch of 600 msec (including base-line) to remove any possible asymmetrical contributionto EEG caused by slow EOG movements due to readingin the Aa condition The effects of the standard tones onthe visual ERPs were analyzed at O1 and O2 for peaksas follows (identied in the individual waveforms withinthe specied latency ranges) P1 75 to 140 msec N1 100

Escera et al 601

to 200 msec P2 180 to 280 msec N2 225 to 350 msecThe visual P3b was identied at Pz between 300 and 500msec from visual-stimulus onset For this analysis visualERP amplitudes were measured against the mean ampli-tude of the 100-msec baseline preceding visual-stimulusonset The fronto-polar negativity was analyzed at Fp1and Fp2 the visual frontal N2 at F3 Fz and F4 and thevisual P3b at P3 Pz and P4 The amplitude measurementsfor visual ERP components in the AV condition werereferred to the 100-msec preauditory stimulus baselinerather than to the 100-msec previsual stimulus baselinewhich was differentially affected by standard tones de-viant tones and novel sounds (see Figure 9)

The ERP and performance data were statistically ana-lyzed by ANOVA with repeated measures Where appro-priate nominal degrees of freedom and epsilon valuesafter the Greenhouse-Geisser correction are reportedScalp-distribution analyses were performed after normal-izing ERP amplitudes to prevent amplitude differencesbetween different components from washing out thegenuine scalp-distribution differences This normalizationwas done by dividing the amplitude at each electrodeby the sum of the squared amplitudes at all electrodes(McCarthy amp Wood 1985)

Acknowledgments

This research was supported by BIOMED-2 contract BMH4-CT96-0819-COBRAIN of the European Union the SpanishMinistry of Science and Education (DGICYT PB93-0802 UE96-0038) the Academy of Finland the scientic exchange programbetween the University of Barcelona and the University ofHelsinki and the National Science Fund of Hungary (OTKAT022681) The authors would like to thank Teemu Rinne andTeija Kujala for their help in making the gures

Reprint requests should be sent to Carles Escera Neurody-namics Laboratory Department of Psychiatry and Clinical Psy-chobiology University of Barcelona P Vall drsquoHebron 171 08035Barcelona Catalonia-Spain or via e-mail cescerapsiubes

Note

1 The MMN amplitude in the unattended channel can behowever attenuated under some conditions by strongly focus-ing attention on other sources of auditory stimulation(Naumlaumltaumlnen 1991 Naumlaumltaumlnen Paavilainen Tiitinen Jiang amp Alho1993 Oades amp Dittmannbalcar 1995 Trejo Ryan-Jones amp Kra-mer 1995 Woldorff Hackley amp Hillyard 1991)

REFERENCES

Alho K (1995) Cerebral generators of mismatch negativity(MMN) and its magnetic counterpart (MMNn) elicited bysound changes Ear and Hearing 16 38ndash51

Alho K Escera C Diacuteaz R Yago E amp Serra J M (1997)Effects of involuntary auditory attention on visual taskperformance and brain activity NeuroReport 8 3233ndash3237

Alho K Winkler I Escera C Huotilainen M Virtanen JJaumlaumlskelaumlinen I Pekkonen E amp Ilmoniemi R (1998) Proc-

essing of novel sounds and frequency changes in the hu-man auditory cortex Magnetoencephalographic record-ings Psychophysiology 35 211ndash224

Alho K Woods D L Algazi A amp Naumlaumltaumlnen R (1992) Inter-modal selective attention II Effects of attentional load onprocessing of auditory and visual stimuli in central spaceElectroencephalography and Clinical Neurophysiology82 356ndash368

Baudena P Halgren E Heit G amp Clarke J M (1995) Intra-cerebral potentials to rare target and distractor auditoryand visual stimuli III Frontal cortex Electroencephalogra-phy and Clinical Neurophysiology 94 251ndash264

Cherry E C (1953) Some experiments on the recognitionof speech with one and with two ears Journal of theAcoustical Society of America 29 975ndash979

Courchesne E (1978) Changes in P3 waves with event repe-tition Long-term effects on scalp distribution and ampli-tudes Electroencephalography and ClinicalNeurophysiology 45 468ndash482

Courchesne E Hillyard S A amp Galambos R (1975) Stimu-lus novelty task relevance and the visual evoked potentialsin man Electroencephalography and Clinical Neurophysi-ology 39 131ndash143

Cowan N Winkler I Teder W amp Naumlaumltaumlnen R (1993) Mem-ory prerequisites of the mismatch negativity in the audi-tory event-related potential (ERP) Journal ofExperimental Psychology Learning Memory and Cogni-tion 19 909ndash921

Cseacutepe V Karmos G amp Molnaacuter M (1987) Evoked potentialcorrelates of stimulus deviance during wakefulness andsleep in catmdashanimal model of mismatch negativity Elec-troencephalography and Clinical Neurophysiology 66571ndash578

Davis H Mast T Yoshie N amp Zerlin S (1966) The slow re-sponse of the human cortex to auditory stimulus Recov-ery process Electroencephalography and ClinicalNeurophysiology 21 105ndash113

Davis H amp Zerlin S (1966) Acoustic relations of the humanvertex potentials Journal of the Acoustical Society ofAmerica 39 109ndash116

Dawson M E (1990) Psychophysiology at the interface ofclinical science cognitive science and neuroscience Psy-chophysiology 27 243ndash255

Dawson M E Fillion D L amp Schell A M (1989) Is elicita-tion of the autonomic orienting response associated withallocation of processing resources Psychophysiology 26560ndash572

Duncan C C amp Kaye W H (1987) Effects of chlonidine onevent-related potential measures of information process-ing In R Johnson J W Rohrbaugh amp R Parasuraman(Eds) Current trends in event-related potentialresearch (EEG supplement 40) (pp 527ndash531) Amster-dam Elsevier

Fillion D J Dawson M E Schell A M amp Hazlett E A(1991) The relationship between skin conductance orient-ing and the allocation of processing resources Psycho-physiology 28 410ndash424

Folk C L Remington R W amp Johnston J C (1992) Involun-tary covert orienting is contingent on attentional controlsettings Journal of Experimental Psychology Human Per-ception and Performance 18 1030ndash1044

Ford J M Roth W T amp Kopell B S (1976) Auditory event-related potentials to unpredictable shifts in pitch Psycho-physiology 13 32ndash39

Friedman D amp Simpson G (1994) ERP amplitude and scalpdistribution to target and novel events Effects of temporalorder in young middle-aged and older adults CognitiveBrain Research 2 49ndash63

602 Journal of Cognitive Neuroscience Volume 10 Number 5

A surprising nding of this study was however thatthe slightly deviant tones degraded the visual perfor-mance accuracy by decreasing the rate of hit responseswhereas the widely deviant novel sounds did not It wasexpected that the amount of distraction caused by devi-ant tones and novel sounds would be proportional tothe magnitude of their deviance from the standard toneand indeed such an effect was found for the RT (17-msecincrease after novel sounds) conrming other recentndings (Jaumlaumlskelaumlinen et al 1996 Schroumlger 1996) How-ever this RT effect was not paralleled by a similar effecton the hit rate Indeed deviant tones were associatedwith a lower hit rate whereas the hit rate after novelsounds was similar to that after standard tones Similarresults were obtained in a related study by Jaumlaumlskelaumlinenet al (1996) and recently replicated for deviant tones(Alho Escera Diacuteaz Yago amp Serra 1997) These ndingssuggest that the distracting effects of an irrelevant acous-tic event on visual performance may depend on thenature of the eliciting sound That is obtrusive novelsounds appear to cause a stronger effect on the RT tosuccessive visual target stimuli than minor changes inthe acoustic stream but in contrast these minor changes

seem to cause a signicant decrement in the number ofhit responses to visual targets not observed after thenovel sounds

The differential effects of distracting novel and deviantsounds on visual performance may be explained by trac-ing these behavioral effects back to the associated neuralevents as reected by ERPs Novel sounds enhanced theN1 amplitude and elicited the MMN as suggested bysignicant differences between the standard-tone andnovel-sound ERP that extended over the N1 latencyrange The present MMN to novel sounds peaked andterminated (returned to baseline) earlier than the MMNto deviant tones in concordance with previous studiesshowing that with larger deviances MMN becomes ear-lier (Tiitinen May Reinikainen amp Naumlaumltaumlnen 1994) over-lapping the N1 peak for wider deviances (Tiitinen et al1994 Scherg et al 1989) The present MMN to novelsounds is also supported by recent magnetoencepha-lograph (MEG) data showing that novel sounds activatedan MMN source in the human auditory cortex (Alhoet al 1998) Thus for novel sounds the attention-switch-ing signal was probably triggered by a combined re-sponse of the transient-detector system reected in thesupratemporal N1 (Naumlaumltaumlnen 1990 1992 Naumlaumltaumlnen ampPicton 1987) or some other N1 component such as thefrontal N1 reported by Giard et al (1994) and the stimu-lus-change detector system reected in the MMN(Naumlaumltaumlnen 1990 1992) This combined signal for atten-tion capture started with the N1 generation and wasoverlapped by the MMN process resulting in an effectiveattentional reorientation as reected by the subsequentlarge P3a wave and the clearly delayed RT to the follow-ing visual targets

Figure 7 Grand-average ERPs to standard tones and novel soundsin the Auditory-alone (Aa left column) and Auditory-Visual (AV rightcolumn) conditions The novel-standard difference waves are shownin the middle column The late P3a was larger in the AV conditionthan in the Aa condition whereas the early P3a was of similar ampli-tude in the two conditions

Figure 8 Grand-average visual ERPs at Pz O1 and O2 in the Visualalone (Va) condition and in the Auditory-Visual (AV) condition afterstandard tones

598 Journal of Cognitive Neuroscience Volume 10 Number 5

Deviant tones in turn elicited a distinct MMN whichwas followed by a small P3a wave and by a secondnegative peak (Figure 5) Thus the neural mechanismresponsible for triggering the attention-switching signalwas based for deviant tones on the stimulus-changedetector system reected in the MMN (Naumlaumltaumlnen 19901992 Schroumlger 1996) This attention capture signal asso-ciated with the MMN was sporadically followed by an

orienting of attention as suggested by the small (aver-aged) P3a and also by the 23 hit rate decrease to visualtargets that followed deviant tones

Although the subjects were instructed to ignore theauditory stimuli in the AV condition the RT to visualtargets following standard tones was slightly (insig-nicantly) shorter than the RT in the Va condition withno auditory stimuli This suggests that the auditory stim-

Figure 9 Grand-averageERPs in the Auditory-Visual(AV) condition to standard-visual (thin line) deviant-visual (thick line) andnovel-visual (dashed line)stimulus pairs Only the mostrelevant electrodes are shownThe epoch starts 100 msec be-fore auditory-stimulus onsetwhich was used as the base-line period to avoid the largeamplitude variability due tothe auditory response in the100 msec preceding the vis-ual stimulus Novel sounds en-hanced the fronto-polarnegativity and the frontal N2to visual stimuli was en-hanced after novel soundsand deviant tones which alsoaffected the offset slope ofP3b Notice that whereas thetraces for standard-visual devi-ant-visual and novel-visualstimulus pairs perfectly matcheach other during the onsetslope of the P3b they start todiffer from the P3b peak forthe novel-visual stimulus pairand about 100 msec later forthe deviant-visual stimuluspair (see P3 Pz P4)

Escera et al 599

uli were used as warning signals for the following visualtargets and thus attention may have been bimodal in theAV condition However because the occurrence of an aud-itory stimulus change or a novel sound was task-irrelevantand unpredictable to the subjects attention was involun-tarily engaged by these events In a more controlledexperiment attention was more effectively withdrawnfrom the auditory stimuli and their occurrence was madenoninformative by the simultaneous presentation of avisual cue stimulus informing about the successive ap-pearance of a visual target (Alho et al 1997) Like in thepresent study a decreased hit rate for visual stimuli thatfollowed MMN- and P3a-eliciting deviant tones was foundThis conrms the involuntary nature of the attention-switching mechanism indicated by the MMN P3a anddeterioration of the visual-task performance

Component Structure of the P3a to Novel Sounds

In both the Aa and AV conditions novel sounds eliciteda large long-lasting and broadly distributed P3a wavewith a double peak at Fz (in the Aa condition) suggestinga contribution of two different subcomponents Theearly part of this P3a had a peak latency of 230 msecand showed a centrally dominant scalp distribution thatwas similar in the Aa and AV conditions (Figure 6) Theearly P3a inverted in polarity at posterior and lateralelectrodes (Figures 3b and 5) suggesting bilateral gener-ators located in the vicinity of the temporal and parietallobes (cf Scherg amp von Cramon 1986) The late part ofthe P3a in turn had a peak latency of 315 msec and didnot invert in polarity at posterior electrodes and its scalpdistribution was centered over the right frontal areas Incontrast to the early P3a the late P3a was enhanced inthe AV condition particularly over the right hemisphere(Figure 6) when subjects apparently used auditory stim-uli as warning signals to prepare for visual targets

The component structure of the present P3a and thescalp distribution of the two subcomponents is in agree-ment with previous results in patients with brain lesionssuggesting a critical role of the temporal-parietal junc-tion in the generation of the auditory (Halgren BaudenaClarke Heit Marinkovic et al 1995 Halgren BaudenaClarke Heit Lieacutegeois et al 1995 Knight Scabini Woodsamp Clayworth 1989) somatosensory (Yamaguchi ampKnight 1991 1992) and visual P3a (Knight 1991 1997)It is also compatible with results suggesting a frontalcontribution to P3a such as those obtained in patientswith frontal lesions (Knight 1984) in topographicalanalysis (Friedman amp Simpson 1994 Friedman Simpsonamp Hamberger 1993) in dipole-source modeling (Meck-linger amp Ullsperger 1995) and in intracerebral record-ings in humans (Baudena Halgren Heit amp Clarke 1995)

The late P3a amplitude in the AV condition was en-hanced with regard to its amplitude in the Aa conditionThis amplitude enhancement of the late P3a reects an

attentional monitoring of auditory stimuli in the AV con-dition these stimuli being used as warning signals for thefollowing visual targets The attentional modulation ofthe P3a amplitude is in agreement with that reported inprevious studies (Holdstock amp Rugg 1995 Woods 1992)and suggests that the late P3a may more closely reectan actual orienting of attention than the involuntary N1and MMN processes triggering this attentional orientingAlso several other ndings support the association of thelate phase of the P3a with the actual reorientation ofattention or orienting response First the right-frontaldominance in the late P3a scalp distribution observed inthe present experiment agrees with ndings suggestingthe involvement of right-frontal areas in the reorienta-tion of attention (Fuster 1989 Mesulam 1981) Secondit has been shown that the P3a amplitude is attenuatedwith the repetition of the eliciting novel event(Courchesne 1978 Friedman amp Simpson 1994 Knight1984) This habituating behavior is a characteristic fea-ture of the orienting response (Oumlhman 1979 1992 Sok-olov 1963)

The early P3a was insensitive to attentional manipula-tions as it was to similar amplitude and scalp distributionin the Aa and AV conditions Its peak latency was rathershort (230 msec) and its scalp distribution suggestsneural generators probably located in the temporal-parietal cortex (Knight 1991 1997 Knight et al 1989Yamaguchi amp Knight 1991) Thus the early P3a mightreect a neural process other than attentional reorienta-tion such as the violation of a polisensorial model of theexternal world maintained in the temporal-parietal asso-ciation cortex (Yamaguchi amp Knight 1991)

Conclusions

The results reported in the present paper support thenotion of two different brain mechanisms of involuntaryattention differentially activated by different acousticevents The ERP component structure to novel soundssuggests that unexpected novelty in the acoustic envi-ronment involuntary captures attention by simultane-ously activating two different neural mechanisms atransient-detector mechanism associated with the su-pratemporal N1 component of the auditory ERPs and achange-detector mechanism reected in the MMN Pre-sumably the attention-switching signal generated by thecombined activation of these two mechanisms resultedin an effective engagement of attention as indicated bythe large P3a component elicited by the novel soundsand the increased RT to subsequent visual targets Devi-ant tones in turn resulted in different behavioral effectsand ERP component structure which was characterizedby the MMN This supports the notion that small changesin the acoustic environment capture attention involun-tarily by activating the stimulus-change detector mecha-nism reected in the MMN

600 Journal of Cognitive Neuroscience Volume 10 Number 5

METHODS

Subjects Stimuli and Procedure

Ten right-handed paid students (21 to 40 years 3 males)with normal hearing and normal or corrected-to-normalvisual acuity participated in the experiment Two furthersubjects were discarded due to their inability to reachthe required 90 hit rate in a practice session

Subjects were presented with four blocks of 400stimulus pairs (AV condition) in a visual forced-choiceRT task Each pair consisted of an auditory stimulusfollowed after 300 msec (onset-to-onset) by a visualstimulus The interpair interval (onset-to-onset) was 12sec (Figure 1) The auditory stimuli were either standardor deviant tones and novel sounds presented in randomorder with probabilities of 08 01 and 01 respectivelyThe standard and deviant tones were sinusoidal tonebursts of 200-msec duration including 10-msec rise andfall times presented binaurally at 75 dB SPL throughheadphones with respective frequencies of 600 and 700Hz Sixty different environmental sounds such as thoseproduced by a drill hammer rain door telephone ring-ing etc were used as novel sounds They were digitallyrecorded then low-pass ltered at 10000 Hz and editedto have a duration of 200 msec including rise and falltimes of 10 msec and an intensity maximum of 70 to 80dB SPL Each different novel sound occurred only oncewithin a stimulus block and was presented twice orthree times during the whole experiment Both devianttones and novel sounds were always preceded by at leastone AV pair containing a standard tone The visual stimuliwere the digits 1 to 8 presented one at a time in randomorder on a computer screen for 200 msec They sub-tended a vertical angle of 17deg and a horizontal angle of11deg (30 mm acute 20 mm 100 cm from the subjectrsquos eyes)

Subjects were comfortably seated in a reclining chairin a dimly lit electrically and acoustically shielded roomThey were instructed to focus on the small xation crossappearing in the middle of the screen and to press oneresponse button with the right index nger for oddnumbers and another button with the right middlenger for even numbers Subjects were also instructedto ignore the auditory stimulation Both speed and accu-racy were emphasized for the primary visual task

In two further conditions visual stimuli were omitted(Va condition Figure 1) or auditory stimuli were omitted(Aa condition Figure 1) In the Va condition subjectsreceived three blocks of 200 visual stimulus each theinstruction being identical to that in the AV condition Inthe Aa condition subjects were instructed to concen-trate on reading a self-selected book and to ignore theauditory stimulation

The Aa and AV blocks were presented in an alternatingorder with half of the subjects starting with the AVcondition The Va blocks were presented one at thebeginning one at the middle and one at the end of the

experiment Before the experimental blocks subjectsparticipated in a practice session consisting of two Vablocks They were required to reach a 90 hit rate beforethe experiment could start Two out of 12 subjects failedto reach this level even after several additional blocksand were nally discarded

EEG Recording and Averaging

The electroencephalogram (EEG) (bandpass 0ndash100 Hz)was continuously digitized at a rate of 500 Hz by aSynAmps amplier (NeuroScan Inc) from 19 scalp elec-trodes positioned according to the 10ndash20 system Fp1Fp2 F7 F3 Fz F4 F8 T3 C3 Cz C4 T4 T5 P3 Pz P4 T6O1 and O2 One additional electrode was placed at theright mastoid (RM) and another at the left mastoid (LM)The horizontal EOG was recorded with electrodes at-tached to the left and right canthi The vertical elec-trooculogram (EOG) was assessed using recordings fromFp1 and Fp2 electrodes The common reference elec-trode was placed on the tip of the nose

ERPs were averaged off-line for each auditory stimulusclass for an epoch of 1300 msec including a preauditorystimulus period of 100 msec in the AV and Aa conditionsIn the Va condition an ERP was also obtained for anepoch of 1300 msec including a previsual stimulus pe-riod of 400 msec Epochs in which the EEG or EOGexceeded plusmn100 m V as well as the ve rst epochs of eachblock were automatically excluded from averaging Inthe Aa and AV conditions standard-tone trials immedi-ately following deviant-tone or novel-sound trials werealso excluded from the averages Individual ERPs wereband-pass ltered between 001 and 30 Hz

Data Analysis

In the AV and Va conditions a correct button presswithin 800 msec after visual-stimulus onset was regardedas a hit the mean RT being computed only for the hittrials An incorrect button press during this period wasclassied as an error and trials with no response withinthis time window were considered misses Hits errorsmisses and RTs were computed across odd and evennumbers

Auditory ERP amplitudes (N1 P2 MMN and P3a) werereferred to the mean amplitude in the 100-msec baselineperiod preceding the auditory stimulus in both the Aaand AV conditions For auditory ERP analysis each indi-vidual waveform in the Aa and AV conditions was linearlydetrended over an epoch of 600 msec (including base-line) to remove any possible asymmetrical contributionto EEG caused by slow EOG movements due to readingin the Aa condition The effects of the standard tones onthe visual ERPs were analyzed at O1 and O2 for peaksas follows (identied in the individual waveforms withinthe specied latency ranges) P1 75 to 140 msec N1 100

Escera et al 601

to 200 msec P2 180 to 280 msec N2 225 to 350 msecThe visual P3b was identied at Pz between 300 and 500msec from visual-stimulus onset For this analysis visualERP amplitudes were measured against the mean ampli-tude of the 100-msec baseline preceding visual-stimulusonset The fronto-polar negativity was analyzed at Fp1and Fp2 the visual frontal N2 at F3 Fz and F4 and thevisual P3b at P3 Pz and P4 The amplitude measurementsfor visual ERP components in the AV condition werereferred to the 100-msec preauditory stimulus baselinerather than to the 100-msec previsual stimulus baselinewhich was differentially affected by standard tones de-viant tones and novel sounds (see Figure 9)

The ERP and performance data were statistically ana-lyzed by ANOVA with repeated measures Where appro-priate nominal degrees of freedom and epsilon valuesafter the Greenhouse-Geisser correction are reportedScalp-distribution analyses were performed after normal-izing ERP amplitudes to prevent amplitude differencesbetween different components from washing out thegenuine scalp-distribution differences This normalizationwas done by dividing the amplitude at each electrodeby the sum of the squared amplitudes at all electrodes(McCarthy amp Wood 1985)

Acknowledgments

This research was supported by BIOMED-2 contract BMH4-CT96-0819-COBRAIN of the European Union the SpanishMinistry of Science and Education (DGICYT PB93-0802 UE96-0038) the Academy of Finland the scientic exchange programbetween the University of Barcelona and the University ofHelsinki and the National Science Fund of Hungary (OTKAT022681) The authors would like to thank Teemu Rinne andTeija Kujala for their help in making the gures

Reprint requests should be sent to Carles Escera Neurody-namics Laboratory Department of Psychiatry and Clinical Psy-chobiology University of Barcelona P Vall drsquoHebron 171 08035Barcelona Catalonia-Spain or via e-mail cescerapsiubes

Note

1 The MMN amplitude in the unattended channel can behowever attenuated under some conditions by strongly focus-ing attention on other sources of auditory stimulation(Naumlaumltaumlnen 1991 Naumlaumltaumlnen Paavilainen Tiitinen Jiang amp Alho1993 Oades amp Dittmannbalcar 1995 Trejo Ryan-Jones amp Kra-mer 1995 Woldorff Hackley amp Hillyard 1991)

REFERENCES

Alho K (1995) Cerebral generators of mismatch negativity(MMN) and its magnetic counterpart (MMNn) elicited bysound changes Ear and Hearing 16 38ndash51

Alho K Escera C Diacuteaz R Yago E amp Serra J M (1997)Effects of involuntary auditory attention on visual taskperformance and brain activity NeuroReport 8 3233ndash3237

Alho K Winkler I Escera C Huotilainen M Virtanen JJaumlaumlskelaumlinen I Pekkonen E amp Ilmoniemi R (1998) Proc-

essing of novel sounds and frequency changes in the hu-man auditory cortex Magnetoencephalographic record-ings Psychophysiology 35 211ndash224

Alho K Woods D L Algazi A amp Naumlaumltaumlnen R (1992) Inter-modal selective attention II Effects of attentional load onprocessing of auditory and visual stimuli in central spaceElectroencephalography and Clinical Neurophysiology82 356ndash368

Baudena P Halgren E Heit G amp Clarke J M (1995) Intra-cerebral potentials to rare target and distractor auditoryand visual stimuli III Frontal cortex Electroencephalogra-phy and Clinical Neurophysiology 94 251ndash264

Cherry E C (1953) Some experiments on the recognitionof speech with one and with two ears Journal of theAcoustical Society of America 29 975ndash979

Courchesne E (1978) Changes in P3 waves with event repe-tition Long-term effects on scalp distribution and ampli-tudes Electroencephalography and ClinicalNeurophysiology 45 468ndash482

Courchesne E Hillyard S A amp Galambos R (1975) Stimu-lus novelty task relevance and the visual evoked potentialsin man Electroencephalography and Clinical Neurophysi-ology 39 131ndash143

Cowan N Winkler I Teder W amp Naumlaumltaumlnen R (1993) Mem-ory prerequisites of the mismatch negativity in the audi-tory event-related potential (ERP) Journal ofExperimental Psychology Learning Memory and Cogni-tion 19 909ndash921

Cseacutepe V Karmos G amp Molnaacuter M (1987) Evoked potentialcorrelates of stimulus deviance during wakefulness andsleep in catmdashanimal model of mismatch negativity Elec-troencephalography and Clinical Neurophysiology 66571ndash578

Davis H Mast T Yoshie N amp Zerlin S (1966) The slow re-sponse of the human cortex to auditory stimulus Recov-ery process Electroencephalography and ClinicalNeurophysiology 21 105ndash113

Davis H amp Zerlin S (1966) Acoustic relations of the humanvertex potentials Journal of the Acoustical Society ofAmerica 39 109ndash116

Dawson M E (1990) Psychophysiology at the interface ofclinical science cognitive science and neuroscience Psy-chophysiology 27 243ndash255

Dawson M E Fillion D L amp Schell A M (1989) Is elicita-tion of the autonomic orienting response associated withallocation of processing resources Psychophysiology 26560ndash572

Duncan C C amp Kaye W H (1987) Effects of chlonidine onevent-related potential measures of information process-ing In R Johnson J W Rohrbaugh amp R Parasuraman(Eds) Current trends in event-related potentialresearch (EEG supplement 40) (pp 527ndash531) Amster-dam Elsevier

Fillion D J Dawson M E Schell A M amp Hazlett E A(1991) The relationship between skin conductance orient-ing and the allocation of processing resources Psycho-physiology 28 410ndash424

Folk C L Remington R W amp Johnston J C (1992) Involun-tary covert orienting is contingent on attentional controlsettings Journal of Experimental Psychology Human Per-ception and Performance 18 1030ndash1044

Ford J M Roth W T amp Kopell B S (1976) Auditory event-related potentials to unpredictable shifts in pitch Psycho-physiology 13 32ndash39

Friedman D amp Simpson G (1994) ERP amplitude and scalpdistribution to target and novel events Effects of temporalorder in young middle-aged and older adults CognitiveBrain Research 2 49ndash63

602 Journal of Cognitive Neuroscience Volume 10 Number 5

Deviant tones in turn elicited a distinct MMN whichwas followed by a small P3a wave and by a secondnegative peak (Figure 5) Thus the neural mechanismresponsible for triggering the attention-switching signalwas based for deviant tones on the stimulus-changedetector system reected in the MMN (Naumlaumltaumlnen 19901992 Schroumlger 1996) This attention capture signal asso-ciated with the MMN was sporadically followed by an

orienting of attention as suggested by the small (aver-aged) P3a and also by the 23 hit rate decrease to visualtargets that followed deviant tones

Although the subjects were instructed to ignore theauditory stimuli in the AV condition the RT to visualtargets following standard tones was slightly (insig-nicantly) shorter than the RT in the Va condition withno auditory stimuli This suggests that the auditory stim-

Figure 9 Grand-averageERPs in the Auditory-Visual(AV) condition to standard-visual (thin line) deviant-visual (thick line) andnovel-visual (dashed line)stimulus pairs Only the mostrelevant electrodes are shownThe epoch starts 100 msec be-fore auditory-stimulus onsetwhich was used as the base-line period to avoid the largeamplitude variability due tothe auditory response in the100 msec preceding the vis-ual stimulus Novel sounds en-hanced the fronto-polarnegativity and the frontal N2to visual stimuli was en-hanced after novel soundsand deviant tones which alsoaffected the offset slope ofP3b Notice that whereas thetraces for standard-visual devi-ant-visual and novel-visualstimulus pairs perfectly matcheach other during the onsetslope of the P3b they start todiffer from the P3b peak forthe novel-visual stimulus pairand about 100 msec later forthe deviant-visual stimuluspair (see P3 Pz P4)

Escera et al 599

uli were used as warning signals for the following visualtargets and thus attention may have been bimodal in theAV condition However because the occurrence of an aud-itory stimulus change or a novel sound was task-irrelevantand unpredictable to the subjects attention was involun-tarily engaged by these events In a more controlledexperiment attention was more effectively withdrawnfrom the auditory stimuli and their occurrence was madenoninformative by the simultaneous presentation of avisual cue stimulus informing about the successive ap-pearance of a visual target (Alho et al 1997) Like in thepresent study a decreased hit rate for visual stimuli thatfollowed MMN- and P3a-eliciting deviant tones was foundThis conrms the involuntary nature of the attention-switching mechanism indicated by the MMN P3a anddeterioration of the visual-task performance

Component Structure of the P3a to Novel Sounds

In both the Aa and AV conditions novel sounds eliciteda large long-lasting and broadly distributed P3a wavewith a double peak at Fz (in the Aa condition) suggestinga contribution of two different subcomponents Theearly part of this P3a had a peak latency of 230 msecand showed a centrally dominant scalp distribution thatwas similar in the Aa and AV conditions (Figure 6) Theearly P3a inverted in polarity at posterior and lateralelectrodes (Figures 3b and 5) suggesting bilateral gener-ators located in the vicinity of the temporal and parietallobes (cf Scherg amp von Cramon 1986) The late part ofthe P3a in turn had a peak latency of 315 msec and didnot invert in polarity at posterior electrodes and its scalpdistribution was centered over the right frontal areas Incontrast to the early P3a the late P3a was enhanced inthe AV condition particularly over the right hemisphere(Figure 6) when subjects apparently used auditory stim-uli as warning signals to prepare for visual targets

The component structure of the present P3a and thescalp distribution of the two subcomponents is in agree-ment with previous results in patients with brain lesionssuggesting a critical role of the temporal-parietal junc-tion in the generation of the auditory (Halgren BaudenaClarke Heit Marinkovic et al 1995 Halgren BaudenaClarke Heit Lieacutegeois et al 1995 Knight Scabini Woodsamp Clayworth 1989) somatosensory (Yamaguchi ampKnight 1991 1992) and visual P3a (Knight 1991 1997)It is also compatible with results suggesting a frontalcontribution to P3a such as those obtained in patientswith frontal lesions (Knight 1984) in topographicalanalysis (Friedman amp Simpson 1994 Friedman Simpsonamp Hamberger 1993) in dipole-source modeling (Meck-linger amp Ullsperger 1995) and in intracerebral record-ings in humans (Baudena Halgren Heit amp Clarke 1995)

The late P3a amplitude in the AV condition was en-hanced with regard to its amplitude in the Aa conditionThis amplitude enhancement of the late P3a reects an

attentional monitoring of auditory stimuli in the AV con-dition these stimuli being used as warning signals for thefollowing visual targets The attentional modulation ofthe P3a amplitude is in agreement with that reported inprevious studies (Holdstock amp Rugg 1995 Woods 1992)and suggests that the late P3a may more closely reectan actual orienting of attention than the involuntary N1and MMN processes triggering this attentional orientingAlso several other ndings support the association of thelate phase of the P3a with the actual reorientation ofattention or orienting response First the right-frontaldominance in the late P3a scalp distribution observed inthe present experiment agrees with ndings suggestingthe involvement of right-frontal areas in the reorienta-tion of attention (Fuster 1989 Mesulam 1981) Secondit has been shown that the P3a amplitude is attenuatedwith the repetition of the eliciting novel event(Courchesne 1978 Friedman amp Simpson 1994 Knight1984) This habituating behavior is a characteristic fea-ture of the orienting response (Oumlhman 1979 1992 Sok-olov 1963)

The early P3a was insensitive to attentional manipula-tions as it was to similar amplitude and scalp distributionin the Aa and AV conditions Its peak latency was rathershort (230 msec) and its scalp distribution suggestsneural generators probably located in the temporal-parietal cortex (Knight 1991 1997 Knight et al 1989Yamaguchi amp Knight 1991) Thus the early P3a mightreect a neural process other than attentional reorienta-tion such as the violation of a polisensorial model of theexternal world maintained in the temporal-parietal asso-ciation cortex (Yamaguchi amp Knight 1991)

Conclusions

The results reported in the present paper support thenotion of two different brain mechanisms of involuntaryattention differentially activated by different acousticevents The ERP component structure to novel soundssuggests that unexpected novelty in the acoustic envi-ronment involuntary captures attention by simultane-ously activating two different neural mechanisms atransient-detector mechanism associated with the su-pratemporal N1 component of the auditory ERPs and achange-detector mechanism reected in the MMN Pre-sumably the attention-switching signal generated by thecombined activation of these two mechanisms resultedin an effective engagement of attention as indicated bythe large P3a component elicited by the novel soundsand the increased RT to subsequent visual targets Devi-ant tones in turn resulted in different behavioral effectsand ERP component structure which was characterizedby the MMN This supports the notion that small changesin the acoustic environment capture attention involun-tarily by activating the stimulus-change detector mecha-nism reected in the MMN

600 Journal of Cognitive Neuroscience Volume 10 Number 5

METHODS

Subjects Stimuli and Procedure

Ten right-handed paid students (21 to 40 years 3 males)with normal hearing and normal or corrected-to-normalvisual acuity participated in the experiment Two furthersubjects were discarded due to their inability to reachthe required 90 hit rate in a practice session

Subjects were presented with four blocks of 400stimulus pairs (AV condition) in a visual forced-choiceRT task Each pair consisted of an auditory stimulusfollowed after 300 msec (onset-to-onset) by a visualstimulus The interpair interval (onset-to-onset) was 12sec (Figure 1) The auditory stimuli were either standardor deviant tones and novel sounds presented in randomorder with probabilities of 08 01 and 01 respectivelyThe standard and deviant tones were sinusoidal tonebursts of 200-msec duration including 10-msec rise andfall times presented binaurally at 75 dB SPL throughheadphones with respective frequencies of 600 and 700Hz Sixty different environmental sounds such as thoseproduced by a drill hammer rain door telephone ring-ing etc were used as novel sounds They were digitallyrecorded then low-pass ltered at 10000 Hz and editedto have a duration of 200 msec including rise and falltimes of 10 msec and an intensity maximum of 70 to 80dB SPL Each different novel sound occurred only oncewithin a stimulus block and was presented twice orthree times during the whole experiment Both devianttones and novel sounds were always preceded by at leastone AV pair containing a standard tone The visual stimuliwere the digits 1 to 8 presented one at a time in randomorder on a computer screen for 200 msec They sub-tended a vertical angle of 17deg and a horizontal angle of11deg (30 mm acute 20 mm 100 cm from the subjectrsquos eyes)

Subjects were comfortably seated in a reclining chairin a dimly lit electrically and acoustically shielded roomThey were instructed to focus on the small xation crossappearing in the middle of the screen and to press oneresponse button with the right index nger for oddnumbers and another button with the right middlenger for even numbers Subjects were also instructedto ignore the auditory stimulation Both speed and accu-racy were emphasized for the primary visual task

In two further conditions visual stimuli were omitted(Va condition Figure 1) or auditory stimuli were omitted(Aa condition Figure 1) In the Va condition subjectsreceived three blocks of 200 visual stimulus each theinstruction being identical to that in the AV condition Inthe Aa condition subjects were instructed to concen-trate on reading a self-selected book and to ignore theauditory stimulation

The Aa and AV blocks were presented in an alternatingorder with half of the subjects starting with the AVcondition The Va blocks were presented one at thebeginning one at the middle and one at the end of the

experiment Before the experimental blocks subjectsparticipated in a practice session consisting of two Vablocks They were required to reach a 90 hit rate beforethe experiment could start Two out of 12 subjects failedto reach this level even after several additional blocksand were nally discarded

EEG Recording and Averaging

The electroencephalogram (EEG) (bandpass 0ndash100 Hz)was continuously digitized at a rate of 500 Hz by aSynAmps amplier (NeuroScan Inc) from 19 scalp elec-trodes positioned according to the 10ndash20 system Fp1Fp2 F7 F3 Fz F4 F8 T3 C3 Cz C4 T4 T5 P3 Pz P4 T6O1 and O2 One additional electrode was placed at theright mastoid (RM) and another at the left mastoid (LM)The horizontal EOG was recorded with electrodes at-tached to the left and right canthi The vertical elec-trooculogram (EOG) was assessed using recordings fromFp1 and Fp2 electrodes The common reference elec-trode was placed on the tip of the nose

ERPs were averaged off-line for each auditory stimulusclass for an epoch of 1300 msec including a preauditorystimulus period of 100 msec in the AV and Aa conditionsIn the Va condition an ERP was also obtained for anepoch of 1300 msec including a previsual stimulus pe-riod of 400 msec Epochs in which the EEG or EOGexceeded plusmn100 m V as well as the ve rst epochs of eachblock were automatically excluded from averaging Inthe Aa and AV conditions standard-tone trials immedi-ately following deviant-tone or novel-sound trials werealso excluded from the averages Individual ERPs wereband-pass ltered between 001 and 30 Hz

Data Analysis

In the AV and Va conditions a correct button presswithin 800 msec after visual-stimulus onset was regardedas a hit the mean RT being computed only for the hittrials An incorrect button press during this period wasclassied as an error and trials with no response withinthis time window were considered misses Hits errorsmisses and RTs were computed across odd and evennumbers

Auditory ERP amplitudes (N1 P2 MMN and P3a) werereferred to the mean amplitude in the 100-msec baselineperiod preceding the auditory stimulus in both the Aaand AV conditions For auditory ERP analysis each indi-vidual waveform in the Aa and AV conditions was linearlydetrended over an epoch of 600 msec (including base-line) to remove any possible asymmetrical contributionto EEG caused by slow EOG movements due to readingin the Aa condition The effects of the standard tones onthe visual ERPs were analyzed at O1 and O2 for peaksas follows (identied in the individual waveforms withinthe specied latency ranges) P1 75 to 140 msec N1 100

Escera et al 601

to 200 msec P2 180 to 280 msec N2 225 to 350 msecThe visual P3b was identied at Pz between 300 and 500msec from visual-stimulus onset For this analysis visualERP amplitudes were measured against the mean ampli-tude of the 100-msec baseline preceding visual-stimulusonset The fronto-polar negativity was analyzed at Fp1and Fp2 the visual frontal N2 at F3 Fz and F4 and thevisual P3b at P3 Pz and P4 The amplitude measurementsfor visual ERP components in the AV condition werereferred to the 100-msec preauditory stimulus baselinerather than to the 100-msec previsual stimulus baselinewhich was differentially affected by standard tones de-viant tones and novel sounds (see Figure 9)

The ERP and performance data were statistically ana-lyzed by ANOVA with repeated measures Where appro-priate nominal degrees of freedom and epsilon valuesafter the Greenhouse-Geisser correction are reportedScalp-distribution analyses were performed after normal-izing ERP amplitudes to prevent amplitude differencesbetween different components from washing out thegenuine scalp-distribution differences This normalizationwas done by dividing the amplitude at each electrodeby the sum of the squared amplitudes at all electrodes(McCarthy amp Wood 1985)

Acknowledgments

This research was supported by BIOMED-2 contract BMH4-CT96-0819-COBRAIN of the European Union the SpanishMinistry of Science and Education (DGICYT PB93-0802 UE96-0038) the Academy of Finland the scientic exchange programbetween the University of Barcelona and the University ofHelsinki and the National Science Fund of Hungary (OTKAT022681) The authors would like to thank Teemu Rinne andTeija Kujala for their help in making the gures

Reprint requests should be sent to Carles Escera Neurody-namics Laboratory Department of Psychiatry and Clinical Psy-chobiology University of Barcelona P Vall drsquoHebron 171 08035Barcelona Catalonia-Spain or via e-mail cescerapsiubes

Note

1 The MMN amplitude in the unattended channel can behowever attenuated under some conditions by strongly focus-ing attention on other sources of auditory stimulation(Naumlaumltaumlnen 1991 Naumlaumltaumlnen Paavilainen Tiitinen Jiang amp Alho1993 Oades amp Dittmannbalcar 1995 Trejo Ryan-Jones amp Kra-mer 1995 Woldorff Hackley amp Hillyard 1991)

REFERENCES

Alho K (1995) Cerebral generators of mismatch negativity(MMN) and its magnetic counterpart (MMNn) elicited bysound changes Ear and Hearing 16 38ndash51

Alho K Escera C Diacuteaz R Yago E amp Serra J M (1997)Effects of involuntary auditory attention on visual taskperformance and brain activity NeuroReport 8 3233ndash3237

Alho K Winkler I Escera C Huotilainen M Virtanen JJaumlaumlskelaumlinen I Pekkonen E amp Ilmoniemi R (1998) Proc-

essing of novel sounds and frequency changes in the hu-man auditory cortex Magnetoencephalographic record-ings Psychophysiology 35 211ndash224

Alho K Woods D L Algazi A amp Naumlaumltaumlnen R (1992) Inter-modal selective attention II Effects of attentional load onprocessing of auditory and visual stimuli in central spaceElectroencephalography and Clinical Neurophysiology82 356ndash368

Baudena P Halgren E Heit G amp Clarke J M (1995) Intra-cerebral potentials to rare target and distractor auditoryand visual stimuli III Frontal cortex Electroencephalogra-phy and Clinical Neurophysiology 94 251ndash264

Cherry E C (1953) Some experiments on the recognitionof speech with one and with two ears Journal of theAcoustical Society of America 29 975ndash979

Courchesne E (1978) Changes in P3 waves with event repe-tition Long-term effects on scalp distribution and ampli-tudes Electroencephalography and ClinicalNeurophysiology 45 468ndash482

Courchesne E Hillyard S A amp Galambos R (1975) Stimu-lus novelty task relevance and the visual evoked potentialsin man Electroencephalography and Clinical Neurophysi-ology 39 131ndash143

Cowan N Winkler I Teder W amp Naumlaumltaumlnen R (1993) Mem-ory prerequisites of the mismatch negativity in the audi-tory event-related potential (ERP) Journal ofExperimental Psychology Learning Memory and Cogni-tion 19 909ndash921

Cseacutepe V Karmos G amp Molnaacuter M (1987) Evoked potentialcorrelates of stimulus deviance during wakefulness andsleep in catmdashanimal model of mismatch negativity Elec-troencephalography and Clinical Neurophysiology 66571ndash578

Davis H Mast T Yoshie N amp Zerlin S (1966) The slow re-sponse of the human cortex to auditory stimulus Recov-ery process Electroencephalography and ClinicalNeurophysiology 21 105ndash113

Davis H amp Zerlin S (1966) Acoustic relations of the humanvertex potentials Journal of the Acoustical Society ofAmerica 39 109ndash116

Dawson M E (1990) Psychophysiology at the interface ofclinical science cognitive science and neuroscience Psy-chophysiology 27 243ndash255

Dawson M E Fillion D L amp Schell A M (1989) Is elicita-tion of the autonomic orienting response associated withallocation of processing resources Psychophysiology 26560ndash572

Duncan C C amp Kaye W H (1987) Effects of chlonidine onevent-related potential measures of information process-ing In R Johnson J W Rohrbaugh amp R Parasuraman(Eds) Current trends in event-related potentialresearch (EEG supplement 40) (pp 527ndash531) Amster-dam Elsevier

Fillion D J Dawson M E Schell A M amp Hazlett E A(1991) The relationship between skin conductance orient-ing and the allocation of processing resources Psycho-physiology 28 410ndash424

Folk C L Remington R W amp Johnston J C (1992) Involun-tary covert orienting is contingent on attentional controlsettings Journal of Experimental Psychology Human Per-ception and Performance 18 1030ndash1044

Ford J M Roth W T amp Kopell B S (1976) Auditory event-related potentials to unpredictable shifts in pitch Psycho-physiology 13 32ndash39

Friedman D amp Simpson G (1994) ERP amplitude and scalpdistribution to target and novel events Effects of temporalorder in young middle-aged and older adults CognitiveBrain Research 2 49ndash63

602 Journal of Cognitive Neuroscience Volume 10 Number 5

uli were used as warning signals for the following visualtargets and thus attention may have been bimodal in theAV condition However because the occurrence of an aud-itory stimulus change or a novel sound was task-irrelevantand unpredictable to the subjects attention was involun-tarily engaged by these events In a more controlledexperiment attention was more effectively withdrawnfrom the auditory stimuli and their occurrence was madenoninformative by the simultaneous presentation of avisual cue stimulus informing about the successive ap-pearance of a visual target (Alho et al 1997) Like in thepresent study a decreased hit rate for visual stimuli thatfollowed MMN- and P3a-eliciting deviant tones was foundThis conrms the involuntary nature of the attention-switching mechanism indicated by the MMN P3a anddeterioration of the visual-task performance

Component Structure of the P3a to Novel Sounds

In both the Aa and AV conditions novel sounds eliciteda large long-lasting and broadly distributed P3a wavewith a double peak at Fz (in the Aa condition) suggestinga contribution of two different subcomponents Theearly part of this P3a had a peak latency of 230 msecand showed a centrally dominant scalp distribution thatwas similar in the Aa and AV conditions (Figure 6) Theearly P3a inverted in polarity at posterior and lateralelectrodes (Figures 3b and 5) suggesting bilateral gener-ators located in the vicinity of the temporal and parietallobes (cf Scherg amp von Cramon 1986) The late part ofthe P3a in turn had a peak latency of 315 msec and didnot invert in polarity at posterior electrodes and its scalpdistribution was centered over the right frontal areas Incontrast to the early P3a the late P3a was enhanced inthe AV condition particularly over the right hemisphere(Figure 6) when subjects apparently used auditory stim-uli as warning signals to prepare for visual targets

The component structure of the present P3a and thescalp distribution of the two subcomponents is in agree-ment with previous results in patients with brain lesionssuggesting a critical role of the temporal-parietal junc-tion in the generation of the auditory (Halgren BaudenaClarke Heit Marinkovic et al 1995 Halgren BaudenaClarke Heit Lieacutegeois et al 1995 Knight Scabini Woodsamp Clayworth 1989) somatosensory (Yamaguchi ampKnight 1991 1992) and visual P3a (Knight 1991 1997)It is also compatible with results suggesting a frontalcontribution to P3a such as those obtained in patientswith frontal lesions (Knight 1984) in topographicalanalysis (Friedman amp Simpson 1994 Friedman Simpsonamp Hamberger 1993) in dipole-source modeling (Meck-linger amp Ullsperger 1995) and in intracerebral record-ings in humans (Baudena Halgren Heit amp Clarke 1995)

The late P3a amplitude in the AV condition was en-hanced with regard to its amplitude in the Aa conditionThis amplitude enhancement of the late P3a reects an

attentional monitoring of auditory stimuli in the AV con-dition these stimuli being used as warning signals for thefollowing visual targets The attentional modulation ofthe P3a amplitude is in agreement with that reported inprevious studies (Holdstock amp Rugg 1995 Woods 1992)and suggests that the late P3a may more closely reectan actual orienting of attention than the involuntary N1and MMN processes triggering this attentional orientingAlso several other ndings support the association of thelate phase of the P3a with the actual reorientation ofattention or orienting response First the right-frontaldominance in the late P3a scalp distribution observed inthe present experiment agrees with ndings suggestingthe involvement of right-frontal areas in the reorienta-tion of attention (Fuster 1989 Mesulam 1981) Secondit has been shown that the P3a amplitude is attenuatedwith the repetition of the eliciting novel event(Courchesne 1978 Friedman amp Simpson 1994 Knight1984) This habituating behavior is a characteristic fea-ture of the orienting response (Oumlhman 1979 1992 Sok-olov 1963)

The early P3a was insensitive to attentional manipula-tions as it was to similar amplitude and scalp distributionin the Aa and AV conditions Its peak latency was rathershort (230 msec) and its scalp distribution suggestsneural generators probably located in the temporal-parietal cortex (Knight 1991 1997 Knight et al 1989Yamaguchi amp Knight 1991) Thus the early P3a mightreect a neural process other than attentional reorienta-tion such as the violation of a polisensorial model of theexternal world maintained in the temporal-parietal asso-ciation cortex (Yamaguchi amp Knight 1991)

Conclusions

The results reported in the present paper support thenotion of two different brain mechanisms of involuntaryattention differentially activated by different acousticevents The ERP component structure to novel soundssuggests that unexpected novelty in the acoustic envi-ronment involuntary captures attention by simultane-ously activating two different neural mechanisms atransient-detector mechanism associated with the su-pratemporal N1 component of the auditory ERPs and achange-detector mechanism reected in the MMN Pre-sumably the attention-switching signal generated by thecombined activation of these two mechanisms resultedin an effective engagement of attention as indicated bythe large P3a component elicited by the novel soundsand the increased RT to subsequent visual targets Devi-ant tones in turn resulted in different behavioral effectsand ERP component structure which was characterizedby the MMN This supports the notion that small changesin the acoustic environment capture attention involun-tarily by activating the stimulus-change detector mecha-nism reected in the MMN

600 Journal of Cognitive Neuroscience Volume 10 Number 5

METHODS

Subjects Stimuli and Procedure

Ten right-handed paid students (21 to 40 years 3 males)with normal hearing and normal or corrected-to-normalvisual acuity participated in the experiment Two furthersubjects were discarded due to their inability to reachthe required 90 hit rate in a practice session

Subjects were presented with four blocks of 400stimulus pairs (AV condition) in a visual forced-choiceRT task Each pair consisted of an auditory stimulusfollowed after 300 msec (onset-to-onset) by a visualstimulus The interpair interval (onset-to-onset) was 12sec (Figure 1) The auditory stimuli were either standardor deviant tones and novel sounds presented in randomorder with probabilities of 08 01 and 01 respectivelyThe standard and deviant tones were sinusoidal tonebursts of 200-msec duration including 10-msec rise andfall times presented binaurally at 75 dB SPL throughheadphones with respective frequencies of 600 and 700Hz Sixty different environmental sounds such as thoseproduced by a drill hammer rain door telephone ring-ing etc were used as novel sounds They were digitallyrecorded then low-pass ltered at 10000 Hz and editedto have a duration of 200 msec including rise and falltimes of 10 msec and an intensity maximum of 70 to 80dB SPL Each different novel sound occurred only oncewithin a stimulus block and was presented twice orthree times during the whole experiment Both devianttones and novel sounds were always preceded by at leastone AV pair containing a standard tone The visual stimuliwere the digits 1 to 8 presented one at a time in randomorder on a computer screen for 200 msec They sub-tended a vertical angle of 17deg and a horizontal angle of11deg (30 mm acute 20 mm 100 cm from the subjectrsquos eyes)

Subjects were comfortably seated in a reclining chairin a dimly lit electrically and acoustically shielded roomThey were instructed to focus on the small xation crossappearing in the middle of the screen and to press oneresponse button with the right index nger for oddnumbers and another button with the right middlenger for even numbers Subjects were also instructedto ignore the auditory stimulation Both speed and accu-racy were emphasized for the primary visual task

In two further conditions visual stimuli were omitted(Va condition Figure 1) or auditory stimuli were omitted(Aa condition Figure 1) In the Va condition subjectsreceived three blocks of 200 visual stimulus each theinstruction being identical to that in the AV condition Inthe Aa condition subjects were instructed to concen-trate on reading a self-selected book and to ignore theauditory stimulation

The Aa and AV blocks were presented in an alternatingorder with half of the subjects starting with the AVcondition The Va blocks were presented one at thebeginning one at the middle and one at the end of the

experiment Before the experimental blocks subjectsparticipated in a practice session consisting of two Vablocks They were required to reach a 90 hit rate beforethe experiment could start Two out of 12 subjects failedto reach this level even after several additional blocksand were nally discarded

EEG Recording and Averaging

The electroencephalogram (EEG) (bandpass 0ndash100 Hz)was continuously digitized at a rate of 500 Hz by aSynAmps amplier (NeuroScan Inc) from 19 scalp elec-trodes positioned according to the 10ndash20 system Fp1Fp2 F7 F3 Fz F4 F8 T3 C3 Cz C4 T4 T5 P3 Pz P4 T6O1 and O2 One additional electrode was placed at theright mastoid (RM) and another at the left mastoid (LM)The horizontal EOG was recorded with electrodes at-tached to the left and right canthi The vertical elec-trooculogram (EOG) was assessed using recordings fromFp1 and Fp2 electrodes The common reference elec-trode was placed on the tip of the nose

ERPs were averaged off-line for each auditory stimulusclass for an epoch of 1300 msec including a preauditorystimulus period of 100 msec in the AV and Aa conditionsIn the Va condition an ERP was also obtained for anepoch of 1300 msec including a previsual stimulus pe-riod of 400 msec Epochs in which the EEG or EOGexceeded plusmn100 m V as well as the ve rst epochs of eachblock were automatically excluded from averaging Inthe Aa and AV conditions standard-tone trials immedi-ately following deviant-tone or novel-sound trials werealso excluded from the averages Individual ERPs wereband-pass ltered between 001 and 30 Hz

Data Analysis

In the AV and Va conditions a correct button presswithin 800 msec after visual-stimulus onset was regardedas a hit the mean RT being computed only for the hittrials An incorrect button press during this period wasclassied as an error and trials with no response withinthis time window were considered misses Hits errorsmisses and RTs were computed across odd and evennumbers

Auditory ERP amplitudes (N1 P2 MMN and P3a) werereferred to the mean amplitude in the 100-msec baselineperiod preceding the auditory stimulus in both the Aaand AV conditions For auditory ERP analysis each indi-vidual waveform in the Aa and AV conditions was linearlydetrended over an epoch of 600 msec (including base-line) to remove any possible asymmetrical contributionto EEG caused by slow EOG movements due to readingin the Aa condition The effects of the standard tones onthe visual ERPs were analyzed at O1 and O2 for peaksas follows (identied in the individual waveforms withinthe specied latency ranges) P1 75 to 140 msec N1 100

Escera et al 601

to 200 msec P2 180 to 280 msec N2 225 to 350 msecThe visual P3b was identied at Pz between 300 and 500msec from visual-stimulus onset For this analysis visualERP amplitudes were measured against the mean ampli-tude of the 100-msec baseline preceding visual-stimulusonset The fronto-polar negativity was analyzed at Fp1and Fp2 the visual frontal N2 at F3 Fz and F4 and thevisual P3b at P3 Pz and P4 The amplitude measurementsfor visual ERP components in the AV condition werereferred to the 100-msec preauditory stimulus baselinerather than to the 100-msec previsual stimulus baselinewhich was differentially affected by standard tones de-viant tones and novel sounds (see Figure 9)

The ERP and performance data were statistically ana-lyzed by ANOVA with repeated measures Where appro-priate nominal degrees of freedom and epsilon valuesafter the Greenhouse-Geisser correction are reportedScalp-distribution analyses were performed after normal-izing ERP amplitudes to prevent amplitude differencesbetween different components from washing out thegenuine scalp-distribution differences This normalizationwas done by dividing the amplitude at each electrodeby the sum of the squared amplitudes at all electrodes(McCarthy amp Wood 1985)

Acknowledgments

This research was supported by BIOMED-2 contract BMH4-CT96-0819-COBRAIN of the European Union the SpanishMinistry of Science and Education (DGICYT PB93-0802 UE96-0038) the Academy of Finland the scientic exchange programbetween the University of Barcelona and the University ofHelsinki and the National Science Fund of Hungary (OTKAT022681) The authors would like to thank Teemu Rinne andTeija Kujala for their help in making the gures

Reprint requests should be sent to Carles Escera Neurody-namics Laboratory Department of Psychiatry and Clinical Psy-chobiology University of Barcelona P Vall drsquoHebron 171 08035Barcelona Catalonia-Spain or via e-mail cescerapsiubes

Note

1 The MMN amplitude in the unattended channel can behowever attenuated under some conditions by strongly focus-ing attention on other sources of auditory stimulation(Naumlaumltaumlnen 1991 Naumlaumltaumlnen Paavilainen Tiitinen Jiang amp Alho1993 Oades amp Dittmannbalcar 1995 Trejo Ryan-Jones amp Kra-mer 1995 Woldorff Hackley amp Hillyard 1991)

REFERENCES

Alho K (1995) Cerebral generators of mismatch negativity(MMN) and its magnetic counterpart (MMNn) elicited bysound changes Ear and Hearing 16 38ndash51

Alho K Escera C Diacuteaz R Yago E amp Serra J M (1997)Effects of involuntary auditory attention on visual taskperformance and brain activity NeuroReport 8 3233ndash3237

Alho K Winkler I Escera C Huotilainen M Virtanen JJaumlaumlskelaumlinen I Pekkonen E amp Ilmoniemi R (1998) Proc-

essing of novel sounds and frequency changes in the hu-man auditory cortex Magnetoencephalographic record-ings Psychophysiology 35 211ndash224

Alho K Woods D L Algazi A amp Naumlaumltaumlnen R (1992) Inter-modal selective attention II Effects of attentional load onprocessing of auditory and visual stimuli in central spaceElectroencephalography and Clinical Neurophysiology82 356ndash368

Baudena P Halgren E Heit G amp Clarke J M (1995) Intra-cerebral potentials to rare target and distractor auditoryand visual stimuli III Frontal cortex Electroencephalogra-phy and Clinical Neurophysiology 94 251ndash264

Cherry E C (1953) Some experiments on the recognitionof speech with one and with two ears Journal of theAcoustical Society of America 29 975ndash979

Courchesne E (1978) Changes in P3 waves with event repe-tition Long-term effects on scalp distribution and ampli-tudes Electroencephalography and ClinicalNeurophysiology 45 468ndash482

Courchesne E Hillyard S A amp Galambos R (1975) Stimu-lus novelty task relevance and the visual evoked potentialsin man Electroencephalography and Clinical Neurophysi-ology 39 131ndash143

Cowan N Winkler I Teder W amp Naumlaumltaumlnen R (1993) Mem-ory prerequisites of the mismatch negativity in the audi-tory event-related potential (ERP) Journal ofExperimental Psychology Learning Memory and Cogni-tion 19 909ndash921

Cseacutepe V Karmos G amp Molnaacuter M (1987) Evoked potentialcorrelates of stimulus deviance during wakefulness andsleep in catmdashanimal model of mismatch negativity Elec-troencephalography and Clinical Neurophysiology 66571ndash578

Davis H Mast T Yoshie N amp Zerlin S (1966) The slow re-sponse of the human cortex to auditory stimulus Recov-ery process Electroencephalography and ClinicalNeurophysiology 21 105ndash113

Davis H amp Zerlin S (1966) Acoustic relations of the humanvertex potentials Journal of the Acoustical Society ofAmerica 39 109ndash116

Dawson M E (1990) Psychophysiology at the interface ofclinical science cognitive science and neuroscience Psy-chophysiology 27 243ndash255

Dawson M E Fillion D L amp Schell A M (1989) Is elicita-tion of the autonomic orienting response associated withallocation of processing resources Psychophysiology 26560ndash572

Duncan C C amp Kaye W H (1987) Effects of chlonidine onevent-related potential measures of information process-ing In R Johnson J W Rohrbaugh amp R Parasuraman(Eds) Current trends in event-related potentialresearch (EEG supplement 40) (pp 527ndash531) Amster-dam Elsevier

Fillion D J Dawson M E Schell A M amp Hazlett E A(1991) The relationship between skin conductance orient-ing and the allocation of processing resources Psycho-physiology 28 410ndash424

Folk C L Remington R W amp Johnston J C (1992) Involun-tary covert orienting is contingent on attentional controlsettings Journal of Experimental Psychology Human Per-ception and Performance 18 1030ndash1044

Ford J M Roth W T amp Kopell B S (1976) Auditory event-related potentials to unpredictable shifts in pitch Psycho-physiology 13 32ndash39

Friedman D amp Simpson G (1994) ERP amplitude and scalpdistribution to target and novel events Effects of temporalorder in young middle-aged and older adults CognitiveBrain Research 2 49ndash63

602 Journal of Cognitive Neuroscience Volume 10 Number 5

METHODS

Subjects Stimuli and Procedure

Ten right-handed paid students (21 to 40 years 3 males)with normal hearing and normal or corrected-to-normalvisual acuity participated in the experiment Two furthersubjects were discarded due to their inability to reachthe required 90 hit rate in a practice session

Subjects were presented with four blocks of 400stimulus pairs (AV condition) in a visual forced-choiceRT task Each pair consisted of an auditory stimulusfollowed after 300 msec (onset-to-onset) by a visualstimulus The interpair interval (onset-to-onset) was 12sec (Figure 1) The auditory stimuli were either standardor deviant tones and novel sounds presented in randomorder with probabilities of 08 01 and 01 respectivelyThe standard and deviant tones were sinusoidal tonebursts of 200-msec duration including 10-msec rise andfall times presented binaurally at 75 dB SPL throughheadphones with respective frequencies of 600 and 700Hz Sixty different environmental sounds such as thoseproduced by a drill hammer rain door telephone ring-ing etc were used as novel sounds They were digitallyrecorded then low-pass ltered at 10000 Hz and editedto have a duration of 200 msec including rise and falltimes of 10 msec and an intensity maximum of 70 to 80dB SPL Each different novel sound occurred only oncewithin a stimulus block and was presented twice orthree times during the whole experiment Both devianttones and novel sounds were always preceded by at leastone AV pair containing a standard tone The visual stimuliwere the digits 1 to 8 presented one at a time in randomorder on a computer screen for 200 msec They sub-tended a vertical angle of 17deg and a horizontal angle of11deg (30 mm acute 20 mm 100 cm from the subjectrsquos eyes)

Subjects were comfortably seated in a reclining chairin a dimly lit electrically and acoustically shielded roomThey were instructed to focus on the small xation crossappearing in the middle of the screen and to press oneresponse button with the right index nger for oddnumbers and another button with the right middlenger for even numbers Subjects were also instructedto ignore the auditory stimulation Both speed and accu-racy were emphasized for the primary visual task

In two further conditions visual stimuli were omitted(Va condition Figure 1) or auditory stimuli were omitted(Aa condition Figure 1) In the Va condition subjectsreceived three blocks of 200 visual stimulus each theinstruction being identical to that in the AV condition Inthe Aa condition subjects were instructed to concen-trate on reading a self-selected book and to ignore theauditory stimulation

The Aa and AV blocks were presented in an alternatingorder with half of the subjects starting with the AVcondition The Va blocks were presented one at thebeginning one at the middle and one at the end of the

experiment Before the experimental blocks subjectsparticipated in a practice session consisting of two Vablocks They were required to reach a 90 hit rate beforethe experiment could start Two out of 12 subjects failedto reach this level even after several additional blocksand were nally discarded

EEG Recording and Averaging

The electroencephalogram (EEG) (bandpass 0ndash100 Hz)was continuously digitized at a rate of 500 Hz by aSynAmps amplier (NeuroScan Inc) from 19 scalp elec-trodes positioned according to the 10ndash20 system Fp1Fp2 F7 F3 Fz F4 F8 T3 C3 Cz C4 T4 T5 P3 Pz P4 T6O1 and O2 One additional electrode was placed at theright mastoid (RM) and another at the left mastoid (LM)The horizontal EOG was recorded with electrodes at-tached to the left and right canthi The vertical elec-trooculogram (EOG) was assessed using recordings fromFp1 and Fp2 electrodes The common reference elec-trode was placed on the tip of the nose

ERPs were averaged off-line for each auditory stimulusclass for an epoch of 1300 msec including a preauditorystimulus period of 100 msec in the AV and Aa conditionsIn the Va condition an ERP was also obtained for anepoch of 1300 msec including a previsual stimulus pe-riod of 400 msec Epochs in which the EEG or EOGexceeded plusmn100 m V as well as the ve rst epochs of eachblock were automatically excluded from averaging Inthe Aa and AV conditions standard-tone trials immedi-ately following deviant-tone or novel-sound trials werealso excluded from the averages Individual ERPs wereband-pass ltered between 001 and 30 Hz

Data Analysis

In the AV and Va conditions a correct button presswithin 800 msec after visual-stimulus onset was regardedas a hit the mean RT being computed only for the hittrials An incorrect button press during this period wasclassied as an error and trials with no response withinthis time window were considered misses Hits errorsmisses and RTs were computed across odd and evennumbers

Auditory ERP amplitudes (N1 P2 MMN and P3a) werereferred to the mean amplitude in the 100-msec baselineperiod preceding the auditory stimulus in both the Aaand AV conditions For auditory ERP analysis each indi-vidual waveform in the Aa and AV conditions was linearlydetrended over an epoch of 600 msec (including base-line) to remove any possible asymmetrical contributionto EEG caused by slow EOG movements due to readingin the Aa condition The effects of the standard tones onthe visual ERPs were analyzed at O1 and O2 for peaksas follows (identied in the individual waveforms withinthe specied latency ranges) P1 75 to 140 msec N1 100

Escera et al 601

to 200 msec P2 180 to 280 msec N2 225 to 350 msecThe visual P3b was identied at Pz between 300 and 500msec from visual-stimulus onset For this analysis visualERP amplitudes were measured against the mean ampli-tude of the 100-msec baseline preceding visual-stimulusonset The fronto-polar negativity was analyzed at Fp1and Fp2 the visual frontal N2 at F3 Fz and F4 and thevisual P3b at P3 Pz and P4 The amplitude measurementsfor visual ERP components in the AV condition werereferred to the 100-msec preauditory stimulus baselinerather than to the 100-msec previsual stimulus baselinewhich was differentially affected by standard tones de-viant tones and novel sounds (see Figure 9)

The ERP and performance data were statistically ana-lyzed by ANOVA with repeated measures Where appro-priate nominal degrees of freedom and epsilon valuesafter the Greenhouse-Geisser correction are reportedScalp-distribution analyses were performed after normal-izing ERP amplitudes to prevent amplitude differencesbetween different components from washing out thegenuine scalp-distribution differences This normalizationwas done by dividing the amplitude at each electrodeby the sum of the squared amplitudes at all electrodes(McCarthy amp Wood 1985)

Acknowledgments

This research was supported by BIOMED-2 contract BMH4-CT96-0819-COBRAIN of the European Union the SpanishMinistry of Science and Education (DGICYT PB93-0802 UE96-0038) the Academy of Finland the scientic exchange programbetween the University of Barcelona and the University ofHelsinki and the National Science Fund of Hungary (OTKAT022681) The authors would like to thank Teemu Rinne andTeija Kujala for their help in making the gures

Reprint requests should be sent to Carles Escera Neurody-namics Laboratory Department of Psychiatry and Clinical Psy-chobiology University of Barcelona P Vall drsquoHebron 171 08035Barcelona Catalonia-Spain or via e-mail cescerapsiubes

Note

1 The MMN amplitude in the unattended channel can behowever attenuated under some conditions by strongly focus-ing attention on other sources of auditory stimulation(Naumlaumltaumlnen 1991 Naumlaumltaumlnen Paavilainen Tiitinen Jiang amp Alho1993 Oades amp Dittmannbalcar 1995 Trejo Ryan-Jones amp Kra-mer 1995 Woldorff Hackley amp Hillyard 1991)

REFERENCES

Alho K (1995) Cerebral generators of mismatch negativity(MMN) and its magnetic counterpart (MMNn) elicited bysound changes Ear and Hearing 16 38ndash51

Alho K Escera C Diacuteaz R Yago E amp Serra J M (1997)Effects of involuntary auditory attention on visual taskperformance and brain activity NeuroReport 8 3233ndash3237

Alho K Winkler I Escera C Huotilainen M Virtanen JJaumlaumlskelaumlinen I Pekkonen E amp Ilmoniemi R (1998) Proc-

essing of novel sounds and frequency changes in the hu-man auditory cortex Magnetoencephalographic record-ings Psychophysiology 35 211ndash224

Alho K Woods D L Algazi A amp Naumlaumltaumlnen R (1992) Inter-modal selective attention II Effects of attentional load onprocessing of auditory and visual stimuli in central spaceElectroencephalography and Clinical Neurophysiology82 356ndash368

Baudena P Halgren E Heit G amp Clarke J M (1995) Intra-cerebral potentials to rare target and distractor auditoryand visual stimuli III Frontal cortex Electroencephalogra-phy and Clinical Neurophysiology 94 251ndash264

Cherry E C (1953) Some experiments on the recognitionof speech with one and with two ears Journal of theAcoustical Society of America 29 975ndash979

Courchesne E (1978) Changes in P3 waves with event repe-tition Long-term effects on scalp distribution and ampli-tudes Electroencephalography and ClinicalNeurophysiology 45 468ndash482

Courchesne E Hillyard S A amp Galambos R (1975) Stimu-lus novelty task relevance and the visual evoked potentialsin man Electroencephalography and Clinical Neurophysi-ology 39 131ndash143

Cowan N Winkler I Teder W amp Naumlaumltaumlnen R (1993) Mem-ory prerequisites of the mismatch negativity in the audi-tory event-related potential (ERP) Journal ofExperimental Psychology Learning Memory and Cogni-tion 19 909ndash921

Cseacutepe V Karmos G amp Molnaacuter M (1987) Evoked potentialcorrelates of stimulus deviance during wakefulness andsleep in catmdashanimal model of mismatch negativity Elec-troencephalography and Clinical Neurophysiology 66571ndash578

Davis H Mast T Yoshie N amp Zerlin S (1966) The slow re-sponse of the human cortex to auditory stimulus Recov-ery process Electroencephalography and ClinicalNeurophysiology 21 105ndash113

Davis H amp Zerlin S (1966) Acoustic relations of the humanvertex potentials Journal of the Acoustical Society ofAmerica 39 109ndash116

Dawson M E (1990) Psychophysiology at the interface ofclinical science cognitive science and neuroscience Psy-chophysiology 27 243ndash255

Dawson M E Fillion D L amp Schell A M (1989) Is elicita-tion of the autonomic orienting response associated withallocation of processing resources Psychophysiology 26560ndash572

Duncan C C amp Kaye W H (1987) Effects of chlonidine onevent-related potential measures of information process-ing In R Johnson J W Rohrbaugh amp R Parasuraman(Eds) Current trends in event-related potentialresearch (EEG supplement 40) (pp 527ndash531) Amster-dam Elsevier

Fillion D J Dawson M E Schell A M amp Hazlett E A(1991) The relationship between skin conductance orient-ing and the allocation of processing resources Psycho-physiology 28 410ndash424

Folk C L Remington R W amp Johnston J C (1992) Involun-tary covert orienting is contingent on attentional controlsettings Journal of Experimental Psychology Human Per-ception and Performance 18 1030ndash1044

Ford J M Roth W T amp Kopell B S (1976) Auditory event-related potentials to unpredictable shifts in pitch Psycho-physiology 13 32ndash39

Friedman D amp Simpson G (1994) ERP amplitude and scalpdistribution to target and novel events Effects of temporalorder in young middle-aged and older adults CognitiveBrain Research 2 49ndash63

602 Journal of Cognitive Neuroscience Volume 10 Number 5

to 200 msec P2 180 to 280 msec N2 225 to 350 msecThe visual P3b was identied at Pz between 300 and 500msec from visual-stimulus onset For this analysis visualERP amplitudes were measured against the mean ampli-tude of the 100-msec baseline preceding visual-stimulusonset The fronto-polar negativity was analyzed at Fp1and Fp2 the visual frontal N2 at F3 Fz and F4 and thevisual P3b at P3 Pz and P4 The amplitude measurementsfor visual ERP components in the AV condition werereferred to the 100-msec preauditory stimulus baselinerather than to the 100-msec previsual stimulus baselinewhich was differentially affected by standard tones de-viant tones and novel sounds (see Figure 9)

The ERP and performance data were statistically ana-lyzed by ANOVA with repeated measures Where appro-priate nominal degrees of freedom and epsilon valuesafter the Greenhouse-Geisser correction are reportedScalp-distribution analyses were performed after normal-izing ERP amplitudes to prevent amplitude differencesbetween different components from washing out thegenuine scalp-distribution differences This normalizationwas done by dividing the amplitude at each electrodeby the sum of the squared amplitudes at all electrodes(McCarthy amp Wood 1985)

Acknowledgments

This research was supported by BIOMED-2 contract BMH4-CT96-0819-COBRAIN of the European Union the SpanishMinistry of Science and Education (DGICYT PB93-0802 UE96-0038) the Academy of Finland the scientic exchange programbetween the University of Barcelona and the University ofHelsinki and the National Science Fund of Hungary (OTKAT022681) The authors would like to thank Teemu Rinne andTeija Kujala for their help in making the gures

Reprint requests should be sent to Carles Escera Neurody-namics Laboratory Department of Psychiatry and Clinical Psy-chobiology University of Barcelona P Vall drsquoHebron 171 08035Barcelona Catalonia-Spain or via e-mail cescerapsiubes

Note

1 The MMN amplitude in the unattended channel can behowever attenuated under some conditions by strongly focus-ing attention on other sources of auditory stimulation(Naumlaumltaumlnen 1991 Naumlaumltaumlnen Paavilainen Tiitinen Jiang amp Alho1993 Oades amp Dittmannbalcar 1995 Trejo Ryan-Jones amp Kra-mer 1995 Woldorff Hackley amp Hillyard 1991)

REFERENCES

Alho K (1995) Cerebral generators of mismatch negativity(MMN) and its magnetic counterpart (MMNn) elicited bysound changes Ear and Hearing 16 38ndash51

Alho K Escera C Diacuteaz R Yago E amp Serra J M (1997)Effects of involuntary auditory attention on visual taskperformance and brain activity NeuroReport 8 3233ndash3237

Alho K Winkler I Escera C Huotilainen M Virtanen JJaumlaumlskelaumlinen I Pekkonen E amp Ilmoniemi R (1998) Proc-

essing of novel sounds and frequency changes in the hu-man auditory cortex Magnetoencephalographic record-ings Psychophysiology 35 211ndash224

Alho K Woods D L Algazi A amp Naumlaumltaumlnen R (1992) Inter-modal selective attention II Effects of attentional load onprocessing of auditory and visual stimuli in central spaceElectroencephalography and Clinical Neurophysiology82 356ndash368

Baudena P Halgren E Heit G amp Clarke J M (1995) Intra-cerebral potentials to rare target and distractor auditoryand visual stimuli III Frontal cortex Electroencephalogra-phy and Clinical Neurophysiology 94 251ndash264

Cherry E C (1953) Some experiments on the recognitionof speech with one and with two ears Journal of theAcoustical Society of America 29 975ndash979

Courchesne E (1978) Changes in P3 waves with event repe-tition Long-term effects on scalp distribution and ampli-tudes Electroencephalography and ClinicalNeurophysiology 45 468ndash482

Courchesne E Hillyard S A amp Galambos R (1975) Stimu-lus novelty task relevance and the visual evoked potentialsin man Electroencephalography and Clinical Neurophysi-ology 39 131ndash143

Cowan N Winkler I Teder W amp Naumlaumltaumlnen R (1993) Mem-ory prerequisites of the mismatch negativity in the audi-tory event-related potential (ERP) Journal ofExperimental Psychology Learning Memory and Cogni-tion 19 909ndash921

Cseacutepe V Karmos G amp Molnaacuter M (1987) Evoked potentialcorrelates of stimulus deviance during wakefulness andsleep in catmdashanimal model of mismatch negativity Elec-troencephalography and Clinical Neurophysiology 66571ndash578

Davis H Mast T Yoshie N amp Zerlin S (1966) The slow re-sponse of the human cortex to auditory stimulus Recov-ery process Electroencephalography and ClinicalNeurophysiology 21 105ndash113

Davis H amp Zerlin S (1966) Acoustic relations of the humanvertex potentials Journal of the Acoustical Society ofAmerica 39 109ndash116

Dawson M E (1990) Psychophysiology at the interface ofclinical science cognitive science and neuroscience Psy-chophysiology 27 243ndash255

Dawson M E Fillion D L amp Schell A M (1989) Is elicita-tion of the autonomic orienting response associated withallocation of processing resources Psychophysiology 26560ndash572

Duncan C C amp Kaye W H (1987) Effects of chlonidine onevent-related potential measures of information process-ing In R Johnson J W Rohrbaugh amp R Parasuraman(Eds) Current trends in event-related potentialresearch (EEG supplement 40) (pp 527ndash531) Amster-dam Elsevier

Fillion D J Dawson M E Schell A M amp Hazlett E A(1991) The relationship between skin conductance orient-ing and the allocation of processing resources Psycho-physiology 28 410ndash424

Folk C L Remington R W amp Johnston J C (1992) Involun-tary covert orienting is contingent on attentional controlsettings Journal of Experimental Psychology Human Per-ception and Performance 18 1030ndash1044

Ford J M Roth W T amp Kopell B S (1976) Auditory event-related potentials to unpredictable shifts in pitch Psycho-physiology 13 32ndash39

Friedman D amp Simpson G (1994) ERP amplitude and scalpdistribution to target and novel events Effects of temporalorder in young middle-aged and older adults CognitiveBrain Research 2 49ndash63

602 Journal of Cognitive Neuroscience Volume 10 Number 5

Friedman D Simpson G amp Hamberger M (1993) Age-re-lated changes in scalp topography to novel and target stim-uli Psychophysiology 30 383ndash396

Fuster J (1989) The prefrontal cortex Anatomy physiologyand neuropsychology of the frontal lobe New York Ra-ven

Gati I amp Ben-Shakhar G (1990) Novelty and signicance inorientation and habituation A feature-matching approachJournal of Experimental Psychology General 119 251ndash263

Giard M H Lavikainen J Reinikainen K Perrin F Ber-trand O Pernier J amp Naumlaumltaumlnen R (1995) Separate repre-sentation of stimulus frequency intensity and duration inauditory sensory memory An event-related potential anddipole-model analysis Journal of Cognitive Neuroscience7 133ndash143

Giard M H Perrin F Echallier J F Theacutevenet M FromentJ C amp Pernier J (1994) Dissociation of temporal and fron-tal components in the auditory N1 wave A scalp currentdensity and dipole model analysis Electroencephalogra-phy and Clinical Neurophysiology 92 238ndash252

Giard M H Perrin F Pernier J amp Bouchet P (1990) Braingenerators implicated in processing of auditory stimulusdeviance A topographic ERP study Psychophysiology 27627ndash640

Grillon C Courchesne E Ameli R Geyer M amp Braff D L(1990) Increased distractibility in schizophrenic patientsArchives of General Psychiatry 47 171ndash179

Halgren E Baudena P Clarke J M Heit G Lieacutegeois CChauvel P amp Musolino A (1995) Intracerebral potentialsto rare target and distractor auditory and visual stimuli ISuperior temporal plane and parietal lobe Electroencepha-lography and Clinical Neurophysiology 94 191ndash220

Halgren E Baudena P Clarke J M Heit G Marinkovic KDevaux B Vignal J P amp Biraben A (1995) Intracerebralpotentials to rare target and distractor auditory and visualstimuli II Medial lateral and posterior temporal lobe Elec-troencephalography and Clinical Neurophysiology 94229ndash250

Hari R Haumlmaumllaumlinen M Ilmoniemi R Kaurokanta E Reini-kainen K Salminen J Alho K Naumlaumltaumlnen R amp Sams M(1984) Responses of the primary auditory cortex to pitchchanges in a sequence of tone pips Neuromagnetic re-cordings in man Neuroscience Letters 50 127ndash132

Hari R Kaila K Katila T Tuomisto T amp Varpula T (1982)Interstimulus interval dependence of the auditory vertexresponse and its magnetic counterpart Implications fortheir neural generation Electroencephalography and Clini-cal Neurophysiology 54 561ndash569

Hikosaka O Miyauchi S amp Shimojo S (1993) Voluntaryand stimulus-induced attention detected as motion sensa-tion Perception 22 517ndash526

Holdstock J S amp Rugg M D (1995) The effect of attentionon the P300 deection elicited by novel sounds Journalof Psychophysiology 9 18ndash31

Jaumlaumlskelaumlinen I Alho K Escera C Winkler I Sillanaukee Pamp Naumlaumltaumlnen R (1996) Effects of ethanol and auditory dis-traction on forced choice reaction time Alcohol 13 153ndash156

Javitt D C Schroeder C E Steinschneider M Arezzo J Camp Vaughan H G (1992) Demonstration of mismatch nega-tivity in the monkey Electroencephalography and Clini-cal Neurophysiology 83 87ndash90

Knight R T (1984) Decreased response to novel stimuli af-ter prefrontal lesions in man Electroencephalographyand Clinical Neurophysiology 59 9ndash20

Knight R T (1991) Evoked potential studies of attention ca-pacity in human frontal lobe lesions In H S Levin H M

Eisenberg amp A L Benton (Eds) Frontal lobe functionand dysfunction (pp 139ndash153) New York Oxford Univer-sity Press

Knight R T (1996) Contribution of human hippocampal re-gion to novelty detection Nature 383 256ndash259

Knight R T (1997) Distributed cortical network for visual at-tention Journal of Cognitive Neuroscience 9 75ndash91

Knight R T Scabini D Woods D L amp Clayworth C C(1989) Contributions of temporal-parietal junction to thehuman auditory P3 Brain Research 502 109ndash116

Kropotov J D Naumlaumltaumlnen R Sevostianov A V Alho K Reini-kainen K amp Kropotova O V (1995) Mismatch negativityto auditory stimulus change recorded directly from the hu-man temporal cortex Psychophysiology 32 418ndash422

Lehtonen J B (1973) Functional differentiation betweenlate components of visual evoked potentials recorded atocciput and vertex Effects of stimulus interval and con-tour Electroencephalography and Clinical Neurophysiol-ogy 35 75ndash82

Lyytinen H Blomberg A P amp Naumlaumltaumlnen R (1992) Event-re-lated potentials and autonomic responses to a change inunattended auditory stimuli Psychophysiology 29 523ndash534

Maumlkelauml J P Hari R amp Leinonen L (1988) Magnetic re-sponses of the human auditory cortex to noisesquarewave transitions Electroencephalography and ClinicalNeurophysiology 69 423ndash430

McCarthy G amp Wood C C (1985) Scalp distributions ofevent-related potentials An ambiguity associated withanalysis of variance models Electroencephalography andClinical Neurophysiology 62 203ndash208

McMillan N A (1973) Detection and recognition of inten-sity changes in tone and noise The detection-recognitiondisparity Perception amp Psychophysics 13 65ndash75

Mecklinger A amp Ullsperger P (1995) The P300 to novel andtarget events A spatio-temporal dipole model analysesNeuroReport 7 241ndash245

Mesulam M M (1981) A cortical network for directed atten-tion and unilateral neglect Annals of Neurology 10 309ndash325

Molnaacuter M Skinner J E Cseacutepe V Winkler I amp Karmos G(1995) Correlation dimension changes accompanying theoccurrence of mismatch negativity and the P3a event-related potential component Electroencephalographyand Clinical Neurophysiology 95 118ndash126

Naumlaumltaumlnen R (1990) The role of attention in auditory informa-tion processing as revealed by event-related potentials andother brain measures of cognitive function Behavioraland Brain Sciences 13 201ndash288

Naumlaumltaumlnen R (1991) Mismatch negativity outside strong atten-tional focus A commentary on Woldorff et al (1991) Psy-chophysiology 28 478ndash484

Naumlaumltaumlnen R (1992) Attention and brain function HillsdaleNJ Erlbaum

Naumlaumltaumlnen R Paavilainen P Alho K Reinikainen K amp SamsM (1989) Do event-related potentials reveal the mecha-nism of the auditory sensory memory in the humanbrain Neuroscience Letters 98 217ndash221

Naumlaumltaumlnen R Paavilainen P Tiitinen H Jiang D amp Alho K(1993) Attention and mismatch negativity Psychophysiol-ogy 30 436ndash450

Naumlaumltaumlnen R amp Picton T (1987) The N1 wave of the humanelectric and magnetic response to sound A review and aanalysis of the component structure Psychophysiology24 375ndash425

Oades R D amp Dittmannbalcar A (1995) Mismatch negativ-ity (MMN) is altered by directing attention NeuroReport6 1187ndash1190

Escera et al 603

Oumlhman A (1979) The orienting response attention andlearning An information-processing perspective In H DKimmel E H van Olst amp J F Orlebeke (Eds) The orient-ing reex in humans (pp 443ndash471) Hillsdale NJErlbaum

Oumlhman A (1992) Preattentive processing and orienting InB A Campbell H Hayne amp R Richardson (Eds) Attentionand information processing in infants and adults Per-spectives from human and animal research (pp 263ndash295) Hillsdale NJ Erlbaum

Pfefferbaum A Buchsbaum M amp Gips J (1971) Enhance-ment of the average evoked response to onset and cessa-tion Psychophysiology 8 332ndash339

Posner M I (1978) Chronometric explorations of mindHillsdale NJ Erlbaum

Ritter W Vaughan H G Jr amp Costa L D (1968) Orientingand habituation to auditory stimuli A study of short termchanges in average evoked responses Electroencephalogra-phy and Clinical Neurophysiology 25 550ndash556

Scherg M Vajsar J amp Picton T W (1989) A source analysisof the late human auditory evoked potentials Journal ofCognitive Neuroscience 1 336ndash355

Scherg M amp von Cramon D (1986) Evoked dipole sourcepotentials of the human auditory cortex Electroencepha-lography and Clinical Neurophysiology 65 334ndash360

Schroumlger E (1996) A neural mechanism for involuntary atten-tion shifts to changes in auditory stimulation Journal ofCognitive Neuroscience 8 527ndash539

Siddle D A T (1991) Orienting habituation and resource al-location An associative analysis Psychophysiology 28245ndash259

Sokolov E N (1963) Perception and the conditioned reexNew York Macmillan

Sokolov E N (1975) The neuronal mechanisms of the orient-ing reex In E N Sokolov amp O S Vinogradova (Eds)Neuronal mechanisms of the orienting reex (pp 217ndash235) Hillsdale NJ Erlbaum

Squires N K Squires K C amp Hillyard S A (1975) Two va-rieties of long-latency positive waves evoked by impredict-able auditory stimuli Electroencephalography andClinical Neurophysiology 38 387ndash401

Theeuwes J (1994) Stimulus-driven capture and attentionalset Selective search for color and visual abrupt onsetsJournal of Experimental Psychology Human Perceptionand Performance 20 799ndash806

Tiitinen H May P Reinikainen K amp Naumlaumltaumlnen R (1994) At-

tentive novelty detection in humans is governed by pre-attentive sensory memory Nature 372 90ndash92

Trejo L J Ryan-Jones D L amp Kramer A F (1995) Atten-tional modulation of the mismatch negativity elicited byfrequency differences between binaurally presented tonesPsychophysiology 32 319ndash328

Vaughan H J Jr amp Ritter W (1970) The sources of the audi-tory evoked responses recorded from the scalp Elec-troencephalography and Clinical Neurophysiology 28360ndash367

Velasco M amp Velasco F (1986) Subcortical correlates of thesomatic auditory and visual vertex activities II ReferentialEEG responses Electroencephalography and ClinicalNeurophysiology 63 62ndash67

Velasco M Velasco F amp Olvera A (1985) Subcortical corre-lates of the somatic auditory and visual vertex activities IBipolar EEG responses and electrical stimulation Elec-troencephalography and Clinical Neurophysiology 61519ndash529

Woldorff M G Hackley S A amp Hillyard S A (1991) The ef-fects of channel-selective attention on the mismatch nega-tivity wave elicited by deviant tones Psychophysiology28 30ndash42

Woods D L (1992) Auditory selective attention in middle-aged and elderly subjects An event-related brain potentialstudy Electroencephalography and Clinical Neurophysiol-ogy 84 456ndash468

Woodward S H Brown W S March J T amp Dawson M E(1991) Probing the time-course of the auditory oddball P3with secondary reaction time task Psychophysiology 28609ndash616

Yamaguchi S amp Knight R T (1991) Anterior and posteriorassociation cortex contributions to the somatosensoryP300 Journal of Neuroscience 11 2039ndash2054

Yamaguchi S amp Knight R T (1992) Effects of temporal-parietal lesions on the somatosensory P3 to lower limbstimulation Electroencephalography and Clinical Neuro-physiology 84 139ndash148

Yantis S (1993) Stimulus-driven attentional capture and at-tentional control settings Journal of Experimental Psy-chology Human Perception and Performance 19676ndash681

Yantis S amp Hillstrom A P (1994) Stimulus-driven atten-tional capture Evidence from equiluminant visual objectsJournal of Experimental Psychology Human Perceptionand Performance 20 95ndash107

604 Journal of Cognitive Neuroscience Volume 10 Number 5