Eight Problems for the Mirror Neuron Theory of Action Understanding in Monkeys and Humans
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Eight Problems for the Mirror Neuron Theory
of Action Understanding in Monkeys
and Humans
Gregory Hickok
Abstract
& The discovery of mirror neurons in macaque frontal cortex rons provide the basis of action understanding. It is argued
has sparked a resurgence of interest in motor/embodied theo- that there is no evidence from monkey data that directly tests
ries of cognition. This critical review examines the evidence this theory, and evidence from humans makes a strong case
in support of one of these theories, namely, that mirror neu- against the position. &
INTRODUCTION
published would be inclined to believe that there was
nothing in consciousness but movement, and that the
. . . we understand action because the motor
presence of sense organs, or of sensory and associatory
representation of that action is activated in
tracts in the cortex was at the least a mistake on the part
our brain. (Rizzolatti, Fogassi, & Gallese, 2001,
of the Creator’’ (Pillsbury, 1911, p. 83).
p. 661)
The mirror neuron theory of action understanding
(Rizzolatti & Craighero, 2004; Rizzolatti et al. 2001;
The [motor] theory is so simple and so easy to
Gallese, Fadiga, Fogassi, & Rizzolatti, 1996; di Pellegrino,
present that every one is glad to believe it. The
Fadiga, Fogassi, Gallese, & Rizzolatti, 1992) is the latest
only question that any one cares to raise is how
in this long line of motor theories—the motor theory
much of it will the known facts permit one to
of speech perception (Liberman, Cooper, Shankweiler,
accept. ( Walter B. Pillsbury, 1911, p. 84)
& Studdert-Kennedy, 1967) being a prominent mid-
century representative—and as with motor theories of
Motor theories of cognition have a long history in psy-
the past, seems to have a firm grasp on the field. In
chology (Scheerer, 1984), dating back at least to Berkeley’s
fact, judging from the frequency of appearance of mirror
(1709) motor interpretation of depth perception, and
neuron-related publications in prominent journals, and
have been proposed as explanations for a wide range of
the range of abilities and disorders to which the theory
mental processes. For example, in the early part of the
has been extended (e.g., speech perception, music per-
20th century, Margaret Floy Washburn proposed a mo-
ception, empathy, altruism, emotion, theory of mind,
tor theory of mental imagery (Washburn, 1914, 1916),
imitation, autism spectrum disorder, among others),
and John B. Watson explained thought as nothing more
the comments of Pillsbury, appropriately updated, are
than speech-related sensory–motor processes: ‘‘accord-
equally applicable today as they were a century ago.
ing to my view, thought processes are really motor hab-
Pillsbury’s goal in his address was ‘‘to attempt a critical
its in the larynx’’ ( Watson, 1913, p. 174). As early as 1910,
if sympathetic survey of the different formulations of the
in the Presidential Address at the American Psychological
theory and to compare it with the facts’’ (p. 84). My goal
Association meeting in Minneapolis, Walter B. Pillsbury
here with respect to mirror neuron theory is the same.
summarized the prevalence of motor theories in the his-
Mirror neurons are an interesting class of cells that de-
tory of psychology succinctly, ‘‘. . . there is nothing in
serve to be thoroughly investigated and their function
the mind that has not been explained in terms of move-
fully understood. My view is that the intense focus on
ment’’ (Pillsbury, 1911, p. 84). He also highlighted the
one interpretation of mirror neuron function, that of
widespread popularity of motor theories in his own time,
action understanding, has impeded progress on mirror
commenting that, ‘‘A reader of some of the texts lately
neuron research. Although the action understanding hy-
pothesis is interesting and worthy of investigation, I will
argue that it fails dramatically on empirical examination
University of California, Irvine (Negri et al., 2007; Mahon & Caramazza, 2005). I will
D 2008 Massachusetts Institute of Technology Journal of Cognitive Neuroscience 21:7, pp. 1229–1243start by providing a brief review of the properties of mir- property distinguishes mirror neurons from well-known
ror neurons, followed by a discussion of eight problems ‘‘set-related’’ neurons in nearby monkey area 6 that dis-
for the mirror neuron theory of action understanding. charge before movement onset (Wise & Mauritz, 1985;
Weinrich, Wise, & Mauritz, 1984). As important controls
for the possibility that ‘‘mirror activity’’ reflected some
MIRROR NEURONS: THE DATA
form of covert movement, Gallese et al. (1996) recorded
Mirror neurons, which famously respond both when the from the hand area of primary motor cortex (F1 or M1),
monkey makes active movements and when it observes and recorded EMG activity from several hand and mouth
the experimenter making meaningful movements, were muscles during action observation. No M1 cells fired, and
discovered in frontal area F5 of the macaque monkey no EMG activity was elicited in response to action obser-
(Macaca nemestrina) (Gallese et al., 1996; di Pellegrino vation. On the basis of this evidence, mirror neurons were
et al., 1992). Studies of F5 before the discovery of mirror hypothesized to support ‘‘action understanding.’’
neurons revealed that most cells in that region respond Since these early studies, mirror neurons have also
during the execution of motor acts such as grasping, hold- been found in monkey parietal cortex (Gallese, Fogassi,
ing, and tearing, and a fraction of these also respond to Fadiga, & Rizzolati, 2002), and problematically (see be-
passive somatosensory (40%) or visual (17%) stimu- low), in M1 (Tkach, Reimer, & Hatsopoulos, 2007).
lation in the absence of action (Rizzolatti et al., 1988).
Accordingly, the region’s function was interpreted as
MIRROR NEURONS: THE THEORY
supporting a motor ‘‘vocabulary where proximal and
distal movement necessary for reaching, grasping, hold- Unlike the majority of the (nonmirror) neurons in macaque
ing and bringing food to the mouth are represented’’ area F5, which are argued to support a ‘‘motor vocabu-
(Rizzolatti et al., 1988, p. 506). In this context, responses lary’’ (Rizzolatti et al., 1988), mirror neurons are claimed
to visual objects or somatosensory stimulation were in- to support ‘‘action understanding’’ (Rizzolatti & Craighero,
terpreted not as the neural basis of object or tactile 2004; Rizzolatti et al., 2001; Gallese et al., 1996; di
understanding, but as a mechanism for sensory stimula- Pellegrino et al., 1992). ‘‘Action understanding’’ is de-
tion to access various motor acts (Rizzolatti et al., 1988). fined somewhat differently in various papers. Gallese
Since the discovery of mirror neurons, interpretation et al. (1996) define it as ‘‘the capacity to recognize that
of nonmirror neurons in F5 has not changed among an individual is performing an action, to differentiate this
most F5 experts. For example, with respect to ‘‘canonical’’ action from others analogous to it, and to use this infor-
(i.e., nonmirror) object-responsive neurons in F5, Nelissen, mation in order to act appropriately’’ (p. 606). Rizzolatti
Luppino, Vanduffel, Rizzolatti, and Orban (2005) state, et al. (2001) propose that action understanding is ‘‘the
‘‘These neurons are known to play an important role capacity to achieve the internal description of an action
in the visuomotor transformation for grasping, but they and to use it to organize appropriate future behaviour’’
do not appear to have any role in objects’ identification’’ (p. 661). Rizzolatti and Craighero (2004) claim, ‘‘Each
(p. 334).1 time an individual sees an action done by another indi-
According to the most detailed early study (Gallese vidual, neurons that represent that action are activated
et al., 1996), mirror neurons constituted 17% of sampled in the observer’s premotor cortex. This automatically
cells in the portion of F5 that was examined, and exhibit induced, motor representation of the observed action
the following properties. The cells were activated when corresponds to that which is spontaneously generated
the monkey observed hand and/or mouth movements during active action and whose outcome is known to the
that were directed toward objects (‘‘goal-directed’’ ac- acting individual. Thus, the mirror system transforms vi-
tions). Roughly half (55%) were selective for one type sual information into knowledge’’ (p. 172). Nelissen et al.
of action, with grasping the most frequently represented (2005) state that ‘‘A mere visual representation [of an
movement across the population of cells (75% of cells). action], without involvement of the motor system, pro-
The majority of cells were either strictly or broadly con- vides a description of the visible aspects of the move-
gruent with their action execution response properties. ment of the agent, but does not give information critical
The cells did not respond to visually presented objects or for understanding action semantics, i.e., what the action
food items, faces, non-goal-directed body movements, is about, what its goal is, and how it is related to other
goal-directed actions made using tools (although see actions’’ (p. 332). The notion ‘‘action understanding’’ has
Ferrari, Rozzi, & Fogassi, 2005), mimicking of grasping in been generalized in humans to include speech percep-
the absence of an object (pantomime), or gestures having tion (Wilson, Saygin, Sereno, & Iacoboni, 2004; Rizzolatti
emotional meaning. The cells do not exhibit movement & Arbib, 1998; Gallese et al., 1996).
preparation activity: They discharge when the monkey It is not obvious from the definitions quoted above
observes an action, stop firing when the action termi- what ‘‘action understanding’’ means. For example, with
nates, and remain quiet even if the object is moved to- respect to the first definition, upon seeing an individual
ward the monkey, firing again only when the monkey producing meaningless, non-goal-directed actions (e.g.,
initiates its own action. This is an important fact as this flailing the arms, which should yield no mirror neuron
1230 Journal of Cognitive Neuroscience Volume 21, Number 7activity), one could presumably ‘‘recognize’’ that actions on the basis of analogy to the motor theory of speech
are being performed, ‘‘differentiate’’ such actions from perception (Liberman et al., 1967). But despite its mirror
other actions (e.g., swinging the arms rhythmically), and neuron-led resurgence in popularity among nonspeech
‘‘act appropriately’’ in response (walk away or call 911), scientists, the motor theory of speech perception ‘‘has
all without ‘‘understanding’’ the meaning of the actions few proponents within the field of speech perception’’
in the goal-directed sense. The nature of the ‘‘internal de- (Galantucci, Fowler, & Turvey, 2006, p. 361). Thus, the
scription’’ in the second definition is itself undefined and theoretical grounding of mirror neuron theory in the
therefore adds little clarity to the nature of action under- speech domain was not particularly strong.
standing. In the third definition, the idea that under- Mirror neurons have also been generalized to explain
standing is achieved by knowing the ‘‘outcome’’ is also imitation (Rizzolatti & Craighero, 2004). This function of
somewhat vague because ‘‘outcome’’ is not defined. The mirror neurons, however, has been restricted to humans
fourth definition also includes concepts that are under- because macaques (at least adult macaques; Ferrari
specified: What is the action of grasping a peanut ‘‘about’’? et al., 2006) do not imitate (Visalberghi & Fragaszy,
What is the ‘‘goal’’ of such an action? And on what level 2001). This means that mirror neuron function, as it is
of analysis is ‘‘relation’’ between actions defined? studied in macaque monkeys, cannot be the basis of imi-
The most reasonable interpretation (in my mind) is tation. Rizzolatti and Craighero (2004) emphasize that
that what is being ‘‘understood’’ by mirror neurons while ‘‘the primary function of mirror neurons cannot be ac-
observing peanut-grasping is something closer to the tion imitation’’ (p. 172). Any evidence regarding the
concept, ‘‘grasping-with-the-hand.’’ However, Nelissen neural basis of imitation in humans, therefore, cannot
et al. suggest that mirror neurons are coding more than be empirically linked to mirror neurons.
the ‘‘essence of grasping’’ (p. 334) which they believe is Although the ‘‘mirror system’’ has been used as the
coded in a more anterior region of F5 (Nelissen et al., basis for understanding a range of behaviors, we will
2005). In short, the concept of action understanding focus our attention on the core function supposed to
has been evolving, but at its core is the idea that self- hold across species, namely, action understanding. If the
generated actions have an inherent semantics and that mirror neuron theory fails to stand up empirically with
observing the same action in others affords access to respect to its core claim, as I will argue, then linkage
this action semantics. between mirror neurons and the many systems and dis-
The existence of mirror neurons has been inferred to orders linked to their function is highly dubious.
exist in humans, beginning with the earliest mirror neu- The perception of a graspable object is sufficient to
ron reports (Gallese et al., 1996; di Pellegrino et al., 1992). trigger the activation of cells in motor area F5 (Rizzolatti
These early claims (Gallese et al., 1996) were based on et al., 1988). Most mirror neuron theorists do not endow
(i) the fact that pantomime recognition deficits exist in cells that respond to the perception of objects with an
aphasia (Gainotti & Lemmo, 1976), (ii) a PET study in object semantics. Instead, they propose that F5 contains
humans showing activation in Broca’s region during ac- a motor vocabulary, and that sensory (object) responses
tion observation (Rizzolatti et al., 1996), and (iii) a trans- in F5 cells reflect a means for grasping-related sensory
cranial magnetic stimulation (TMS) study that showed information to access that vocabulary.2 When consider-
enhanced distal muscle motor-evoked potentials (MEPs) ing mirror neuron function, it is helpful to adopt this
during action observation (Fadiga, Fogassi, Pavesi, & view of F5 function as the null hypothesis, namely, that
Rizzolatti, 1995). However, the empirical basis for the gen- F5 is fundamentally a motor area that is capable of sup-
eralization of the mirror neurons to humans was dubious porting sensory–motor associations. In order to make
from the start based on the very data that was claimed to a serious case for mirror neurons as the basis of action
support it: (i0) Mirror neurons do not respond to panto- understanding, one has to show that they are qualita-
mimed actions and so pantomime recognition should not tively different from other sensory–motor cells in F5,
rely on the mirror system. Further, pantomime recogni- specifically, that they are coding more than just a sensory–
tion deficits were not associated with frontal lesions, but motor association (they have a semantics that other
rather were predominantly associated with posterior le- sensory cells in F5 do not). In what follows, I will detail
sions (Heilman, Rothi, & Valenstein, 1982). (ii0) The PET eight problems that undermine the claim that mirror
study showing Broca’s region activation during action ob- neurons go beyond other sensory–motor cells in F5 and
servation failed to show overlapping activation during support action understanding.
grasping production (Rizzolatti et al., 1996), in contrast
to the central mirror neuron observation. And (iii0) the
TMS finding of peripheral motor activation during action
1. There Is No Evidence in Monkeys That Mirror
observation directly contradicted the early demonstration
Neurons Support Action Understanding
in monkeys that M1 and the peripheral motor system did
not exhibit mirror properties (Gallese et al., 1996). The mirror neuron theory of action understanding pre-
Mirror neuron findings were also quickly generalized dicts that disruption of motor areas in F5 should pro-
to speech (Gallese et al., 1996; di Pellegrino et al., 1992) duce deficits in action perception. Although functional
Hickok 1231disruption of macaque area F5 has been shown to dis- sponse, but rather the knowledge of the ‘‘meaning’’ of rupt grasping behavior (Fogassi et al., 2001), the pre- the action. Again, following the logic, the results of the dicted corresponding decrement in action perception study indicate that half of all mirror neurons are not cod- has never been reported. Rizzolatti and Craighero (2004) ing action meaning, and again there is a simpler explana- argue that such studies are not feasible. This is because tion. The monkey can represent the object in working (i) the mirror system is bilateral and involves parietal memory which, according to popular views, involves structures, (ii) there are other mechanisms that mediate the same systems that represent the object when it is action recognition, and (iii) if one lesioned the entire physically present (Postle, 2006; Pasternak & Greenlee, mirror neuron system, more general cognitive deficits 2005; Ruchkin, Grafman, Cameron, & Berndt, 2003; would result, making interpretation difficult (Rizzolatti Fuster, 1995). This information can then be used in the & Craighero, 2004). However, if the claim is that motor normal manner as if the object was visible. systems underlie action understanding, and if it is pos- Rizzolatti and Craighero (2004) claim that these studies sible to impair motor behavior by disruption of motor show that ‘‘the activity of mirror neurons correlates with systems in F5 (Fogassi et al., 2001), then it should follow action understanding’’ (p. 174). However, action under- that action understanding will be commensurately im- standing was never actually measured, and there is a paired. If, on the other hand, motor behavior and action simpler explanation of both results, one that fits well with understanding dissociate in macaque following F5 dis- the hypothesized function of the nonmirror neurons in ruption, this would constitute evidence against a critical F5, namely, that perceptual information—including ob- role for motor systems (and area F5) in action under- jects, tactile stimulation, sounds, and actions—can be standing, independently of whether the mirror system associatively linked to and can prime a ‘‘motor vocabu- extends beyond F5 or not. lary’’ in F5 (Rizzolatti et al., 1988). In place of the standard lesion method, three studies The third study by Fogassi et al. (2005) uses a differ- are held up as evidence that mirror neurons in monkeys ent approach to argue for abstract, action understanding support action understanding. One involves the dem- properties of mirror neurons. These authors present onstration that some mirror neurons (15%) respond to very interesting data from the inferior parietal lobule action-associated sounds presented in isolation (cracking (IPL) of monkeys, which also contains mirror neurons, peanut shell, ripping paper) (Kohler et al., 2002). The as noted above. Monkeys were trained either to grasp logic here is that ‘‘If mirror neurons mediate action un- a piece of food and put it in his (the monkey’s) mouth, derstanding, their activity should reflect the meaning of or to pick up an object and put it in a container. In the observed action, not its visual features’’ (Rizzolatti & some conditions, the container was next to the mon- Craighero, 2004, p. 173). According to this logic, Kohler key’s mouth such that the mechanics of the movement et al.’s findings indicate that 85% of mirror neurons do were very similar between grasping-to-eat and grasping- not mediate action understanding because their activity to-place. In addition, a condition was also implemented does not reflect the meaning of the perceived action.3 in which the monkey grasped and placed a piece of This still leaves a population of 15% of mirror neurons— food in the container to control for differences between the audiovisual type—that may code the meaning of ac- food items and objects, both visually and tactilely. In all tions. Does the existence of these audiovisual mirror neu- variants of the experiment, the authors report that some rons prove that they are coding meaning? No. A more IPL cells preferentially responded to the goal of the ac- straightforward interpretation of this result is that sounds tion: grasping-to-eat versus grasping-to-place. Again, this can be associated with actions in F5 neurons, just as was true even when the placing-action terminated in objects can be associated with actions in F5 neurons close proximity to the mouth and involved grasping (Rizzolatti et al., 1988). Framed in terms of a priming ex- a piece of food. Some of these cells also responded planation, we might argue that the animal has associated selectively and congruently during the observation of the action of breaking a peanut with the sound of break- grasping-to-eat and grasping-to-place. Thus, both in per- ing a peanut, and when hearing only the sound, the activa- ception and action, there are IPL cells that seem to be tion spreads to F5; a form of partial cue retrieval. Right or selective for the specific goal of an action rather than wrong, the point is that we do not need to endow these the sensory or motor features of an action—a very in- cells with semantic properties to explain the finding. triguing result. Fogassi et al. discuss their motor findings The second experiment showed that although mirror in the context of ‘‘intentional chains’’ in which differ- neurons do not respond to pantomimed actions (ac- ent motor acts forming the entire action are linked in tions without the object present), they do respond if an such a way that each act is facilitated in a predictive and action is directed toward an object that is hidden behind goal-oriented fashion by the previous ones. They give a screen such that the monkey knows it is there (Umiltà an example of IPL neurons observed in another unpub- et al., 2001). In this scenario, more than half of the mir- lished study that respond to flexion of the forearm, have ror neurons that were tested also responded in the hid- tactile receptive fields around the mouth, and respond den condition. The logic here is the same, that it is not during grasping actions of the mouth and suggest that the physical features of the action that drives the re- ‘‘these neurons appear to facilitate the mouth opening 1232 Journal of Cognitive Neuroscience Volume 21, Number 7
when an object is touched or grasped’’ (p. 665). Re- neural systems in the ventral visual stream support ob-
garding the action perception response properties of ject recognition/understanding. Object information, pro-
the IPL neurons in their study, Fogassi et al. conclude, cessed for ‘‘meaning’’ in the temporal lobe, can gain
‘‘that IPL mirror neurons, in addition to recognizing the access to motor programs as appropriate for behaviors
goal of the observed motor act, discriminate identical such as grasping, thus explaining the object response
motor acts according to the action in which these acts properties of F5 cells, even though the meaning of the
are embedded. Because the discriminated motor act is objects is not coded in these motor areas (Nelissen
part of a chain leading to the final goal of the action, this et al., 2005). If there is a neural network outside of the
neuronal property allows the monkey to predict the goal mirror system that can support action understanding, as
of the observed action and, thus, to ‘read’ the intention Rizzolatti and colleagues suggest, then we can propose
of the acting individual’’ (p. 666). an identical form of interaction. Actions are processed
According to Fogassi et al., IPL mirror neurons code for ‘‘meaning’’ in this other system, which via the same
action goals and can ‘‘read the intention’’ of the acting associative mechanisms can gain access to motor pro-
individual. Perhaps Fogassi et al.’s notion of predictive grams in F5, thus producing ‘‘mirror’’ responses, anal-
coding and their example of the IPL neuron with re- ogous to object responses.
ceptive fields on the face can provide a simpler expla- A candidate region for an action understanding alter-
nation. Suppose the abstract goal of an action and/or native to mirror neurons is the superior temporal sulcus
its meaning is coded outside of the motor system. And (STS). Cells in portions of the macaque STS respond to
suppose that Fogassi et al. are correct in that a com- a wide range of actions in a manner that appears more
plex motor act leads to some form of predictive coding sophisticated than that found in mirror neurons. STS
(anticipatory opening of the mouth, salivation, perhaps neurons respond to actions such as walking toward or
even forward modeling of the expected somatosensory away, head turning, movement into or out of view, arm
consequences of the action). The predictive coding in movements, and hand–object interaction where there
the motor system is now going to be different for the is selectivity for specific actions including reaching, re-
grasping-to-eat versus grasping-to-place actions. For eat- trieving, manipulating, picking, tearing, presenting to
ing, there may be anticipatory opening of the mouth, sal- the monkey, and holding (Perrett, Mistlin, Harries, &
ivation, and perhaps anticipatory activity associated with Chitty, 1990; Perrett et al., 1985). These cells do not have
the expected somatosensory consequences of the action. motor properties in that they do not appear to fire
For placing, there will be no mouth-related coding, but during action execution (although this has not been
there may be other kinds of coding such as expectations investigated thoroughly). Interestingly, the region of in-
about the size, shape, or feel of the container, or the ferior parietal cortex that contains mirror neurons (PF),
sound that will result if the object is placed in it. If cells and which projects to F5, receives input from the STS
in the IPL differ in their sensitivity to feedback from (Rizzolatti & Craighero, 2004). This would seem to be an
these different systems, then it may look like the cells ideal circuit for representing actions (STS) and coordi-
are coding goals, when in fact they are just getting nating their interaction (PF) with the motor system (F5).
differential feedback input from the forward models.
Observing an action may activate this system with similar
3. M1 Contains Mirror Neurons
electrophysiological consequences, not because it is read-
ing the intention of the actor, but simply because the It was recently observed that mirror neurons exist in pri-
sensory event is associated with particular motor acts. mary motor cortex of macaque monkeys (Tkach et al.,
2007). Although this is consistent with the MEP work in
humans (Fadiga et al., 1995), it undermines an impor-
2. Action Understanding Can Be Achieved via
tant control observation in the original mirror neuron
Nonmirror Neuron Mechanisms
reports. Recall that the lack of mirror neurons in M1 was
Rizzolatti and Craighero (2004) noted that the mirror taken as evidence against the possibility that the mon-
neuron system may not be the only mechanism that keys were covertly generating movement responses dur-
can support action understanding. Rizzolatti et al. (2001) ing the perception of actions. In other words, it ruled
also emphasize that ‘‘these [mirror neuron] findings do out the possibility that ‘‘mirror’’ responses were merely
not exclude the possibility that other areas are involved some kind of unimplemented motor command, and
in the description of biological movement and the un- opened the door to a more interesting, higher-level
derstanding of action’’ (p. 662). The existence of other function. Now with the demonstration of ‘‘mirror’’ re-
mechanisms for action understanding is a problem for sponses in low-level motor circuitry (M1 in macaque, and
the mirror neuron theory of action understanding be- distal muscles in humans, as demonstrated with TMS), it
cause, it places action understanding on par with ‘‘object is entirely possible that ‘‘mirror’’ responses are nothing
understanding.’’ Object responses in F5 are not generally more than the facilitation of the motor system via learned
interpreted as the neural basis for object understand- associations. Tkach et al. (2007) suggest a similar inter-
ing (Rizzolatti et al., 1988), presumably because other pretation of their data, namely, ‘‘that the neural activity
Hickok 1233during observation is attributable to the covert generation enriched system exists in humans, it is then an easy and
of a motor command and that [the reason] we observe prima facie logical inference that the human mirror sys-
congruent neural activity during observation [is] because tem can support higher-order functions such as lan-
the visual goal, and thus the motor command generated, guage and empathy. However, this inference falls apart
is the same as during active movement’’ (p. 13247). if any of the assumptions about mirror neurons are in-
correct. Thus, my caution here is not that we cannot or
even should not use mirror neurons to guide human
4. The Relation between Macaque Mirror Neurons
research, but that we have to first validate our assump-
and the ‘‘Mirror System’’ in Humans Is Either
tions before making inferences regarding human behav-
Nonparallel or Undetermined
iors, especially those that do not exist in monkeys.
As noted above, mirror neuron function has been gen- Let me illustrate the problem with an abstract argu-
eralized to a wide range of human behaviors. Indeed, ment. Suppose that Rizzolatti and colleagues are correct,
much of the excitement over mirror neurons is directly namely, that mirror neurons in monkeys are the basis for
related to their potential to explain complex human action understanding, but not imitation (because adult
capacities and disorders. A statement by Oberman et al. macaques don’t imitate). In humans, the mirror system
(2005) illustrates both the extent of the generaliza- behaves differently than in monkeys such that it appears
tion and the excitement: ‘‘Mirror neurons are primarily to support imitation (Rizzolatti & Craighero, 2004): It
thought to be involved in perception and comprehen- activates during the perception and execution of even
sion of motor actions, but they may also play a critical meaningless movements (Iacoboni et al., 1999). This ob-
role in higher order cognitive processes such as imi- servation has led some mirror neuron theorists to argue
tation, theory of mind, language, and empathy, all of that the mirror system in humans has evolved to sup-
which are known to be impaired in individuals with port not only action understanding (based on inferences
autism spectrum disorders’’ (pp. 190–191, citation num- from monkey data), but also imitation (based on human
bers omitted). data) (Rizzolatti & Craighero, 2004). The assumption
The problem with statements such as this, and many made by these authors is that in the evolution of this
like it, is that the species that has been shown to possess system, old properties of mirror neurons are fully con-
mirror neurons does not, to our knowledge, possess served. But what if the mirror system evolved in humans
any of these higher-order cognitive processes, and the such that it now supports imitation but no longer sup-
species that possesses the higher-order cognitive pro- ports action understanding? Perhaps humans evolved a
cesses has not been shown conclusively to possess mir- more sophisticated semantic system, distinct from the
ror neurons (Chong, Cunnington, Williams, Kanwisher, motor system, that freed the mirror system to support
& Mattingley, 2008; Dinstein, 2008; Dinstein, Thomas, imitation. Possibilities such as this are not considered in
Behrmann, & Heeger, 2008; Dinstein, Hasson, Rubin, & mainstream mirror neuron theorizing. Instead, monkey
Heeger, 2007). To be sure, there have been a host of data and theories are typically imported to human work
studies aimed at investigating the ‘‘mirror system’’ in hu- without empirical validation of the assumptions.
mans, but much of this work has investigated behaviors Here is a concrete example of how monkey data are
that mirror neurons could not possibly support given assumed to hold, problematically, in human work. In the
their response properties in monkeys,4 and therefore, context of studying the human mirror system, a number
the connection between these behaviors and mirror of functional imaging experiments have investigated the
neurons is tenuously based on a chain of assumptions: perception of meaningless gestures, pantomimed ges-
Mirror neurons exist in humans (there are individual tures, and imitation (Koski, Iacoboni, Dubeau, Woods,
cells that respond both during action execution and ac- & Mazziotta, 2003; Koski et al., 2002; Iacoboni et al.,
tion perception), mirror neurons have evolved to sup- 1999; Grezes, Costes, & Decety, 1998; Decety et al., 1997).
port functions in humans that they do not support in These studies, which often implicate portions of the
monkeys, this evolution has conserved the functional inferior frontal gyrus (IFG) and the inferior precentral
properties found in monkeys, and mirror neurons are gyrus, are cited as evidence for the existence of a hu-
responsible for the behavior in question. There is noth- man mirror system that has evolved to support imitation
ing wrong with using animal models to generate testable (Rizzolatti & Craighero, 2004). However, this is not the
hypotheses in humans—indeed, this is a productive only interpretation. There are at least three logical pos-
and important research strategy. The problem in the sibilities. (i) Mirror neurons do not exist in humans,
case of mirror neurons is that the system has been gen- and the activation in these studies results from the func-
eralized to humans without systematic validation, and tion of some other system. (ii) Mirror neurons exist in
with the wholesale adoption of the mirror neuron doc- humans exactly as they do in monkeys (with the same
trine concerning action understanding. When a human properties), and the activations in these studies result
study starts with the assumption that mirror neurons from the function of some other system. (iii) Mirror neu-
support action understanding (see above quote from rons exist in humans, but have evolved such that they
Oberman et al.), and that a homologous and functionally now support pantomime recognition and imitation. The
1234 Journal of Cognitive Neuroscience Volume 21, Number 7third interpretation is typically adopted, whereas the tant to confirm that this area has overlapping sensory–
other possibilities are not even considered. Why? In motor response properties. Surprisingly, many investiga-
monkey mirror neuron research, other possibilities tions of the mirror system fail to confirm this fundamental
were considered. Gallese et al. (1996) considered both property of mirror neurons—another example of unver-
the possibility that mirror neurons reflected ‘‘set-related’’ ified generalization from monkey to human work. In-
responses and the possibility that mirror neurons were deed, Morin and Grezes’ review, which aimed explicitly
reflecting a nonimplemented motor plan (see above). to identify the human homologue of the macaque mir-
Because these possibilities were ruled out empirically in ror neuron system, focused exclusively on perceptual
monkeys, it is assumed (probably implicitly) that there responses. It will be important to determine whether
is no need to rule them out in humans. But this is faulty the response properties of this ventral BA 6 region can
logic. If mirror neurons exist in humans as is claimed, the be linked directly to action processing, or whether it
system is demonstrably different from that in the mon- might be performing a more general function, on which
key. One therefore cannot assume that monkey data will action processes rely. For example, recent fMRI and le-
hold in the human system. The alternative possibilities sion evidence has implicated this region in predicting
have to be ruled out empirically again. Consider in this sequences of abstract nonbiological stimuli, suggesting
respect a highly cited study of imitation in the human a more general functional role involving sequence pro-
mirror system (Iacoboni et al., 1999), which found equiv- cessing (Schubotz, Sakreida, Tittgemeyer, & von Cramon,
alent activation during the passive perception of an ac- 2004; Schubotz & von Cramon, 2004).
tion (a moving hand), a static hand, and a rectangle with Other evidence often cited as support for the exis-
a spatial cue (to which subjects were previously trained tence of a human ‘‘mirror system’’ is the demonstration
to make a hand movement). The authors explain the that viewing actions can result in peripheral motor po-
activation to the latter nonaction stimulus is this way: tentiation. TMS of motor cortex produces MEPs in distal
‘‘During all scans the participants knew that the task was muscles. The amplitude of MEPs in distal muscles is en-
either to move a finger or to refrain from moving it. hanced during action observation (Fadiga et al., 1995),
Thus mental imagery of their finger (or of the finger which has been interpreted as evidence for a human
movement) should have been present even during sim- mirror system. Although this work shows clearly that
ple observation’’ (Iacoboni et al., 1999, pp. 2526–2527). associations between observed actions and motor exe-
This would suggest that it is not action perception that cution systems exist, it does not indicate that these as-
is driving these ‘‘mirror activations,’’ but simply the in- sociations are mediated by anything like macaque mirror
ternal activation of a motor act. Indeed, there is evidence neurons. For example, TMS/MEP data cannot rule out
that human area 44, a presumed component of the hu- the possibility that the link between action observation
man mirror system, is involved in movement preparation and action execution could be mediated by a nonmotor
(Krams, Rushworth, Deiber, Frackowiak, & Passingham, conceptual representation.
1998). The relation between the macaque mirror neuron sys-
There are also a number of studies that have investi- tem and the hypothesized human homologue remains
gated the human mirror system using behaviors that do to be elucidated (for recent discussion and evidence on
hold of mirror neurons, namely, object-directed actions. this debate, see Chong et al., 2008; Dinstein et al., 2007,
A recent meta-analyses of fMRI studies of the human 2008). This in itself, however, is not an argument against
‘‘mirror system’’ (Morin & Grezes, 2008) has suggested the mirror neuron theory of action understanding. By
that human BA 44 is not the homologue of the macaque the same token, even if a network with mirror-neuron-
mirror neuron F5 region, as this region is insensitive to like properties can be fully outlined in humans, this in
the presence or absence of target objects in action per- itself would not be an argument for the mirror neuron
ception. That is, BA 44 does not distinguish between theory of action understanding. Such an argument re-
object-directed actions and actions that are non-object- quires a different sort of evidence; this is the topic of the
directed. Instead, Morin and Grezes (2008) point out next section.
that a more posterior region, ventral premotor cortex
(BA 6), is activated significantly more often during the
5. Action Understanding in Humans Dissociates
perception of object-directed action than actions without
from Neurophysiological Indices of the Human
object goals. Accordingly, these authors propose ventral
‘‘Mirror System’’
BA 6 as the human homologue of the mirror system.
Although Morin and Grezes’ hypothesis is quite reason- There are examples in the human ‘‘mirror system’’ lit-
able and is based on direct parallels with the macaque erature of dissociations between action understanding
mirror neuron system, it remains to be experimentally and ‘‘mirror system’’ function. One study (Buccino et al.,
verified. For example, a nontrivial fraction (36%) of the 2004) examined functional activations during the per-
studies reviewed by Morin and Grezes reported that per- ception of biting actions or communicative gestures per-
ception of non-object-directed actions activated ventral formed by a human, a monkey, or a dog. Independent of
BA 6. What drove these activations? Also, it will be impor- the species performing the action, viewing biting actions
Hickok 1235activated regions thought to be part of the human mir- activity is not mirroring anything, but rather reflects
ror system, the left IFG and the precentral gyrus (among adaptive task-dependent sensory–motor associations.
other areas). Viewing communicative gestures elicited
activation of these frontal mirror systems for actions
6. Action Understanding and Action
performed by a human (lip-reading) and a monkey (lip
Production Dissociate
smacking), but not a dog (barking). On the assumption
that the study participants ‘‘understood’’ all three com- As noted above, there is no evidence that deactivation of
municative actions, it is interesting that only the human the monkey mirror system disrupts action understand-
and monkey actions resulted in ‘‘mirror system’’ activa- ing. The issue has been taken up in human research,
tion. (At least subjects very likely understood that the however, where there are now several published studies
lip-reading action was associated with speech and the that investigate action recognition. This work is suitable
barking action was associated with barking. The monkey for testing several predictions of the (human extrapo-
action arguably contained less semantic information.) lated) mirror neuron theory of action understanding.
This result clearly shows that actions can be understood One such prediction is that action understanding and ac-
without the mirror system, or more to the point, that tion production should be strongly correlated. Although
mirror system activity is not particularly correlated with it has been found that these two abilities can be corre-
action understanding. Of interest is that the STS was ac- lated in group studies, there is strong evidence that they
tivated across all conditions. also are quite dissociable.
Another demonstration of the dissociability of the Several recent studies have investigated the issue. One
mirror system from action understanding comes from assessed a sample of 21 patients with limb apraxia and
a TMS/MEP study (Catmur, Walsh, & Heyes, 2007). The found a strong correlation between gesture production
authors used TMS to induce MEPs in the abductor mus- (imitation of meaningful gestures) and gesture recogni-
cles of the hand. When subjects watched a video of a tion (determining which of two sequentially presented
hand with index finger abduction, the MEPs were greater gestures match a gesture name) (Buxbaum, Kyle, &
in the subjects own index finger, whereas when the Menon, 2005). However, because the production mea-
video showed movement of a hand with little finger ab- sure has a perceptual component, a deficit affecting only
duction, MEPs were greater in the little finger of the perception could lead to correlated deficits on the
observer. This is the standard ‘‘mirror’’ MEP effect. The recognition and production tasks. Further, the recogni-
investigators then trained subjects to move their fingers tion task involved some form of working memory:
in a manner incongruent with the hand in the video: Subjects had to remember the gesture name (an audi-
Move little finger when index finger movement is shown torily and visually presented verb, e.g., ‘‘hammering’’),
and vise versa. After training, MEPs were greater in the and two gestures that were presented sequentially with
little finger when index finger movement was observed, a 2-sec interstimulus interval. If working memory for
and vise versa. ‘‘Mirror’’ effects can be trained simply by gestures recruits some form of motor-related rehearsal
sensory–motor association. The important implication of component, as is the case for speech (Baddeley, 1992),
this result is that study participants who exhibited in- then both tasks shared a production component, which
congruent MEP responses presumably did not mistake may also have contributed to the correlation.
the perception of index finger movement for little finger Another study (Pazzaglia, Smania, Corato, & Aglioti,
movement and vise versa. This indicates that a prominent 2008) also tested a sample of 21 patients with limb
indicator of human ‘‘mirror system’’ activity (Fadiga et al., apraxia and found a correlation (r .5) between a
1995) dissociates from action understanding. gesture discrimination task (judging whether or not an
It should not be surprising that measures of human action is performed correctly) and a gesture production
‘‘mirror system’’ function dissociate from action under- task (asking subjects ‘‘to perform seven complex actions
standing, as we are fully capable of understanding ac- that required the use of real objects,’’ p. 3031). How-
tions we have never produced. For example, musically ever, a cluster analysis showed that while 14 of the 21
untrained people can recognize, say, saxophone playing patients with limb apraxia had ‘‘a severe gesture rec-
even if they have never touched the instrument, just as ognition deficit,’’ 7 patients ‘‘presented with no deficit’’
one can recognize actions of non-conspecifics (barking, (p. 3034), indicating that the two abilities are dissociable.
flying). Similarly, it would be surprising, maladaptive even, A third study (Tessari, Canessa, Ukmar, & Rumiati,
if all observed actions resulted in the activation of the 2007) of unselected left hemisphere damage patients
exact same motor program in the observer. Indeed, most (n = 22) reported a weaker correlation between gesture
sports would be impossible to play, as the observation of imitation and action (pantomime) recognition (r = .32;
an object-directed action (throwing a ball) would result in again, not surprising because imitation involves a recog-
the activation of the same action in the observer when nition component), but no correlation between action
a very different action is required (catching or blocking). recognition and real object use (r = .13), which
The same is arguably true in many daily activities. The arguably provides a better assessment of ‘‘mirror sys-
results of Catmur et al. show that presumed mirror system tem’’ function. Importantly, double dissociations were
1236 Journal of Cognitive Neuroscience Volume 21, Number 7evident across patients in the latter relation: Case 23 per- motor system, BA 44/6, do not support action under-
formed at 20% accuracy on action recognition, but 100% standing. This is clearly contrary to the central claim of
on object use, whereas Case 15 performed at 100% accu- Rizzolatti et al. that it is motor representations that un-
racy on object recognition and 57% on object use. Other derlie action understanding.
cases showed similar dissociations. More recent studies using modern lesion analysis meth-
Similar findings of group-level gesture perception– ods have provided mixed results regarding the anatomi-
production correlation, but case-level dissociations were cal correlate of action understanding deficits. One such
obtained by Negri et al. (2007). This study tested an un- study (Buxbaum et al., 2005) confirmed earlier observa-
selected group of 37 patients with unilateral brain le- tions showing an association between deficits in object-
sions on several tasks including pantomime recognition, related gesture recognition and lesions to the inferior
pantomime imitation, object use, and object recogni- parietal lobe, whereas another study (Saygin, Wilson,
tion. Significant correlations were found between ob- Dronkers, & Bates, 2004) reported that action compre-
ject use and pantomime recognition (r = .58), object hension is associated with lesions to BA 44/6/4. However,
use and object recognition (r = .37), pantomime imita- this latter study examined a sample of aphasic patients
tion and recognition (r = .59), and pantomime imitation which may have biased their findings compared to studies
and object use (r = .79). However, despite these group that use unselected patients or patients selected on the
trends, subsets of patients demonstrated dissociations basis of gesture-related deficits. Further, Saygin et al. did
between each of the correlated pairs of tests including not use dynamic actions for their stimuli, but rather static
double dissociations between object use and panto- pictures of pantomimed actions (the subject then pointed
mime recognition, and object use and object recogni- to the pictured object that best fit the action). The relation
tion. The authors of this study conclude that ‘‘. . .(a) The between action understanding in dynamic actions and
ability to use objects is not necessary in order to be able static actions is unknown, so interpretation of this study
to recognize object-associated pantomimes; (b) the abil- is further compromised. However, it is relevant that de-
ity to imitate pantomimes is not necessary in order to be ficits in the understanding of linguistically specified ac-
able to recognize object-associated pantomimes; and (c) tions (written phrases such as, ‘‘She is sweeping the. . .’’
the ability to use objects is not necessary in order to be followed by the same picture choices used in the ‘‘action’’
able to recognize objects’’ (p. 806). condition) dissociated behaviorally from understanding
Sign language provides additional evidence for the dis- of pictured actions, and were not associated with lesions
sociation between action production and action under- to BA 44/6/4, but with portions of the superior temporal
standing. For example, Case ‘‘Gail D.’’ presented with gyrus, insula, and inferior parietal lobe. One can conclude
very severe deficits in sign language production asso- from the behavioral and neural dissociation between pic-
ciated with a large left frontal lobe lesion, yet her com- tured actions and linguistically specific actions that what is
prehension of sign language was well-preserved (Poizner, being mapped in this study, and associated with BA 44/6/4
Klima, & Bellugi, 1987). in the picture condition, is not ‘‘action semantics,’’ as ac-
In summary, although gesture production and gesture cess to this information is available via other routes. Thus,
recognition can be correlated in groups of both apraxic this study provides evidence against the view that the
and unselected patients with focal brain lesions, these meaning of actions is encoded in motor representations
abilities double dissociate, contrary to the prediction of in motor cortex.
the mirror neuron theory of action understanding. Another recent study (Pazzaglia et al., 2008) appears
to provide compelling evidence for an association be-
tween IFG damage and deficits in action understanding.
7. Damage to the Inferior Frontal Gyrus Is Not
Lesions in patients with limb apraxia and gesture discrim-
Correlated with Action Understanding Deficits
ination deficits were compared with lesions in patients
If the human homologue of F5 is BA 44/6, then damage with limb apraxia but without gesture discrimination defi-
to this region should result in action understanding defi- cits. Subtraction of the lesions in these two groups of
cits. Available evidence does not support this predic- patients identified the left IFG as being associated with
tion. For example, based on earlier research, Heilman the limb apraxia plus gesture discrimination deficits. A
et al. (1982) have argued that lesions to the parietal lobe voxel-based lesion–symptom mapping analysis showed
are associated with both production and comprehension the same result. However, an examination of the relation
deficits, whereas frontal lesions produce only production between the amount of damaged tissue in the IFG and
deficits. A mirror neuron proponent may counter that gesture discrimination scores in the group of patients
the parietal lobe also contains mirror neurons, and thus, who had gesture discrimination deficits showed no rela-
the association between parietal lobe damage and ac- tion (Figure 1, circles). For example, the four patients
tion understanding deficits could be viewed as consistent with the most IFG involvement (Figure 1, right solid rect-
with ‘‘mirror system’’ claims. Following this line of ar- angle) had gesture discrimination scores that are in-
gument, one would have to conclude that portions of distinguishable from the three patients with the least
the mirror system that are more closely aligned with the IFG involvement (left solid rectangle), and the latter are
Hickok 1237action judgments, vPMc stimulation yielded longer reac-
tion times than EBA stimulation, and the reverse held for
form judgments, longer reaction times for EBA stimula-
tion than vPMc stimulation. Stimulation had no effect on
accuracy. In the other study (Urgesi, Calvo-Merino, et al.,
2007), which seemed to involve more difficult stimuli
and only asked subjects to judge body configuration, an
effect of accuracy was observed with vPMc stimulation
associated with more errors on the configuration match-
ing task than with EBA stimulation. Oddly, there were
no reaction time effects.
Thus, two studies show that interference stimulation to
vPMc negatively affects performance on a body configu-
ration delayed matched-to-sample task. Again, because
Figure 1. Scatterplot showing the relation between gesture
recognition and the amount of lesioned tissue in the IFG of left these studies did not assess action understanding, they
hemisphere damaged patients with apraxia. Dark square points are cannot speak to the question of whether the mirror sys-
patients without gesture recognition deficits; lighter circle points tem supports action understanding. However, they do
are patients with gesture recognition deficits. Solid rectangles are suggest that processing of body configurations at least in
aligned on the y-axis and show that patients at the extremes of the
the delayed match-to-sample task involves vPMc to some
distribution of IFG tissue damage have indistinguishable scores
on gesture recognition. Dotted rectangle outlines patients without extent. Given that the tasks involved working memory,
gesture recognition deficits for comparison. Figure modified from it seems possible that this region may support some sort
Pazzaglia et al. (2008). TGR = Transitive Gesture Recognition score. of working memory for body configurations. This is con-
sistent with many claims regarding the sensory–motor
nature of working memory systems (Pa, Wilson, Pickell,
themselves well within the distribution of patients with- Bellugi, & Hickok, in press; Buchsbaum & D’Esposito,
out gesture discrimination deficits (left dashed rectan- 2008; Postle, 2006; Hickok, Buchsbaum, Humphries, &
gle) in terms of the amount of IFG involvement. Clearly, Muftuler, 2003; Ruchkin et al., 2003; Wilson, 2001).
IFG involvement is not predicting gesture discrimina- More work is needed to characterize the neural basis
tion performance. It is unclear why the lesion subtrac- of ‘‘action understanding.’’ Available evidence, however,
tion and voxel-based mapping analyses identified the IFG leads us to conclude that the IFG does not play a cen-
in this study, but the fact that these analyses were cal- tral role.
culated using a measure that was not corrected for re-
sponse bias may be a factor (the task was implemented
8. Generalization of the Mirror System to Speech
using a signal detection paradigm, but percent correct
Recognition Fails on Empirical Grounds
rather than the bias-corrected d0 statistic was used for
lesion analyses). Mirror neuron function has been generalized to speech
Two recent rTMS experiments have studied the ef- perception from the earliest reports (Rizzolatti & Arbib,
fects of functional disruption of ventral premotor cortex 1998; Gallese et al., 1996). Basing their speculation on
(vPMc) on visual discrimination of action-related pictures the motor theory of speech perception (Liberman &
(Urgesi, Calvo-Merino, Haggard, & Aglioti, 2007; Urgesi, Mattingly, 1985; Liberman et al., 1967), Rizzolatti et al.
Candidi, Ionta, & Aglioti, 2007). In both of these stud- suggested that mirror neurons may underlie the per-
ies, subjects were asked to make two-choice, match-to- ception of speech gestures. The motor theory of speech
sample judgments: A picture of a body configuration was perception had been all but abandoned among the
presented (the sample) followed by a mask (500 msec), majority of speech scientists when mirror neurons were
and then a picture of two body configurations; the sub- discovered, but has enjoyed a healthy revival since. How-
ject was asked to indicate which of the two matched the ever, there is exceptionally strong evidence against the
sample. First, it is important to note that neither of these motor theory of speech perception, and consequently,
studies actually tested action understanding. That is, dis- the mirror neuron generalization of action understand-
crimination performance did not depend on understand- ing to the speech domain.
ing the meaning of the actions, but could be performed A motor theory of speech perception makes a very
based on configural information alone. One study (Urgesi, clear and strong prediction. Damage to the motor speech
Candidi, et al., 2007) compared the effects of interference areas should produce deficits in speech recognition. How-
stimulation of vPMc with interference stimulation of a ever, damage to motor speech areas, evidenced in many
ventral temporal–occipital location (the extrastriate body cases by large left frontal lesions and severe speech pro-
area, EBA) during action discrimination (which action duction deficits, do not typically lead to speech recog-
matches the sample?) versus form discrimination (which nition deficits. Paul Broca’s original case, Leborne, is
actor matches the sample, independent of action?). For representative of this pattern in that the patient could
1238 Journal of Cognitive Neuroscience Volume 21, Number 7You can also read
