Mirror neurons and the simulation theory of mind reading

Gallese and Goldman – Mirror neurons and mind-reading Opinion

Mirror neurons and the
simulation theory of
mind-reading
Vittorio Gallese and Alvin Goldman

A new class of visuomotor neuron has been recently discovered in the monkey’s
premotor cortex: mirror neurons. These neurons respond both when a particular action
is performed by the recorded monkey and when the same action, performed by another
individual, is observed. Mirror neurons appear to form a cortical system matching
observation and execution of goal-related motor actions. Experimental evidence
suggests that a similar matching system also exists in humans. What might be the
functional role of this matching system? One possible function is to enable an organism
to detect certain mental states of observed conspecifics. This function might be part of,
or a precursor to, a more general mind-reading ability. Two different accounts of mind-
reading have been suggested. According to ‘theory theory’, mental states are
represented as inferred posits of a naive theory. According to ‘simulation theory’, other
people’s mental states are represented by adopting their perspective: by tracking or
matching their states with resonant states of one’s own. The activity of mirror neurons,
and the fact that observers undergo motor facilitation in the same muscular groups as
those utilized by target agents, are findings that accord well with simulation theory but
would not be predicted by theory theory.

H ow do we understand other people’s behavior? How
can we assign goals, intentions, or beliefs to the inhabitants
cortex (referred to also as inferior area 6) is composed of two
distinct areas, designated as F4 and F5 (Ref. 2) (Fig. 1A).
of our social world? A possible way to answer these chal- Area F5 occupies the most rostral part of inferior area 6, ex-
lenging questions is to adopt an evolutionary frame of ref- tending rostrally within the posterior bank of the inferior
erence, both in phylogenetical and ontogenetical terms, limb of the arcuate sulcus. Area F5 is reciprocally connected
V. Gallese is at the
envisaging these ‘mind-reading’ capacities as rooted in ante- with the hand field of the primary motor cortex3–5 and has
Istituto di Fisiologia
cedent, more ‘ancient’ and simple mechanisms. This ap- direct, although limited, projections to the upper cervical
Umana, Università
proach can capitalize on the results of different fields of segments of the spinal cord6. Microstimulation in F5 evokes di Parma, Via
investigation: neurophysiology can investigate the neural hand and mouth movements at thresholds generally higher Volturno 39,
correlates of precursors of these mechanisms in lower than in the primary motor cortex7,8. The functional properties I-43100 Parma,
species of social primates such as macaque monkeys. of F5 neurons were assessed in a series of single unit record- Italy.
Developmental psychology can study how the capacity to ing experiments9–11. These experiments showed that the
tel: +39 521 903879
attribute propositional attitudes to others develops. activity of F5 neurons is correlated with specific hand and fax: +39 521 903900
In the present article we will propose that humans’ mouth motor acts and not with the execution of individual e-mail: fisioum@
mind-reading abilities rely on the capacity to adopt a simu- movements like contractions of individual muscle groups. symbolic.pr.it
lation routine. This capacity might have evolved from an What makes a movement into a motor act is the presence of
A. Goldman is at the
action execution/observation matching system whose a goal. This distinction is very important since it allows one
Department of
neural correlate is represented by a class of neurons recently to interpret the role of the motor system not just in terms of
Philosophy, University
discovered in the macaque monkey premotor cortex: mirror the control of the dynamic variables of movement (like joint
of Arizona, Tucson,
neurons (MNs). torques, etc.), but rather as a possible candidate for the AZ 85721-0027,
instantiation of mental states such as purpose or intention. USA.
The macaque monkey premotor area F5 and mirror Using the effective motor act as the classification criterion,
neurons the following types of neurons were described: ‘Grasping tel: +1 520 621 3120
fax: +1 520 621 9559
Converging anatomical evidence (see Matelli and Luppino1 neurons’, ‘Holding neurons’, ‘Tearing neurons’ and e-mail: goldman@u.
for review) supports the notion that the ventral premotor ‘Manipulation neurons’. Grasping neurons discharge when arizona.edu

1364-6613/98/$ – see front matter © 1998 Elsevier Science. All rights reserved. PII: S1364-6613(98)01262-5
493
Trends in Cognitive Sciences – Vol. 2, No. 12, December 1998

Opinion Gallese and Goldman – Mirror neurons and mind-reading

A PE
F7 F2 F1
S1 VIP LIP
F4 AIP PG
46 F5
PF

Fig. 1 Anatomical location and functional properties of mirror neurons. (A) Lateral view of the macaque brain showing the cytoarchitectonic parcellation of
the agranular frontal cortex and of the posterior parietal cortex. Motor and premotor areas, indicated by the letter F, are defined according to Matelli et al.2 Mirror
neurons were all recorded from area F5 (shown in bold). (B) Visual and motor responses of a mirror neuron. In the upper part of each panel the behavioral context
in which the neuron was studied is shown. In the lower part of each panel a series of consecutive rasters and the relative peristimulus response histograms are shown.
In the upper panel the experimenter grasps a piece of food with his hand and moves it towards the monkey, who grasps it. The neuron discharges during grasping
observation, is silent when the food is moved, and discharges again when the monkey grasps it. In the middle panel the experimenter grasps the food with a tool.
Subsequent series of event as in the previous panel. During grasping observation the neuron is silent. In the lower panel the monkey grasps the food in complete
darkness. In the upper and middle panels rasters and histograms are aligned (vertical bar) with the moment in which the experimenter grasps the food. In the lower
panel alignment is with the beginning of the grasping movement. (Histograms bin width, 20 ms. Ordinates, spikes/bin. Abscissae, time.) (C) Visual and motor re-
sponses of a mirror neuron. In the upper panel the recorded monkey observes another monkey grasping food. In the middle panel the recorded monkey observes the
experimenter grasping food. In the lower panel the recorded monkey actively grasps food. Each panel illustrates five consecutive trials. The spontaneous activity of
the neuron was virtually absent. (Panels B and C are modified from Ref. 15.)

494


the monkey performs movements aimed to take possession during four different conditions: observation of an experi-
of objects with the hand (‘Grasping-with-the-hand neur- menter grasping objects; observation of an experimenter
ons’), with the mouth (‘Grasping-with-the-mouth neurons’), doing aimless movements in the air with his arm; observa-
or with both. Grasping-with-the-hand neurons form the tion of objects; detection of the dimming of a small spot of
largest class of F5 neurons. Most neurons of this class are light. The results of this study showed that during grasping
selective for different types of grip. The role of these neur- observation MEPs recorded from the hand muscles
ons has been conceptualized by Rizzolatti12 as a ‘motor markedly increased with respect to the other conditions, in-
vocabulary’ of actions related to prehension. cluding the attention-demanding dimming detection task.
The study of F5 neurons’ responsiveness to visual stim- Even more intriguing was the finding that the increase of
uli led to the discovery of two distinct classes of neurons: excitability was present only in those muscles that subjects
canonical neurons13, which are activated during observation would use when actively performing the observed move-
of graspable objects, and MNs (see Refs 14,15) which ments. This study provided for the first time evidence that
discharge when the monkey observes another individual humans have a mirror system similar to that in monkeys.
performing an action. We will describe in more detail the Every time we are looking at someone performing an ac-
functional properties of this class of neurons. Figure 1B and tion, the same motor circuits that are recruited when we
C illustrate two examples of the activity of MNs. ourselves perform that action are concurrently activated.
MNs respond both when a particular action is per- These results posed the question of the anatomical loca-
formed by the recorded monkey and when the same action tion of the mirror system within the human brain. This
performed by another individual is observed. All MNs, as issue has been addressed by two brain-imaging experiments
mentioned above, discharge during specific goal-related utilizing the technique of Positron Emission Tomography
motor acts. Grasping, manipulating and holding objects are (PET) (Refs 17,18). These two experiments, although dif-
by far the most effective actions triggering their motor re- ferent in many respects, shared a condition in which normal
sponse. About half of them discharge during a specific type human subjects observed the experimenter grasping 3-D
of prehension, precision grip (prehension of small objects objects. Both studies used the observation of objects as a
by opposing the thumb and the index finger) being the control condition. The results showed that grasping obser-
most common one. The most effective visual stimuli trig- vation significantly activates the cortex of the left superior
gering MNs’ visual responses are actions in which the ex- temporal sulcus (Brodmann’s area 21), of the left inferior
perimenter (Fig. 1B), or a second monkey (Fig. 1C), inter- parietal lobule (Brodmann’s area 40) and of the anterior
acts with objects with their hand or with their mouth. part of Broca’s region (Brodmann’s area 45). The acti-
Neither the sight of the object alone nor of the agent alone vation, during action observation, of a cortical sector of the
is effective in evoking the neuronal response. Mimicking human brain traditionally linked with language raises the
the action without a target object, or performing the action problem of the possible homologies between Broca’s region
by using tools (middle panel of Fig. 1B) is similarly ineffec- and the premotor area F5 of the monkey, in which MNs
tive. In over 90 percent of MNs a clear correlation between have been discovered. This issue is outside the scope of the
the most effective observed action and their motor response present article and will not be dealt with here (for discussion,
was observed. In many neurons this correlation was strict see Ref. 19).
both in terms of the general goal of the action (e.g. grasp-
ing) and in terms of the way in which it was executed (e.g. Mirror neurons and mind-reading
precision grip)14,15. What is the function of the mirror system? One possible
On the basis of their functional properties, here sum- function could be to promote learning by imitation. When
marized, MNs appear to form a cortical system that new motor skills are learned, one often spends the first
matches observation and execution of motor actions. What training phases trying to replicate the movements of an ob-
could be the possible functional role of this matching sys- served instructor. MNs could in principle facilitate that
tem? Before addressing this issue it is important to stress kind of learning. We do not favor this possible role of MNs,
that the existence of an equivalent system has also been at least in non-human primates (see Box 1). Here we ex-
demonstrated in humans. plore another possibility: that MNs underlie the process of
‘mind-reading’, or serve as precursors to such a process.
The mirror system in humans Mind-reading is the activity of representing specific
Two lines of evidence strongly suggest that an action/obser- mental states of others, for example, their perceptions,
vation matching system similar to that discovered in mon- goals, beliefs, expectations, and the like. It is now agreed
keys also exists in humans. The first refers to an elegant that all normal humans develop the capacity to represent
study by Fadiga et al.16 in which the excitability of the mental states in others, a system of representation often called
motor cortex of normal human subjects was tested by using folk psychology. Whether non-human primates also deploy
Transcranic Magnetic Stimulation (TMS). The basic as- folk psychology is more controversial (see last section of
sumption underlying this experiment was the following. If this article), but it certainly has not been precluded. The
the observation of actions activates the premotor cortex in hypothesis explored here is that MNs are part of – albeit
humans, as it does in monkeys, this mirror effect should perhaps a rudimentary part of – the folk psychologizing
elicit an enhancement of the motor evoked potentials (MEPs) mechanism.
induced by TMS of the motor cortex, given its strong Like imitation learning, mind-reading could make a
anatomical links to premotor areas. TMS was performed contribution to inclusive fitness. Detecting another agent’s

495


Box 1. Mirror neurons and imitation
The ability of non-human primates to imitate the behavior of which the broader, more highly structured component of a com-
conspecifics is a highly controversial issue. Tomasello et al.a plex skill is retained, with subjective solutions to the low-level
identify three strict criteria to delimit imitational learning: specifications. Byrne and Russonb conclude that imitational
(1) the imitated behavior should be novel for the imitator; learning in non-human primates might have been overlooked
(2) it should reproduce the behavioral strategies of the model; by the exclusive application of the action-level strategy as the
(3) it should share with it the same final goal. Behaviors not sat- defining criterion.
isfying these criteria should not be considered as true imita- What is the relevance of MNs for imitation in non-human
tional ones, and are rather to be explained by means of other primates? First of all, it should be stressed that imitation behav-
mechanisms such as stimulus enhancement, emulation, or re- ior has never been observed in association with MN activity.
sponse facilitation. By applying these strict criteria to the extant Furthermore, even adopting Byrne and Russon’s criteria, we are
literature, Tomasello et al.a exclude the possibility that wild not aware of any clearcut evidence of imitation of grasping be-
animals may display true imitative behavior. A different perspec- havior among adult macaque monkeys, although this possibility
tive is offered by Byrne and Russonb. These authors start from is not precluded for young monkeys during development. On
the concept that behaviors display a hierarchical structure, and the basis of these considerations we are inclined not to favor the
can be therefore described at several levels of increasing com- hypothesis that MNs in area F5 promote grasping imitation
plexity. Manual skills represents a good example. Because com- learning in adult monkeys.
plex behaviors are hierarchically structured, ‘…there exists a
range of possibilities for how imitation might take place, be- References
yond the simple dichotomy of imitation versus no imitation’. a Tomasello, M., Kruger, A.C. and Ratner, H.H. (1993) Cultural
Byrne and Russonb single out an action-level imitation in which learning Behav. Brain Sci. 16, 495–511
a detailed specification of the various motor sequences compos- b Byrne, R.W. and Russon, A.E. Learning by imitation: a hierarchical
ing a complex action is made, and a program-level imitation in approach Behav. Brain Sci. (in press)

goals and/or inner states can be useful to an observer be- the skill changes are construed as manifestations of changes
cause it helps him anticipate the agent’s future actions, in theory30,31. Theory theorists differ among themselves as
which might be cooperative, non-cooperative, or even to whether theory of mind is acquired by a general-purpose
threatening. Accurate understanding and anticipation en- scientizing algorithm32 or by the maturation of a domain-
able the observer to adjust his responses appropriately. Our specific module or set of modules33,34. This debate will not
discussion of mind-reading will initially and primarily focus concern us here.
on humans; later we will return to its possible realization in ST arose partly from doubts about whether folk
non-human primates. psychologizers really represent, even tacitly, the sorts of
causal/explanatory laws that TT typically posits. ST sug-
Two theories of mind-reading gests that attributors use their own mental mechanisms to
There is a large literature concerned with the nature of calculate and predict the mental processes of others. For ex-
(human) mind-reading. Two types of approaches have ample, Kahneman and Tversky35 gave subjects a description
dominated recent discussion: theory theory (TT) and simu- of two travellers who shared the same limousine en route to
lation theory (ST) (Refs 20–22). The fundamental idea of the airport and were caught in a traffic jam. Their planes
TT is that ordinary people accomplish mind-reading by ac- were scheduled to depart at the same time, but they arrived
quiring and deploying a commonsense theory of the mind, 30 minutes late. Mr A was told that his flight left on time;
something akin to a scientific theory. Mental states attrib- Mr B was told that his flight was delayed and just left five
uted to other people are conceived of as unobservable, the- minutes ago. Who was more upset? Ninety-six percent of
oretical posits, invoked to explain and predict behavior in the experimental subjects said that Mr B was more upset.
the same fashion that physicists appeal to electrons and How did they arrive at this answer? According to TT there
quarks to predict and explain observable phenomena. On must be some psychological law they exploited to infer the
the standard presentation, the theory of mind possessed by travellers’ relative upsetness. According to ST, on the other
ordinary people consists of a set of causal/explanatory laws hand, each subject would have put himself in each of the
that relate external stimuli to certain inner states (e.g. per- imaginary traveller’s ‘shoes’ and imagined how he would
ceptions), certain inner states (e.g. desires and beliefs) to have felt in their place36. Another example concerns the pre-
other inner states (e.g. decisions), and certain inner states diction of decisions. To predict White’s next move in a
(e.g. decisions) to behavior. This picture has been articu- chess match ST suggests that you try to simulate White’s
lated by functionalist philosophers of mind23–26 as well as by thought processes and arrive at a decision which you then
developmental psychologists27,28. According to TT, attribut- attribute to him36–38. First you create in yourself pretend
ing particular mental states to others arises from theoretical desires, preferences, and beliefs of the sort you take White
reasoning involving tacitly known causal laws. to have; for example, preferences among chess strategies.
Much on this subject has been done by developmental- These pretend preferences and beliefs are fed into your own
ists, eager to determine how the mind-reading capacity is decision-making mechanism, which outputs a (pretend)
acquired in childhood29. Many interpret children’s changes decision (see Fig. 2). Instead of acting on that decision, it
in mind-reading skills as evidence in favor of TT because is taken ‘off-line’ and used to predict White’s decision.

496


According to this simulation account,
you need not know or utilize any psycho- Perceptual Body monitoring
logical laws. processes system
If simulation is going to make accu-
Inference
rate predictions of targets’ decisions, pre-
mechanisms
tend desires and beliefs must be suffi-
ciently similar to genuine desires and
beliefs that the decision-making system
operates on them the same way as it op- Beliefs Desires
erates on genuine desires and beliefs. Are
pretend states really similar enough to
the genuine articles that this will happen?
Homologies between pretend and natu- Decision-making
ral (i.e. non-pretend) mental states are (practical reasoning)
well documented in the domains of vis- system
ual and motor imagery39–43. (We assume
here that visual and motor imaging Behavior-predicting ‘Pretend’ belief and
consist, respectively, in pretending to see and -explaining system desire generator
and pretending to do; see Currie and
Ravenscroft44.) These visual and motor
homologies do not show, of course, that Action control
other pretend mental states, for example, system
desires and beliefs, also functionally
resemble their natural counterparts, but
informal evidence suggests this (see Behavior
Goldman45). Fig. 2 The basic elements of the simulation routine. Cognitive steps in predicting or explaining someone’s
decision by means of simulation are shown in the lower part of the figure. A dedicated pretend-state generator
The difference between TT and ST generates pretend beliefs and desires suited to the target agent. These pretend beliefs and desires are fed into
the attributor’s decision-making system (the same system that normally operates on natural, non-pretend beliefs
The core difference between TT and ST,
and desires). The output of the decision-making system is taken ‘off-line’. That is, instead of being fed into the
in our view, is that TT depicts mind- action control system, the output decision is sent to the behavior-predicting and -explaining system, which
reading as a thoroughly ‘detached’ theo- outputs a prediction that the target will make that very decision. (Modified from Ref. 59.)
retical activity, whereas ST depicts mind-
reading as incorporating an attempt to
replicate, mimic, or impersonate the mental life of the target out this conjecture by pretending to have g as well as certain
agent46. This difference can be highlighted diagrammatically, beliefs about the effectiveness or ineffectiveness of the action
as shown in Fig. 3. m vis-à-vis goal g. This simulation leads him to form a (pre-
In the simulation scenario there is a distinctive match- tend) decision to do m. He therefore uses this result to con-
ing or ‘correspondence’ between the mental activity of the clude that the target did indeed have goal g. In this fashion,
simulator and the target. This is highlighted by the similar the attributor ultimately makes a ‘backward’ inference from
state-sequences the two undergo (Fig. 3, A and B), the only the observed action to a hypothesized goal state.
exception being that the simulator uses pretend states rather
than natural states. The attributor in the TT scenario (Fig. Mirror neurons and simulation
3C) does not utilize any pretend states that mimic those of In a similar fashion, it is conceivable that externally-gener-
the target; nor does he utilize his own decision-making sys- ated MN activity serves the purpose of ‘retrodicting’ the tar-
tem to arrive at a prediction. Thus, ST hypothesizes that a get’s mental state, moving backwards from the observed
significant portion of mind-reading episodes involves the action. Let us interpret internally generated activation in
process of mimicking (or trying to mimic) the mental activ- MNs as constituting a plan to execute a certain action, for
ity of the target agent. TT predicts no such mimicking as example, the action of holding a certain object, grasping it,
part of the mind-reading process. This contrast presents a or manipulating it. When the same MNs are externally ac-
potential basis for empirically discriminating between ST tivated – by observing a target agent execute the same action
and TT. If there is evidence of mental mimicry in the mind- – MN activation still constitutes a plan to execute this ac-
reading process, that would comport nicely with ST and tion. But in the latter case the subject of the MN activity
would not be predicted by TT. knows (visually) that the observed target is concurrently
Before turning to such evidence, however, we should performing this very action. So we assume that he ‘tags’ the
note that simulation can be used to retrodict as well as pre- plan in question as belonging to that target. In fact, exter-
dict mental states, that is, to determine what mental states nally generated MN activity does not normally produce
of a target have already occurred. Figure 4 depicts a retro- motor execution of the plan in question. Externally gener-
dictive use of simulation. The attributor starts with the ated plans are largely inhibited, or taken ‘off-line’, precisely
question, ‘What goal did the target have that led him to per- as ST postulates. Thus MN activity seems to be nature’s
form action m?’ He conjectures that it was goal g, and tries way of getting the observer into the same ‘mental shoes’ as

497


tion advocated by Liberman47, in which
A the common link between the sender and
g
the receiver is not sound but the neural
Decision
mechanism, shared by both, allowing the
making m
system production of phonetic gestures.
A proponent of TT might say that
m g TT also has ways of accounting for retro-
dictive attributions of mental states. Is it
clear that anything similar to simulation
B T occurs in externally generated MN ac-
tivity? The point is that MN activity is
g T not mere theoretical inference. It creates
Decision T will in the observer a state that matches that
making m decide of the target. This is how it resembles the
T system to do m
simulation heuristic. Nothing about TT
m g leads us to expect this kind of matching.
It should be emphasized that the hypoth-
esis being advanced here is not that MNs
themselves constitute a full-scale realiz-
C ation of the simulation heuristic. In par-
T desires g ticular, we do not make this conjecture
Theoretical T will for MNs in monkeys. Our conjecture is
T believes (m g) reasoning decide only that MNs represent a primitive
system to do m version, or possibly a precursor in phyl-
Decision-making law ogeny, of a simulation heuristic that
might underlie mind-reading.
Fig. 3 Two possible ways of predicting someone’s decision. (A) A simple decision by an agent. His desire A further link between mirror
for goal g and belief that action m would be a good means to g are fed into his decision-making system, which
outputs a decision to perform m. (B) shows how an attributor can successfully predict this agent’s decision using
neuron activity and simulation can be
a simulation routine. After learning of the target’s (T) desire and belief (for example, from previous applications inferred from the fact that, as the TMS
of the simulation routine), the attributor creates similar pretend states in himself. These states are ‘tagged’ as be- experiment by Fadiga et al.16 demon-
longing to the target, and then fed into the attributor’s decision-making system, which outputs a (pretend) deci-
strates, the human equivalent matching
sion to do m. The attributor takes this decision ‘off-line’ and predicts that the target will decide to do m. (C) rep-
resents the way an attributor might predict the target’s decision using theoretical reasoning. The attributor starts
system facilitates in the observer the same
with knowledge that the target has a desire for goal g and a belief that m would achieve g. He also believes some muscle groups as those utilized by the
psychological law about human decision-making. These beliefs are all fed into his own theoretical-reasoning sys- target. This supports the idea that even
tem, which outputs the belief that the target will decide to do m. Squares represent desires; ellipses represent be-
when one is observing the action of an-
liefs; diamonds represent decisions; and hexagons represent cognitive mechanisms. Shading indicates that the
mental state is a pretend state. other, one undergoes a neural event that
is qualitatively the same as an event that
triggers actual movement in the observed
the target – exactly what the conjectured simulation heuris- agent. It is as if the tracking process in the observer is not
tic aims to do. taken entirely off-line. This might appear to be a violation
Although we compare externally generated MN activity of ST, but actually it is wholly within ST’s spirit. ST
with what transpires in Fig. 4, there clearly are differences. postulates mental occurrences in the mind-reader that are
One difference is that the real attributor does not go back to analogous to mental occurrences in the target, so it is not
a distal goal or set of beliefs. He only goes back to a motoric surprising that downstream motor activity is not entirely in-
plan. Still, this seems to be a ‘primitive’ use of simulation hibited. If TT were correct, and an observer represents a tar-
with the same structure as that depicted in Fig. 4. It also get’s behavior in purely theoretical fashion, it would not be
bears a resemblance to the motor theory of speech percep- predicted that the same muscle groups would be facilitated
in the observer as in the target. But if ST
were correct, and a mind-reader repre-
What goal T
sents an actor’s behavior by recreating in
did T have g himself the plans or movement inten-
that led him T
Decision tions of the actor, then it is reasonable to
to do m? T had
making m predict that the same muscular activation
T goal g
system will occur in the mind-reader. As match-
ing muscular activation is actually ob-
m g
served in the observer, this lends support
to ST as opposed to TT.
Fig. 4 A retrodictive use of simulation. After observing the target agent (T) perform action m, the attributor
Clinical evidence of a similar phe-
uses simulation to test whether goal g would have fitted with the choice of m. Goal g is re-created and fed into
his decision-making system, which does output m.
nomenon is found in so-called ‘imitation
behavior’48. A group of patients with

498


Box 2. Neural coding of complex biological stimuli
Neurons responding to complex biological stimuli have been possibility, that we favor, is that these two ‘action detector’ sys-
previously described in the macaque brain. A series of studies tems could represent distinct stages of the same analysis. The
showed that in the inferior temporal cortex there are neurons STS neurons would provide an initial ‘pictorial’ description of
that discharge selectively to the presentation of faces or handsa–c. actions that would be then fed (most likely through an inter-
More recently it has been shown that some of these neurons re- mediate step in the posterior parietal cortex) to the F5 motor
spond to specific features of these stimulid. Neurons responding vocabulary where it would acquire a meaning for the individual.
to complex biological visual stimuli such as walking or climbing The latter hypothesis stresses the role of action in providing
were reported also in the amygdalae. Even more relevant to the meaning to what is perceived.
issues addressed in the present paper is the work of Perrett and
co-workersb,f,g. These authors showed that in the cortex buried References

within the superior temporal sulcus (STS) there are neurons se- a Gross, C.G. et al. (1972) Visual properties of neurons in
inferotemporal cortex of the monkey J. Neurophysiol. 35, 96–111
lective to the observation of hand movements. These properties
b Perrett, D.I., Rolls, E.T. and Caan, W. (1982) Visual neurones
resemble the visual properties of F5 MNs very closely: both
responsive to faces in the monkey temporal cortex Exp. Brain Res.
populations of neurons code the same types of actions; they
47, 329–342
both generalize their responses to the different instances of the c Gross, C.G. et al. (1985) Inferior temporal cortex and pattern
same action; they both are not responsive to mimicked hand ac- recognition, in Pattern Recognition Mechanisms (Chagas, C.,
tions without the target object. However, the unique feature of Gattass, R. and Gross, C., eds), Springer-Verlag
F5 MNs resides in the fact that they also discharge during active d Tanaka, K. et al. (1991) Coding visual images of objects in the
movements of the observer. An observed action produces the inferotemporal cortex of the macaque monkey J. Neurophysiol.
same neural pattern of activation as does the action actively 66, 170–189

made by the observer. e Brothers, L., Ring, B. and Kling, A. (1990) Response of neurons in
the macaque amygdala to complex social stimuli Behav. Brain Res.
The presence of two brain regions with neurons endowed
41, 199–213
with similar complex visual properties raises the question of their
f Perrett, D.I. (1989) Frameworks of analysis for the neural
possible relationship. Two possibilities might be suggested.
representation of animate objects and actions J. Exp. Biol. 146,
One is that F5 MNs and STS neurons have different functional 87–113
roles: STS neurons would code the semantic properties, the g Perrett, D.I. (1990) Understanding the visual appearance and
meaning, of hand–object interactions, while F5 MNs would be consequence of hand actions, in Vision and Action: the Control of
engaged in the pragmatic coding of the same actions. A second Grasping (Goodale, M.A., ed.), pp. 163–180, Ablex

prefrontal lesions compulsively imitate gestures or even and diversified types of social interactions51. Within such a
complex actions performed in front of them by an experi- complex and hierarchically organized social structure, indi-
menter. This behavior is explained as an impairment of the viduals are able to recognize kinship, hierarchical ranks, to
inhibitory control normally governing motor schemas, or discriminate allies from enemies. Stammbach52 showed that
plans. It may be inferred from this that normal humans, dominant macaque monkeys modified their social relation-
when observing someone else perform an action, generate a ships with lower-ranking individuals who had previously
plan to do the same action, or an image of doing it, them- learned how to retrieve food by pressing a lever. Dominant
selves. Normally this plan is inhibited so that it does not individuals started grooming the low-ranking ones more
yield motor output, but such inhibition is impaired in the often than before, once they ‘understood’ that the newly ac-
patient population in question48. quired skills of the low-ranking individuals could be more
easily triggered, and therefore exploited, by using this sort of
Non-human primates: behaviorists or mind-readers? social upgrading. All these examples, although not provid-
A mind-reading capacity for non-human primates is a hotly ing conclusive evidence of mind-reading abilities, neverthe-
debated issue among primatologists and behavioral scientists. less, in our view, provide a strong argument supporting the
In a recent paper Heyes49 argued that a survey of empirical hypothesis that non-human primates are endowed with
studies of imitation, self-recognition, social relationships, cognitive abilities that cannot be easily dismissed as the re-
deception, role-taking and perspective-taking fails to support sult of simple stimulus–response operant conditioning.
the theory of mind hypothesis over non-mentalist alterna- Being a ‘cognizer’, nevertheless, does not necessarily imply
tives. Although, for sake of concision, it is not possible here being a mind-reader, or a possessor of the ability to detect
to address this issue thoroughly (for reviews see Refs 49 and intentional states in others. The argument that seems to
50), a few points are worth making. suggest the presence, in non-human primates, of elementary
Let us consider first the social nature of non-human pri- forms of mind-reading abilities comes from the discovery
mates. Social organization is by no means a distinctive fea- of deceptive behavior. In a series of field experiments,
ture of primates: within the realm of insects several species Hauser53,54 showed that rhesus monkeys can withhold in-
(ants are one example) are endowed with a clear social struc- formation about food location in order to deceive con-
ture. The distinctive hallmark of the social organization of specifics and obtain more food for themselves. Deception is
non-human primates is its sophisticated complexity. Non- particularly relevant here, since deceptive behavior calls for
human primates live in groups that can comprise as many as the existence of second-order intentionality, and therefore
100 individuals. These groups are characterized by intense for the capability to attribute mental states to conspecifics.

499


inferior area 6 in the macaque monkey J. Comp. Neurol. 251, 281–298
Outstanding questions 6 He, S.Q., Dum, R.P. and Strick, P.L. (1993) Topographic organization of
corticospinal projections from the frontal lobe: motor areas on the
• Is mirror-neuron activity innate or learned, and what is the relevance of lateral surface of the hemisphere J. Neurosci. 13, 952–980
this to the simulation-theory interpretation of mirror-neuron activity? 7 Gentilucci, M. et al. (1988) Functional organization of inferior area 6 in
• Is the motor system involved in the semantic mode of internally coding the macaque monkey: I. Somatotopy and the control of proximal
actions? movements Exp. Brain Res. 71, 475–490
• Can any evidence be found of ‘matching’ events for observed agent’s 8 Hepp-Reymond, M-C. et al. (1994) Force-related neuronal activity in
beliefs (as well as their plans or intentions) in non-human primates? two regions of the primate ventral premotor cortex Can. J. Physiol.
Pharmacol. 72, 571–579
9 Rizzolatti, G. et al. (1981) Afferent properties of periarcuate neurons
The relevance of the data on deceptive behavior has in macaque monkey. II. Visual responses Behav. Brain Res. 2, 147–163
been questioned on the basis of two main arguments. First, 10 Okano, K. and Tanji, J. (1987) Neuronal activities in the primate motor
fields of the agranular frontal cortex preceding visually triggered and
field reports of ethologists are anecdotal and therefore in-
self-paced movement Exp. Brain Res. 66, 155–166
trinsically ambiguous. Second, alternative non-mentalistic
11 Rizzolatti, G. et al. (1988) Functional organization of inferior area 6 in
explanations, such as chance behavior, associative learning, the macaque monkey: II. Area F5 and the control of distal movements
and inferences about observable features of the situation Exp. Brain Res. 71, 491–507
have been proposed as more parsimonious explanations of 12 Rizzolatti, G. and Gentilucci, M. (1988) Motor and visual-motor
functions of the premotor cortex, in Neurobiology of Neocortex
deceptive behavior (see Heyes49).
(Rakic, P. and Singer, W., eds), pp. 269–284, John Wiley & Sons
However, according to Byrne55, who surveyed the
13 Murata, A. et al. (1997) Object representation in the ventral premotor
literature thoroughly, there are at least 18 independent re- cortex (Area F5) of the monkey J. Neurophysiol. 78, 2226–2230
ports of intentional deception in non-human primates sup- 14 Gallese, V. et al. (1996) Action recognition in the premotor cortex
porting the notion that they can represent the mental states Brain 119, 593–609
15 Rizzolatti, G. et al. (1996) Premotor cortex and the recognition of
of other conspecifics. On the basis of this evidence, Byrne
motor actions Cognit. Brain Res. 3, 131–141
and Whiten56 suggested that primates act according to a
16 Fadiga, L. et al. (1995) Motor facilitation during action observation: a
manipulative strategy very similar to that put forward in the magnetic stimulation study J. Neurophysiol. 73, 2608–2611
sixteenth century by Niccolò Machiavelli in his masterpiece 17 Rizzolatti, G. et al. (1996) Localization of grasp representations in
Il Principe57. humans by PET: 1. Observation versus execution Exp. Brain Res. 111,
246–252
Our speculative suggestion is that a ‘cognitive continu-
18 Grafton, S.T. et al. (1996) Localization of grasp representations in
ity’ exists within the domain of intentional-state attribution
humans by PET: II. Observation compared with imagination Exp. Brain
from non-human primates to humans, and that MNs rep- Res. 112, 103–111
resent its neural correlate (see also Box 2). This continuity is 19 Rizzolatti, G. and Arbib, M.A. (1998) Language within our grasp Trends
grounded in the ability of both human and non-human Neurosci. 21, 188–194
20 Davies, M. and Stone, T., eds (1995) Folk Psychology, Blackwell
primates to detect goals in the observed behavior of con-
21 Davies, M. and Stone, T., eds (1995) Mental Simulation, Blackwell
specifics. The capacity to understand action goals, already
22 Carruthers, P. and Smith, P., eds (1996) Theories of Theories of Mind,
present in non-human primates, relies on a process that Cambridge University Press
matches the observed behavior to the action plans of the ob- 23 Sellars, W. (1963) Empiricism and the philosophy of mind, in Science,
server. It is true, as pointed out by Meltzoff and Moore58, Perception and Reality, pp. 127–194, Routledge
24 Lewis, D. (1972) Psychophysical and theoretical identifications
that the understanding of action goals does not imply a full
Australas. J. Philos. 50, 249–258
grasp of mental states such as beliefs or desires. Action-goal
25 Fodor, J.A. (1987) Psychosemantics, MIT Press
understanding nevertheless constitutes a necessary phylo- 26 Churchland, P.N. (1988) Matter and Consciousness, MIT Press
genetical stage within the evolutionary path leading to the 27 Wellman, H. (1990) The Child’s Theory of Mind, MIT Press
fully developed mind-reading abilities of human beings. 28 Gopnik, A. (1993) How we know our minds: the illusion of first-person
knowledge of intentionality Behav. Brain Sci. 16, 1–14
29 Astington, J., Harris, P. and Olson, D., eds (1988) Developing Theories
Acknowledgements of Mind, Cambridge University Press
This work was supported by a Human Frontier Scientific Program grant 30 Perner, J. (1991) Understanding the Representational Mind, MIT Press
to V.G. The authors wish to thank Giacomo Rizzolatti, Elisabeth Pacherie 31 Gopnik, A. and Wellman, H. (1992) Why the child’s theory of mind
and the anonymous referees for their valuable comments and criticisms. really is a theory Mind Lang. 7, 145–171
32 Gopnik, A. and Meltzoff, A. (1997) Words, Thoughts, and Theories,
MIT Press
References 33 Leslie, A. (1994) Pretending and believing: issues in the theory of
1 Matelli, M. and Luppino, G. (1997) Functional anatomy of human TOMM Cognition 50, 211–238
motor cortical areas, in Handbook of Neuropsychology (Vol. 11) 34 Baron-Cohen, S. (1995) Mindblindness, MIT Press
(Boller, F. and Grafman, J., eds), pp. 9–26, Elsevier 35 Kahneman, D. and Tversky, A. (1982) The simulation heuristic, in
2 Matelli, M., Luppino, G. and Rizzolatti, G. (1985) Patterns of Judgment Under Uncertainty (Kahneman, D., Slovic, P. and Tversky, A.,
cytochrome oxidase activity in the frontal agranular cortex of eds), pp. 201–208, Cambridge University Press
macaque monkey Behav. Brain Res. 18, 125–137 36 Goldman, A. (1989) Interpretation psychologized Mind Lang. 4, 161–185
3 Matsumura, M. and Kubota, K. (1979) Cortical projection of hand-arm 37 Gordon, R. (1986) Folk psychology as simulation Mind Lang. 1, 158–171
motor area from postarcuate area in macaque monkey: a histological 38 Heal, J. (1986) Replication and functionalism, in Language, Mind and
study of retrograde transport of horseradish peroxidase Neurosci. Lett. Logic (Butterfield, J., ed.), pp. 135–150, Cambridge University Press
11, 241–246 39 Kosslyn, S. (1978) Measuring the visual angle of the mind’s eye Cognit.
4 Muakkassa, K.F. and Strick, P.L. (1979) Frontal lobe inputs to primate Psychol. 10, 356–389
motor cortex: evidence for four somatotopically organized ‘premotor’ 40 Decety, J. et al. (1991) Vegetative response during imagined
areas Brain Res. 177, 176–182 movement is proportional to mental effort Behav. Brain Res. 42, 1–5
5 Matelli, M. et al. (1986) Afferent and efferent projections of the 41 Farah, M., Soso, M. and Dasheiff, R. (1992) Visual angle of the mind’s

500

Gallese and Goldman – Mirror neurons and mind-reading Books
etcetera
eye before and after unilateral/occipital lobectomy J. Exp. Psychol. 51 Cheney, D.L. and Seyfarth, R.M., eds (1990) How Monkeys See the
Hum. Percept. Perform. 18, 241–246 World, Chicago University Press
42 Yue, G. and Cole, K. (1992) Strength increases from the motor 52 Stammbach, E. (1988) Group responses to specially skilled individuals
program: comparison of training with maximal voluntary and in Macaca fascicularis Behaviour 107, 241–266
imagined muscle contractions J. Neurophysiol. 67, 1114–1123 53 Hauser, M.D. (1992) Costs of deception: cheaters are punished in
43 Jeannerod, M. (1994) The representing brain: neural correlates of rhesus monkeys Proc. Natl. Acad. Sci. U. S. A. 89, 12137–12139
motor intention and imagery Behav. Brain Sci. 17, 187–202 54 Hauser, M.D. and Marler, P. (1993) Food-associated calls in rhesus
44 Currie, G. and Ravenscroft, I. (1997) Mental simulation and motor macaques (Macaca mulatta): II. Costs and benefits of call production
imagery Philos. Sci. 64, 161–180 and suppression Behav. Ecology 4, 206–212
45 Goldman, A. (1992) In defense of the simulation theory Mind Lang. 7, 55 Byrne, R.W. (1995) The Thinking Ape. Evolutionary Origins of
104–119 Intelligence, Oxford University Press
46 Goldman, A. The mentalizing folk, in Metarepresentation (Sperber, D., 56 Byrne, R.W. and Whiten, A. (1988) Machiavellian Intelligence: Social
ed.), Oxford University Press (in press) Expertise and the Evolution of Intellect in Monkeys, Apes and Humans,
47 Liberman, A. (1996) Speech: A Special Code, MIT Press Clarendon Press
48 Lhermitte, F., Pillon, B. and Serdaru, M. (1986) Human autonomy and 57 Machiavelli, N. (1532) Il Principe (English transl., 1979, The Prince),
the frontal lobes: I. Imitation and utilization behavior: a neuro- Penguin Books
psychological study of 75 patients Ann. Neurol. 19, 326–334 58 Meltzoff, A. and Moore, M.K. (1995) Infants’ understanding of people
49 Heyes, C.M. (1998) Theory of mind in nonhuman primates Behav. Brain and things: from body imitation to folk psychology, in The Body and
Sci. 21, 101–148 the Self (Bermudez, J.L. et al., eds), pp. 43–69, MIT Press
50 Barresi, J. and Moore, C. (1996) Intentional relations and social 59 Stich, S. and Nichols, S. (1992) Folk psychology: simulation versus tacit
understanding Behav. Brain Sci. 19, 107–154 theory Mind Lang. 7, 29–65

Books Received D.A. Norman The Design of Everyday
Things MIT Press, 1998. £9.95 (xiii + 257
Review copies of the following books have been received. Books that have pages) ISBN 0 262 64037 6
been reviewed in Trends in Cognitive Sciences are not included.
M. Paradis (ed.) Pragmatics in
The appearance of a book in the list does not preclude the possibility of it Neurogenic Communication Disorders
being reviewed in the future. Pergamon, 1998. $122.50/Dfl 213.00
(x + 257 pages) ISBN 0 08 0430651

M.A. Arbib (ed.) The Handbook of University Press, 1998. £9.95/$15.95 D. Scarborough and S. Sternberg (eds)
Brain Theory and Neural Networks (viii + 329 pages) ISBN 0 521 63755 4 Methods, Models and Conceptual Issues
MIT Press, 1998. $75.00/$49.95 (xi + MIT Press, 1998. £24.95 (xiv + 950 pages)
1118 pages) ISBN 0 262 51102 9 S. Healy (ed.) Spatial Representation in ISBN 0 262 65046 0
Animals Oxford University Press, 1998.
P. Barbosa, D. Fox, P. Hagstrom, M. HB £45.00, PB £22.50 (x + 188 pages) N.J. Wade A Natural History of Vision
McGinnis and D. Pesetsky (eds) Is The ISBN HB 0 19 850007 6, PB 0 19 850006 8 MIT Press, 1998. £34.95/$55.00 (xvi +
Best Good Enough? Optimality and 466 pages) ISBN 0 262 23194 8
Competition in Syntax MIT Press, 1998. F. Jongejan, W. Goff and E. Camus (eds)
£21.95 (450 pages) ISBN 0 262 52249 7 Tropical Veterinary Medicine: Molecular T. Watanabe (ed.) High-level Motion
Epidemiology, Hemoparasites and their Processing: Computational, Neuro-
A.G. Cairns-Smith Evolving the Mind Vectors, and General Topics The New biological, and Psychophysical Per-
on the Nature of Matter and the York Academy of Sciences, 1998. $140.00 spectives MIT Press, 1998. £47.95
Origin of Consciousness Cambridge (xiv + 503 pages) ISBN 1 57331 142 1 (x + 417 pages) ISBN 0 262 23195 6

And finally,
Trends in Cognitive Sciences relies on the goodwill and cooperation of all its contributors. We
should like to acknowledge and thank the efforts of the authors, the referees and members of the
Advisory Editorial Board in assisting the journal in all stages in its production.

We hope that readers have enjoyed 1998 as
much as we have and look forward to your
continued support in the coming year.

Peter Collins, Editor

Julian Ogilvie, Assistant Editor

Lucy Greenberry and Tracy-Ann
Smith, Editorial Administrators

Margaret Newell,
Production Coordinator

501

Mirror neurons and the simulation theory of mind reading

More Related Content

What's hot (8)

Viewers also liked (9)

Similar to Mirror neurons and the simulation theory of mind reading (20)

More from merzak emerzak (17)

Recently uploaded (20)

Mirror neurons and the simulation theory of mind reading