Speech production is a remarkable skill that appears effortless; however, it involves a complex set of hierarchically organized cognitive processes (Hickok, 2012, 2022; Piai & Zheng, 2019). These processes encompass concept retrieval and selection, retrieval of syntactic, morphological, and phonological information, post-lexical planning of articulation, as well as self-monitoring and language control mechanisms (Dell, 1986; Hickok, 2012; Levelt et al., 1999). Neuroimaging research has identified several key neural structures responsible for these functions in healthy individuals, such as the inferior frontal (IFG), superior temporal (STG), and supramarginal gyri (SMG), which are critical for linguistic functions, as well as the cerebellum, primary motor, and somatosensory cortices, which control the mapping between sensory and motor representations (Hickok, 2012). While extensively studied in monolinguals (Hickok, 2012, 2022; W. Zhang et al., 2023), research on the neural architecture underlying speech production in bilinguals remains limited. The present study aims to fill this knowledge gap by investigating the neural networks involved in bilingual speech production, offering potential insights into language recovery in bilingual patients with brain pathologies.
Research investigating first (L1) and second (L2) language production in healthy bilinguals has yielded inconsistent results. Some studies have shown specificities across languages (Gurunandan et al., 2019, 2020; Quiñones et al., 2021; Rodriguez-Fornells et al., 2002; Sierpowska et al., 2018; Xu et al., 2017) while others have revealed overlapping neural substrates (Consonni et al., 2013; Geng et al., 2022; Hernandez et al., 2001; Willms et al., 2011). To shed light on this contradictory scenario, (Tao et al., 2021) conducted a systematic review to summarize functional and structural neuroplasticity findings associated with bilingualism. They concluded that the experience of bilingualism has an impact on brain activity in domain-general control regions (see also Bruin et al., 2021; Pliatsikas & Luk, 2016). These areas include the right caudate nucleus, the anterior cingulate cortex, the left parietal lobe, and the bilateral cerebellum. However, some fMRI studies have also revealed functional changes in language-specific regions, including the left STG, left SMG, fusiform gyrus, and IFG (Xu et al., 2017). While this evidence suggests a shared neural substrate albeit with certain specificities for each language, uncertainty persists regarding whether a dual-language experience alters general, non-language-specific mechanisms, as proposed by Tao’s et al. (2021) meta-analysis, or if it also shapes language-specific neural architecture.
Understanding the bilingual brain becomes even more pertinent as the number of bilingual individuals at risk of neurological diseases rises. Brain damage to the language network, such as in the case of brain tumors, can have a significant impact on an individual's quality of life, affecting one or both languages they speak. The statistics on this matter are alarming: according to the World Health Organization, central nervous system tumors affect around 300,000 individuals per year, many of whom have linguistic profiles that involve multiple languages (Patel et al., 2019). Taking this into account, healthcare centers have incorporated preoperative language mapping protocols into their clinical routines, using functional magnetic resonance imaging (fMRI) along with simple production tasks (De Witte et al., 2015; Shapiro et al., 2005). This non-invasive technique provides valuable information on the configuration and functional lateralization of the language network, enabling neurosurgeons to design personalized interventions to minimize the risk of postsurgical neurological sequelae while maximizing tumor resection (Stopa et al., 2020).
However, there is currently no established and validated protocol for preoperative language mapping in the bilingual brain. Most studies on preoperative mapping in brain tumor patients have been conducted in the official language of the country where the patient was treated, regardless of the patient's L1 or the different languages they speak (Fernandez-Coello et al., 2021; Pascual et al., 2023; ReFaey et al., 2020). To address this gap, we developed and validated a picture-naming task called MULTIMAP, which consists of an open-source database of pictures representing objects and actions, normed for ten different languages while controlling for linguistic measures such as name agreement, frequency, length, and substitution neighbors. This task minimizes the linguistic distance between language pairs allowing multilingual pre- and intraoperative mapping in brain tumor patients (Gisbert-Muñoz et al., 2021).
In the present study, we created an fMRI sentence completion task based on the MULTIMAP pictures. This was used to measure brain activity in healthy Spanish-Basque balanced bilinguals and bilingual patients with brain tumors while performing the task in either their L1 or L2. This study pursues two specific objectives. Firstly, we aim to explore whether the L1 and L2 in healthy individuals engage similar or different neural networks in terms of regional activation and lateralization patterns. By analyzing this, we can gain valuable insights into the neurobiology of bilingual speech production and provide normative data to better understand potential neuroplastic changes in patients. Secondly, we utilize the same task to investigate how neuroplastic compensatory mechanisms, triggered by brain tumors affecting the language-dominant hemisphere, differentially impact the L1 and L2 in bilingual patients.
Given the contradictions in previous fMRI evidence, we will evaluate two possible hypotheses regarding language production in the group of healthy balanced bilinguals. First, if the L1 and L2 rely on a common neural substrate, we should not find differences across languages. Alternatively, considering studies that have shown differences between languages, it is plausible that certain parts of the healthy language network are modulated depending on the language being used. Specifically, we predict a coadjuvant recruitment of domain-general networks involved in executive functions and language control mechanisms in the L2 production task (Abutalebi & Green, 2016; Bice et al., 2020; Sulpizio et al., 2020; Tao et al., 2021). This effect could be reflected in the activation pattern of areas such as the basal ganglia, and the cerebellum, which are crucial components of these networks (Burgaleta et al., 2016; Camerino et al., 2022; Murphy et al., 2022; Pliatsikas et al., 2017). Testing these hypotheses will enable us to clarify whether the distinctions between languages are limited to domain-general circuits or if they also impact language-specific areas.
Then, taking the data from the healthy individuals as a baseline, we will investigate how the two languages are distributed in a group of bilingual patients with gliomas affecting critical regions for language processing. Considering previous evidence in clinical populations, we expect to uncover how functional neuroplasticity mediates the engagement of the non-dominant contralateral hemisphere as well as potential ipsilateral changes induced by the lesion (Deverdun et al., 2020; Quiñones et al., 2021; Van Dokkum et al., 2019). These findings will contribute to the discussion on the need for personalized pre- and intra-operative multilingual functional mapping strategies, which are crucial for avoiding language-specific cognitive impairments that may differentially affect the L1 and L2.