RESEARCH ARTICLES

74. J. Garcia-Fernàndez, P. W. H. Holland, Nature 370, 86. R. C. Hardison, Proc. Natl. Acad. Sci. U.S.A. 93, 5675 tional Laboratory under contract no. W-7405-ENG-
563 (1994). (1996). 36; and by MEXT, Japan (grants 12201001 to Y.K.,
75. D. E. Ferrier, C. Minguillon, P. W. H. Holland, J. Garcia- 87. K. E. van Holde, K. I. Miller, H. Decker, J. Biol. Chem. 12202001 to N.S.), Japan Society for the Promotion
Fernandez, Evol. Dev. 2, 284 (2000). 276, 15563 (2001). of Science (to Y.S.), Human Frontier Science Program
76. A. Di Gregorio et al., Gene 156, 253 (1995). 88. Y. Satou et al., Development 128, 2893 (2001). (to N.S. and M.L.), and NIH (HD-37105 and NSF
77. M. Gionti et al., Dev. Genes. Evol. 207, 515 (1998). 89. T. Kusakabe et al., Dev. Biol. 242, 188 (2002). IBN-9817258 to M.L.)
78. D. Chourrout, R. Di Lauro, personal communication. 90. N. Harafuji, D. N. Keys, M. Levine, Proc. Natl. Acad. Supporting Online Material
79. O. Hobert, H. Westphal, Trends Genet. 16, 75 Sci. U.S.A. 99, 6802 (2002). www.sciencemag.org/cgi/content/full/298/5601/2157/
(2000). 91. D. N. Keys et al., in preparation. DC1
80. S. I. Tomarev, Int. J. Dev. Biol. 41, 835 (1997). 92. This work was performed under the auspices of the SOM Text
81. G. Krishnan, Indian J. Exp. Biol. 13, 172 (1975). U.S. Department of Energy’s Office of Science, Bio- Tables S1 to S9
82. S. M. Read, T. Bacic, Science 295, 59 (2002). logical and Environmental Research Program; by the Figures S1 and S2
83. J. Zuo et al., Plant Cell 12, 1137 (2000). University of California, Lawrence Livermore National References
84. N. Lo et al., Curr. Biol. 10, 801 (2000). Laboratory under Contract No. W-7405-Eng-48,
85. D. R. Nobles, D. K. Romanovicz, R. M. Brown Jr., Plant Lawrence Berkeley National Laboratory under con-
Physiol. 127, 529 (2001). tract no. DE-AC03-76SF00098, and Los Alamos Na- 1 November 2002; accepted 20 November 2002

The Cortical Topography of context violations on one or more compo-

nents of event-related brain responses that

Tonal Structures Underlying index the presence and magnitude of con-

text violations. Overall, the cognitive dis-
tance of the probe event from the estab-
Western Music lished harmonic context correlates positive-
ly with the amplitudes of such components.
Petr Janata,1,2* Jeffrey L. Birk,1 John D. Van Horn,2,3 These effects appear even in listeners with-
Marc Leman,4 Barbara Tillmann,1,2 Jamshed J. Bharucha1,2 out any musical training (9, 11). The per-
ceptual and cognitive structures that facil-
Western tonal music relies on a formal geometric structure that determines itate listening to music may thus be learned
distance relationships within a harmonic or tonal space. In functional magnetic implicitly (2, 12–15).
resonance imaging experiments, we identified an area in the rostromedial The prefrontal cortex has been implicated
prefrontal cortex that tracks activation in tonal space. Different voxels in this in the manipulation and evaluation of tonal
area exhibited selectivity for different keys. Within the same set of consistently information (10, 11, 16 –18). However, the
activated voxels, the topography of tonality selectivity rearranged itself across regions that track motion on the tonality sur-
scanning sessions. The tonality structure was thus maintained as a dynamic face have not been identified directly. When
topography in cortical areas known to be at a nexus of cognitive, affective, and presented with a stimulus that systematically
mnemonic processing. moves across the entire tonality surface, will
some populations of neurons respond selec-
The use of tonal music as a stimulus for one listens to music. Patterns of expecta- tively to one region of the surface and other
probing the cognitive machinery of the hu- tion elicitation and fulfillment may underlie populations respond selectively to another
man brain has an allure that derives, in part, our affective responses to music (5). region of the surface?
from the geometric properties of the theo- Two lines of evidence indicate that the Identification of tonality-tracking
retical and cognitive structures involved in tonality surface is represented in the human brain areas. In order to identify cortical
specifying the distance relationships among brain. First, when one subjectively rates sites that were consistently sensitive to acti-
individual pitches, pitch classes (chroma), how well each of 12 probe tones, drawn vation changes on the tonality surface, eight
pitch combinations (chords), and keys (1– from the chromatic scale (6 ), fits into a musically experienced listeners (see “sub-
3). These distance relationships shape our preceding tonal context that is established jects” in supporting online text) underwent
perceptions of music and allow us, for ex- by a single chord, chord progression, or three scanning sessions each, separated by 1
ample, to notice when a pianist strikes a melody, the rating depends on the relation- week on average, in which they performed
wrong note. One geometric property of ship of each tone to the instantiated tonal two perceptual tasks during separate runs.
Western tonal music is that the distances context. Nondiatonic tones that do not oc- During each run, they heard a melody that
among major and minor keys can be repre- cur in the key are rated as fitting poorly, systematically modulated through all 12 ma-
sented as a tonality surface that projects whereas tones that form part of the tonic jor and 12 minor keys (see “stimuli and
onto the doughnut shape of a torus (1, 4 ). A triad (the defining chord of the key) are tasks” in supporting online text) (Fig. 1 and
piece of music elicits activity on the tonal- judged as fitting best (2). Probe-tone pro- audio S1). A timbre deviance detection task
ity surface, and harmonic motion can be files obtained in this manner for each key required listeners to respond whenever they
conceptualized as displacements of the ac- can then be correlated with the probe-tone heard a note played by a flute instead of the
tivation focus on the tonality surface (3). profile of every other key to obtain a matrix standard clarinet timbre, whereas a tonality
The distances on the surface also help gov- of distances among the 24 major and minor violation detection task required listeners to
ern expectations that actively arise while keys. The distance relationships among the respond whenever they perceived notes that
keys readily map onto the surface of the violated the local tonality (Fig. 1D). The use
torus (4 ). Thus, there is a direct correspon- of two tasks that required attentive listening
1
Department of Psychological and Brain Sciences, dence between music-theoretic and cogni- to the same melody but different perceptual
2
Center for Cognitive Neuroscience, 3Dartmouth
Brain Imaging Center, Dartmouth College, Hanover,
tive descriptions of the harmonic organiza- analyses facilitated our primary goal of iden-
NH 03755, USA. 4Institute for Psychoacoustics and tion of tonal music (7 ). tifying cortical areas that exhibit tonality
Electronic Music, Ghent University, Ghent, Belgium. Second, electroencephalographic stud- tracking that is largely independent of the
*To whom correspondence should be addressed. E- ies of musical expectancy (8 –11) have ex- specific task that is being performed (see
mail: petr.janata@dartmouth.edu amined the effect of melodic and harmonic “scanning procedures” in supporting online

www.sciencemag.org SCIENCE VOL 298 13 DECEMBER 2002 2167

 RESEARCH ARTICLES
Fig. 1. Properties of the tonality surface and
behavioral response profiles. In the key names,
capital letters indicate major keys and lower-
case letters indicate minor keys. (A) Unfolded
tori showing the average tonality surfaces for
each of the 24 keys in the original melody. The
top and bottom edges of each rectangle wrap
around to each other, as do the left and right
edges. ␪ and ␾ refer to the angular position
along each of the circles comprising the torus.
The color scale is arbitrary, with red and blue
indicating strongest and weakest activation, re-
spectively. Starting with C major and shifting
from left to right, the activation peak in each
panel reflects the melody’s progression through
all of the keys. (B) The circle of fifths. Major
keys are represented by the outside ring of
letters. Neighboring keys have all but one of
their notes in common. The inner ring depicts
the (relative) minor keys that share the same
key signature (number of sharps and flats) with
the adjacent major key. The color code refers to
the three groups of keys into which tonality
tracking voxels were categorized (Fig. 3). (C)
Correlations among the average tonality sur-
face topographies for each key. The topogra-
phies of keys that are closely related in a
music-theoretic sense are also highly positively
correlated, whereas those that are distantly
related are negatively correlated. Three groups
of related keys, indicated in (B), were identified
by singular value decomposition of this corre-
lation matrix. (D). Average response profiles
(eight listeners, three sessions each) from the
tonality deviance detection task illustrate the
propensity of specific test tones to pop out and
elicit a response in some keys but not in others
over the course of the melody. Error bars reflect 1 SEM.

text). Using a regression analysis with sepa- Fig. 2. Group conjunction maps
rate sets of regressors to distinguish task ef- showing the consistency with
fects from tonality surface tracking, we iden- which specific structures were
activated across listeners. Con-
tified task- and tonality-sensitive areas (see junction maps of individual lis-
“fMRI analysis procedures” in supporting on- teners, containing the voxels
line text). Tonality regressors were construct- that were activated significantly
ed from the output of a neural network model (P ⬍ 0.001) in all scanning ses-
of the moment-to-moment activation changes sions for that listener, were nor-
on the tonality surface (see “tonality surface malized into a common space
and summed together across lis-
estimation” in supporting online text). teners (see “spatial normaliza-
Our tasks consistently activated several tion” in supporting online text).
regions in the temporal, parietal, frontal, Voxels that were consistently
and limbic lobes as well as the thalamus activated by at least four of the
and cerebellum. The most extensive consis- eight listeners are projected onto
tent activation was along the superior tem- the group’s mean normalized T1
image. (A) Areas sensitive to the
poral gyrus (STG) of both hemispheres, two task regressors (Table 1). (B)
though the extent was greater in the right The only areas whose activity
hemisphere, stretching from the planum patterns were significantly and
temporale to the rostral STG and middle consistently correlated with the
temporal gyrus (Fig. 2A and Table 1). Both tonality regressors both within
the task and the tonality regressors corre- and across listeners were the
rostral portion of the ventrome-
lated significantly and consistently with ac- dial superior frontal gyrus and
tivity in the rostromedial prefrontal cortex, the right orbitofrontal gyrus.
primarily in the rostral and ventral reaches
of the superior frontal gyrus (SFG) (Figs. 2
and 3). The consistent modulation of this individual level, we reconstructed and cate- reconstructed surfaces from each session in-
area in all of our listeners led us to focus on gorized the tonality sensitivity surface (TSS) dicated that the medial prefrontal cortex
this region as a possible site of a tonality for each voxel that exhibited significant re- maintains a distributed topographic represen-
map. sponses (P ⬍ 0.001) in every one of the three tation of the overall tonality surface (Fig. 3).
Tonality-specific responses in the scanning sessions (see “tonality surface Although some voxels exhibited similar TSSs
rostromedial prefrontal cortex. At the estimation” in supporting online text). The from session to session, the global tonality

2168 13 DECEMBER 2002 VOL 298 SCIENCE www.sciencemag.org

 RESEARCH ARTICLES
given a preceding musical input. Given the
diversity of the music we hear, the situa-
tions in which we hear it, and our affective
and motoric responses to it, it is likely that
tonal contexts are maintained in cortical
regions predisposed to mediating interac-
tions between sensory, cognitive, and af-
fective information. The medial prefrontal
cortex is a nexus for such functions (20, 21)
and is therefore an ideal region for main-
taining a tonality map. In the macaque,
connections to the medial prefrontal cortex
from unimodal sensory cortices are wide-
spread for the auditory modality and sparse
for the other sensory modalities (22). In our
experiments, we observed significant task-
related activity in auditory association ar-
eas and the anterior STG, primarily in the
right hemisphere. Reciprocal projections
between these areas and the ventral medial
prefrontal cortex help explain how and why
a tonality map might be maintained in the
medial prefrontal cortex. This region has
already been implicated in assessing the
degree of musical consonance or disso-
nance caused by a harmonic accompani-
ment to a melody (23). Our results suggest
that the rostromedial prefrontal cortex not
only responds to the general degree of con-
sonance but actively maintains a distributed
topographic representation of the tonality
surface. The perception of consonance and
dissonance depends on intact auditory cor-
tices (24, 25). However, even with bilateral
auditory cortex ablations, the ability to gen-
erate expectancies based on tonal contexts
remains, suggesting that the cognitive
structures maintaining tonal knowledge
largely reside outside of temporal lobe au-
ditory structures (24 ).
Dynamic topographies. In contrast to
Fig. 3. Topography of tonality sensitivity of rostroventral prefrontal cortex in three listeners distributed cortical representations of classes
across three scanning sessions each. Each voxel’s color represents the key group with which the of complex visual objects that appear to be
voxel’s TSS was maximally correlated (Fig. 1B). The minority of voxels that were maximally topographically invariant (26), we found that
correlated with the average tonality surface are shown in white. A TSS represents how sensitive the mapping of specific keys to specific neu-
the voxel is to each point on the torus. The TSSs of selected voxels are displayed as unfolded ral populations in the rostromedial prefrontal
tori. Figure 1A serves as a legend for assigning keys to the individual TSSs. The highlighted cortex is relative rather than absolute. Within
voxels were chosen to display both the consistency and heterogeneity of the tonality surfaces
across sessions. For each listener, the activity of all voxels shown was significantly correlated a reliably recruited network, the populations
with the tonality regressors in all sessions. Thus, what changed between sessions was not the of neurons that represent different regions of
tonality-tracking behavior of these brain areas but rather the region of tonal space (keys) to the tonality surface are dynamically allocated
which they were sensitive. This type of relative representation provides a mechanism by which from one occasion to the next. This type of
pieces of music can be transposed from key to key, yet maintain their internal pitch dynamic topography may be explained by the
relationships and tonal coherence. properties of tonality structures. In contrast to
categories of common visual objects that dif-
topography varied across sessions in each of in the right hemisphere; the temporal pole; fer in their spatial features, musical keys are
the listeners. The number of voxels falling the anterior and posterior superior temporal abstract constructs that share core properties.
into each of the tonality categories (Fig. 1B) sulci; the precuneus and superior parietal gy- The internal relationships among the pitches
was evenly distributed within each session rus; the posterior lingual gyrus; and the cer- defining a key are the same in each key,
(table S1), but the relative pattern of tonality ebellum (19). thereby facilitating the transposition of musi-
sensitivity changed. For all listeners, we also Discussion. Central to our ability to cal themes from one key to another. Howev-
found tonality-sensitive voxels outside of the hear music coherently are cognitive struc- er, the keys themselves are distributed on a
medial prefrontal region (table S2). The pre- tures that maintain perceptual distance re- torus at unique distances from one another. A
cise constellations of sensitive areas differed lationships among individual pitches and dynamic topography may also arise from the
across listeners. We found tonality-sensitive groups of pitches. These structures shape interplay of short-term and long-term memo-
foci in the orbital and frontal gyri, primarily expectations about pitches we will hear, ry stores of tonal information and may serve

www.sciencemag.org SCIENCE VOL 298 13 DECEMBER 2002 2169

 RESEARCH ARTICLES
Table 1. Loci consistently showing a main effect of task in a majority of listeners. MTG, middle temporal gyrus; IFG, inferior frontal gyrus; SPG, superior parietal
gyrus.

Left hemisphere Right hemisphere

Lobe Region (Brodmann area) Location (mm) Listeners Cluster Location (mm) Listeners Cluster
at peak size at peak size
x y z (no.) (voxels) x y z (no.) (voxels)

Temporal
STG (22) – 64 –11 10 6 74
STG/Heschl’s gyrus (41/42) –56 –19 9 7 74 52 –11 5 8 163
STG/planum temporale (22) – 68 – 41 15 5 14 64 –30 15 6 163
64 –26 5 6 163
Rostromedial STG 38 15 –35 5 36
Rostroventral MTG (21) 52 0 –35 6 36
Middle MTG/superior temporal sulcus (21) 56 –15 –15 5 163
Ventral MTG (21) 60 –11 –25 6 163
Frontal
Rostroventromedial SFG (10/14) 0 49 0 5 27 4 64 0 5 27
Superior frontal sulcus/frontopolar gyrus (10) 26 64 30 5 3
Lateral orbital gyrus (11) 49 41 –10 5 4
IFG, pars orbitalis (47) 49 45 4 5 3
IFG, pars opercularis (44) 56 19 5 6 3
60 22 20 4 11
Precentral gyrus (6) 49 4 55 5 10
Parietal
Postcentral gyrus (1) 64 –11 25 6 163
Supramarginal gyrus (40) 64 –30 35 6 3
Precuneus (7) 0 – 45 55 5 42 0 – 45 55 5 42
–4 –56 75 6 42
SPG (7) 11 –56 80 5 3
19 – 49 75 6 5
SPG/transverse parietal sulcus (7) –4 –71 60 6 22
Limbic
Collateral sulcus –30 –8 –30 5 10
Hippocampus/collateral sulcus 26 –11 –25 5 23
Other
Cerebellum –4 – 82 –35 5 11
–38 –79 –25 6 19 26 – 86 –30 5 10
45 – 64 – 45 5 8
Mediodorsal thalamic nucleus 0 –11 9 5 3

a beneficial role in coupling the moment-to- probe tone ratings depend, partially, on the pitch 23. A. J. Blood, R. J. Zatorre, P. Bermudez, A. C. Evans,
moment perception of tonal space with cog- distribution statistics of the contexts that precede Nature Neurosci. 2, 382 (1999).
probes (33). 24. M. J. Tramo, J. J. Bharucha, F. E. Musiek, J. Cognit.
nitive, affective, and motoric associations, 8. P. Janata, J. Cognit. Neurosci. 7, 153 (1995). Neurosci. 2, 195 (1990).
which themselves may impose constraints on 9. M. Besson, F. Faı̈ta, J. Exp. Psychol. Hum. Percept. 25. I. Peretz, A. J. Blood, V. Penhune, R. Zatorre, Brain 124,
the activity patterns within rostral prefrontal Perf. 21, 1278 (1995). 928 (2001).
10. A. D. Patel, E. Gibson, J. Ratner, M. Besson, P. J. 26. J. V. Haxby et al., Science 293, 2425 (2001).
regions (21, 27–29). Holcomb, J. Cognit. Neurosci. 10, 717 (1998). 27. A. R. Damasio, Cognition 33, 25 (1989).
11. S. Koelsch, T. Gunter, A. D. Friederici, E. Schröger, J. 28. J. J. Eggermont, Neurosci. Biobehav. Rev. 22, 355
References and Notes Cognit. Neurosci. 12, 520 (2000). (1998).
1. R. N. Shepard, Psychol. Rev. 89, 305 (1982). 12. J. J. Bharucha, K. Stoeckig, J. Exp. Psychol. Hum. 29. S. Funahashi, Neurosci. Res. 39, 147 (2001).
2. C. L. Krumhansl, Cognitive Foundations of Musical Percept. Perf. 12, 403 (1986). 30. E. Bigand, B. Poulain, B. Tillmann, D. D’Adamo, J. Exp.
Pitch (Oxford Univ. Press, New York, 1990). 13. H. G. Tekman, J. J. Bharucha, J. Exp. Psychol. Hum. Psychol. Hum. Percept. Perf., in press.
3. F. Lerdahl, Tonal Pitch Space (Oxford Univ. Press, New Percept. Perf. 24, 252 (1998). 31. D. Huron, R. Parncutt, Psychomusicology 12, 154
York, 2001). 14. R. Francès, La Perception de la Musique (Vrin, Paris, (1993).
4. C. L. Krumhansl, E. J. Kessler, Psychol. Rev. 89, 334 1958).
32. M. Leman, Music Percept. 17, 481 (2000).
(1982). 15. B. Tillmann, J. J. Bharucha, E. Bigand, Psychol. Rev.
33. N. Oram, L. L. Cuddy, Psychol. Res. 57, 103 (1995).
5. L. B. Meyer, Emotion and Meaning in Music (Univ. of 107, 885 (2000).
34. We thank T. Laroche for assistance with data collec-
Chicago Press, Chicago, 1956). 16. R. J. Zatorre, A. C. Evans, E. Meyer, A. Gjedde, Science
tion. Supported by NIH grant P50 NS17778-18. The
6. The chromatic scale consists of 12 equally sized inter- 256, 846 (1992).
data and stimuli from the experiment are available
vals into which an octave is divided. On a piano, a 17. R. J. Zatorre, A. C. Evans, E. Meyer, J. Neurosci. 14,
on request from the fMRI Data Center at Dartmouth
chromatic scale starting at middle C would be played by 1908 (1994).
College (www.fmridc.org) under accession number
striking adjacent keys until the note C, either one octave 18. B. Maess, S. Koelsch, T. C. Gunter, A. D. Friederici,
2-2002-1139B.
above or below middle C, was reached. Nature Neurosci. 4, 540 (2001).
7. The extent to which tonality representations are 19. The existence of a tonal map that is distributed Supporting Online Material
maintained in long-term or short-term memory within and across cortical areas rather than focused www.sciencemag.org/cgi/content/full/298/5601/2167/
stores, or a combination of the two, is a matter of within a small cortical area may seem paradoxical, DC1
debate. Self-organizing neural network models of im- yet this representational form is predicted by some SOM Text
plicit learning accurately mimic results from a wide models of functional brain organization (27). Figs. S1 to S3
array of experiments that assess tonal knowledge 20. H. Barbas, Brain Res. Bull. 52, 319 (2000). Tables S1 and S2
(15), and harmonic priming experiments directly 21. D. Tranel, A. Bechara, A. R. Damasio, in The New References
highlight the influence of learned tonal structures Cognitive Neurosciences, M. S. Gazzaniga, Ed. (MIT Audio S1
(13, 30). However, models of short-term sensory Press, Cambridge, MA 2000), pp. 1047–1061.
memory account for significant proportions of the 22. H. Barbas, H. Ghashghaei, S. M. Dombrowski, N. L.
variance in probe-tone experiments (31, 32), and Rempel-Clower, J. Comp. Neurol. 410, 343 (1999). 17 July 2002; accepted 27 September 2002

2170 13 DECEMBER 2002 VOL 298 SCIENCE www.sciencemag.org