Demonstrating the causal role of the motor and the inhibitory neural networks in the comprehension of action language

Abstract

Extended summary In the last few decades, the notion that cognition is embodied, namely that it is grounded on the sensory and motor experiences has become an important approach in cognitive science. In the field of language, the embodied cognition (EC) approach posits that meaning understanding requires sensory and motor simulations of objects, events and situations described by words and sentences. With the term “simulation”, we refer to a reactivation of perceptual, motor and introspective states that are acquired during experience with the world, the body and the mind. At the level of neural mechanisms, this means that language comprehension involves, to some extent, the activation of the same brain regions associated with the real world experiences that words and sentences refer to (Fischer & Zwaan, 2008; Gallese & Lakoff, 2005; García & Ibáñez, 2016; Jirak et al., 2010; Taylor & Zwaan, 2008). The embodied cognition approach to language has received considerable empirical support (see below), particularly from studies showing that understanding action-related language is associated with increased activation of brain motor regions. However, this neural embodiment is considerably reduced when a negation marker is inserted into action sentences, leading to a disembodiment effect. In other words, the negation acts like a gate that reduces the accessibility of the negated concept, expressed at neural level by a decrease of motor activity. Most of the literature on embodiment and disembodiment (in negation) effects reports correlational results, that is, it offers evidence of an association between language comprehension and neural activity of the motor cortex. By contrast, using non-invasive brain stimulation methodologies, the experiments included in the present thesis provide evidence that such activity of the motor cortex is causally involved in the comprehension of action language. Let us briefly summarize the state of the art of research on embodiment and disembodiment of language processes. To date, a large number of behavioral studies using a dual-task paradigm, reported that the processing of sentences referring to a directional action significantly influences the execution of a directionally compatible motor response (action-compatibility effect paradigm, ACE), suggesting that the motor areas activated in the action execution are also implicated in the comprehension of action-language (Borreggine & Kaschak, 2006; Boulenger et al., 2006; Buccino et al., 2005; Dalla Volta et al., 2009; de Vega et al., 2013; Glenberg & Kaschak, 2002; Kaschak & Borreggine, 2008; Sato et al., 2008; Taylor & Zwaan, 2008; Zwaan & Taylor, 2006). Further evidence was provided by several functional Magnetic Resonance Imaging (fMRI) experiments that revealed a somatotopic activation of the motor areas involved in the execution of the same movement expressed by action-related word (Olaf Hauk et al., 2004; Kemmerer et al., 2008; van Dam et al., 2010) and sentences (Aziz-Zadeh et al., 2006; de Vega et al., 2014; Desai et al., 2010; Tettamanti et al., 2005), and that such activation is context dependent rather than automatic (Raposo et al., 2009). Another set of studies using electroencephalography (EEG), provide additional information, indicating that the motor cortices are recruited in action-language processing at an early temporal window after the word presentation (~ 200 ms) (Dalla Volta et al., 2014; Olaf Hauk & Pulvermüller, 2004; Pulvermüller et al., 2001), and that, this involvement is expressed by a suppression of mu and beta rhythms similar to the one observed during action observation or execution (Moreno et al., 2013, 2015; van Elk et al., 2010). Finally, neurophysiological studies suggest that, single-pulse transcranial magnetic stimulation (sp-TMS) applied to the primary motor cortex (M1) selectively influences the behavioural response for action words (Pulvermüller, Hauk, et al., 2005; Tomasino et al., 2008) and the comprehension of action-related language modulates motor cortical excitability (Buccino et al., 2005; Candidi et al., 2010; Innocenti et al., 2014; Papeo et al., 2009). Overall, these evidences mainly point out that understanding language embedded in an action context engages the sensory and motor systems; however, they failed to assess the “necessity question”. This question refers to the debate about whether motor activation is causally necessary in the processing of action-related meaning or is just an epiphenomenon, which occurs as a post-lexical simulation process with no relation to meaning (Leshinskaya & Caramazza, 2016; Mahon, 2015; Mahon & Caramazza, 2008; Papeo et al., 2013). One way to resolve the question is by means of causal techniques like repetitive Transcranial Magnetic Stimulation (rTMS) or transcranial direct current stimulation (tDCS), with the appropriate experimental designs. For instance, temporarily perturbing M1 activity by offline low-frequency rTMS delayed the processing of action words in a morphological task (Gerfo et al., 2008) and in a semantic task (Repetto et al., 2013), which indicates a functional relation between M1 activation and action language understanding. In the same line, online rTMS on M1 impaired the comprehension of action-related language in a priming semantic task (Kuipers et al., 2013) and in a concreteness judgment task (Vukovic et al., 2017). Compatibly, inhibitory tDCS on M1 reduced the learning of a novel action word (Liuzzi et al., 2010). However, some “paradoxical” effects have been reported; that is, the comprehension of action language improved, rather than getting worse, after the perturbation of the premotor cortex via rTMS (Willems et al., 2011) or tDCS (Gijssels et al., 2018; Niccolai et al., 2017). The above results considerably support the idea that the motor system plays a causal role in the processing of action-related meaning. However, it should be noted that the experiments conducted so far not always give consistent results in the direction of the stimulation effects (facilitatory vs. inhibitory). In addition, they neglected some important issues. First, they focused mainly on evaluating the effect of inhibitory protocols (“virtual lesions”), paying less attention to the impact of the enhancement of motor activation on action language processing, which could be done using facilitatory stimulation protocols. Second, they usually collected online behavioral measures (i.e., reaction times to semantic judgments), while they did not test the involvement of the motor system in long-term cognitive processes such as memory. Third, when they measure behavioral effects of non-invasive stimulation, they do not monitor at the same time the physiological changes induced by the stimulation. These crucial issues are addressed in detail in the current dissertation. Specifically, one objective is to provide further evidence for the functional role of M1 in the action-related language processing, by assessing for the first time its impact on memory performance for action language. To this aim, the first experiment induced the enhancement of motor activity, through a facilitatory tDCS protocol, to test whether it improves the performance in a language memory task consisting of memorizing action-related and attentional sentences. The participants performed two separated sessions; in one they received offline active tDCS (anodal/facilitatory or cathodal/inhibitory) on M1 and in the other they received sham tDCS, before undertaking the memory task. Additionally, since the stimulation, especially cathodal tDCS, is highly variable between subjects (Batsikadze et al., 2013; Jamil et al., 2017; Wiethoff et al., 2014), it was useful to assess the physiological effects of the tDCS on motor excitability, by measuring the motor-evoked potentials induced by sp-TMS. The results showed that applying off-line anodal tDCS, compared to the sham stimulation, led to a better memory for manual-action sentences but no for attentional sentences. No significant effect was observed for cathodal stimulation. Remarkably, the improvements in memory performance were predicted by the physiological changes induced by tDCS. In fact, the increase on motor excitability positively correlates with a better performance in the memory task, selectively for the action-language. Assuming that, as mentioned before, the comprehension of action-related language requires the sensory motor simulation of the referred actions, sentential negation poses an interesting challenge for this embodiment view. How action sentences with a negation marker are represented or simulated? As neuroimaging studies have shown, negation applied to an action context, reduced the activation in motor and premotor cortices, compared to affirmative action sentences (Tettamanti et al., 2008; Tomasino et al., 2010). Consistently with these results, studies using dual task paradigms (Aravena et al., 2012; Bartoli et al., 2013), typing word paradigms (García-Marco et al., 2019) and passive reading (Foroni & Semin, 2013) revealed that negative action sentences reduced the activation of peripheral muscles, relative to their affirmative counterparts. In addition, processing negative action sentences selectively modulates motor cortex excitability compared to affirmative action sentences, while no similar effect of polarity was detected for abstract sentences (Liuzza et al., 2011; Papeo et al., 2016). Altogether, the evidence suggests that the negation inserted in an action context reduces or blocks the access to the motor simulation of the negated concept, inducing a “disembodiment” effect. However, the neural mechanisms that produce this negation effect are still poorly understood. A recent proposal in this regard is the Reusing Inhibition for Negation (RIN) hypothesis, which, as its name suggests, asserts that the processing of negation reuses the mechanism of inhibitory control (Beltrán et al., 2018; de Vega et al., 2016; Liu et al., 2020). Some EEG studies, with dual-task paradigms, indirectly support the RIN hypothesis, showing that the processing of negative action sentences, compared to the affirmative ones, reduced the power of fronto-central theta rhythms (de Vega et al., 2016) and enhanced the amplitude of the ERP N1 component (Beltrán et al., 2018), in an immediately following inhibition-demanding task (NoGo or Stop trials). These facts are important because both theta rhythms and the N1component are indices of inhibitory activity. On the other hand, pre-setting an inhibitory state (NoGo trials) selectively affects the processing of negative action sentences (Liu et al., 2020). Notably, these EEG effects have an estimate source in the right inferior frontal gyrus (rIFG), which plays a crucial role in the response inhibition system (see Aron et al., 2014; Chambers et al., 2009 for reviewes). Further evidence supporting the RIN hypothesis showed that negative action language slowed the typing of manual action verbs (García-Marco et al., 2019), reflecting an inhibitory effect on motor programs, and increased the cortical silent period (Papeo et al., 2016), an index of the GABAergic inhibitory interneurons activity. Although there are robust empirical results indicating a relationship between the inhibitory system and the “disembodied” effect of negation, they only allow to establish correlational conclusions. Indeed, to ensure the functional influence of the response inhibition mechanism in negation processing, a causal methodology is necessary. In order to fulfil this gap, in the second experiment of this dissertation we used an innovative stimulation technique that combined the registration of motor evoked potentials (MEPs) via sp-TMS on M1 and off-line rTMS on the inhibitory system. This protocol allows to test the causal influence of a target area transiently perturbed on M1 excitability. Specifically, to assess whether the inhibitory mechanism is functionally involved during the process of negation, we perturbed the rIFG activity, a key region of the inhibitory network, and then we measure the left M1 excitability during the presentation of action and attentional sentences, both constructed in affirmative or negative polarity format. In a control group, instead of active stimulation, sham rTMS was delivered over the vertex. Moreover, to evaluate whether the negation effect was exclusively associated with the verb presentation or it extended to later stages of sentences processing, we recorded the MEPs at two loci: the verb and the object. The results showed that in the sham group (baseline), the amplitude of MEP, recorded at the verb, was significantly lower for negative action-related sentences than for their affirmative counterparts and, more importantly, the rTMS over rIFG suppressed this inhibitory effect, leading to a comparable motor excitability for both conditions. No similar negation effect was observed on MEP either for the verb of attentional sentences nor for the object locus of any sentence. This pattern of results indicates the critical role of rIFG as regulator of M1 activation during the processing of negated action sentences. In sum, this dissertation deals with the causal role of neural processes in embodied meaning: 1. The first experiment demonstrated, for the first time that the motor system activity is functionally involved in the memory for action language. 2. The second study provides novel evidence of a functional role of the neural mechanism of response inhibition in the “disembodied” effect of negative action language.