This project aims at uncovering the cognitive and neural bases of the process whereby speakers inspect their own speech for errors both during speech planning (internal monitoring) and articulation (external monitoring) at different linguistic levels (ranging from semantics to articulation). There theoretical accounts will be systematically assessed: comprehension-based monitoring, internal modelling based monitoring and conflict-based monitoring.
By taking into account the complexity of linguistic representations (through various linguistic levels), contexts of monitoring load (varying the tasks) and components of monitoring (internal and external) while assessing simultaneously three hypothesis and using a multi-method approach we expect to obtain the most complete blueprint of speech error monitoring to date. This will constitute the rudiments of a new neurocognitive model of fluent language production.
We rely on a combination of EEG, fMRI and TMS to obtain fine grained temporal, spatial and functional information about different effects and brain structures that have been associated to the three monitoring theories that we assess in the project.
So far, we have found evidence for a common mechanism of internal and external monitoring through the cognitive process of internal modelling sustained by the cerebellum. However, we have also found dissociations between internal and external monitoring since the latter seem to be more reliant on auditory and vocal feedback control sustained by temporal and medial frontal structures in the cerebral cortex.
We are currently examining how the different monitoring mechanisms might be modulated by using a different task (such as learning instead of error-priming) and by manipulating different linguistic levels (semantics and articulatory-phonetic instead of lexical).
1. Elin Runnqvist, Valérie Chanoine, Kristof Strijkers, Chotiga Patamadilok, Mireille Bonnard, et al.. Cerebellar and cortical correlates of internal and external speech error monitoring. Cerebral Cortex Communications. doi.org/10.1093/texcom/tgab038
2. Cristina Baus, Mikel Santesteban, Elin Runnqvist, Kristof Strijkers, Albert Costa. Characterizing lexicalization and self-monitoring processes in bilingual speech production. Journal of Neurolinguistics, Elsevier, 2020, 56, pp.100934. ?10.1016/j.jneuroling.2020.100934?. ?hal-02913144?
3. CIRILLO, Giusy, RUNNQVIST, Elin, STRIJKERS, Kristof, et al. Conceptual alignment in a joint picture-naming task performed with a social robot. 2021.https://psyarxiv.com/cjy24
The aim of this project is to investigate the neuro-cognitive basis of self-monitoring in language production. Specifically, the main objective consists in assessing whether and to what extent this ability may be shared with other cognitive domains. One possibility is that language production errors, just as errors related to motor actions, are monitored through internal modeling. In the domain of motor control it is widely held that, in parallel to any motor command, the sensory consequences of actions are predicted before they occur by means of internal modeling involving the cerebellum, and the predicted sensory response is suppressed in the relevant area of somato-sensory cortex (e.g., auditory cortex in the case of speech). By contrasting the internal model with the actual sensory input, any mismatch between the two results in an attenuated suppression, signaling that an imminent error needs to be corrected. It has been hypothesized that for language processes prior to articulation, sensory predictions would be compared with the output of the production process proper at different internal stages, accounting in this way for inner monitoring of language production. Another possibility is that language production errors are monitored in the same way we deal with other types of cognitive conflict. That is, there would be a system located in medial frontal structures such as the anterior cingulate cortex (ACC) and the pre supplementary motor area (pre-SMA) sensitive to levels of conflict between simultaneously active responses, in which the presence of conflict would trigger cognitive control in pre-frontal areas. The present project will test the following four (not mutually exclusive) hypotheses concerning a potential involvement of internal modeling and conflict-based monitoring in language production: (A) Internal Modeling is restricted to monitoring of the end stages of speech production (speech motor control); (B) Internal Modeling is used for the complete language production process (e.g., from semantics to articulation); (C) Conflict-based Monitoring is used to detect errors internally through the presence of competing responses; (D) Conflict-based Monitoring is used to detect overt errors. Five experiments are programmed contrasting (1) Different sub-processes of monitoring (inner error detection will be isolated by examining the effect of monitoring load on correct trials, and detection of overt errors will be isolated by contrasting correct trials and errors); (2) Different linguistic levels (each experiment will manipulate monitoring load at two different levels more or less distant from articulation); and (3) Contexts involving conflict of different explicitness (achieved by using three different tasks; the SLIP task, the picture-word interference task and the tongue twister task). Three different techniques will be used (EEG, fMRI and TMS), each of which will provide information from a complementary angle for the questions of interest (component-specific evidence, spatial evidence, and temporal-functional evidence). Neurophysiological patterns related to internal modeling and conflict-based monitoring (i.e., N100 and ERN in EEG, auditory cortex suppression in fMRI) and the implication of brain areas characteristic to internal modeling (cerebellum, pSTG) and conflict-based monitoring (ACC, pre-SMA) respectively will be examined for the different conditions. In this way, the results obtained in this project should provide the most direct and complete spatio-temporal blueprint of self-monitoring to date. As such, it will not only constrain and potentially falsify existing models, but also go a long way towards the creation of a new integrative model of language production monitoring.
Madame Elin Runnqvist (Laboratoire Parole et Langage)
The author of this summary is the project coordinator, who is responsible for the content of this summary. The ANR declines any responsibility as for its contents.
LPL Laboratoire Parole et Langage
Help of the ANR 254,631 euros
Beginning and duration of the scientific project: December 2018 - 48 Months