16 OCTOBER 2009 VOL 326 SCIENCE www.sciencemag.org
372
PERSPECTIVES
How does intention to
speak become the
action of speaking?
It involves the generation of
a preverbal message that is
tailored to the requirements
of a particular language, and
through a series of steps, the
message is transformed into
a linear sequence of speech
sounds ( 1, 2). These steps
include retrieving different
kinds of information from
memory (semantic, syntac-
tic, and phonological), and
combining them into larger
structures, a process called
unification. Despite general
agreement about the steps that
connect intention to articu-
lation, there is no consensus
about their temporal profi le or
the role of feedback from later
steps ( 3, 4). In addition, since
the discovery by the French
physician Pierre Paul Broca
(in 1865) of the role of the
left inferior frontal cortex in
speaking, relatively little prog-
ress has been made in under-
standing the neural infrastruc-
ture that supports speech pro-
duction ( 5). One reason is that
the characteristics of natural language are
uniquely human, and thus the neurobiology
of language lacks an adequate animal model.
But on page 445 of this issue, Sahin et al. ( 6)
demonstrate, by recording neuronal activity
in the human brain, that different kinds of
linguistic information are indeed sequentially
processed within Broca’s area.
Sahin et al. had the unique opportunity to
record from three patients with epilepsy dur-
ing presurgical preparation. Depth electrodes
were implanted in Broca’s area and the ante-
rior temporal cortex, and local fi eld poten-
tials were recorded while the patients were
engaged in a language production task. The
subjects were asked either to read silently
words presented on a screen, or to silently
produce the infl ected form of the presented
nouns and verbs in accordance with the
syntactic requirements imposed by a short
sentence fragment (e.g., Yesterday they…
walked). This latter process has two compo-
nents (see the fi gure). One is to determine the
correct tense of the target word and to gen-
erate (for regular infl ections) or retrieve (for
irregular infl ections) the correct mor phologi-
cal form. The other is the generation of the
concomitant phonological code and prepara-
tion for articulation.
Particularly in Broca’s area, more spe-
cifi cally Brodmann area 45, a clear triphasic
local fi eld potential response was observed.
At about 200 ms after presentation of the
word, word identifi cation had taken place,
with a stronger response for low-frequency
words than for high-frequency words. Mor-
phological composition and retrieval for
nouns and verbs happened at around 320 ms.
Finally, at about 450 ms, phonological encod-
ing had been completed. All these operations
were not only temporally separated, but also
spatially segregated at a scale of only a few
millimeters, which is below the effective spa-
tial resolution of standard functional mag-
netic resonance imaging of brain activity.
These data are relevant for both cognitive
models of speech production and for accounts
on the role of Broca’s area. The time course
is clear evidence for the sequentiality of dif-
ferent access and unifi cation operations in
speaking, and is consistent with the few esti-
mates in the literature ( 7, 8). Moreover, both
the anatomical and the temporal segregation
of word-encoding operations in Broca’s area
are in line with the view that this region is
involved with each of these encoding oper-
ations and their unifi cation over time. Feed-
back operations among these processes can-
not be excluded. However, the fi ne-grained
temporal and spatial separation of these steps
suggests that we are witnessing the “fi rst go”
process at work here.
Both functional magnetic resonance
imaging and lesion studies have shown that
Broca’s area is also involved in processing
infl ectional morphology during comprehen-
sion ( 9). In combination with the fi ndings of
Sahin et al., this suggests that Broca’s area is
recruited during both language production
and comprehension. Whether these recruit-
ments can be separated at the scale of the
microcircuitry within Broca’s area remains
to be seen.
Broca’s area has been proposed to have a
more specialized role in language process-
ing—facilitating linguistically motivated
operations of syntactic movement ( 10) and
processing hierarchical structures ( 11). The
The Speaking Brain
NEUROSCIENCE
Peter Hagoort
1, 2 and Willem J. M. Levelt
1
Recordings of electrical activity in the human
brain reveal the fi ne-tuned, stepwise neuronal
processing of language and speech.
Times at which
Broca’s area
contributes to
the different
processing steps
Feedback for self-monitoring
Broca’s area
Lexical concept
Lemma selection
Lemma
Retrieving
morphemic codes
Phonological code
Phonological and
phonetic encoding
Articulatory score
Word recognition
(visual, auditory)
~200 ms
~320 ms
~450 ms
From intention to articulation. Shown is an adapted
version of the lexical encoding model for speech pro-
duction ( 2), specifying steps in the paradigm used
by Sahin et al. Based on the visual input, a lemma
is selected that specifi es the syntactic features of a
lexical concept. For instance, for the lemma horse, it
specifi es that it is a count noun. In addition, the mor-
phemic codes are retrieved. For instance, when the
speaker wants to produce the plural form of horse,
the codes for both the stem and the plural suffi x are
retrieved. Next, the phonological codes for each mor-
pheme are retrieved, combined, and transformed
into a motor command to the articulatory system. The
approximate times (in milliseconds) at which Broca’s
area contributes to the different processing steps are
shown. The late (i.e., at 500 to 600 ms) monophasic
component observed in the temporal lobe ( 6) might
refl ect self-monitoring of the speech output.
CREDIT: Y. GREENMAN/SCIENCE
1Max Planck Institute for Psycholinguistics, NL-6500 HB
Nijmegen, Netherlands. 2Donders Institute for Brain, Cog-
nition and Behaviour, Radboud University Nijmegen, Neth-
erlands. E-mail: [email protected]
Published by AAAS
on October 16, 2009 www.sciencemag.orgDownloaded from
