Our study aimed at providing novel insights on the neural correlates underlying pantomimed movements at two different levels. First, we wanted to characterize how action- and goal-related information are encoded within the tool network during the planning and execution of this type of movements. Second, we aimed at providing a description of the functional interactions between the ventral stream and parieto-frontal motor pathways happening during pantomimed movements.
With respect to the second aim, the different spatial distribution of goal encoding during the two phases of the task suggested a possible exchange of information from the ventral stream and/or the parietal regions to the frontal cortex. In addition, our connectivity analysis showed an increased bidirectional coupling between pMTG and PMv during the execution of pantomime of tool use, suggesting a possible supportive role of ventral stream in orchestrating the normal performance of this type of movements. We will discuss further the implications of our findings in the next sections.
Representation of action information during pantomimed movements
A recent MVPA investigation distinguished the functional selectivity of the regions within the tool network, by decoding specific movements (grasping vs. reaching) performed either with a real tool or only with the hand (Gallivan et al.
2013c). During planning, Gallivan et al. (
2013c) showed that the intention to perform hand and tool movements are represented differently within the parieto-frontal networks and the ventral stream. Several regions within the network hosted overlapping but separate representations for planned hand and tool movements (M1, aIPS) or showed representations of shared goals across effectors (PMd, PMv, posterior IPS). The authors identified cortical areas representing only planned hand movements [i.e. SPOC, extrastriate body area (EBA)] from areas encoding just tool movements (i.e. pMTG, SMG)—suggesting a specialization of these last regions for real tool actions.
Our MVPA study complemented these results showing that—during planning—even specific exemplars of pantomimed movements were represented within similar regions of parieto-frontal networks (PMv, PMd, aIPS, SPL, SPOC). Moreover, upcoming pantomimes were also represented within areas of the tool network (SMG and pMTG) previously shown to be selective only for movements performed with real tool—even if our task did not involve interactions with real objects.
Of interest for the present study is the description of action encoding within the temporal lobe during movement planning. This is in line with the increasing number of evidence showing that the lateral occipito-temporal cortex (LOTC), comprising pMTG, represented specific motor features of upcoming actions, such as the adopted effector and the type of hand-object interaction (Lingnau and Downing
2015; Gallivan and Culham
2015). Following this interpretation, decoding of action information within pMTG might represent the specific parameters about upcoming (pantomimed) tool movements (Gallivan et al.
2013c) and/or a specific hand-tool relationships (Bracci et al.
2012; Bracci and Peelen
2013) which might be used to define a final desired state template to be achieved during action execution (Buxbaum
2017).
Overall, action-related encoding within the tool network suggested that pantomimes are a specific hand movement category which relies, at least partly, on the same cortical regions engaged during real tool actions by recruiting tool-selective regions of the temporal and parietal cortex.
During the planning and the execution of pantomimes, our cross-decoding MVPA showed goal encoding within the temporal (pMTG) and parietal nodes (aIPS, SPL) of the tool network. Other regions (PMv, SMG) showed decoding of goal-related information only during the execution phase of the task. To be fully appreciated, our results need to be considered within a broader perspective to understand the possible role of temporal and parietal regions in pantomimed movements.
A general framework to interpret our results might come from a recent investigation on the simple observation of hand and tool images (Bracci et al.
2016). Bracci et al. (
2016) showed that LOTC, possibly comprising also pMTG, hosted neural representations of action and object category, whereas the aIPS hosted only action information. The authors proposed that LOTC might integrate object and action information, representing a critical hub for linking more posterior temporal regions, involved only in perceptual processing, with parietal areas—involved in representing action information (Bracci et al.
2016).
Following this line of interpretation, decoding of goal information during action planning in pMTG might be related to the representation of the final goal to be obtained irrespective of the specific means and adopted tool. In parallel, the recruitment of the ventral stream during action preparation might also support the retrieval of semantic information about tools—i.e. their general function irrespective of the specific means and memory-based experiences about their typical use—which needs to be integrated with goal information into the to-be-performed pantomime (Lingnau and Downing
2015). Bidirectional exchange of task-relevant information between IPL (aIPS, SMG) and LOTC (pMTG) support the idea of an integrative role of LOTC (Almeida et al.
2013; Garcea et al.
2019). LOTC might be particularly relevant in planning pantomimes, as in this type of movement, the information about the tool is internally generated.
With respect to the parietal cortex, we showed encoding of goal information also within two regions of the IPL (aIPS and SMG) which might subtend different functional roles during pantomimes. Goal encoding in aIPS was evident during both planning and execution—in line with previous investigations (Gallivan et al.
2013b; Turella et al.
2020)—whereas SMG hosted goal-related information only during execution.
Recent fMRI studies adopting MVPA (Chen et al.
2016,
2018) reported the representation of similar goal-related information within IPL during the execution of pantomimes of tool movements. The first investigation (Chen et al.
2016) showed that tool-preferring regions of both parietal and temporal cortex encode goal-related information. As in our study, it was possible to decode goal information—invariant to the identity of the adopted tool—within IPL (aIPS and SMG) during pantomime execution. A further confirmation of goal encoding within IPL was evident in a subsequent study of the same group (Chen et al.
2018). Here, cross-decoding for goal-related information was possible within both SMG and aIPS during two different tasks: the execution of tool pantomime and the identification of visually presented tools. Nevertheless, only within SMG, decoding of goal information was independent from the performed task (Chen et al.
2018). This last result suggested a possible dissociation in the representational content hosted within aIPS and SMG.
Hints for a possible dissociation between the causal role of aIPS and SMG during pantomimed movements come from a recent lesion study (Watson and Buxbaum
2015). In this investigation, aIPS and pMTG were associated with a general impairment in the performance of tool pantomime (Watson and Buxbaum
2015), whereas SMG seemed to play a different role, more related to action selection (Buxbaum
2017). Indeed, lesions in SMG (and IFG) were associated with impairment in the selection of the to-be-performed pantomimed action (e.g. move vs. use) among possible candidates (Watson and Buxbaum
2015). Information from pMTG and aIPS might be the input to SMG, providing the possible candidates to be selected (Buxbaum
2017). This transfer of information might be supported by the temporal difference in encoding—from the planning to the execution phase—shown in our data, suggesting a transfer of goal information from pMTG and/or aIPS to SMG.
Significant decoding of goal information was also shown in superior parietal cortex, i.e. within SPL and SPOC. This information might be related to the maintenance of a representation of the intended general outcome to be subsequently performed (move vs. use) which might be adopted for subsequent online monitoring within both SPL and SPOC.
Still, SPOC showed a unique pattern of results, as it showed an opposite pattern with respect to other regions (SMG, PMv), shifting from representing action and goal information during planning to representing only action information during execution. This finding seems to support the central role of SPOC in the online control and guidance of an action and in the transformation of object-related information into a possible motor program (Gallivan et al.
2011; Vesia et al.
2017) in line with neurophysiological work on caudal parietal areas in monkeys (Filippini et al.
2018; Hadjidimitrakis et al.
2019).
Our results confirmed and extended the role of parietal (SMG, aIPS, SPL, SPOC) and of the ventral stream (pMTG) as crucial cortical hubs for processing goal information, further characterizing the different contribution of these regions during pantomimed movements.
Functional interactions between ventral stream and parieto-frontal motor pathways during pantomimed movements
It is difficult to directly compare MVPA and DCM results, as these two methods capture different aspects of fMRI signal. Nevertheless, these methods allowed to describe the communication between ventral stream and parieto-frontal motor pathways from two different but complementary perspectives. Our MVPA results showed a possible transfer of abstract goal information from the ventral stream and/or the parietal cortex to PMv, whereas DCM showed a task-specific bidirectional functional interplay between the temporal and frontal cortex.
Overall, our findings are in line with recent investigations supported the possible exchange of information between ventral stream and parieto-frontal motor pathways during hand actions (van Polanen and Davare
2015; Milner
2017; Hutchison and Gallivan
2018). Nevertheless, the dynamics of this interplay are still poorly understood (Cloutman
2013). Our study contributed to better understand these interactions in two ways.
First, we showed differences in the spatial patterns of goal encoding during the planning and execution of pantomimed movements. Our results suggested a possible transfer of information from the posterior nodes of the network (pMTG, aIPS), where the encoding of goal information is stable through time, to the more anterior one (PMv). There are two possible, and not mutually exclusive, alternatives for information to reach the frontal cortex: in one case, information could be transferred from pMTG to aIPS and from there (or indirectly from SMG) to PMv, in the other case, information could be transferred from temporal regions—comprising pMTG—to the IFG and then to the premotor cortex (PMv).
Our findings showed that the representational content within the tool network changed flexibly according to the evolution of the movement, supporting a dynamic representation of goal-related information. Goal information might be transferred from posterior to anterior regions during the unfolding of the movement. This exchange of information might be possible also through the functional interactions between ventral stream and parieto-frontal motor networks.
Second, our connectivity analysis provided complementary evidence for an interaction between these pathways. In this case, we described a reciprocal exchange of information between the frontal, parietal and temporal nodes of the network occurring during the execution phase of the task (intrinsic connectivity). Looking at the specific effect of our grasp-to-use condition (modulatory effect), we showed that pantomiming the use of a tool selectively enhanced the functional coupling between the pMTG and the PMv in both directions. This interplay might subtend a continuous exchange of information between these two pathways. Communication in one direction (PMv > pMTG) could consist in comparing the actual state of the movement stored in premotor cortices with the planned end-state stored in temporal regions. The other interaction (pMTG > PMv) might represent a feedback providing information on the comparison between the action which is performed and the originally planned end-state.
The modulatory effect of the grasp-to-use task demonstrated enhanced communication between the ventral and the dorsolateral pathway, but mainly through the connections between temporal and frontal regions. Our results supported the indirect conclusions of a study (Vry et al.
2015) combining fMRI and tractography. Based on fMRI activation patterns, tractography identified a ventral pathway, connecting temporal cortex to frontal regions, as the main cortical route specific for object-directed pantomime. Moreover, our data are in line with a recent fMRI investigation (Garcea and Buxbaum
2019), even if the authors considered different nodes of the tool network and adopted a different connectivity measure. This study (Garcea and Buxbaum
2019) showed increased functional connectivity between parietal, temporal and frontal nodes of the network during the planning and execution of pantomime of tool use. Our results extended these findings providing a description of the nature of the bidirectional functional interplay between temporal and frontal regions during tool use pantomime.
Overall, if we look at our MVPA and DCM results from a general perspective, we provided evidence for the involvement of pMTG—and possibly of other LOTC regions—in the planning and execution of pantomimed movements, but it is still difficult to define their functional role in motor control. During planning, LOTC and the dorsal stream might exchange information about the properties of the to-be-grasped objects (e.g. its function), of the upcoming action and/or of its expected sensory consequences (Gallivan and Culham
2015).
During execution, information hosted within LOTC and exchanged with the parieto-frontal pathways might be different. A possibility could be that an efference copy of the executed action hosted within LOTC might be represented and adopted for online monitoring and possible corrections. A complementary interpretation might point towards the possible transfer of information related to the intention behind the pantomimed action, mediating the communicative side of pantomime (Goldenberg
2017; Finkel et al.
2018).
Irrespective of the possible interpretations, our data highlighted the pivotal role of temporo-frontal bidirectional interactions for the performance of meaningful pantomimed movements. Further studies are needed to better understand the interactions within the tool network underlying this type of movements, unveiling also the possible role of homologous regions within the right hemisphere, as hinted by a recent investigation (see Watson et al.
2019).