asymmetries started in 1861, when French
anatomist Pierre Paul Broca astounded his
Parisian colleagues with the observation
that speech abilities are perturbed after le-
sions in the left, but not right, brain hemi-
sphere ( 5 ). This seminal insight not only
ended the long-held view of symmetrical
brain organization; it also launched the
quest for asymmetries in other cognitive
domains. One field that lent itself particu-
larly well to comparison was music; it has
numerous structural parallels to speech ( 6 ),
and deficits in one but not the other domain
were shown to have developed in several
neurological patients (although shared defi-
cits in both domains were also reported) ( 7 ).
What was missing for a long time was a
cogent theory that could explain why speech
and music should show a divergent hemi-
spheric lateralization. Today, influential
neuroacoustic models ( 2 , 3 ) seek reasons in
the specific computational requirements im-
posed by the structure of speech and musical
sounds. For example, proper speech percep-
tion hinges strongly (but not solely) on the
ability to process short-lived temporal modu-
lations that are decisive for discriminating
similar-sounding words, such as “bear” from
“pear.” By contrast, proper music perceptionrequires, among others, the ability to process
the detailed spectral composition of sounds
(frequency fluctuations). Biophysical laws
constrain the simultaneous extraction of
both fast temporal and fine spectral infor-
mation. The relative specialization of the left
and right auditory cortices to better process
one or the other cue appears to be an effi-
cient solution to this dilemma. What may be
a plausible computational explanation for au-
ditory asymmetries of speech and music has,
however, been vigorously contested ( 8 ), and
although there is empirical support, it is of-
ten confined to synthetic stimuli ( 9 ) or acous-
tically manipulated speech ( 10 ), not music.
The study of Albouy et al. presents a ma-
jor step forward in this debate by demon-
strating direct links among (i) the amount
of temporal or spectral detail in the sound
structure of songs, (ii) listeners’ compre-
hension of the verbal or melodic content
of these songs, and (iii) corresponding left-
right asymmetries in the anterior auditory
cortex known to represent complex sound
categories ( 11 ). Applying special temporal
and spectral filters to recordings of 100
carefully constructed unaccompanied (a
cappella) songs, the authors established
that the stepwise removal of temporal de-
tail impairs recognition of the words (but
not recognition of the melodies), whereas
the degradation of spectral detail weakens
recognition of the melodies (but not rec-
ognition of the words). They used machine
learning to identify patterns of brain ac-
tivity (determined by functional magnetic
resonance imaging) that correlated with the
words or the melody being heard by partici-
pants. This approach was most successful
in decoding the words from neural patterns
in left (but not right) auditory regions, par-
ticularly for songs with high temporal de-
tail. The decoding of the melodies was most
accurate from neural patterns in right (but
not left) auditory regions and scaled with
the spectral detail of the songs. The behav-
ioral (mis)understanding of words or melo-
dies was mirrored in the lateralized neural
activity patterns, which suggests that the
computations carried out in left or right au-
ditory regions on opposite acoustic features
are relevant for participants’ split percep-
tion of words and melodies in songs.
The conclusions drawn from the find-
ings of Albouy et al. make it tempting to
think of speech and music as two cogni-
tive domains with sharp borders. However,
this dichotomy should be taken with cau-
tion. Both speech and music have multiple
acoustic and cognitive facets ( 8 ), many
of which are shared by both domains:SCIENCE sciencemag.org 28 FEBRUARY 2020 • VOL 367 ISSUE 6481 975
The melody and words in music are processed
asymmetrically in the human brain.Max Planck Institute for Human Cognitive and Brain
Sciences, Leipzig, Germany. Email: [email protected]
Published by AAAS