Neurons active during both action execution and observation [mirror neurons (MirNs)] are central to theories of action understanding, yet what they represent remains debated: abstract goals, static grips, or movement kinematics? We recorded 433 neurons from macaque premotor cortex during execution and observation of reach-to-grasp actions. Population analyses revealed grasp-specific information in both conditions, broadly distributed across neurons and dynamically reconfigured over time. Generalization across task phases was limited, indicating time-specific and evolving population codes rather than static representations. Execution and observation were linked by a shared, partially overlapping population geometry that supported reliable cross-condition classification, with the strongest alignment emerging during movement and hold. Neural activity was systematically related to multidimensional hand kinematics, and these relationships generalized across neuronal populations and across agents. Together, these findings support a dynamic population-level account of MirN function in which premotor circuits integrate visual and motor signals to represent and anticipate the unfolding structure of others’ actions.

We recently demonstrated that the dorsal premotor cortex (PMd) of macaque monkeys houses mirror neurons (MirNs). MirNs, originally described in the ventral premotor cortex, are sensorimotor neurons that fire both when an animal performs an action and when the same animal observes another agent performing an identical or similar action. Here, we investigate whether a decoder trained on the activity induced during action-observation can reliably distinguish hand grips using the activity induced during action-execution. To this end, we use neuronal spiking cortical activity recorded from the PMd of macaques engaged in execution and observation of reaching-to-grasp actions. Decoding of actions is done by a Support Vector Machine with a linear kernel function using the activity occurred in 200 ms periods (starting 800 ms before and ending 1400 ms after movement onset) at various temporal combinations between observation and execution. Classification performance, using either the whole set of units (n=140) or subsets of units of various sizes, was at 50% regardless of the number of included units. To overcome this issue, a greedy-approach was employed: the unit with the best performance was initially selected and at each subsequent step a unit was added to those of the previous step so that to maximize the performance of the resulting population. Following this approach we reached cross-decoding accuracy above 90% using less than half of the population. Thus, a selected subset of units, that shares the same computations in the execution and observation, is sufficient for effective and consistent inter-condition decoding of grips. These results demonstrate that PMd-MirNs can provide a reliable signal for the development of grip decoders which could be trained even when the required registration of neural activity and behavior can’t be obtained due to the incapability of generating overt movements.

Motor cognition is related to the planning and generation of actions as well as to the recognition and imagination of motor acts. Recently, there is evidence that the motor system participates not only in overt actions but also in mental processes supporting covert actions. Within this framework, we have investigated the cortical areas engaged in execution, observation, and imagination of the same action, by the use of the high resolution quantitative 14C-deoxyglucose method in monkeys and by fMRI in humans, throughout the entire primate brain. Our data demonstrated that observing or imagining an action excites virtually the same sensory-motor cortical network which supports execution of that same action. In general agreement with the results of five relevant meta-analyses that we discuss extensively, our results imply mental practice, i.e. internal rehearsal of the action including movements and their sensory effects. We suggest that we actively perceive and imagine actions by selecting and running off-line restored sensory-motor memories, by mentally simulating the actions. We provide empirical evidence that mental simulation of actions underlies motor cognition, and conceptual representations are grounded in sensory-motor codes. Motor cognition may, therefore, be embodied and modal. Finally, we consider questions regarding agency attribution and the possible causal or epiphenomenal role the involved sensory-motor network could play in motor cognition.

We identified neurons in dorsal premotor cortex (PMd) of the macaque brain that respond during execution and observation of reaching-to-grasp actions, thus fulfilling the mirror neuron (MirN) criterion. During observation, the percentage of grip-selective MirNs in PMd and area F5 were comparable, and the selectivity indices in the two areas were similar. During execution, F5-MirNs were more selective than PMd-MirNs for grip, which was reflected in the higher selectivity indices in F5 than in PMd. PMd displayed grip-related information earlier than F5 during both conditions. In both areas, the number of neurons exhibiting congruent visual and motor selectivity did not differ from that expected by chance. However, both the PMd and F5 neuronal ensembles provided observation-execution matching, suggesting that the congruency may be achieved in a distributed fashion across the selective elements of the population. Furthermore, representational similarity analysis revealed that grip encoding in PMd and F5 is alike during both observation and execution. Our study provides direct evidence of mirror activity in PMd during observation of forelimb movements, and suggests that PMd is a node of the MirN circuit.

Several studies have shown that the information related to grip type, object identity and kinematics of monkey grasping actions is available in macaque cortical areas of F5, MI, and AIP. In particular, these studies show that the neural discharge patterns of the neuron populations from the aforementioned areas can be used for accurate decoding of action parameters. In this study, we focus on single neuron decoding capacity of neurons in a given region, F5, considering their functional classification, i.e. as to whether they show the mirror property or not. To this end, we recorded neural spike data and arm kinematics from a monkey that performed grasping actions. The spikes were then used as a regressor to predict the kinematic parameters. Results show that single neuron real-time decoding of the kinematics is not perfect, but reasonable performance can be achieved with selected neurons from both populations. Considering the neurons that we have studied (N:32), non-mirror neurons seem to act as better single-neuron decoders. Although it is clear that population-level activity is needed for robust decoding, single-neuron decoding capacity may be used as a quantitative means to classify neurons in a given region.

Mirror neurons (MirNs) are sensorimotor neurons that fire both when an animal performs a goal-directed action and when the same animal observes another agent performing the same or a similar transitive action. It has been claimed that the observation of intransitive actions does not activate MirNs in a monkey's brain. Prompted by recent evidence indicating that the discharge of MirNs is modulated also by non-object-directed actions, we investigated thoroughly the efficacy of intransitive actions to trigger MirNs' discharge. Using representational similarity analysis, we also studied whether the elements constituting the visual scene presented to the monkey during the observation of actions (both transitive and intransitive) are represented in the discharge of MirNs. For this purpose, the moving hand was modeled by its kinematics and the object by features of its geometry. We found that MirNs respond to the observation of both transitive and intransitive actions and that the discharge differences evoked by the observation of object- and non-object-directed actions are correlated more with the kinematic differences of these actions than with the differences of the objects' features. These findings support the view that observed action kinematics contribute to action mirroring.NEW & NOTEWORTHY Mirror neurons in the monkey brain are thought to respond exclusively to the observation of object-directed actions. Here, we show that mirror neurons also respond to the observation of intransitive actions and that the kinematics of the observed movements are represented in their discharge. This finding supports the view that mirror neurons provide also a kinematics-based representation of actions.

We used the (14)C-deoxyglucose method to reveal changes in activity, in the lateral sulcus of monkeys, elicited by reaching-to-grasp in the light or in the dark and by observation of the same action executed by an external agent. Both visually-guided execution and observation of the same action activated the secondary somatosensory cortex, the ventral somatosensory area, the somatorecipient parietal ventral area, the retroinsula and the caudo-medial area of the auditory belt. These matching activations indicate that the somesthetic consequences of movements, generated bottom-up during action execution, may also be triggered top-down during action observation to represent the predicted sensory consequences of the perceived movement. The posterior granular part of insula found to be activated only for action execution and its anterior agranular part activated only for action observation may contribute to the attribution of action to the correct agent. Also, execution in the dark implicated all components activated by execution in the light but the retroinsula. In conclusion, activation of the somatorecipient parietal areas, not only for action-execution but also for action-observation, indicates that perception of actions performed by an external agent presupposes knowledge about the action-effect relationships, and that understanding others' actions consists of running off-line previously stored sensory-motor programs.

In an attempt to shed light on the role of the prefrontal cortex in action perception, we used the quantitative 14C-deoxyglucose method to reveal the effects elicited by reaching-to-grasp in the light or in the dark and by observation of the same action executed by an external agent. We analyzed the cortical areas in the principal sulcus, the superior and inferior lateral prefrontal convexities and the orbitofrontal cortex of monkeys. We found that execution in the light and observation activated in common most of the lateral prefrontal and orbitofrontal cortical areas, with the exception of 9/46-dorsal activated exclusively for observation and 9/46-ventral, 11 and 13 activated only for execution. Execution in the dark implicated only the ventral bank of the principal sulcus and its adjacent inferior convexity along with areas 47/12-dorsal and 13, whereas execution in the light activated both banks of the principal sulcus and both superior and inferior convexities along with areas 10 and 11. Our results demonstrate that the prefrontal cortex integrates information in the service of both action generation and action perception, and are discussed in relation to its contribution in movement suppression during action observation and in attribution of action to the correct agent.

Mirror neurons fire during both action execution and observation of a similar action performed by another individual [1]. However, this definition does not account for the existence of representational equivalence between execution and observation. To investigate this issue we recorded 68 neurons from area F5 of a macaque monkey trained either to execute reaching-to-grasp actions towards objects or to observe the experimenter performing the same actions [2], and adopted a decoding framework to find whether neurons effective in decoding the object/grip type in (1) execution and (2) observation conditions do exist, and most critically, to (3) assess whether transfer between execution and observation decoders (i.e. cross-decoding) can be employed. By 'transfer' we mean the application of the decoder parameters estimated using the neural discharge in observation to the neural firing recorded in execution, and vice versa. The success rate of such a decoder indicates the equivalence of representations in the two conditions. Our analysis indicates that, at the level of single neurons, object/grip-specific decoders can be constructed, i.e. the type of the object/grip employed in either execution or observation can be decoded (success rate: 80%-100%, chance level: 25%). However, only in 10% of the cases (corresponding to the congruent type mirror neurons [1]) the decoder based on the execution discharge was effective when transferred to the observation discharge. The same was true for the reverse transfer. To extend this analysis at the population level we examined all pair performance of a 10-neuron set, consisted of 4 neurons having the best decoding performance and 6 neurons randomly selected. Out of the 45 possible pairs, 7 displayed high success rates (80% on average) in cross-decoding. Remarkably, high performing pairs were constituted only when one of the neurons displaying reliable decoding performance was paired with a randomly selected -poor solo decoder- neuron, which acted as a helper. These results strongly point to a population based representation where good and poor decoders may cooperate to form a robust recognition system.

The role of the superior temporal sulcus (STs) in action execution and action observation remains unsettled. In an attempt to shed more light on the matter, we used the quantitative method of (14)C-deoxyglucose to reveal changes in activity, in the cortex of STs and adjacent inferior and superior temporal convexities of monkeys, elicited by reaching-to-grasp in the light or in the dark and by observation of the same action executed by an external agent. We found that observation of reaching-to-grasp activated the components of the superior temporal polysensory area [STP; including temporo-parieto-occipital association area (TPO), PGa, and IPa], the motion complex [including medial superior temporal area (MST), fundus of superior temporal area (FST), and dorsal and ventral parts of the middle temporal area (MTd and MTv, respectively)], and area TS2. A significant part of most of these activations was associated with observation of the goal-directed action, and a smaller part with the perception of arm-motion. Execution of reaching-to-grasp in the light-activated areas TS2, STP partially and marginally, and MT compared with the fixation but not to the arm-motion control. Consequently, MT-activation is associated with the arm-motion and not with the purposeful action. Finally, reaching-to-grasp in complete darkness activated all components of the motion complex. Conclusively, lack of visibility of our own actions involves the motion complex, whereas observation of others' actions engages area STP and the motion complex. Our previous and present findings together suggest that sensory effects are interweaved with motor commands in integrated action codes, and observation of an action or its execution in complete darkness triggers the retrieval of the visual representation of the action.

Given that prerequisite of activating the mirror neuron system is the preshaping of the hand and its interaction with the object during observation of a reaching-to-grasp-an-object action, the effects of viewing the object, the reaching forelimb and the static hand may obscure the effects of observing the grasping action per se. To disentangle these effects, we employed the (14)C-deoxyglucose quantitative autoradiographic method to map the functional activity in the entire cortex of monkeys (Macaca mulatta) which observed the experimenter performing non-goal-directed (purposeless) forelimb movements towards an object that was previously presented but no longer visible. Thus, our monkeys were exposed to the view of an object, a moving arm and a static hand with extended wrist and fingers. The distribution of metabolic activity was analyzed in 20μm thick brain sections, and two dimensional maps were reconstructed in the occipital operculum, the temporal, the lateral and medial parietal, the lateral and medial frontal, the lateral prefrontal and orbitofrontal cortices, including the cortex within the lunate, superior temporal, lateral, parietoccipital, intraparietal, central, arcuate and principal sulci. Increased metabolic activity, as compared to fixation-control monkeys, was measured in the forelimb representation of the primary motor and somatosensory cortices, the premotor cortices F2 and F5, cingulate motor areas, the secondary somatosensory cortex SII, the posterior intraparietal area 5 and areas TPOc and FST, in the hemisphere contralateral to the moving arm. Moreover, bilateral activations were elicited in areas pre-SMA, 8m, SSA and the somatorecipient area VS, the retroinsula, the auditory belt area CM, motion areas MT, MST, LOP/CIP, area 31, visual areas TEO, V6, V6Av and the parafoveal and peripheral visual representations of areas V1 and V2, respectively. Few parietal, auditory and visual areas were bilaterally depressed. In brief, a surprisingly wide cortical network is recruited even by mere observation of an arm executing goalless movements, which partially overlaps with the cortical network supporting the execution and observation of goal-directed forelimb actions. Interestingly, this overlap concerns mainly lower order sensory-motor rather than higher order association prefrontal and parietal cortices. Our results demonstrate that in order to reveal the net effects specifically induced by observation of a purposeful reaching-to-grasp action, the use of an appropriate control taking into account the effects of viewing the object to be grasped, the reaching arm and the static hand is crucial.

BACKGROUND:In the recent past several invasive cortical neuroprostheses have been developed. Signals recorded from the motor cortex (area MI) have been decoded and used to control computer cursors and robotic devices. Nevertheless, few attempts have been carried out to predict different grips.A Support Vector Machines (SVMs) classifier has been trained for a continuous decoding of four/six grip types using signals recorded in two monkeys from motor neurons of the ventral premotor cortex (area F5) during a reach-to-grasp task.FINDINGS:The results showed that four/six grip types could be extracted with classification accuracy higher than 96% using window width of 75-150 ms.CONCLUSIONS:These results open new and promising possibilities for the development of invasive cortical neural prostheses for the control of reaching and grasping.

Area V6A encodes hand configurations for grasping objects (Fattori et al., 2010). The aim of the present study was to investigate whether V6A cells also encode three-dimensional objects, and the relationship between object encoding and grip encoding. Single neurons were recorded in V6A of two monkeys trained to perform two tasks. In the first task, the monkeys were required to passively view an object without performing any action on it. In the second task, the monkeys viewed an object at the beginning of each trial and then they needed to grasp that object in darkness. Five different objects were used. Both tasks revealed that object presentation activates ∼60% of V6A neurons, with about half of them displaying object selectivity. In the Reach-to-Grasp task, the majority of V6A cells discharged during both object presentation and grip execution, displaying selectivity for either the object or the grip, or in some cases for both object and grip. Although the incidence of neurons encoding grips was twofold that of neurons encoding objects, object selectivity in single cells was as strong as grip selectivity, indicating that V6A cells were able to discriminate both the different objects and the different grips required to grasp them. Hierarchical cluster analysis revealed that clustering of the object-selective responses depended on the task requirements (view only or view to grasp) and followed a visual or a visuomotor rule, respectively. Object encoding in V6A reflects representations for action, useful for motor control in reach-to-grasp.

Many neurons in the monkey ventral premotor area F5 discharge selectively when the monkey grasps an object with a specific grip. Of these, the motor neurons are active only during grasping execution, whereas the visuomotor neurons also respond to object presentation. Here we assessed whether the activity of 90 task-related F5 neurons recorded from two macaque monkeys during the performance of a visually-guided grasping task can be used as input to pattern recognition algorithms aiming to decode different grips. The features exploited for the decoding were the mean firing rate and the mean interspike interval calculated over different time spans of the movement period (all neurons) or of the object presentation period (visuomotor neurons). A support vector machine (SVM) algorithm was applied to the neural activity recorded while the monkey grasped two sets of objects. The original set contained three objects that were grasped with different hand shapes, plus three others that were grasped with the same grip, whereas the six objects of the special set were grasped with six distinctive hand configurations. The algorithm predicted with accuracy greater than 95% all the distinct grips used to grasp the objects. The classification rate obtained using the first 25% of the movement period was 90%, whereas it was nearly perfect using the entire period. At least 16 neurons were needed for accurate performance, with a progressive increase in accuracy as more neurons were included. Classification errors revealed by confusion matrices were found to reflect similarities of hand grips used to grasp the objects. The use of visuomotor neurons' responses to object presentation yielded grip classification accuracy similar to that obtained from actual grasping execution. We suggest that F5 grasping-related activity might be used by neural prostheses to tailor hand shape to the specific object to be grasped even before movement onset.

The lateral intraparietal area (LIP) of monkeys is known to participate in the guidance of rapid eye movements (saccades), but the means it uses to specify movement variables are poorly understood. To determine whether area LIP devotes neural space to encode saccade metrics spatially, we used the quantitative [14C]deoxyglucose method to obtain images of the distribution of metabolic activity in the intraparietal sulcus (IPs) of rhesus monkeys trained to repeatedly execute saccades of the same amplitude and direction for the duration of the experiment. Different monkeys were trained to perform saccades of different sizes and in different directions. A clear topography of saccade metrics was found in the cytoarchitectonically identified area LIP ventral (LIPv) contralateral to the direction of the eye movements. We demonstrate that the representation of the vertical meridian runs parallel to the fundus of the IPs and that it is not orthogonal to the representation of the horizontal meridian. Instead, the latter runs through the middle of LIPv parallel to its border with area LIP dorsal (LIPd). The upper part of oculomotor space is represented rostrally and dorsally relative to the horizontal meridian toward the LIPv–LIPd border, whereas the lower part of oculomotor space is represented caudally and ventrally toward the caudal edge of the IPs. Saccade amplitude is also represented in an orderly manner.

We have previously demonstrated that the primary motor and somatosensory cortices of monkeys are somatotopically activated for action-observation as are for action-generation, indicating that the recruitment of learned somatosensory-motor representations underlies the perception of others' actions. Here we examined the effects of seen and unseen actions on the early visual cortices, to determine whether stored visual representations are employed in addition to the somatosensory-motor ones. We used the quantitative (14)C-deoxyglucose method to map the activity throughout the cortex of the occipital operculum, lunate, and inferior occipital sulci of "rhesus monkeys" who reached to grasp a 3D object either in the light or in the dark or who observed the same action executed by another subject. In all cases, the extrastriate areas V3d and V3A displayed marked activation. We suggest that these activations reflect processing of visuospatial information useful for the reaching component of action, and 3D object-related information useful for the grasping part. We suggest that a memorized visual representation of the action supports action-recognition, as well as action-execution in complete darkness when the object and its environment are invisible. Accordingly, the internal representation that serves action-cognition is not purely somatosensory-motor but also includes a visual component.

We have previously demonstrated that the forelimb representations of the primary motor and somatosensory cortices, as well as several premotor and parietal areas, are activated by both action–execution and action–observation, indicating that the spectator mentally simulates the observed action. Moreover, several studies demonstrated repeatedly that corticospinal excitability is modulated during action observation, providing evidence of an activation of the observer's motor system. However, evidence for the involvement of the spinal cord in action observation is controversial. The aim of the present study was to explore whether and how action–observation affects the spinal cord. To this end, we analyzed the spinal cord of eight monkeys (Macaca mulatta) trained to either execute reaching-to-grasp movements or observe the experimenter performing the same movements. Observation of grasping induced a bilateral decrease of glucose consumption in the spinal forelimb representation, whereas execution of grasping induced an increase of glucose utilization in the same area, ipsilaterally to the grasping hand. The depression of overall activity in the cervical enlargement of the spinal cord for action–observation may explain the suppression of overt movements, despite the activation of the observer's motor system.

Brain control of prehension is thought to rely on two specific brain circuits: a dorsomedial one (involving the areas of the superior parietal lobule and the dorsal premotor cortex) involved in the transport of the hand toward the object and a dorsolateral one (involving the inferior parietal lobule and the ventral premotor cortex) dealing with the preshaping of the hand according to the features of the object. The present study aimed at testing whether a pivotal component of the dorsomedial pathway (area V6A) is involved also in hand preshaping and grip formation to grasp objects of different shapes.Two macaque monkeys were trained to reach and grasp different objects. For each object, animals used a different grip: whole-hand prehension, finger prehension, hook grip, primitive precision grip, and advanced precision grip. Almost half of 235 neurons recorded from V6A displayed selectivity for a grip or a group of grips. Several experimental controls were used to ensure that neural modulation was attributable to grip only.These findings, in concert with previous studies demonstrating that V6A neurons are modulated by reach direction and wrist orientation, that lesion of V6A evokes reaching and grasping deficits, and that dorsal premotor cortex contains both reaching and grasping neurons, indicate that the dorsomedial parieto-frontal circuit may play a central role in all phases of reach-to-grasp action. Our data suggest new directions for the modeling of prehension movements and testable predictions for new brain imaging and neuropsychological experiments.

We used the (14)C-deoxyglucose method to map the functional activity in the cortex of the lateral and medial parietal convexity, the intraparietal and the parietoccipital sulci of monkeys which either reached and grasped a 3D-object or observed the same reaching-to-grasp movements executed by a human. Execution of reaching-to-grasp induced activations in the superior parietal areas SI-forelimb/convexity, PE, PE caudal (PEc); in the intraparietal areas PE intraparietal (PEip), medial intraparietal (MIP), 5 intraparietal posterior, ventral intraparietal (VIP), anterior intraparietal (AIP), lateral intraparietal dorsal; in the inferior parietal areas PF, PFG, PG; in the parietoccipital areas V6, V6A-dorsal; in the medial cortical areas PGm/7m and retrosplenial cortex. Observation of reaching-to-grasp activated areas SI-forelimb/convexity, PE lateral, PEc, PEip, MIP, VIP, AIP, PF, V6, PGm/7m, 31, and retrosplenial cortex. The common activations were stronger for execution than for observation and the interhemispheric differences were smaller for observation than for execution, contributing to the attribution of action to the correct agent. The extensive overlap of parietal networks activated for action execution and observation supports the "mental simulation theory" which assigns the role of understanding others' actions to the entire distributed neural network responsible for the execution of actions, and not the concept of "mirroring" which reflects the function of a certain class of cells in a couple of cortical areas.