Neural correlates of manual action language: Comparative review, ALE meta-analysis and ROI meta-analysis

Highlights

• We compared action language to motor observation, imagery and execution.

• A systematic review, meta-analysis and ROI analysis were conducted.

• Action language is more similar to observation than to imagery and execution.

• The core network for motor execution is not involved in action language processing.

Abstract

Despite decades of research, the nature of the involvement of the motor system in action language processing is still controversial, and little is known about how processing action language relates to observing, imaging and executing motor actions. This study combines a systematic review of the literature, an ALE meta-analysis and a region-of-interest (ROI) meta-analysis to provide the first complete (qualitative and quantitative) account of the motor-related functional network involved in action language processing in comparison to activation reported during motor observation, motor imagery and motor execution. The review of the literature revealed that the methodology of action language studies differed considerably from the methodology of other motor-related processes which may have contributed to blurring the interpretations over the years. The ALE and ROI meta-analyses showed that the functional network of action language was more similar to observation than imagery and finally execution, following a motor gradation. Overall, our results point towards a more cognitive, as opposed to purely motoric, involvement of the motor system during action language processing.

Introduction

The cerebral network of action language has been the focus of extensive research in the last decades and a wide range of theories on language embodiment have been developed (e.g. Binder and Desai, 2011; Gallese and Lakoff, 2005; Glenberg, 2015; Mahon and Caramazza, 2008; Pulvermüller, 2013; Zwaan, 2014). However, the specific role of the motor system in action language processing and the extent to which action language is embodied remain controversial (Meteyard et al., 2012; Zwaan, 2014). Functional magnetic resonance imaging (fMRI) studies have played an important part in demonstrating the involvement of the motor system during the processing of action language (e.g.Hauk et al., 2004; Kemmerer et al., 2008; Labruna et al., 2011; van Dam et al., 2012). Evidence has accumulated that suggests that the relationship between action language and the motor system varies as a function of several factors, including context, expertise, attentional focus and semantic control demand (Davey, 2015; Moody and Gennari, 2010; Yang, 2014; Zwaan, 2014). This suggests that the involvement of the motor system in action language processing is complex but important questions remain. What is the nature of this motor involvement? What are the motor-related neural mechanisms at play during action language? Of particular interest are the studies that directly compared action language with other motor-related processes. Such comparisons have revealed similarities between the action language network and the networks of motor observation (Aziz-Zadeh et al., 2006; Meister and Iacoboni, 2007), motor imagery (Yang and Shu, 2014) and motor execution (Moody-Triantis et al., 2014; Papeo et al., 2012; Peck et al., 2009). However, a lack of overlap between the neural networks of action language and these other motor-related processes has also been reported (Tomasino et al., 2007; Tremblay and Small, 2011; Willems et al., 2009; Zhang et al., 2018). Thus, agreement about the relationship between the neural networks supporting action language processing and those supporting other motor-related processes has not been reached yet.

Meta-analyses provide a quantitative methodology to summarize empirical data, which may help in reaching a consensus. Meta-analyses have been conducted to characterize the cerebral network of action language (Jirak et al., 2010), motor observation (Caspers et al., 2010) and motor imagery (Hétu et al., 2013), providing a comprehensive overview of the neural networks involved in each of these processes. A more recent meta-analysis compared motor observation, imagery, and execution (Hardwick et al., 2018) and revealed that motor execution shared more activation sites with motor imagery than with motor observation. A systematic meta-analysis comparing action language, motor observation, imagery, and execution would provide unique and important insights into the motor processes involved in action language processing.

The main objective of the present study was to provide a comprehensive analysis of the fMRI literature on action language processing and its links with other motor-related processes (motor observation, imagery, and execution) through a systematic review of the literature and a two-fold meta-analysis (ALE and ROI). The aims of the systematic review were to generate a clear portrait of research on the neural correlates of action language processing and to examine whether this literature has explored action/movement-related parameters that have been explored in other motor-related fields of research. The overall aim of the ALE meta-analysis was to characterize the motor-related network of action language processing in comparison to those of motor observation, imagery, and execution. The first specific objective was to compare action language, motor observation, and motor imagery to motor execution in order to confirm the existence of a gradation in the similarity of the neural networks associated with these processes to the motor execution network. Considering that the motor execution network is more similar to the network of motor imagery than motor observation (Hardwick et al., 2018), and that the network of action language is more similar to motor observation than execution (Rueschemeyer et al., 2014), we expected action language to be the least “motoric” of the following motor-related processes on a motor gradation: motor execution – motor imagery – motor observation – action language. The second specific objective was to determine to which motor-related process the action language network was most similar. Since the action language network has been shown to be more similar to motor observation than to motor execution (Rueschemeyer et al., 2014), we expected the degree of similarity between action language and other motor-related processes to follow the previously mentioned motor gradation: the neural network of action language would be more similar to that of motor observation, then to motor imagery and finally to motor execution. The third specific objective was to determine whether the action language network was more closely related to movement or action processing. We hypothesized that the action language network would be more similar to the action compared to the movement processing network. Finally, an anatomical ROI meta-analysis aimed to provide complementary information regarding the functional network for action language by comparing, region by region, the bilateral language activation to the profile for motor imagery, observation, and execution.