Proc. Nati. Acad. Sci. USAVol. 88, pp. 8996-9000, October 1991Neurobiology

Human auditory evoked gamma-band magnetic fieldsC. PANTEV*t, S. MAKEIGt, M. HOKE§, R. GALAMBOS¶, S. HAMPSON II, AND C. GALLENII*Naval Health Research Center, San Diego, CA 92186; institute of Experimental Audiology, Munster, Germany; 1Department of Neuroscience, University ofCalifornia at San Diego, La Jolla, CA 92093; and ItDepartment of Neuropharmacology, Scripps Clinic and Research Foundation, San Diego, CA 92037

Contributed by R. Galambos, June 14, 1991

ABSTRACT We have discovered a ca. 40-Hz transientmagnetic oscillatory response, evoked in the human brain bythe onset of auditory stimuli, consisting of four or more cycleslocked in phase to stimulus onset in approximately the 20- to130-ms poststimulus interval. The response originates in thesupratemporal auditory cortex, some millimeters deeper andanterior to the source of the larger-amplitude slow-wave M100component of the evoked magnetic field and moves in aposterior arcing trajectory 1 cm or more in length. Theoscillatory cortical activation elicited by auditory stimuli maybe similar to the gamma-band cortical oscillations elicited byolfactory and visual stimuli and may represent an essentialcomponent of auditory perceptual processing.

Electroencephalographic oscillations in the 20- to 90-Hz fre-quency band were first seen by Berger (1), who hypothesized thatthe high-frequency potentials he observed during performance ofmental arithmetic "must be the material concomitant of mentalprocesses." In the past few years brain activity in this firquencyband, for which several authors advocate the term "gammaband" (2), has been the subject of intense scientific interest. Inthe central olfactory system, oscillatory sniff-evoked gamma-band responses (GBR), first studied by Adrian in the early 1940s(3), have been the subject ofdistinguished work by Freeman andco-worker (4-6), who observed that the oscillations vary intemporal and spatial coherence depending on the animal's pastexperience of an odor. Recently Gray et aL (7, 8) and Eckhornet al. (9) have reported that in cats a spatially disparate subset ofvisual cortical cells fires in coherent bursts (ca. 40-60 Hz) afterappropriate visual stimuli. These synchronous gamma-band os-cillations are now being proposed as a possible fundamentalphysiological mechanism underlying object recognition and at-tention (10, 11). Both auditory- and somatosensory-evoked ac-tivity in the gamma band also have been noted in direct record-ings from cortex, reticular formation, and cerebellum of variousanimals (12-14).

In man, electric and magnetic steady-state responses(SSRs) can be driven by periodically presented clicks (15-19)or tone pips (20) or by amplitude modulation of continuoustone (21) and reach their maximum amplitude in adults atrepetition rates of 35-40 Hz. Both cortical and subcorticalsources for the SSR have been proposed (18). Makeig andGalambos (22, 23) have identified a transient electric GBRlocked in time to the onset of auditory stimuli that increasesin amplitude as the stimulus rate is decreased below one persecond. We report here measurements of a comparablemagnetic GBR, using a new-generation large-array biomag-netometer. Large-array magnetic measurements allow accu-rate and reliable localization of the source of the magneticGBR in relation to the better studied M100 and M200 slow-wave (low-frequency; see below) components of the fieldevoked by auditory stimuli (24).

METHODS

The auditory stimulus-evoked magnetic brain activity gener-ated in the left temporal cortex of 20 right-handed adultsubjects was recorded in a magnetically shielded room witha 37-channel biomagnetometer (Biomagnetic Technologies,San Diego, CA). The response of one subject was recordedin 30 blocks of 128 stimuli. Three to six blocks were recordedfrom each of the other subjects. During the measurementssubjects lay on their right side with eyes open and were askedto remain alert. Stimuli were 1000-Hz tone bursts of 60 dBrelative to normal hearing level (nHL) (80-dB nHL in onesession) with 500-ms duration and 10-ms rise and fall times,presented to the right ear at interstimulus intervals of 4 s.Blocks of 128 stimulus-related epochs of 1000 ms wererecorded with a 200-ms prestimulus baseline. Electroenceph-alographic epochs were recorded simultaneously from onederivation (C,-earlobe).The neuromagnetic field pattern was recorded over a

circular area 144 mm in diameter above the auditory cortexthrough 37 axially symmetric first-order gradiometer pickupcoils (diameter, 20 mm; baseline, 50 mm). Each coil wasconnected to a superconducting quantum interference device(SQUID) sensor that produced a voltage proportional to themagnetic field radial to the coil. The intrinsic noise level was<10 ff/Hzl/2 in all but one ofthe channels. A sensor positionindicator system determined the spatial locations of thesensors relative to the head and insured that no head move-ments occurred during the measurements. In successivesessions, the magnetometer was carefully repositioned toallow response comparison and grand averaging. Blocks of128 stimulus-related epochs of 1000 ms were recorded with a200-ms prestimulus baseline in the passband of0.1-95 Hz anda sampling rate of 215 Hz. The wide-band responses werefirst averaged with artifact rejection and then digitally filteredin two passbands, 1-20 Hz and 28-48 Hz, to separate theevoked gamma-band and slow-wave field components.For each subject a spherical model was fit to the digitized

head shape, and the location, orientation, and amplitude ofanequivalent current dipole tangential to the surface of themodel sphere (25, 26) were estimated in each passband foreach point in time. The origin of the coordinate system usedis the midpoint between the preauricular points; the x-axisjoins the origin to the nasion; the y-axis passes between thepreauricular points; the z-axis is perpendicular to the x-yplane. Correlations between the theoretical fields generatedby the model and the measured field were used to estimate thegoodness of fit of the model parameters. Only estimates witha correlation coefficient above 0.90 were accepted.

Abbreviations: GBR, gamma-band response; SSR, steady-state re-sponse; MR, magnetic resonance; MRI, MR imaging; nHL, normalhearing level.*Visiting Scientist at Scripps Clinic and Research Foundation andBiomagnetic Technologies, Inc., San Diego, CA 92121.tTo whom reprint requests should be addressed at: Institute ofExperimental Audiology, MEG Laboratory, University of MonsterKardinal-von-Galen-Ring 10, D4400 Monster, Germany.

8996

The publication costs of this article were defrayed in part by page chargepayment. This article must therefore be hereby marked "advertisement"in accordance with 18 U.S.C. §1734 solely to indicate this fact.

Dow

nloa

ded

by g

uest

on

Apr

il 6,

202

0

Proc. Natl. Acad. Sci. USA 88 (1991) 8997

Orthogonal coronal, sagittal, and axial MR images with3-mm slice thickness were obtained. The T1-weighted contrastbetween cortical gray matter and adjacent white matter wasmaximized through inversion recovery sequences. The nasionand the bilateral preauricular points were identified on MRimages by affixing cod liver oil capsules with an inner diameterof 3 mm. The spatial coordinates of these MR imaging (MRI)markers were used forjuxtaposing the coordinate systems ofthe MRI and the magnetic measurements, allowing referenceto a common Cartesian coordinate system.

RESULTSIn the auditory wide-band response (0.1-95 Hz), detectorchannels recording large-amplitude slow-wave componentswith peak latencies near 50, 90, and 170 ms also containnoticeable gamma-band activity during at least the first 100ms after stimulus onset (Fig. 1). We present here data of theseven subjects with the best GBR signal-to-noise ratio. Thewide-band response contains spectral energy in two distinctfrequency ranges (Fig. 2). One peak near 10 Hz derivespredominantly from the slow-wave components, while an-other peak in the 30- to 40-Hz range reflects the GBR. Thespectral segregation of the two responses facilitates theirseparation by band-pass filtering for the purpose of studyingtheir spatiotemporal characteristics.A plot of the distribution of the gamma-band waveforms

reveals that the response consists of at least four oscillations(Fig. 3). Further, the maximum and minimum field regionssubstantially overlap those of the larger-amplitude slow-wave components of the evoked field (Fig. 1), suggestingsimilarly oriented dipolar sources for the evoked slow-waveand gamma-band activities. Examination of wide-band andgamma-band data for several channels located along a lineconnecting the maximum and minimum ofthe neuromagneticfields (Fig. 4A) reveals the peaks of the GBR superimposedon the early slow-wave components and confirms that theGBR burst begins before the MS0 peak and is largest justbefore the M100 peak.

40

30

20

10

Q)-o--a-E0

20

15

10

5-

frequency (Hz)

FIG. 2. Spectrum of the wide-band responses averaged overthree consecutive data blocks from seven subjects. The ordinate axisshows amplitude in 20 loglo fT. The upper two traces were con-structed by averaging the Fourier amplitude spectra of 200-msHanning-windowed epochs and smoothing the result with a rectan-gular 3-Hz window. Epochs selected were -35 to 165 ms for thewide-band response (WBR, top trace) and -200 to 0 ms and 300 to500 ms for the baseline estimate (BL, middle trace). The bottom traceis the logarithmic difference between the response and baselinespectra, clearly showing separate spectral peaks for the slow-waveauditory evoked field (SAEF) and the GBR components.

During the 20- to 130-ms period after stimulus onset, themagnetic and electric GBR records are strongly locked inphase (Fig. 4B), though they are uncorrelated before and after.Since biomagnetic recordings are sensitive only to sourcesoriented tangentially to the detector coils whereas bioelectricrecordings are sensitive to both radial and tangential sources,the phase-locking of the magnetic and electric GBR suggeststhat both responses have a common brain source orientedpredominantly tangential to the surface of the head.

2037 h .21

36T

11,9 Tb8

1r9

35t Tal31

34 7=333 6~~~3

.,2928

IVV.4m 100IfT |-100

-100 500msec

FIG. 1. Averaged wide-band auditoryevoked responses (n = 384) recorded through 37magnetic sensors from the left hemisphere of aright-handed subject to 1000-Hz tone bursts of80-dB nHL with 500-ms duration. The wide-band response is plotted from 100 ms before'to500 ms after stimulus onset (marked by thevertical line on the time axis)'. .Data from channel36 were noisy and are not shown.

Neurobiology: Pantev et A

Dow

nloa

ded

by g

uest

on

Apr

il 6,

202

0

Proc. Natl. Acad. Sci. USA 88 (1991)

20

-1.A^

36T1

35

1 171 1

33 F1PV

32 .-T

1F- 1 'v -

1 9 I ABAAAS 9 --lglh1.LA - - T V Vv'TO 23 L

2 10 T~ V

7 F-~VV3 S'g 24.^A,.ILAIAAA.ToAL6 1 4 1-r11' 5.IAA.X. IJAIIA id.L h 5 .~1} 12 I-.NIfiv--Ivvrvs~~~ rl-www26

15 13 ^A.T^^ 4 14 .^_, i rw~~~~~~~~~~~~~~~I A31 -Twlo30 28r V2

ivfT^^A^^^.29 techInTT

30

-30 1-100 200

msec

FIG. 3. Grand average of three data blocks(n = 384) from the same subject and under thesame experimental conditions as described inFig. 1, after digital band-pass filtering between28 and 48 Hz with a 2-pole Butterworth filterwith zero phase shift.

The best-fitting equivalent current dipole source locationsfor the M100 and M200 peaks of the evoked slow-wave fieldfor 30 separate data blocks from one subject are shown in Fig.5A. The M100 and M200 peak locations are clearly separated,the M200 peak being 8 mm more anterior (P < 0.001, t test)and 4 mm more medial (P < 0.001, t test) than the M100 peak.Superimposition of these locations on the subject's brainmagnetic resonance (MR) images confirmed that both com-ponents originate in the supratemporal auditory cortex (24).The estimated GBR peak locations were mostly intermediateto those ofM100 and M200 (Fig. 5A), also suggesting a sourcein the supratemporal auditory cortex. However, the means ofthe source locations for the individual GBR peaks (Fig. SB)

A

I-

-)a)

._

-200 02 0-200 400

progress in an arcing trajectory beginning close to the originof M200, then curving first deeper into the sulcus and thenoutward again in a posterior movement towards the origin ofM100. The location of the equivalent current dipole for thelargestGBR peak (near 60 ms) was distinct from the locationsof M100 (4 mm medial; P < 0.001, t test) and M200 (4 mmposterior; P < 0.001, t test). Significant differences (P < 0.05,t test) in the same directions but slightly larger (5 mm) wereobtained for the dipole location estimates for the sevensubjects (Fig. 5C).The grand mean dipole source location estimates for the

GBR peaks (Fig. 5B) and the M100 slow-wave field compo-nent (Fig. 5A) were superimposed upon MRI sections of our

B

01)N

zEL-oc1-

a)

ca.Etv

*- 0

600 BOO

latency (ms)-200 -100 0 100 200 300 400 500

latency (ms)

FIG. 4. Evidence for the magnetic GBR. (A) Wide-band response (WBR) and GBR from five channels located along a posterior-anteriorline between channels 24 and 33 (same data as in Figs. 1 and 3). The coincidence of the GBR peaks in the unfiltered and filtered data is visibleas well as their polarity reversal across the recording array. The initial WBR peaks (M50, M100, and M200) are followed by a sustained fieldand then by an off-response that contains no noticeable GBR components. (B) Superimposition of simultaneously recorded magnetic and electricGBRs from six subjects.

8998 Neurobiology: Pantev et al.

Dow

nloa

ded

by g

uest

on

Apr

il 6,

202

0

Proc. Natl. Acad. Sci. USA 88 (1991) 8999

A best-studied subject (Fig. 6). Coronal, sagittal, and axial MRIviews (Fig. 6 a, b, and d) show that the equivalent dipolesources are located in cortical regions on the floor of theSylvian fissure (the transverse temporal gyrus and the trans-verse part of the superior temporal gyrus), which encompassthe primary auditory cortex and the primary integrationregion (27). In particular, the arcing trajectory of GBRactivity follows the cortical surface. This is best seen in theenlarged coronal slice shown in Fig. 6c.

DISCUSSIONWe propose that the auditory GBR is an event-relatedrhythmic response that occurs in parallel with the M50 and

4.S < } z < oM100 components of the evoked slow-wave field. This con-0 cept is based on five observations. (i) Individual peaks of the

Oss~<>4°5response are visible in each subject's wide-band data super-Iea6 X t(cry) imposed on the slow-wave activity. (ii) The spectrum of the

t at poS wide-band response contains a peak near 10 Hz, and a(cry,) tit second, clearly identifiable peak at 30-40 Hz. (iii) After

separation into slow-wave and gamma-band responses,B equivalent current dipoles account for each data set almost

completely. (iv) The source locations of the slow-wave andgamma-band activity differ significantly, although superim-

6 S / . position of the source estimates on MRI images shows thatboth responses originate within the auditory cortex. (v) The. posterior movement of the GBR source estimates is contrary

E6)GBR, , l¢ ; . to the anterior progression of the slow-wave response sourcelocations.While the initial peaks of the magnetic GBR have been

observed in magnetic recordings and identified with peaks ofthe electric middle latency response (28, 29), the full temporalextent of the response has not been recognized previously.S 6Thi's may be in part because the relatively small, high-frequency GBR waveforms were regarded as noise and

4.5 > < H e m >( X Hefiltered out of previous magnetic recordings. In additional4.S experiments on the same subjects, we have observed similar

0 GBRs to tone bursts of different frequencies, short Gaussiantone pips, or clicks. Since it is generated by a variety of

rnea..at 6 X (crT) stimuli and occurs very early in auditory cortical information(crt) ,p° processing, the GBR appears to reflect the activity of a

Artll fundamental auditory perceptual mechanism.The observed trajectory of the magnetic GBR peak-

C location estimates, a posterior-moving arc across the audi-tory cortex, is amenable to at least four interpretations: (i)continuous moving dipole: the GBR is produced in a small

6 s ^ ffi ) < > | ~>locus of cortex whose location moves continuously across

>Lo g the cortex along a posterior arc; (ii) discrete moving dipole:* If gthe GBR peaks are produced by a string of sequentially

_6sGBR1lQ|755 activated cortical loci; (iii) multiple dipoles: the apparentmovement ofthe GBR arises from a continuous change in therelative amplitudes of several fixed dipoles located at or

beyond the two ends and outside the point of maximumcurvature of the observed arc; and (iv) mass action: the

C... Ss6311> fi _|observedarc of movement of the source location of thesuccessive GBR peaks represents the movement of the

4.5_0

A dd5peaks M100 and M200 for 30 blocks of data (n = 128) from one

0 0;fly ~ - subject. The opposite orientation of the arrows illustrates the oppo-5 0 site polarity of the two peaks. (B) Source locations of the individual

rri,0. 5 GBR peaks for the same subject. To increase the signal-to-noise ratio

ee-16 I) of the GBR data, 10 subaverages of 3 blocks each were derived fromCt the 30 data blocks. Both grand mean (circles) and subaverage (dots)

(crrn) Pnt' responses are shown. The numeric values in the bottom plane*fl indicate the mean peak latencies. (C) Source locations for the GBR

peaks averaged across seven subjects. The three-dimensional crossFIG. 5. Equivalent dipole source locations of the major compo- in the upper-right corner shows the standard deviation of the

nents of the evoked slow-wave field and of the successive GBR localization estimates across subjects. The numerical values in thepeaks. (A) Source location estimates of the evoked slow-wave field bottom plane indicate the mean peak latencies.

Neurobiology: Pantev et A

Dow

nloa

ded

by g

uest

on

Apr

il 6,

202

0

Proc. Natl. Acad. Sci. USA 88 (1991)

B XZ-Plane (Sagittal View)

C YZ-Plane (Enlargement) 1) YX-Plane (Axial Views

FIG. 6. Coronal (A), sagittal (B), enlarged view of the coronal(including the Sylvian fissure) (C), and axial (D) MRI sections for thesame subject as in Fig. 5 A and B with overlays showing the meansource-location estimates of several of the GBR peaks and the peakof the evoked slow-wave field M100 component (white dots at thetips of arrows). The sagittal and axial slices shown include thelocation estimate for the 60-ms GBR peak. The coronal slice includesthe location estimates for the M100 peak (right arrow in C) andseveral peaks in the GBR (left arrow in C; compare Fig. SB).

centroid of a spatially distributed pattern of cortical activa-tion.The continuous and discrete moving dipole models are

straightforward interpretations of the data but have no sup-port from direct observations in the cortex. A multiple dipolemodel has been suggested by Scherg et al. (29) to explain theirdata on the early GBR components. The model they fitassumes two dipoles oriented at right angles to each other,but their results also lack support from anatomical evidence.Finely sampled observations of the spatial distribution ofgamma-band activity in olfactory cortex have been reportedby Freeman (4, 5), who has observed that GBR activity in theolfactory bulb spreads out across the cortical surface rapidlyin all directions from an initial focus with only a small phaselag across a recording array spanning several millimeters.Until direct observations with finer spatial resolution are

available, the most plausible interpretation of our results isthat GBR activity may also spread out from an anterior focusacross the surface of the superior temporal plane, and theobserved movement of the equivalent source representsmovement of the centroid of this gamma-band activity. Aconceivable function of a global GBR excitation could be tosynchronize and dynamically link separate subregions ofauditory cortex in order to combine separately representedauditory features into unitary auditory percepts-for exam-

ple, the pitch of complex tones (30).The fact that electric and magnetic SSRs are largest for

rates in the frequency range of the GBR (35-40 Hz) alsosuggests that the SSR to auditory stimuli is essentially the

overlapping sum of the successive GBRs. The relative loca-tions of the sources for the largest GBR peak and the M100peak are also consistent with previous reports on the relativespatial locations of the M100 and the SSR (18, 19). Manyother facts about the auditory cortex GBR, including confir-mation of its spatio-temporal extent, await further measure-ments. In general, the auditory evoked GBR appears toreflect the general observation that perception in most or allsense modalities involves coherent rhythmic brain activity atgamma-band frequencies.

We thank Drs. F. Bloom, C. Gray, F. Crick, W. Black, E.Hirschkoff, and B. Ldtkenhoner for reviewing earlier versions ofthismanuscript and S. Cobb for assistance in overlaying the MRIs. Thework was supported by grants from the Deutsche Forshungsgemein-schaft (Klinische Forschergruppe "Biomagnetismus und Biosig-nalanalyse"), from the Armstrong MacDonald Foundation and thePasarow Foundation, and by a grant to Dr. Makeig from the NationalAlliance for Research in Schizophrenia and Depression.

1. Berger, H. (1937) Arch. Psychiatr. Nervenkrankh. 106, 165-187.2. Bressler, S. L. (1990) Trends Neurosci. 13, 161-162.3. Adrian, E. D. (1942) J. Physiol. (London) 100, 459-473.4. Freeman, W. (1975) MassAction in the Nervous System (Academic,

New York).5. Freeman, W. (1985) in Handbook of Electroencephalography and

Clinical Neurophysiology, eds. Gevins, A. & Remond, A. (Elsevier,Amsterdam), Vol. 3A, part 2, chap. 18.

6. Freeman, W. & Skarda, C. (1985) Brain Res. Rev. 10, 147-175.7. Gray, C. & Singer, W. (1989) Proc. Natl. Acad. Sci. USA 86,

1698-1702.8. Gray, C., Konig, P., Engel, A. & Singer, W. (1989) Nature (London)

338, 334-337.9. Eckhorn, R., Bauer, R., Jordan, W., Brosch, M., Kruse, W., Munk,

M. & Reitboeck, H. J. (1988) Biol. Cybernet. 60, 121-130.10. Crick, F. & Koch, C. (1991) Neurosciences 2, in press.11. Barinaga, M. (1990) Science 249, 856-858.12. Steriade, M. & Dimetrescu, M. (1962) Electroencephalogr. Clin.

Neurophysiol. 14, 21-36.13. Basar, E. (1980) EEG Brain Dynamics (Elsevier, Amsterdam).14. Bouyer, J. J., Montaron, M. F. & Rougeul, A. (1981) Electroen-

cephalographr. Clin. Neurophysiol. 51, 244-252.15. Galambos, R., Makeig, S. & Talmachoff, P. J. (1981) Proc. Natl.

Acad. Sci. USA 78, 2643-2647.16. Galambos, R. & Makeig, S. (1988) in Dynamics of Sensory and

Cognitive Processing by the Brain, eds. Basar, E. & Bullock, T.(Springer, Berlin), pp. 103-122.

17. Makeig, S. (1985) Studies in Musical Psychobiology (UniversityMicrofilms, Ann Arbor, MI).

18. Makela, J. P. & Hari, R. (1987) Electroencephalogr. Clin. Neuro-physiol. 66, 539-546.

19. Hari, R., Hamalainen, M. & Joustiniemi, S. L. (1989) J. Acoust.Soc. Am. 86, 1033-1039.

20. Ribary, U., Llinas, R., Suk, J. & Ferris, S. H. (1990) in Advancesin Biomagnetism, eds. Williamson, S., Hoke, M., Stroink, G. &Kotani, M. (Plenum, New York), pp. 311-314.

21. Romani, G. L., Williamson, S. J. & Kaufman, L. (1982) Science216, 1339-1340.

22. Makeig, S. & Galambos, R. (1989) Soc. Neurosci. Abstr. 15, 113.23. Makeig, S. (1990) in Psychophysiological Brain Research, eds.

Brunia, C., Gaillard, A. & Kok, A. (Tilburg Univ. Press, TheNetherlands), Vol. 2, pp. 60-64.

24. Pantev, C., Gallen, C., Hampson, S., Buchanan, S. & Sobel, D.(1991) Am. J. EEG Technol. 31, 83-101.

25. Geselowitz, D. B. (1970) IEEE Trans. Mag. 6, 346-347.26. Sarvas, J. (1987) J. Phys. Med. Biol. 32, 11-22.27. Talairach, J. & Tournoux, P. (1988) Co-Planar Stereotaxic Atlas of

the Human Brain (Thieme, Leipzig).28. Pelizone, M., Hari, R., Makela, J. P., Huttunen, J., Ahlfors, S. &

Himaidinen, M. (1987) Neurosci. Lett. 82, 303-307.29. Scherg, M., Hari, R. & Hfmilainen, M. (1990) in Advances in

Biomagnetism, eds. Williamson, S., Hoke, M., Stroink, G. &Kotani, M. (Plenum, New York), pp. 97-100.

30. Terhardt, E. (1989) Naturwissenschaften 76, 496-504.

A YZ- Plane (Coronal View)

9000 Neurobiology: Pantev et al.

Dow

nloa

ded

by g

uest

on

Apr

il 6,

202

0