See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/7784335

Brain Mechanisms Linking Language and Action

Article  in  Nature reviews Neuroscience · August 2005

DOI: 10.1038/nrn1706 · Source: PubMed

CITATIONS READS

1,008 1,031

1 author:

Friedemann Pulvermüller
Freie Universität Berlin
300 PUBLICATIONS   18,070 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Modelling the cortical mechanisms underlying language processes in the brain View project

MEG biomarkers of incipient dementia View project

All content following this page was uploaded by Friedemann Pulvermüller on 02 June 2014.

The user has requested enhancement of the downloaded file.

PERSPECTIVES

Taking into account well-known facts

OPINION
about the cortical basis of learning — namely
that frequently co-activated neurons

Brain mechanisms linking strengthen their mutual connections — it is

likely that the cortical systems for language
and action develop specific links between
language and action each other whenever actions correlate with
specific language processes. From this we can
predict that whenever language and action
Friedemann Pulvermüller information processing correlate with each
other in the different cortical areas, distrib-
Abstract | For a long time the cortical have been interpreted as strong evidence uted functional systems are being established
systems for language and actions were for a strict modular organization of both that allow for fast, interactive processing of
believed to be independent modules. the language and the action systems. The multimodal information across cortical
However, as these systems are reciprocally conceptual roots of local encapsulated mod- areas. The existence of distributed interac-
connected with each other, information ules date back to the functional centres of tive systems has been proposed by several
about language and actions might interact nineteenth century connectionist models3, researchers who have called them cell
in distributed neuronal assemblies. A and the results of many brain imaging stud- assemblies5, neuronal ensembles6,8, distrib-
critical case is that of action words that ies are largely consistent with the modular uted functional networks7, and, if they have
are semantically related to different parts perspective4. specific cognitive functions, neurocognitive
of the body (for example, ‘lick’, ‘pick’ and Modern theoretical perspectives offer a networks9 or cognits10.
‘kick’): does the comprehension of these different view: cortical functions might be
words specifically, rapidly and automatically served by distributed interactive functional Dissociations and distributed systems
activate the motor system in a somatotopic systems rather than local encapsulated For some time, especially in the realm of
manner, and does their comprehension rely modules5–10. Evidence for this idea has been language, modular models were preferred
on activity in the action system? provided by neuroanatomical investigations because it was thought that they were neces-
that have shown neuronal connections both sary for explaining neuropsychological double
The cortical systems for language and within and between the cortical systems for dissociations. An example of a double disso-
action control were traditionally thought action and the homologues of the human ciation is the predominant loss of speech that
to be paradigmatic examples of independ- language system in monkeys. For exam- coincides with relatively intact comprehension
ent and autonomous functional systems or ple, there are links between the dorsal and in patients with Broca’s aphasia and the reverse
modules1,2. These systems have different ventral premotor cortex and between the pattern of comprehension deficits with fluent
cortical bases in circumscribed areas (motor left inferior frontal (Broca’s area) and supe- speech output in Wernicke’s aphasia. Double
and premotor cortex versus left perisylvian rior temporal (Wernicke’s area) language dissociations also occur in the processing of
‘language regions’), are fully dissociable by regions11,12. Importantly, many links have conceptual and lexico–semantic word catego-
neurological disease (paralysis and apraxic also been shown between the premotor and ries; for example, meaningful content versus
action deficits versus aphasic language defi- language areas where they are adjoined, in grammatical function words16, nouns versus
cits) and can themselves be subdivided into the inferior frontal cortex, and through long- verbs17,18 or animal versus tool names19, with
finer functional systems — for movement distance cortico-cortical connections11–15 the complementary category being relatively
of different body parts and linguistic func- (FIG. 1b). The dorsal and ventral prefrontal spared. Therefore, local encapsulated modules
tions of different types (for example, speech and premotor cortex connects with auditory for speech production, comprehension and
production versus comprehension, or pho- areas in the belt and parabelt region of the the processing of specific lexico–semantic
nology versus syntax versus semantics). superior temporal gyrus and sulcus, thereby categories were proposed2,19.
The specific inability of patients who have providing multiple links between the supe- However, this line of argument was weak-
had a stroke to move one extremity while rior temporal language area and the motor ened as it became clear that double disso-
all other motor and language functions system. These links indicate that information ciations can also be explained by accounts of
remain relatively intact, or the predomi- flow is possible between the cortical systems distributed interactive systems2,20,21. If they are
nant loss of usage of one category of words for language and action. structured according to the neuroanatomical

576 | JULY 2005 | VOLUME 6 www.nature.com/reviews/neuro

© 2005 Nature Publishing Group
 PERSPECTIVES

a That the links between superior temporal

Leg
perceptual circuits and frontocentral speech
Arm M1 production machinery are functionally
Face effective has been shown using transcranial
PMC magnetic stimulation (TMS). During the
PFC
perception of spoken words and language
sounds (phonemes) that strongly involve the
b tongue, TMS applied to the inferior motor
cortex elicits stronger muscular responses
BPT BPO of the articulators compared with control
A1 conditions33. Interestingly, this effect is most
WB prominent when the critical phonemes are
WPB presented in a meaningful word context,
which indicates that cell assemblies for
meaningful words have a role in linking
articulatory gestures and auditory signals
at the cortical level22. Converging evidence
c Leg-related words Arm-related words Face-related words from functional connectivity studies based
on positron emission tomography (PET)
and functional MRI (fMRI) data indicates
that the links between superior temporal
and inferior frontal language areas depend
on the amount of meaningful information
being transmitted by words34.
Although the documented tight functional
Figure 1 | Schematic illustration of the cortical systems for language and action. a | Somatotopy links between action and perception circuits
of the motor and premotor cortex: the approximate location of the face/articulators, arm/hand and foot/ of the left perisylvian language array20,35 can-
leg representations. b | Connections between the language and action systems. Inferences about cortico- not be explained in a straightforward man-
cortical links in humans are based on neuroanatomical studies in monkeys. The arrows indicate long- ner within a modular approach, they meet
distance cortico-cortical links. c | Semantic somatotopy model of action word processing: distributed
neuronal assemblies bind information about word forms and the actions to which they are semantically
the predictions of a distributed interaction
linked. Because action words can relate to different body parts (for example, ‘lick’, ‘pick’ or ‘kick’), the model that has been used to postulate senso-
cortical distributions of their neurocognitive networks differ. A1, core region of the primary auditory cortex; rimotor links and binding between specific
BPO, Broca’s area, pars opercularis; BPT, Broca’s area, pars triangularis; M1, primary motor cortex; acoustic speech patterns and the articulatory
PFC, prefrontal cortex, posterior part adjacent to motor system; PMC, premotor cortex; WB, auditory belt gestures that generate them. However, these
region in Wernicke’s area; WPB, auditory parabelt region in Wernicke’s area. Data from REFS 1115,20,36. results do not answer the question of the
Panel c modified, with permission, from REF. 83 © (2001) Elsevier Science.
universality and functional specificity of the
sensorimotor links: are there one or more
specific cortical areas dedicated to storing
connectivity of the left perisylvian language active, as are the superior temporal areas in and processing different neuronal ensem-
areas, such models can be used to explain the the vicinity of the auditory cortex25–27 (FIG. 2). bles that bind information about actions and
emergence of double dissociations that occur During speaking, the superior temporal cor- language? Or is it a general property of the
between aphasia syndromes as a result of focal tex is active, as are areas in the inferior motor, language and action systems, and possibly
cortical damage22. In principle, any double premotor and prefrontal cortex, even when the entire cortex, to host topographically
dissociation between categories of knowl- self-produced sounds are prevented from specific and dissociable distributed networks
edge that can be attributed to separate local being perceived through the auditory chan- that process linguistic information together
encapsulated modules can also be explained nel28. This indicates that interactive neural with motor programs? Such questions can
by distributed functional systems with spe- systems that are distributed over the infe- be addressed experimentally by using words
cific topographies that involve neurons in the rior frontal and superior temporal cortex that are semantically related to actions.
critical areas to different degrees20,23,24. contribute to both speech production and
Imaging studies have been used to perception. During spoken word recogni- Semantic somatotopy of action words
directly address the question of whether the tion and understanding, these systems are Model and predictions. Action words are
left hemispheric inferior frontal and supe- activated near-simultaneously and largely in defined by abstract semantic links between
rior temporal language areas are modules parallel, with a peak activation delay in the language elements and motor programs. In
specialized in either speech perception or inferior frontal cortex of ~20 ms after the infancy, action words are learned in the con-
production, or whether they represent two activation peak in superior temporal areas26. text of action performance. A child performs
local areas that house neural elements par- These results indicate tight and rapid func- an action and, in close temporal vicinity,
ticipating in interactive distributed cortical tional links between speech perception and the caretaker typically uses the action word
processes that contribute to both speech speech production processes, as postulated (usually a verb). In the cortex, the motor
production and comprehension. Such studies by established psycholinguistic theories, program and the neural representation
have shown that while listening to syllables including the motor29,30 and direct realist31,32 of the word are therefore activated near-
and words, the left inferior frontal cortex is theories of speech perception. simultaneously, so that synaptic connections

NATURE REVIEWS | NEUROSCIENCE VOLUME 6 | JULY 2005 | 577

© 2005 Nature Publishing Group
PERSPECTIVES

a b lies in the premotor cortex in the lateral

Z = 46 precentral gyrus and resembles the map in
the primary motor cortex14,38. If action words
are semantically related to the movements of
the face or articulators, arm or hand, or leg
or foot, the distributed neuronal ensembles
80 90 100 110 120
would include semantic neurons in peri-
3
sylvian (face-related words), lateral (arm-
related words) or dorsal (leg-related words)
nAm motor and premotor cortex20. Therefore, this
semantic somatotopy model of action words
0
implies that there are differently distributed
130 140 150 160 170
Time (ms) networks for the English words ‘lick’, ‘pick’
Figure 2 | Functional links between the superior temporal speech perception and inferior frontal and ‘kick’ (FIG. 1d). The model allows general
motor systems. Links between speech perception and motor systems are demonstrated by functional predictions to be made about action word-
imaging studies that show the activation of frontal areas in response to speech input. related cortical activity within the limits of
a | Magnetoencephalography data (n = 12) showing that superior temporal activation elicited by spoken
the well-known inter-individual variation
words is immediately followed by inferior frontal activation (time difference = 22 ms). Source strengths are
given in nanoampère-metres (nAm). b | Functional MRI data from one representative participant that show of cortical maps, most notably as a result of
premotor activation while listening to speech. The activated motor circuits might contribute to the speech practice-related reorganization39. This is open
perception and comprehension process, as proposed by psycholinguistic theories. Panel a modified, with to further elaboration, taking into account
permission, from REF. 26 © (2003) Elsevier Science. Panel b modified, with permission, from REF. 27 © additional mapping rules — for example,
(2004) Macmillan Magazines Ltd. the topography of coordinated actions in a
body-centred workspace that was suggested
by the authors of recent work40.
between neurons in specific motor and pre- parts. The motor cortex has a somatotopic Crucial predictions about the semantic
motor areas and those in the language areas organization (FIG. 1c), with the mouth and somatotopy model are as follows. First, the
become stronger. This would be expected articulators represented close to the Sylvian perception of spoken and written action
to lead to neuronal ensembles with specific fissure, the arms and hands at dorsolateral words activates cortical areas involved in
cortical distributions. sites and the feet and legs projected to the action control and execution in a category-
The postulated specific cortical topog- vertex and interhemispheric sulcus36. There specific somatotopic fashion, which depends
raphies are best illustrated using the case are additional somatotopic maps in the fronto- on the semantics of the action words. Second,
of action words that refer to different body central cortex37, a prominent one of which owing to the internal connections of the
distributed neuronal ensembles through
fast-conducting axons, the spread of activ-
a Movement b Passive reading of action words
ity is fast, so that specific sensorimotor areas
are activated early in the course of spoken
and written word comprehension. Third,
because of the strong within-assembly con-
nections that link language and action rep-
resentations, activation of the sensorimotor
cortex should not require people to attend
to language stimuli, but should instead be
automatic. And, finally, functional changes
in the motor and premotor cortex influence
the processing of action words in a category-
specific manner.

Experimental evidence. In functional imag-

ing experiments, elementary repetitive move-
ments of single body parts activate the motor
and premotor cortex. For example, Hauk et al.
reported fMRI data showing that tongue,
finger and foot movements lead to the soma-
Foot Finger Tongue Leg-related Arm-related Face-related totopic activation pattern illustrated in FIG. 3.
movements movements movements words words words
When the same participants were instructed
Figure 3 | Cortical activation during movement and during passive reading of action words. to silently read action words that related to
Functional MRI data (n = 14) showing that a degree of overlap in activation is elicited by corresponding
the face, arm and leg that were otherwise
actions (a) and action words (b). Leg-related words activate areas overlapping and adjacent to areas
involved in foot movements, and there are similar relationships between arm-related words and finger
matched for important psycholinguistic
movements and face-related words and tongue movements. This indicates that a common neural variables (such as word frequency, length and
substrate is involved in the processing of actions and the meaning of action words. Modified, with imageability), a similar pattern of activation
permission, from REF 41 © (2004) Elsevier Science. emerged along the motor strip41. Consistent

578 | JULY 2005 | VOLUME 6 www.nature.com/reviews/neuro

© 2005 Nature Publishing Group
 PERSPECTIVES

Face/arm-related word Leg-related word between two interpretations of this finding. words ‘hotki’ (eat) and ‘potki’ (kick) — which
Action word recognition could either auto- included the same recording of the syllable
matically and immediately trigger the acti- [kI] spliced to the end of each word’s first
vation of specific action-related networks, as syllable — were compared. In this way, any
predicted by the semantic somatotopy model, differential activation elicited by the final syl-
or motor activation could be the consequence lable [kI] in the context of [hot] or [pot] could
110 ms
of a late, postlexical strategy to imagine or be uniquely attributed to its lexico–semantic
plan an action. Earlier fMRI research showed context. Magnetoencephalography (MEG)
that the observation of action-related pic- results showed that an early brain response,
tures, and also mere mental imagery of the mismatch negativity50, was elicited by
actions, activates the premotor cortex in a face/arm- and leg-related word contexts
140 ms somatotopic fashion43. Neurophysiological (FIG. 4). Relatively stronger activation was
experiments were conducted to reveal the present in the left inferior frontal cortex
time course of cortical activation in action for the face/arm-related word, but signifi-
word recognition, and to investigate whether cantly stronger activation was seen in supe-
specific motor areas are activated immedi- rior central areas, close to the cortical leg
ately or after some delay. Late postlexical representation, for the leg-related word49.
170 ms meaning-related processes are reflected These MEG results were confirmed with
Figure 4 | Cortical activation dynamics elicited by late components of the event-related a different method, electroencephalography
by face/arm- and leg-related words. potential (ERP) and field, which are maximal (EEG), using words from different languages,
Magnetoencephalography (n = 16) was used to ~400 ms after word onset44. However, neuro- including, for example, the English word pair
measure the activation elicited at different times
after spoken action words could be uniquely
physiological differences between word cat- ‘pick’ versus ‘kick’48. It is remarkable that the
recognized. Note the slight upward movement of egories that reflect lexico–semantic processes activation peak of the superior central source
the inferior central source for the face/arm-related have been found as early as 100–200 ms45. followed that of the inferior frontal source
word and the delayed appearance of the superior ERP experiments that focused on silent with an average delay of only 30 ms, which is
central source for the leg-related word. These reading of the face-, arm- and leg-related consistent with the spread of activation being
activation time courses might reflect the words used in the fMRI study by Hauk et al. mediated by fast-conducting cortico-corti-
movement of neuronal activity in distributed
showed that category-specific differential cal fibres between the perisylvian and dorsal
neuronal assemblies that represent and process
words with different action-related meanings. activation was present ~200 ms after word sensorimotor cortex.
Modified, with permission, from REF. 49 © (2005) onset46,47. Consistent with the fMRI results, Even if action word processing activates
MIT Press. distributed source localization performed on the motor system in a specific somatotopic
stimulus-triggered ERPs revealed an inferior fashion, this does not necessarily imply that
frontal source that was strongest for face- the motor and premotor cortex influence
with earlier findings, all words equally acti- related words, and a superior central source the processing of action words. In another
vated areas in the temporal cortex and also that was maximal for leg-related items47. study, different parts of the motor system
the inferior frontal cortex25–27. The additional This dissociation in brain activity patterns were stimulated with weak magnetic pulses
category-specific somatotopic activation that supports the idea of stimulus-triggered early while participants processed action words
was seen in the motor system in response to lexico–semantic processes. To investigate in a lexical decision task 51. To minimize
face-, arm- and leg-related words was close to whether motor preparation processes co- interference between word-related activa-
and overlapped with the motor and premotor determined this effect, experiments were tion of the motor system and response
representations of the tongue, fingers and carried out in which the same response — a execution processes, lip movements were
feet, respectively. These results indicate that button press with the left index finger — was required while arm- and leg-related words
specific action representations are activated required for all words. The early activation were presented. When subthreshold TMS
during action word understanding. difference between face- and leg-related was applied to the arm representation in the
A similar experiment was carried out words persisted, which indicates that lexico– left hemisphere, causing strong magnetic
with action words embedded in sentences. semantic processes, rather than postlexical pulses to elicit muscle contractions in the
In this case, participants heard action motor preparation, were reflected46. right hand, faster processing of arm-related
descriptions such as ‘the boy kicked the The earliness of word category-specific words occurred relative to leg-related words.
ball’ or ‘the man wrote the letter’ while their semantic activation in the sensorimotor When TMS was applied to the cortical leg
brain metabolism was monitored42. Specific cortex in passive reading tasks indicates area, the opposite pattern of faster leg-related
premotor areas reflecting the differential that this feature might be automatic. To than arm-related word responses emerged51.
involvement of body part information in the further investigate this possibility, partici- Processing speed did not differ between
semantic analysis of the language input were pants were actively distracted while action stimulus word groups in control conditions in
again found to be active. Taken together, words were being presented, and their brain which ineffective ‘sham’ stimulation or TMS
these fMRI results indicate that somatotopic responses were measured48,49. Participants to the right hemisphere was applied. This
activation of motor circuits reflects aspects were instructed to watch a silent video film shows a specific influence of activity in the
of word and sentence meaning. and ignore the language input while spoken motor system in response to the processing
Although these studies showed language- face/arm- and leg-related action words were of action-related words (FIG. 5).
related somatotopic cortical activation, the presented. Care was taken to exactly control Further evidence for specific functional
low temporal resolution of haemodynamic for physical and psycholinguistic features of links between the cortical language and
imaging makes it impossible to decide the word material. For example, the Finnish action systems comes from TMS-induced

NATURE REVIEWS | NEUROSCIENCE VOLUME 6 | JULY 2005 | 579

© 2005 Nature Publishing Group
PERSPECTIVES

Arm-related words some of the classic dissociations — for

Leg-related words
example, between nouns and verbs or tool
and animal names — are open to alternative
explanations in terms of psycholinguistic
variables (such as word frequency and
imageability)58 or conceptual–perceptual
Arm Leg Leg Arm structure (high/low similarity between
members of semantic categories)59–61. In light
of this discussion, the action word experi-
ments are of great theoretical relevance. In
620 620 620 these cases, confounding psycholinguistic
600 600 600
factors could be excluded owing to meticu-
lous stimulus matching, and the supported
Response time (ms)

Response time (ms)

Response time (ms)

580 580 580
a priori prediction of semantic somatotopy
560 560 560 of motor and premotor areas leaves little
540 540 540
room for explanations that deny the link
between activated areas and motor function.
520 520 520 Somatotopic semantic activation cannot be
500 500 500 explained by a unitary semantic–conceptual
480 480 480
system that processes all word meanings in
Arm site Leg site Arm site Leg site the same cortical locus.
TMS to left hemisphere TMS to right hemisphere Sham stimulation
Figure 5 | Effects of transcranial magnetic stimulation of cortical motor areas on action word One semantic–conceptual binding site?
processing. Transcranial magnetic stimulation (TMS; n = 11) was applied to arm and leg loci over the left Although the results discussed here cannot
and right hemispheres and compared with sham control stimulation during the processing of arm- and be explained if all semantic processes are
leg-related words. Response times to arm- versus leg-related words were differentially affected only restricted to one cortical area, they might
during the application of TMS to the left hemisphere. Those parts of the motor systems that reflect
be compatible with the idea of a central
aspects of the semantic meaning of action words might, therefore, make a crucial contribution to the
processing of these words. Reproduced, with permission, from REF. 51 © (2005) Blackwell Publishing.
semantic system, if this system is thought to
manage dynamic functional links between
various cortical areas that process word
motor responses33. Listening to Italian sen- a word seem to determine the cortical distri- forms and conceptual–semantic information.
tences describing actions performed with bution of the neuronal network that the word Apart from the motor system, action words
the arm or leg differentially modulates the activates. During action word processing, have also been shown to activate inferior
motor responses brought about by mag- the language regions and motor cortex are temporal areas (FIG. 3), which makes these
netic stimulation of the hand and leg motor activated in parallel, following finely-tuned data compatible with a semantic role for this
cortex52. It seems that effective specific con- spatio–temporal patterns, and activity pat- region4,53. In the study by Hauk et al., the
nections of language and action systems can terns in motor and language systems seem most pronounced activation elicited by all
be documented for spoken or written lan- to interact, as shown by TMS experiments. action words was in the inferior frontal cor-
guage, at the word and sentence levels, and So, these results support a distributed inter- tex, close to the human homologue of mon-
for various languages (for example, English, active systems account in general and, more key area F5, where mirror neurons are most
Italian, German and Finnish) using various particularly, a distributed model of semantic common35. This is in line with the idea that
neuroscientific methods (fMRI, MEG, EEG somatotopy of action word processing, and this inferior frontal area has a key role in the
and TMS). might have further implications for our semantic binding of action, perception and
understanding of cognitive brain functions language-related information35. The debate
Implications for semantics in the brain BOX 1. about one or many cortical systems control-
These results show that action words activate The results with action words show that ling semantic processing must, therefore,
the cortical system for action processing in semantic processing can engage many cor- continue, although the data reviewed here on
a somatotopic fashion and that this soma- tical areas. They contradict the view that action words show topographically-specific
totopy reflects referential word meaning. meaning processing is localized in a unitary activation in different areas that both reflect
However, they do not imply that all aspects cortical locus — for example an area that is and influence lexico–semantic processes.
of the meaning of a word are necessarily anterior, inferior or posterior to Wernicke’s The bilateral nature of neural degeneration
reflected in the brain activation pattern that area in the left temporal lobe4,53,54 or in the that is usually seen in patients with semantic
it elicits. It seems that such cortical–semantic inferior frontal cortex55. Category-specific dementia might indicate that one focal lesion
correspondence can be postulated for words semantic systems in a range of cortical areas is insufficient to cause general semantic defi-
that refer to concrete entities related to action that normally process sensory or action cits53. The existence of many semantic binding
or perception patterns. It remains to be deter- information have previously been postulated sites is also supported by the specific semantic
mined whether it might be possible to read on the basis of neuropsychological18,19,24 and deficit in action word processing that is seen
aspects of the meaning of other words, such neurophysiological double dissociations56,57 in patients with motor neuron disease62, and
as abstract items, from the cortex in a similar between words and concepts that relate to by double dissociations between semantic
manner. The cortical systems that process actions and objects. However, this idea has word categories that arise from lesions in
information about the referential meaning of been challenged by the observation that right-hemispheric fronto-parietal versus

580 | JULY 2005 | VOLUME 6 www.nature.com/reviews/neuro

© 2005 Nature Publishing Group
 PERSPECTIVES

Box 1 | Action in language comprehension Friedemann Pulvermüller is at the Medical

Research Council Cognition and Brain Sciences
The results of investigations into the brain connections between language and action systems Unit, Cambridge CB2 2EF, UK.
reviewed in the main text might change the way we think about language and meaning. e-mail:
[email protected]
Hearing a word seems to be associated with activation of its articulatory motor program, and
doi:1038/nrn1706
understanding an action word seems to lead to the immediate and automatic thought of the
Published online 15 June 2005
action to which it refers. A tight functional link between speech production and perception
1. Fodor, J. A. The Modularity of Mind (MIT Press,
mechanisms has been proposed through psycholinguistic theories — for example, the Cambridge, Massachusetts, USA, 1983).
motor30,65 and the direct realist31,32 theories of speech perception. A neurobiological account 2. Shallice, T. From Neuropsychology to Mental Structure
(Cambridge Univ. Press, New York, USA, 1988).
relates specific strengthening of neuronal links between inferior frontal speech motor 3. Lichtheim, L. On aphasia. Brain 7, 433–484 (1885).
circuits and the superior temporal speech perception machinery to correlations between 4. Price, C. J. The anatomy of language: contributions from
functional neuroimaging. J. Anat. 197, 335–359 (2000).
perceptual and motor information that give rise to distributed cell assemblies that contribute
5. Hebb, D. O. The Organization of Behavior. A
to both production and comprehension of speech22,66. The articulatory perception–action Neuropsychological Theory (John Wiley, New York, USA,
loops might also be important as a cortical basis of short-term verbal memory67. In the same 1949).
6. Palm, G. Neural Assemblies (Springer, Berlin, Germany,
way that the mirror neuron theory implies that action understanding arises from an 1982).
association between the perceived actions of others and one’s own action control system67–72, 7. Roland, P. E. & Gulyas, B. Visual imagery and visual
representation. Trends Neurosci. 17, 281–287;
the comprehension of action-related language seems to require that words are mapped onto discussion 287–294, 514–516 (1994).
actions that one can perform oneself 57,73–76. This grounding of language in action is essential, 8. Braitenberg, V. & Schüz, A. Cortex: Statistics and
as comprehension of action-related language and concepts is impaired after lesions of the Geometry of Neuronal Connectivity (Springer, Berlin,
Germany, 1998).
motor system62,63. 9. Mesulam, M. M. From sensation to cognition. Brain 121,
Action meaning seems to be not only necessary, but also highly relevant for language. Verbs 1013–1052 (1998).
10. Fuster, J. M. Cortex and Mind: Unifying Cognition (Oxford
form the grammatical backbone of sentences, and the majority explicitly refer to actions. Univ. Press, Oxford, UK, 2003).
Furthermore, there are several classes of words without action reference that are, nevertheless, 11. Makris, N. et al. MRI-based topographic parcellation of
human cerebral white matter and nuclei II. Rationale and
semantically linked to actions. Tool words, for example, relate to actions for which the tools are applications with systematics of cerebral connectivity.
made56,77, and words that denote internal states, such as ‘pain’ or ‘disgust’, can be understood Neuroimage 9, 18–45 (1999).
only because both speaker and listener can relate them to similar motor programs that are, by 12. Pandya, D. N. & Yeterian, E. H. in Cerebral Cortex. Vol. 4
(eds Peters, A. & Jones, E. G.) 3–61 (Plenum, London,
genetic endowment, associated with the expression of pain or disgust78. Understanding UK, 1985).
language means relating language to one’s own actions, possibly because the automatic and 13. Romanski, L. M. et al. Dual streams of auditory afferents
target multiple domains in the primate prefrontal cortex.
extremely rapid linkage of sensory and motor information in our brains benefits Nature Neurosci. 2, 1131–1136 (1999).
comprehension and learning processes29,79. These insights, which are supported by 14. Rizzolatti, G. & Luppino, G. The cortical motor system.
neuroscience research, have important implications for constructing life-like perception– Neuron 31, 889–901 (2001).
15. Young, M. P., Scannell, J. W., Burns, G. & Blakemore, C.
action systems and robots with brain-like control systems72,80–82. Analysis of connectivity: neural systems in the cerebral
cortex. Rev. Neurosci. 5, 227–249 (1994).
16. Gardner, H. & Zurif, E. Bee but not Be: oral reading of
single words in aphasia and alexia. Neuropsychologia 13,
181–190 (1975).
temporo-occipital areas63. Some such lesions From here, various questions emerge. 17. Miceli, G., Silveri, M., Villa, G. & Caramazza, A. On the
basis of agrammatics’ difficulty in producing main verbs.
are so focal that they affect only the motor Does the semantic brain topography of Cortex 20, 207–220 (1984).
and premotor cortex, but, nevertheless, action words persist if they are embedded 18. Daniele, A., Giustolisi, L., Silveri, M. C., Colosimo, C. &
Gainotti, G. Evidence for a possible neuroanatomical
specifically degrade the processing of action in sentences with idiomatic or metaphoric basis for lexical processing of nouns and verbs.
words in psychological experiments. These expressions in which any action relationship Neuropsychologia 32, 1325–1341 (1994).
19. Warrington, E. K. & McCarthy, R. A. Categories of
dissociations indicate that there is more than is lost (‘pull my arm versus leg’)? Would the knowledge: further fractionations and an attempted
one semantic integration system in both semantic topography be influenced by the integration. Brain 110, 1273–1296 (1987).
20. Pulvermüller, F. Words in the brain’s language. Behav.
cerebral hemispheres64. grammatical status of words as either nouns Brain Sci. 22, 253–336 (1999).
or verbs? Can the time course of semantic 21. Plaut, D. C. & Shallice, T. Deep dyslexia: a case study of
connectionist neuropsychology. Cognit. Neuropsychol.
Outlook and lexical activation be correlated with spe- 10, 377–500 (1993).
Investigations of words and sentences that cific features of meaning processing, so that 22. Pulvermüller, F. & Preissl, H. A cell assembly model of
language. Network Comput. Neural Syst. 2, 455–468
are semantically related to actions involv- aspects of the comprehension process could (1991).
ing the face, arm or leg show somatotopic be read from the spatio–temporal pattern 23. Kiefer, M. & Spitzer, M. The limits of a distributed account
of conceptual knowledge. Trends Cogn. Sci. 5, 469–471
activation of the sensorimotor cortex. of brain activation? What are the reliability (2001).
Activation is rapid and largely independent limits of investigations of lexico–semantic 24. Humphreys, G. W. & Forde, E. M. Hierarchies, similarity,
of attention. Action-related word meaning brain processes in individual participants? and interactivity in object recognition: ‘category-specific’
neuropsychological deficits. Behav. Brain Sci. 24,
is not only reflected in the cortical activity Is there a perspective on monitoring 453–509 (2001).
pattern, but stimulation of the motor system these processes in particular neurological 25. Zatorre, R. J., Evans, A. C., Meyer, E. & Gjedde, A.
Lateralization of phonetic and pitch discrimination in
also produces differential effects on the rec- populations, such as elderly patients with speech processing. Science 256, 846–849 (1992).
ognition of action words of different seman- degenerative brain disease? Finally, can the 26. Pulvermüller, F., Shtyrov, Y. & Ilmoniemi, R. J. Spatio-
temporal patterns of neural language processing: an
tic type. These results are best explained by learning of word meaning be traced with MEG study using minimum-norm current estimates.
distributed interactive neuronal assemblies, large-scale neurophysiological techniques? Neuroimage 20, 1020–1025 (2003).
27. Wilson, S. M., Saygin, A. P., Sereno, M. I. & Iacoboni, M.
the cortical topographies of which reflect The brain connections between language Listening to speech activates motor areas involved in
the loci where relevant semantic and form- and action systems might provide a fruitful speech production. Nature Neurosci. 7, 701–702
(2004).
related information is being processed. model for cognitive neuroscience, not least 28. Paus, T., Perry, D. W., Zatorre, R. J., Worsley, K. J. &
These differentially distributed cell assem- because these systems are ideally placed for Evans, A. C. Modulation of cerebral blood flow in the
human auditory cortex during speech: role of motor-to-
blies might bind multimodal lexico–semantic monitoring with multimodal neuroimaging sensory discharges. Eur. J. Neurosci. 8, 2236–2246
information. techniques. (1996).

NATURE REVIEWS | NEUROSCIENCE VOLUME 6 | JULY 2005 | 581

© 2005 Nature Publishing Group
PERSPECTIVES

29. Liberman, A. M., Cooper, F. S., Shankweiler, D. P. & 50. Näätänen, R., Tervaniemi, M., Sussman, E., Paavilainen, P. 70. Rizzolatti, G., Fadiga, L., Gallese, V. & Fogassi, L.
Studdert-Kennedy, M. Perception of the speech code. & Winkler, I. ‘Primitive intelligence’ in the auditory cortex. Premotor cortex and the recognition of motor actions.
Psychol. Rev. 74, 431–461 (1967). Trends Neurosci. 24, 283–288 (2001). Brain Res. Cogn. Brain Res. 3, 131–141 (1996).
30. Liberman, A. M. & Whalen, D. H. On the relation of 51. Pulvermüller, F., Hauk, O., Nikulin, V. V. & Ilmoniemi, R. J. 71. Jeannerod, M. Neural simulation of action: a unifying
speech to language. Trends Cogn. Sci. 4, 187–196 Functional links between motor and language systems. mechanism for motor cognition. Neuroimage 14,
(2000). Eur. J. Neurosci. 21, 793–797 (2005). S103—S109 (2001).
31. Fowler, C. A. An event approach to the study of speech 52. Buccino, G. et al. Listening to action-related sentences 72. Hari, R. et al. Activation of human primary motor cortex
perception from a direct realist perspective. J. Phonetics modulates the activity of the motor system: a combined during action observation: a neuromagnetic study.
14, 3–28 (1987). TMS and behavioral study. Cogn. Brain Res. (in the Proc. Natl Acad. Sci. USA 95, 15061–15065 (1998).
32. Fowler, C. A., Brown, J. M., Sabadini, L. & Weihing, J. press). 73. Barsalou, L. W. Perceptual symbol systems. Behav.
Rapid access to speech gestures in perception: evidence 53. Patterson, K. & Hodges, J. R. in International Brain Sci. 22, 577–609; discussion 610–660 (1999).
from choice and simple response time tasks. J. Mem. Encyclopaedia of the Social and Behavioural Sciences 74. Barsalou, L. W., Kyle Simmons, W., Barbey, A. K. &
Lang. 49, 396–413 (2003). (eds Thompson, R. F. & McClelland, J. L.) 3401–3405 Wilson, C. D. Grounding conceptual knowledge in
33. Fadiga, L., Craighero, L., Buccino, G. & Rizzolatti, G. (Pergamon, New York, USA, 2001). modality-specific systems. Trends Cogn. Sci. 7, 84–91
Speech listening specifically modulates the excitability of 54. Scott, S. K. & Johnsrude, I. S. The neuroanatomical and (2003).
tongue muscles: a TMS study. Eur. J. Neurosci. 15, functional organization of speech perception. Trends 75. Lakoff, G. & Johnson, M. Philosophy in the Flesh: the
399–402 (2002). Neurosci. 26, 100–107 (2003). Embodied Mind and its Challenge to Western Thought
34. Horwitz, B. & Braun, A. R. Brain network interactions in 55. Posner, M. I. & Pavese, A. Anatomy of word and (Basic Books, New York, USA, 1999).
auditory, visual and linguistic processing. Brain Lang. 89, sentence meaning. Proc. Natl Acad. Sci. USA 95, 76. Shastri, L., Grannes, D., Narayana, S. & Feldman, J. in
377–384 (2004). 899–905 (1998). Hybrid Information Processing in Adaptive Autonomous
35. Rizzolatti, G. & Craighero, L. The mirror-neuron system. 56. Chao, L. L., Haxby, J. V. & Martin, A. Attribute-based Vehicles (eds Kraetzschmar, G. K. & Palm, G.) (Springer,
Annu. Rev. Neurosci. 27, 169–192 (2004). neural substrates in temporal cortex for perceiving and Berlin, Germany, 2005).
36. Penfield, W. & Rasmussen, T. The Cerebral Cortex of knowing about objects. Nature Neurosci. 2, 913–919 77. Pulvermüller, F., Mohr, B. & Schleichert, H. Semantic or
Man (Macmillan, New York, USA, 1950). (1999). lexico–syntactic factors: what determines word-class
37. He, S. Q., Dum, R. P. & Strick, P. L. Topographic 57. Pulvermüller, F., Lutzenberger, W. & Preissl, H. Nouns and specific activity in the human brain? Neurosci. Lett. 275,
organization of corticospinal projections from the frontal verbs in the intact brain: evidence from event-related 81–84 (1999).
lobe: motor areas on the lateral surface of the potentials and high-frequency cortical responses. Cereb. 78. Wittgenstein, L. Philosophical Investigations (Blackwell,
hemisphere. J. Neurosci. 13, 952–980 (1993). Cortex 9, 498–508 (1999). Oxford, UK, 1953).
38. Matelli, M., Camarda, R., Glickstein, M. & Rizzolatti, G. 58. Bird, H., Lambon-Ralph, M. A., Patterson, K. & 79. Hommel, B., Musseler, J., Aschersleben, G. & Prinz, W.
Afferent and efferent projections of the inferior area 6 in Hodges, J. R. The rise and fall of frequency and The theory of event coding (TEC): a framework for
the macaque monkey. J. Comp. Neurol. 251, 281–298 imageability: noun and verb production in semantic perception and action planning. Behav. Brain Sci. 24,
(1986). dementia. Brain Lang. 73, 17–49 (2000). 849–878; discussion 878–937 (2001).
39. Elbert, T., Pantev, C., Wienbruch, C., Rockstroh, B. & 59. Humphreys, G. W. & Riddoch, M. J. On telling your fruit 80. Wermter, S. et al. Towards multimodal neural network
Taub, E. Increased cortical representation of the fingers from your vegetables: a consideration of category- robot learning. Rob. Auton. Syst. 47, 171–175 (2004).
of the left hand in string players. Science 270, 305–307 specific deficits after brain-damage. Trends Neurosci. 10, 81. Wermter, S., Weber, C., Elshaw, M., Gallese, V. &
(1995). 145–148 (1987). Pulvermüller, F. in Biomimetic Learning for Neural Robots
40. Graziano, M. S., Taylor, C. S. & Moore, T. Complex 60. Tyler, L. K., Moss, H. E., Durrant-Peatfield, M. R. & (eds Wermter, S., Palm, G. & Elshaw, M.) (Springer,
movements evoked by microstimulation of precentral Levy, J. P. Conceptual structure and the structure of Berlin, Germany, in the press).
cortex. Neuron 34, 841–851 (2002). concepts: a distributed account of category-specific 82. Knoblauch, A., Markert, H. & Palm, G. in International
41. Hauk, O., Johnsrude, I. & Pulvermüller, F. Somatotopic deficits. Brain Lang. 75, 195–231 (2000). Work-Conference on the Interplay Between Natural and
representation of action words in the motor and premotor 61. McClelland, J. L. & Rogers, T. T. The parallel distributed Artificial Computation 2005 (eds Mira, J. & Alvarez, J. R.)
cortex. Neuron 41, 301–307 (2004). processing approach to semantic cognition. Nature Rev. 405–414 (Springer, Berlin, Germany, 2005).
42. Tettamanti, M. et al. Listening to action-related sentences Neurosci. 4, 310–322 (2003). 83. Pulvermüller, F., Hummel, F. & Härle, M. Walking or talking?:
activates fronto-parietal motor circuits. J. Cogn. 62. Bak, T. H., O’Donovan, D. G., Xuereb, J. H., Boniface, S. Behavioral and neurophysiological correlates
Neurosci. 17, 273–281 (2005). & Hodges, J. R. Selective impairment of verb processing of action verb processing. Brain Lang. 78, 143–168 (2001).
43. Buccino, G. et al. Action observation activates premotor associated with pathological changes in Brodmann areas
and parietal areas in a somatotopic manner: an fMRI 44 and 45 in the motor neurone disease-dementia- Acknowledgements
study. Eur. J. Neurosci. 13, 400–404 (2001). aphasia syndrome. Brain 124, 103–120 (2001). This work was supported by the Medical Research Council (UK)
44. Holcomb, P. J. & Neville, H. J. Auditory and visual 63. Neininger, B. & Pulvermüller, F. Word-category specific and by the European Community under the Information Society
semantic priming in lexical decision: a comparision using deficits after lesions in the right hemisphere. Technologies Programme. For discussions related to this work,
event-related brain potentials. Lang. Cognit. Process. 5, Neuropsychologia 41, 53–70 (2003). I am grateful to R. Assadollahi, T. Bak, V. Gallese, R. Ilmoniemi,
281–312 (1990). 64. Pulvermüller, F. & Mohr, B. The concept of transcortical O. Hauk, M. Kiefer, A. Knoblauch, W. Marslen-Wilson, B. Mohr,
45. Sereno, S. C., Rayner, K. & Posner, M. I. Establishing a cell assemblies: a key to the understanding of cortical B. Neininger, G. Palm, K. Patterson, G. Rizzolatti, Y. Shtyrov and
time line for word recognition: evidence from eye lateralization and interhemispheric interaction. Neurosci. S. Wermter.
movements and event-related potentials. Neuroreport Biobehav. Rev. 20, 557–566 (1996).
13, 2195–2200 (1998). 65. Liberman, A. M. & Mattingly, I. G. The motor theory of Competing interests statement
46. Pulvermüller, F., Härle, M. & Hummel, F. speech perception revised. Cognition 21, 1–36 (1985). The author declares no competing financial interests.
Neurophysiological distinction of verb categories. 66. Braitenberg, V. & Schüz, A. in Language Origin:
Neuroreport 11, 2789–2793 (2000). A Multidisciplinary Approach (eds Wind, J., Chiarelli, B.,
47. Hauk, O. & Pulvermüller, F. Neurophysiological distinction Bichakjian, B. H., Nocentini, A. & Jonker, A.) 89–102 Online links
of action words in the fronto-central cortex. Hum. Brain (Kluwer, Dordrecht, Germany, 1992).
Mapp. 21, 191–201 (2004). 67. Baddeley, A. Working memory: looking back and looking FURTHER INFORMATION
48. Shtyrov, Y., Hauk, O. & Pulvermüller, F. Distributed forward. Nature Rev. Neurosci. 4, 829–839 (2003). Cognition and Brain Sciences Unit EEG laboratory:
neuronal networks for encoding category-specific 68. Fadiga, L., Fogassi, L., Pavesi, G. & Rizzolatti, G. Motor http://www.mrc-cbu.cam.ac.uk/eeg/index.shtml
semantic information: the mismatch negativity to action facilitation during action observation: a magnetic Pulvermüller’s homepage: http://www.mrc-cbu.cam.ac.uk/
words. Eur. J. Neurosci. 19, 1083–1092 (2004). stimulation study. J. Neurophysiol. 73, 2608–2611 (1995). common/people/people-pages/friedemann.pulvermuller.shtml
49. Pulvermüller, F., Shtyrov, Y. & Ilmoniemi, R. J. Brain 69. Gallese, V., Fadiga, L., Fogassi, L. & Rizzolatti, G. Action The Neuroscience of Language:
signatures of meaning access in action word recognition. recognition in the premotor cortex. Brain 119, 593–609 http://www.cambridge.org/us/features/pulvermuller/
J. Cogn. Neurosci. (in the press). (1996). Access to this interactive links box is free online.

582 | JULY 2005 | VOLUME 6 www.nature.com/reviews/neuro

© 2005 Nature Publishing Group
View publication stats