What electrodes can be used to measure mu rhythm (de)synchronization in the context of speech comprehension studies? An insight from theoretical analysis

The electroencephalographic (EEG) mu rhythm, or sensorimotor rhythm, is an oscillatory activity that consists of two nonharmonic components in the alpha (8-13 Hz) and beta (~13-30 Hz) frequency ranges. It is considered a neural marker for studying sensorimotor integration during speech perception and comprehension tasks (Saltuklaroglu et al., 2018;Inamoto et al., 2023).Mu rhythm suppression (or desynchronization) is registered primarily over the sensorimotor cortex and serves as an indirect indicator of activity of mirror neuron system (MNS), which is a key neural network mediating action observation/execution and speech processing (Hickok, 2010;Gatti et al., 2017). However, the mu rhythm, as well as its suppression, is not limited exclusively to sensorimotor regions of the brain.As more attention is being paid to mu rhythm reactivity as a marker of speech comprehension, one principal aspect of research is to select the optimal EEG electrodes for capturing mu activity.A substantial body of research that investigated the linkage between speech processing and mu rhythm reactivity was performed on 10-20 systems and typically considered central (C3, C4, and corresponding sites in high-density systems), parietal (P3, P4, Pz), and frontal (F3, F4, Fz) electrodes as regions of interest (Cuellar et al., 2012;Antognini and Daum, 2019;Patzwald et al., 2020;Mikhailova et al., 2021;Salo et al., 2023). Whereas the selection of temporal sites, particularly T3, T4, T5, and T6, has been reported only in limited studies (Belalov et al., 2020).The relatively poor spatial resolution of scalp-recorded EEG complicates accurate localization of mu rhythm signals and their isolation from other oscillatory activity (e.g., occipital alpha rhythm). Besides, speech comprehension tasks suggest evaluating the activity of widely represented cortical networks, including temporal, parietal, and frontal areas, requiring a search for methodologies to define optimal sets of electrodes.This paper aims to shape the arguments and create a comprehensive rationale behind electrode selection in high-density EEG systems for mu rhythm research in speech perception studies.The manuscript is organized in several concise sections that together focus on the interrelation between specific Brodmann areas (BAs), their certain cytoarchitectonic characteristics, and the neurotransmitter systems involved in mu rhythm generation as well as MNS activity and speech processing, thus providing the theoretical support for the strategy of electrode selection. The results of our analysis led to the identification of the temporo-parietal electrode cluster of electrodes (over BA22) as the most optimal for measuring mu reactivity in speech comprehension tasks, whereas frontal, central, and part of temporo-parietal clusters were assigned auxiliary roles.Reviewing and highlighting the brain regions (BAs) crucially associated with language comprehension in this section is the first step in building a rationale for EEG electrode selection.Certain aspects of the physiological roles of the canonical cortical language areas and their anatomo-functional interactions remain a subject of debate.Wernicke's area (mainly BA 22,BA 21,BA 41,and BA 42) has long been postulated to have the highest priority on speech comprehension. At present, additional BAs (BA20, BA37, BA38, BA39, and BA40) are considered to be a part of the so-called "extended Wernicke's area" (Ardila et al., 2016).There is a compelling reason to believe that the classical concept positing Broca's area (BA44, BA45) as an exclusively motor speech center is too simplistic. Together with supplying motor speech, this region is responsible for various functions in language comprehension. For instance, there is data indicating that the inferior frontal gyrus (IFG) pars opercularis (BA44) and especially pars triangularis (BA45), parts of Broca's area, are heavily implicated in processing syntactic information and semantic analysis (Liuzzi et al., 2024). Thus, the "classical" roles of Wernicke's and Broca's areas are not isolated; rather, they function as interrelated high-order hubs within a broader linguistic system. This integrated view is formalized in contemporary neurobiological models of language. A widely accepted dual-stream model of speech and language processing proposed by Hickok and Poeppel posits that sensory information is routed into two distinct but interacting neural pathways (Hickok, 2022). Within the dualstream model, the temporo-frontal extreme capsule fasciculus (TFexcF) constitutes the ventral stream of language processing. It is involved in the mapping of auditory speech signals into conceptual and semantic representations, thus primarily contributing to speech comprehension (López-Barroso and de Diego-Balaguer, 2017; Weiller et al., 2021;Barbeau et al., 2024). TFexcF represents the anatomical connection between the pars triangularis (i.e., BA45) and the superior temporal sulcus (BA21, BA22) and the medial temporal gyrus, which is primarily associated with BA22 (Barbeau et al., 2024).Anatomically, the dorsal network comprises suprasylvian projections (primarily arcuate fasciculus) between the posterior superior temporal/inferior parietal (roughly BA40 and BA22 to some extent) and inferior frontal gyrus (BA44/45) and premotor cortex (BA6) (Ries et al., 2019). But limited research reports that cortical terminations of the arcuate fasciculus reach BA44, BA45, BA46, BA47, BA6, and BA9 in the frontal lobe (Rilling et al., 2008).While the dorsal stream is predominantly involved in speech production, its activity is considered most important for speech perception during the language acquisition phase. It's also thought to support sensorimotor integration for auditory sequences and supply syntactic processing (Rauschecker, 2011;Ries et al., 2019). This significantly blurs the line of a strict functional separation between motor and sensory speech systems and suggests that language comprehension relies substantially on the dorsal stream in addition to the ventral network. In this regard, there is an intriguing paper by Ono et al. that demonstrated the bidirectional neural activity between Broca's and Wernicke's areas during interactive verbal communication in listeners, whereas speakers were characterized by only a unidirectional relationship (Ono et al., 2022).The BAs initially identified in this section (and summarized in Table 1) represent basic anatomical components, each a potential target for selection strategy. Table 1 could serve as a central source of reference that contains information about the rationale discussed in more details in the following sections.Table 1. Domains of evidence supporting the rationale behind the strategy of electrode selection.The table shows a set of criteria with different priority levels was established to guide the selection of cortical areas and their corresponding EEG electrodes. The anatomofunctional criterion (called "relation to the linguistic system") has been determined as the primary and highest priority criterion. This prioritization was driven by two factors: (1) the nature of the issue under investigation and (2) the relatively localized distribution of these anatomic substrates. The remaining domains are regarded as lower-order criteria that provide supportive justification for the selection. Following the reasoning outlined in the sections below, the severity of the BAs' mirror properties has been defined as the most important lower-order criterion, while neurochemical profile was assigned lesser importance.Thus, the dual-stream model by Hickok and Poeppel indicates that speech comprehension is mediated by the coordinated work of both ventral (including Wernicke's extended area) and dorsal (including Broca's area) processing networks. Keeping this in mind, and given the (1) close anatomical proximity of mirror neurons to cortical language areas, (2) strong association of the dorsal stream with sensorimotor integration, there is a substantial functional link that might reasonably be expected between the dorsal and ventral networks and the MNS (described in more details below).There is a lack of consensus within the scientific community regarding whether mu rhythm suppression reliably reflects the functioning of the MNS.However, since sensorimotor (mu) rhythm is still considered a potential index of MNS activity, the identification of cortical areas integral to both the MNS and the language network could provide an important part of a theoretical basis for our strategy, emphasizing the focus on language-relevant BAs discussed in the last section.That is, in this section we try to prove that areas with "mirror" properties can be considered as nodes where language comprehension and sensorimotor integration (indexed by the mu rhythm) converge.Regions with mirror properties might be involved in simulating the action during understanding of action-related language, which is discussed in literature under the term "embodied language processing" (Hoedemaker and Gordon, 2014;Tian et al., 2020). At present the relevance of this concept has not diminished; rather, it has intensified. For instance, mental imagery of words with a motor component is thought to be implicated in the comprehension of figurative speech (Garello et al., 2024).Consequently, the embodied view of language comprehension could provide a conceptual link between the functional properties of the MNS and the cognitive processes underlying speech comprehension.According to Pineda et al., three cortical areas (i.e., inferior frontal gyrus, inferior parietal lobule, and superior temporal sulcus) could be conceptualized as the "core" of the MNS in humans (Pineda, 2008). Similar to this view, but more specifically, the meta-analysis by Molenberghs et al. identified BA44, BA7, BA9, BA6, BA40, and, to a large extent, BA22, BA45 as the most frequently reported lateral cortical areas (BAs) exhibiting mirror neuron activity (Molenberghs et al., 2012). As discussed above, BA22, BA40, BA44, and BA45 are considered the cortical areas directly involved in speech comprehension, whereas the involvement of BA9 and BA6 in this function is not as obvious. Nevertheless, there is sufficient evidence supporting the importance of BA9 for strategic inference processes during language comprehension (Chow et al., 2008). And BA9/46 is regarded as critical for understanding idiomatic expression (Sela et al., 2012).Neuroimaging studies provide substantial support for the centrality of BA6, 44, 9, 46, and 40 in working memory processing (Cabeza and Nyberg, 2000;Ramsey et al., 2004). Scientific literature emphasizes the critical role of working memory in language comprehension by maintaining semantic and phonological information (for example, see Martin, 2021). In this context it is important to note that MNS not only mediates action-related semantics but is also known to contribute to phonological working memory (Jairew et al., 2025).Despite the fact that human BA6 is traditionally thought of as a "motor" area, data are gradually being collected indicating its involvement in speech understanding. Distinct neuroimaging studies demonstrate that motor areas, namely BA4a/6, are activated by listening to speech (for example, Wilson et al., 2004). Likewise, available literature discussed the role of the ventral part of BA6 in phonological processing (Hagoort, 2005).Thus, a significant portion of the Brodmann areas within the MNS overlaps with the regions that form the main cortical substrates for speech comprehension.Altogether, these arguments support the belief that MNS is fundamentally integrated with the language comprehension network, and assessing mirror activity with mu rhythm over the identified Bas could indirectly estimate the contribution of MNS to speech perception and comprehension processes.It is generally accepted that mu/alpha activity relies heavily on the laminar architecture of the cortex (i.e., mu activity seems to be significantly layer-specific). Therefore, in the context of our discussion, regional cytoarchitectonic differences can be employed as a filter to guide the selection of electrodes.Although multiple cortical layers contribute to EEG signals, layer (L) 5 pyramidal neurons play a major role due to their size and the perpendicular orientation of their apical dendrites to the cortical surface (Kirschstein and Köhling, 2009). Likewise, neurons of layer 5 are considered to be critical for the generation of alpha/mu activity (Haegens et al., 2015;Scheeringa et al., 2016), especially given their extensive connections with thalamic nuclei. However, the significance of supragranular cortical layers for alpha activity is still subject to discussion (Scheeringa et al., 2016). Particular emphasis in this regard is given to L3 by reason that it contains a significant number of pyramidal cells and the thalamus sending strong projections to this layer of the cortex (especially primary motor and somatosensory cortices). Since mu rhythm is a thalamocortical phenomenon and considering that pyramidal cells in L3 can modulate firing of L5 neurons, activity of L3 neurons at least contributes to the alpha/mu-range oscillations.L3 pyramidal neurons form horizontal excitatory connections between different cortical regions and mediate higher-level cognitive functions, including speech (Larsen et al., 2022). For instance, a study performed by Moerel et al. illustrates that neuronal populations in superficial layers of primary auditory cortex displayed an increased complexity of sound processing but were characterized by slower responses than neurons of L4 (Moerel et al., 2019). It was therefore suggested that primary auditory cortex supply complex auditory processing in humans together with physical sound analysis and indicate the importance of L3 neurons for processes underlying speech understanding. A paper authored by Zachlod et al. illustrates that BA22 is characterized by dense layers 5 and 2/3 (in the upper bank of the superior temporal sulcus -STS1) and well-defined large pyramidal cells in L3 in the temporal area Te3a posterior part of BA22 (Zachlod et al., 2020). The BAs 20 and 21 have a smaller layer 3 than BA22 but a wide L5. L3 and L5 are fairly noticeable in BA39 and BA40 but less prominent than in BA22 and L5 of BA40 is smaller than that of BA39 (Zilles, 2004).The thickest L3 among all BAs is observed in BA10, which is only indirectly (through the maintaining of working memory) involved in speech perception (Burgess and Wu, 2013). L3 in BA10 consists of large pyramidal neurons, and it has an extensive dendritic branching network, highlighting its role in higher-order (associative) functions. Together with BA10, adjacent BA9 also has well-developed L3 with large pyramidal cells and is characterized by extensive associative projections (Prkačin et al., 2024). Interestingly, BA4, which corresponds to the part of the central electrodes, is characterized by prominent L3 and 5, comprising together 70% of the cortical thickness of this area (Alan et al., 2023).BA44 is a dysgranular area characterized by large pyramidal cells in L3 and in L5. BA45 differs from BA44 by the presence of a well-developed L4 and strikingly large pyramidal neurons in the deeper part of L3 (Petrides, 2005).Thus, it may be concluded that cytoarchitectonic features of MNS-and language-related BAs, namely the thickness of L3 and/or L5, can also guide the selection of BAs and associated electrodes as appropriate for capturing mu activity during a specific task.This section represents an additional justification for strengthening the logical framework of our conception by delineating the contributions of specific neurotransmitter systems supporting speech and language processing and underlying the modulation (rather than of mu It be that of reasoning are in evidence has been to a comprehensive of the of different involvement in the processes of mu rhythm generation and and systems are major to the in this section we on two other neurotransmitter and This is due to several (1) both systems the mu activity, (2) the and systems are (for example, see and of these systems in the lateral cortex is more localized than and and systems are with MNS et al., and and contribute to speech comprehension et al., et al., 2020).The evidence and linkage between the language network (especially and the system. is data from a study more of comprehension of action-related words (i.e., embodied in than that this specific could be a of the of the pathways to the cortex et al., neurons from the extensive connections with cortical neurons of the primarily BA and BA et al., 2024). In there is limited data indicating that the inferior frontal is to et al., The study by et al. could serve as an example, highlighting the between the and the fasciculus and the role of the system in the neurochemical of speech et al., is widely that a role in the functioning of the working which is for language, especially phonological and semantic processing et al., and study a performed by et al. demonstrated increased in BA45 and in both and ventral parts of in the but not in parietal areas et al., et al. available data from research on the distribution of neurotransmitter in the human that are relatively in several regions of the lateral BA46, BA7, BA39, BA and et al., are with some For instance, the paper authored by et al. reports about the of in of the area) cortex in humans et al., study a expression of in the and temporal (BA20, BA22) with of in deeper layers et al., studies demonstrate that cells are in L5 of the primary motor cortex known as and the and cortex while the of neurons were observed in 2/3 of these et al., et al., 2024). and in their have identified areas and and 21 as regions with of and is, the relatively expression of in the MNS regions associated with language comprehension the role of the system in the functioning of both neural networks (i.e., the MNS and language to be directly involved in the of the system. evidence suggests that this is by distinct of in the of in activity, whereas (in pathways (for more see et al., 2004). are the pathways contributing to the of sensorimotor that in significantly sensorimotor rhythm et al., 2012). It be that for alpha/mu the of levels to less of et al., distribution of and their on is also with the that L5 is for mu rhythm whereas L3 is more important to its consensus is that is a to this it is that research in the perception of in speech et al., 2023). evidence to the fact that a significant role in neuronal For instance, of activity et al., 2022). It is also known that can modulate neuron et al., 2017). Interestingly, layers and of than other layers in et al., 2017). has been to EEG activity in the alpha/mu Hz) and beta Hz) the of the activity of mirror neurons by et al., The system is with the cortex and there are projections from neurons to the medial cortex 2020). are limited data between of system and lateral However, some reports indicate that can relatively activity in regions of the lateral including the superior temporal sulcus (BA21, inferior frontal gyrus (IFG) pars triangularis and inferior parietal BA40) and premotor cortex et al., et al., 2013). has also been to modulate mirror neuron activity, especially in the context of et al., the and systems can modulate activity in the mu over the regions in speech comprehension, and the of this modulation to be The expression levels of and in L3 and L5 support the rationale for these areas as of the our strategic to electrode the context of our this line of serves as an important Nevertheless, work is to whether the discussed neurotransmitter systems are significantly to mu reactivity over the suggested to suggest optimal electrodes for of mu reactivity within speech comprehension we an strategy from several theoretical of to the cortical areas (BAs), a set of selection criteria has been in of (1) The BA is a component of the dual-stream model of speech processing priority to the ventral and it has an established role in speech comprehension. (2) The BA has been identified as a cortical region exhibiting mirror properties in The BA a well-developed cortical layer and which are considered to be critical for the generation and modulation of thalamocortical in the mu BA shows a significant presence of and systems involved in to from cortical regions to EEG sites, we employed data from mapping we the of the studies by et al. and and which provide the of EEG and their underlying brain regions et al., and 2022). electrode was considered a for it was associated with one of our BAs in table that BAs not the and may be suggested for the our view, and and since they are to and can be as a this with the discussed strategy. (1) BA22 is a of Wernicke's area and an integral part cortical of the ventral stream that primarily auditory comprehension as well as part of (2) BA22 has a significant of and it has a prominent L3 and L5 with a of which play an role in mu and may given that these electrodes are adjacent to and C4, the classical mu rhythm And also into that the BA40 on which these electrodes are to key regions of However, we suggest these sites as the of the given that the BA40 has priority on speech comprehension and L3 and L5 than electrodes can be as P4, represent a and F4, and a The of frontal electrodes as of is by the fact that BAs where they are (i.e., areas and are primarily associated with higher-order functions (e.g., working cognitive that support language comprehension The underlying the of selection can be represented as a Thus, we the view discussed (for instance, see et al., that electrode selection for of mu activity be determined primarily by the And regions of interest be on the functions of cortical areas to a than a of areas with mu a several important be the of EEG electrodes on the (for instance, is as they are by motor for a specific the of a large electrode set is proposed strategy is theoretical and The main interest is the of the electrode sites, the in with classical central (C3, C4, and as the most for mu of suggests a for as it the of activity within the ventral and dorsal with the distribution of over the proposed sites within speech comprehension of in this study is a that as it may significantly the and potential of our For instance, distribution of neurotransmitter and laminar data substantially from primary and their to the temporal and parietal association in humans is thus