What type of neurons are active when we observe someone else perform an action?

Abstract

Mirror neurons have generated intense interest since their discovery in the early 1990s because they offer a potential neural mechanism for linking the observation of a conspecific’s action to the representation of the motor plan for that action in the observer’s brain. Much progress has been made in the last two and one-half decades, but much remains mysterious as well. In this chapter, we discuss research in macaque monkeys and what has been revealed about the functional, anatomical and connectivity characteristics of mirror neurons (MNs). We also discuss the use of noninvasive brain imaging to measure MNs, considering the pros and cons. Further discussion concerns what role MNs play in action understanding as well as various models of MN function.

Action Representation in the Monkey Mirror Neuron System

Mirror neurons were originally classified into three categories based on the correspondence between their visual and motor responses (Gallese et al., 1996): the majority, presenting a link but not identity between observed and executed actions, were qualified as broadly congruent, whereas neurons presenting strict correspondence between observed and executed actions were qualified as strictly congruent. A third class is represented by noncongruent mirror neurons with no clear relation between observed and executed actions (Figure 2).

Independently of the degree of congruence, mirror neurons are widely assumed to play a role in action understanding, by matching the goal of observed and executed motor acts (Rizzolatti & Sinigaglia, 2010). The goal-coding property of mirror neurons was firstly evidenced by testing neuron activation during the observation of a grasping act from which the final part was hidden (Umiltà et al., 2001). Results showed that mirror neurons were found to respond with a discharge peak close to the time of hand–object contact, both when the hand and object were in full view and when the last part of the action was hidden and could thus only be inferred. This result supported the idea that, even under limited visual cues, the mirror neuron system authorizes an internal motor representation of the entire motor act, including its final goal. In addition to coding single motor acts, the mirror neuron system is tuned differently according to the overall goal of a complex action. Indeed, the discharge of the majority of grasping mirror neurons in areas F5 and PFG varies depending on how the grasping act is embedded into different actions (i.e., grasp to eat or grasp to place; Fogassi et al., 2005).

Specific features of the visual stimulus were shown to modulate the mirror response, suggesting that mirror neurons encode observed motor acts not only for action understanding but also to analyze these motor acts in terms of the features that are relevant to generate appropriate behaviors. For example, a high number of F5 mirror neurons are modulated by the distance at which the observed action is performed (i.e., inside or outside of the monkey's peripersonal space; Caggiano, Fogassi, Rizzolatti, Thier, & Casile, 2009). These neurons display changing properties according to the possibility that the monkey will interact with the object (Caggiano et al., 2009). In addition, the majority of F5 mirror neurons are modulated by the view from which motor acts are observed, whereas only a minority show view-independent discharge (Caggiano et al., 2011). Finally, mirror neuron responses might also be influenced by the reward contingency (food or no-food) associated with an observed action (Caggiano et al., 2012), shifting the role of the mirror neuron system to the subjective evaluation of the observed action.

Mirror neuron responses can be triggered by effectors other than the hand, such as the mouth (Ferrari, Gallese, Rizzolatti, & Fogassi, 2003) or tools (Rochat et al., 2010) performing goal-directed actions. The ventral region of F5 contains mainly mouth mirror neurons responding to goal-directed feeding-related actions, such as grasping with the mouth, breaking, biting, and sucking. These neurons can be classified as broadly – or strictly – congruent in a percentage similar to hand mirror neurons. Moreover, a fraction of mouth-related mirror neurons respond to intransitive communicative gestures, such as lip smacking and lip or tongue protrusion.

Considering tool-responding mirror neurons, it was assumed that these cells had a higher response during the observation of actions made with tools (e.g., grasping food with pliers) as compared to actions made with a biological effector (e.g., hand or mouth; Ferrari, Rozzi, & Fogassi, 2005). By contrast, in a more recent study, Rochat et al. (2010) reported that hand-grasping observation triggered stronger discharge than tool-use observation in monkeys trained to grasp objects with reverse pliers (i.e., which opens by closing the hand and vice versa, such in the case of escargot-holding tools). Moreover, in this study, tool-responding mirror neurons displayed a stronger discharge during observation of actions already present in the monkey's motor repertoire (grasping with hand and with reverse pliers) as compared to actions never practiced by the monkey (spearing food with a stick; Rochat et al., 2010). It is worth noting that these neurons were activated, albeit to different degrees, by a whole family of stimuli directed to the same goal, which suggests that the mirror neuron system encodes the goal of executed and observed actions independently of how it is achieved. This tool-responding mirror property could also reflect an impact of motor experience, as it is evidenced only after exhaustive training.

Regarding the sensory modality, visual stimuli are not exclusive to trigger a mirror response. Auditory stimuli were also found effective, as revealed by F5 neurons discharging both when the monkey performs a specific action and when it sees or hears the same action performed by another individual (Keysers et al., 2003; Kohler et al., 2002). Mirror neurons can thus be classified based on the type of sensory input eliciting a response, namely, visual (the monkey observes an action performed in the absence of auditory input), auditory (the monkey listens to the sound associated with an action), and audiovisual (the monkey observes an action and listens to the associated sound). The audiovisuomotor property was taken as an argument in favor of a mirror neuron function in action understanding that would be based on encoding the abstract meaning of action independently of whether it is performed, seen, or heard (Kohler et al., 2002). Moreover, the auditory property presents some degree of selectivity, with stronger neuron response to one sound in comparison to others.

To fully understand the mirror neuron function, it is crucial to consider the more specific properties and corticocortical connections of the considered subpopulation. An example comes from mirror neurons located in parietal area VIP (strongly connected with premotor area F4) that respond to tactile and visual stimuli delivered in the peripersonal space of the monkey, as well as to stimuli presented in the peripersonal space of an individual facing the recorded monkey (Ishida et al., 2009). This population displays a mirror mechanism encoding body-directed rather than object-directed motor acts. Another mirror phenomenon was found in parietal area LIP (connected with FEF) where some neurons fire both when the monkey looks in the neuron-preferred direction and when it sees another monkey looking in the same direction. Two sets of LIP neurons were detected, either increasing or decreasing activity when the observed monkey looked in the same direction as the recorded monkey (Shepherd et al., 2009). This finding highlighted a role of LIP mirror neurons in the sharing of attention between individuals. The last set of experiments refers to premotor area F5. Although the evidence reported earlier supports the idea that actions are organized with respect to a goal–object, F5 neurons contributing to the corticospinal pyramidal tract were found to present a mirror response also during the observation of intransitive mimicked actions, that is, performed in the absence of a goal–object (Kraskov, Dancause, Quallo, Shepherd, & Lemon, 2009). Among these pyramidal tract mirror neurons, in parallel to those that increase their firing rate during action observation and execution, a class of neurons exhibits an increase in firing rate during action execution and a decrease during action observation. It was suggested that these neurons are involved in the inhibition of self-movement during action observation.

84.3.1 Macaque

Mirror neurons were discovered serendipitously during single-cell recordings of hand motor representations in the macaque. Rizzolatti and colleagues found neurons firing in premotor cortex (area F5) in a motionless monkey during observation of the experimenter (Di Pellegrino, Fadiga, Fogassi, Gallese, & Rizzolatti, 1992). Individual neurons were active during observation and execution for hand and mouth movements (Ferrari, Gallese, Rizzolatti, & Fogassi, 2003; Gallese, Fadiga, Fogassi, & Rizzolatti, 1996). Additional mirror neurons possessing visuomotor properties were subsequently identified in the inferior parietal region of the macaque (Fogassi, Gallese, Fadiga, & Rizzolatti, 1998), primarily in subcomponents PF and PFG (Rozzi, Ferrari, Bonini, Rizzolatti, & Fogassi, 2008), which have strong anatomical projections to the ventral premotor cortex (F5). These findings led to the suggestion of a functional “mirror” network (Rozzi et al., 2006).

The existence of mirror neurons immediately prompted hypotheses about their role in action recognition (Rizzolatti, Fadiga, Gallese, & Fogassi, 1996). Further support for this has been provided by the discovery that some mirror neurons in macaque F5 have auditory as well as visual and motor properties (Kohler et al., 2002), firing in response to observation and execution of actions, and for sounds associated with those actions. This multimodal integration at the level of a single cell may form the basis for action understanding and motor learning (Jeannerod, 1994).

ABSTRACT

Mirror neurons are neurons that fire both when an agent performs an action and when the agent observes the same action performed by someone else. The Mirror System Hypothesis claims that brain mechanisms for language evolved atop a mirror system for grasping through the successive emergence of systems for imitation, pantomime, and protosign (a system of conventionalized gestures). To focus the discussion of neurolinguistics, we relate praxis and language to the action-oriented perception of scenes. Computational models of the canonical and mirror systems for grasping introduce a variety of macaque brain regions of use in determining homologs in the human brain that can ground a processing approach to neurolinguistics.

Mirror neurons for grasping, each active during both execution of some actions and observation of similar actions, are contained in premotor area F5 of the macaque brain; a homologous mirror system for grasping is found in or near Broca’s area in humans. The mirror system hypothesis holds that brain mechanisms supporting language evolved from an F5-like mirror system to support language parity and the multimodality of language performance. The article examines how macaque F5 and human Broca’s area are embedded in frontal-parietal-temporal networks, and explains the role of complex imitation in bridging from the brain of the macaque–human common ancestor to the human brain.

6.2.7.2 Parietal Cortex

Mirror neurons were also recorded from IPL and, in particular, from area PFG (Fogassi et al., 2005; Gallese, Fadiga, Fogassi, & Rizzolatti, 2002; Rozzi, Ferrari, Bonini, Rizzolatti, & Fogassi, 2008). There is evidence, however, that mirror neurons are also present in AIP (personal data). PFG grasping neurons have been specifically studied to elucidate whether their discharge was modulated by the overarching action intention (Fogassi et al., 2005). For this purpose their activity was recorded while the monkey executed a motor task in which the same motor act (grasping) was embedded into two different actions (grasping to eat and grasping to place). The neurons were then tested with the monkey observing the same task, performed by an experimenter.

The results showed that a high percentage of parietal neurons discharge with different intensity during grasping execution, depending on overarching goal of the actions. On the basis of these findings, it was proposed (Fogassi et al., 2005) that parietal neurons form prewired chains in which a neuron coding a given motor act is facilitated by the neuron coding that previously executed. Any time an agent has the intention (overarching goal) to perform an action, a specific neuronal chain is activated. This model accounts for the fluidity with which the different motor acts of an action are executed one after another (Jeannerod, Paulignan, & Weiss, 1998; Rosenbaum, Cohen, Jax, Weiss, & van der Wel, 2007).

In the visual task, as in the motor task, it was found that most mirror neurons discharged differently during grasping, depending on overarching goal of the actions. Because in this case grasping was performed by the observed agent, it was suggested that the neuronal selectivity for the action goal during grasping observation activated the chain of motor neurons corresponding to a specific intention. Similar results were also obtained in area F5, where the same paradigm was applied (Bonini et al., 2010).

2.2 Mirror Neurons

F5 mirror neurons are neurons that discharge when the monkey makes a specific hand action and when it observes another individual making a similar action. The simple presentation of 3-D objects does not activate them. Similarly, the observation of actions that are not directed toward an object, or the observation of actions made using tools, fail to activate them (for a review, see Rizzolatti et al. 1999).

The visual and motor properties of mirror neurons are usually congruent. A neuron that discharges when the monkey grasps an object discharges also when the monkey observes another individual making the same action. In some neurons the way in which the object is precisely grasped does not matter (broad congruence). Other neurons fire only if the observed action is identical to that coded by that neuron (e.g., executed precision grip and observed precision grip; strict congruence).

There is evidence that mirror neurons are mostly located in the caudal part of F5. This sector is anatomically connected with parietal area PF, where, unlike in area AIP, there are neurons with mirror properties. It appears, therefore, that mirror neurons are part of a parieto-premotor circuit separate from that of canonical F5 neurons.

The discharge evoked by the observation of actions is most likely, as in the case of canonical neurons, not a pictorial description of the stimuli, but the motor representation of the action coded by the neuron. However, while canonical neurons use this representation to guide successive motor actions, this is not true for mirror neurons. This, obviously, raises the question of what this motor representation is for. Monkeys, unlike humans, are unable to imitate hand actions. Thus, the imitation hypothesis can be ruled out (at least for hand F5 neurons). The most accepted interpretation of mirror neuron activity is that it is used for action understanding. The assumptions underlying this hypothesis are: (a) individuals understand actions made by other individuals because they are able to react to them internally; and (b) the individuals know the outcome of their actions.

In conclusion, mirror neurons data indicate that the parieto-premotor circuits intervene in higher order cognitive functions and, very interestingly, that the same mechanisms that are at the basis of sensorimotor transformation underlie cognitive functions.

Mirror Neurons and Action Understanding

Mirror neurons are motor neurons. This leads immediately to two questions concerning their possible role in action understanding: (a) In virtue of which mechanism could the motor system play a role in understanding the goal of actions of others? (b) Why should the motor system be endowed with such a function?

As far as the first question is concerned, the hypothesis put forward since the first studies on mirror neurons is the following: the observation of other people executing a given motor act elicits a motor activation in the observers' brain similar to that occurring when observers themselves plan and execute that action. Because of this goal similarity, the observers recognize the goal of others' actions as their own. They do not need any inferential processing (Rizzolatti et al., 2001; Rizzolatti & Sinigaglia, 2008).

Two studies, in which the experimental paradigm allows the monkey to understand the goal of motor acts done by the experimenter in the absence of visual information, supported this interpretation. In both studies, neurons were recorded from area F5. In one study, monkeys heard the sounds typical of a motor act, such as the ripping a piece of paper or the breaking of a peanut, without seeing it (Kohler et al., 2002). In another study, the monkeys were tested in two situations: In the first, they saw the experimenter's hand moving toward an object in order to grasp it; in the second, the experimenter performed the same act but the monkey saw only the initial part of it (reaching), but not the final one (grasping) because of an interposed opaque screen (Umiltà et al., 2001). The data showed that in both experiments, mirror neurons discharged in the absence of proper visual information. The neuron activation reflected therefore the comprehension of the goal of another's motor act, not the sensory information describing that motor act.

Granted that mirror neurons are involved in encoding the goal of others' actions, why should the motor system be endowed in such a function? Before addressing this question, one must make some preliminary considerations. The claim that the mirror mechanism is involved in understanding the goal of others' motor acts does not necessarily imply that there are no other mechanisms that might perform this function. Some rely on the association between a stimulus and its effect. There is little doubt, for example, that one can understand a gesture conveying a menace, without transforming this gesture into a motor format. As shown by Wood, Glynn, Phillips, and Hauser (2007) and Wood and Hauser (2008), a monkey is frightened when it sees an individual throwing a stone toward it. This occurs even when the stone is thrown in a way different from that of the monkeys. This is not surprising because what counts here is the pain caused by stone hitting the monkey, rather than the mirroring of the precise stone-throwing gesture.

On an opposite cognitive side, there is a long philosophical tradition that maintains that action understanding is based on the capability of individuals to ‘read’ the mind of others, that is, to attribute a causal role to their mental states (such as beliefs and desires) in representing and executing actions. While the nature and the format of this ‘mindreading’ is controversial (Carruthers & Smith, 1996; Goldman, 2006; Hutto & Ratcliffe, 2007; Malle et al., 2001), there is no doubt that humans are endowed with this capability.

It is important to stress, at this point, that there is a capital difference between goal understanding based on the mirror mechanism and goal understanding relying either on a lower-order associative mechanism or on a higher-order metarepresentational capabilities. An experiment carried out by Buccino et al. (2004) clarifies this point. In an fMRI study, video clips showing motor acts done by individuals of different species (a young man, a monkey, and a dog) were presented to normal volunteers. Two types of actions were shown: biting and communicative actions typical of the different species (i.e., silent reading, lip smacking, and barking). The results showed that biting, regardless of who was the agent of the action, activated bilaterally the parietofrontal mirror network. The activation was virtually identical for the actions done by individuals of the three species. In contrast, communicative actions produced activations that depended on the species. In particular, speech produced an activation in Broca's area and related areas, while barking produced only visual activation. These data show that a motor behavior (in this case, biting) of human species is represented in the observer's motor system via a mirror mechanism even when performed by nonconspecifics. In contrast, action that does not belong to human repertoire (as is the case of barking) does not produce mirror activations, even if the observers fully understand the observed actions.

If we accept the classical view, it is difficult to understand why propositional attitudes are used to understand that a dog is barking, but not that a dog is biting. The explanation lies in our capability to understand the goal-relatedness of the observed motor using or not our own motor repertoire. The mirror understanding is an understanding from the inside (Rizzolatti & Sinigaglia, 2010), when the others' actions activate our experiences. When this is impossible, we are compelled to use alternative cognitive strategies, but to understand the others as ‘ourselves’ is a unique function of mirror neurons.

IV.D. Physiology and Anatomy

“Mirror” neurons in premotor cortex of the macaque monkey become active when the monkey performs an action and when it observes a similar action performed by the experimenter. The triggering of mirror neurons requires interaction between the agent and the object of an action. When the same movements are performed without the appropriate object or when the object is displayed without an action, the neurons do not react. Mirror neurons may subserve recognition of biologically significant actions. Their existence indicates that the monkey appreciates the communality between actions executed by another subject and by itself and thus possesses an essential prerequisite for imitation, but imitation is not within the behavioral repertoire of these monkeys. It would be a bold speculation to consider mirror neurons as correlates to the neural substrate of human imitation of body-centered gestures.

Lesion data from patients with autotopagnosia, finger agnosia, and defective imitation of meaningless gestures tell a rather straightforward story about the cerebral basis of body perception for imitation. The lesions in cases of pure autotopagnosia are remarkably uniform. They always affect the left inferior parietal lobe. That pointing to proximal body parts depends on integrity of the left hemisphere was confirmed by group studies of patients with left or right brain lesions. Regardless of whether the body parts were designated verbally or nonverbally, only patients with left brain damage committed errrors in pointing to them. In contrast, errors in selecting fingers have been found with about equal frequency in patients with left or right brain damage.

Patients with pure visuoimitative apraxia had either parietal lobe degeneration or vascular lesions in the left inferior parietal lobe. The combination of autotopagnosia and visuoimitative apraxia following left parietal lesions has been documented. The difference between the cerebral substrates of autotopagnosia and finger agnosia is paralleled by a difference between the cerebral substrates of disturbed imitation of hand and finger postures. Imitation of hand postures is disturbed exclusively in patients with left brain damage, whereas imitation of finger postures can be impaired in left and in right brain-damaged patients.

Previously, I identified the inferior parietal lobes as the neural substrate of conscious perception of the opposite half of one's own body. I now ascribe to the left inferior parietal lobe a central role for imitation on both sides of the body. It will be an interesting question for further research whether there is a difference in location between both kinds of body representations within the left inferior parietal lobe.

The Development and Ongoing Integration of Bodily Experience and Cognition

We are built for action—for intentional movement. It is part of our evolutionary heritage as animals. Our considerable intelligence depended on the development and augmentation of the frontal lobes of the brain, which evolved from the motor cortex. The frontal lobes confer the capacity to shape action through anticipation and planning. All cognition can be viewed as an adaptation to support more successful inter-action with the environment, which includes other people, to satisfy survival and other needs.

Motor-related activity and cognition continue to integrate with each other throughout life. Observing man-made objects such as tools, seeing pictures of them, imagining them, or silently reading their name all activate the premotor region of the frontal lobe, which is crucial to the control of movement.6 Moreover the specific areas activated are the same ones that are activated during the actual manipulation of these objects. Substantial evidence also suggests that language comprehension about action uses some of the same neural circuits as the action itself.7,8 For example, subjects reading a phrase that implies outward motion, such as “open the drawer,” comprehend it more quickly when they are required to make a motion away from their body, a concordant motion, as opposed to one toward their body, a discordant motion.9 Sensorimotor circuits are also activated during the creation of novel metaphors that extends action into the abstract realm.10,11

When we imagine or think about action, our brain seems to simulate it. Some researchers also believe that to perceive purposeful behavior in others, we have to simulate it below awareness in our own brains.11 This view rests in large part on the neurophysiology of groups of neurons in the premotor region and the parietal lobe that map the body’s relationship to objects in space.10,11 (With simulated action, the connections between the actual motor neurons that would carry it out are thought to be inhibited.)

Take grasping, a basic form of intentional behavior that is present in infants. Gallese and others have shown that certain cells in the monkey’s premotor cortex fire whenever the animal successfully grasps something, such as a cup or a slice of apple, whether by mouth or with one or two hands.12,13 Another group of grasping-responsive cells in the premotor region has visual properties in addition to motor properties. Called mirror neurons, they are active whenever the monkey grasps an object or sees the experimenter grasp one. They also fire when the monkey or the experimenter grasps something the monkey saw placed out of sight behind a barrier. If a habitual action involves a sound (for example, cracking a nut or tearing a piece of paper), special cells in the premotor region called audiovisual motor cells respond to the sound as well as to the sight and motion of the action. Another group of premotor neurons with both visual and motor properties fire in response to the presence of any graspable object with a particular shape and size, and/or kind of grip, even when neither the monkey nor the experimenter reaches for it. Moreover, a group of cells in the parietal lobe fires when the monkey turns its head a certain amount (say, 15 degrees to the right) or when an object is placed in that position.

Together these grasping-responsive and ancillary cells encode the monkey’s multimodal sensorimotor generalizations—in other words, their meaning making—around grasping. They communicate in a nonlinguistic manner that either the monkey or someone else in their immediate vicinity, grasps or can grasp an object with a certain particular grip and size and shape, in a certain position relative to its head. Various other groups of cells in the premotor and parietal areas encode the many additional details required for successful action, such as the specifics of the motion and the force required for the beginning, middle, and end stages of grasping. Bound into networks, the various cell types crucial to organizing complex grasping behaviors in the monkey through stimulation also likely have an essential role in perceiving grasping behaviors and making inferences about them in others.

The monkey mirror neuron studies reveal the important point that the close equivalence between internal and external meaning making, at least in monkeys, is hardwired in the brain. By the same token, the multisensory, grasping-related premotor and parietal cells indicate that the monkey brain is genetically prepared to generalize from experience and encode critical abstract regularities. Referring specifically to the audiovisual motor neurons, but applicable to many of the sensorimotor neurons, Gallese writes:

[T]hey seem to suggest that it is possible to have the sameness of informational content at a quite “abstract” level, the level of conceptual content, without being endowed with the cognitive faculty of language.15(p1238)

Evidence also points to the existence mirror neurons in humans, but there is limited opportunity to record from single cells in the human brain.10 Some argue that with humans as opposed to monkeys, associative learning mechanisms rather than mirror neurons forge the connections between cells in the sensory and motor areas that respond to one’s own action and to similar action by others.16 In either case, our brains build a model of the external world that integrates our sensorimotor experience. Because of its adaptive value, the equivalence between inside and outside that supports concept formation is, in one way or another, part of our evolutionary heritage.

As Gallese goes on to say, the growth of conceptual knowledge occurs automatically because of the dynamic nature of the brain’s model of the environment. It is continually reshaped by interaction with the world. Even in the absence of language, the model increasingly incorporates abstract knowledge that supports the implicit inferences so central to intuition:

[T]he level of conceptual knowledge that I am proposing to ascribe to monkeys (but that I also take to be quite alive in our human mind), thus enabling them with the possibility to entertain certain abstract contents, is heavily dependent on “implicit inferences.” Inferences are just more or less reliable predictions about fact. And prediction is a product of the constant shaping/rewiring of our model of the world as we interact with it.15(p1238)

Human nonverbal conceptual abilities, as Gallese intimates, are much greater than those of monkeys. Even before our ancestors developed language, we were already a highly intelligent and innovative evolutionary line.17 A portion of our behavioral insights still may not require verbal abilities. The embodied mind has a reservoir of knowledge that is not directly available to consciousness. Intuition works in our attempts to understand “objective” reality, in part because it allows us to exploit and build on the body’s inherent nonverbal understanding of the world in which it exists.

Unlike monkeys, we do have language, which greatly magnifies our conceptual abilities. Moreover we can manipulate language in both associative and rigorous ways. The rich multidimensional knowledge that comes from our experience in the world, in conjunction with our internal mental activity, sometimes can be grasped and expressed only via metaphors, which give voice to these intuitions.

For both humans and monkeys, sensorimotor concepts such as grasping embed complex combinations of the more basic kinesthetic schema highlighted by Lakoff—some of the elementary components of our interaction with our environment. Survival requires that the newborn grasp the milk-giving breast or nipple. As intentional activity, it represents the most basic form of the source–path–goal schema. To grasp something also means to form a link with it, which is essential for the infant at many different levels. Grasping the breast also involves the container schema, because it forms a link with something outside and brings it inside. In humans, via metaphorical extension this sensorimotor knowledge eventually supports language formation and reasoning. Then we use verbal metaphors to grasp for, in the sense of reaching for, abstractions we sense below awareness but cannot express directly.