Periodic Reporting for period 1 - MetaBil (Metacognition and bilingualism in linguistic and non-linguistic domains)
Reporting period: 2018-12-01 to 2020-11-30
Metacognition is the ability to monitor one’s own cognitive processing, behavioural performance and conscious decisions. Earlier evidence suggested that individual experience in a particular operational domain (e.g. spatial visual attention, memory) can lead to improvements in metacognitive monitoring in this domain: while making a decision in a particular task, people can better estimate the likelihood of making an error, and, for example, they assign lower confidence rating to those decisions when they are more likely to make a mistake. In other words, people can better evaluate available evidence they can use for decision, and their own ability to give correct answer based on this evidence. However, we did not have evidence to what extent this effect extends beyond low-level perceptual tasks and memory to high-level cognitive performance (e.g. in language processing, calculus, etc.) Moreover, it is hotly debated whether metacognitive enhancement in one domain can be transferred to a different domain, or whether metacognitive skills are specific to the domain in which they are forged. In this project, we addressed the question whether enhanced individual experience with language processing, honed by bilingualism, leads to enhanced metacognitive skills in language domain, and if so, whether this enhancement can be transferred to non-language domains (in other words, whether bilinguals, when compared to monolinguals, can better estimate the probability of giving a correct answer or making an error when they are doing a language task; and if so, whether bilinguals’ metacognitive enhancement can also be observed in non-language domain.
I looked at the effect of bilingualism on metacognitive efficiency in artificial grammar learning tasks on linguistic material and non-linguistic material, in visual and auditory perceptual modalities (four experiments in total). I used within-subject design to be able to directly compare metacognitive efficiency across modalities (metacognitive efficiency was selected as a target measure of interest because it is comparable between modalities and tasks of varying degrees of complexity and difficulty). As participants, I recruited Basque-Spanish bilinguals and Spanish monolinguals. Participants were exposed to a long stream of images (non-linguistic material) or letters (in the visual modality) or environmental sounds (non-linguistic materials) and syllables (in the auditory modality), with statistical regularities (i.e. artificial grammar) embedded into the stream. After familiarization each streams, a recognition test was performed: participants listened to or watched a sequence of the syllables/letters/images/environmental sounds, and responded whether the sequence was grammatical (i.e. whether they thought this short sequence could belong to a long familiarization stream) or ungrammatical. On each response, participants responded whether they were confident in their decision. Both during the familiarization and during the recognition test, EEG signal was recorded. In total, 36 bilinguals and 36 monolinguals were recorded in the auditory modality, 26 bilinguals and 26 monolinguals were recorded in the visual modality. During the analysis of the EEG signal, I estimated error-related negativity and error-related positivity on each response. For the EEG analysis of the familiarization phase, I performed frequency-analysis and correlated power of neural rhythms in different frequency bands during learning with overall confidence, metacognitive efficiency and ERP components observed during the recognition test, at individual level.
Main results:
1. In language domain, bilinguals indeed outperform monolinguals in metacognitive efficiency.
2. Bilinguals better estimate the likelihood of making an error in language tasks compared to monolinguals. This leads to lower confidence in their response when the probability of making an error is high. Therefore, bilinguals’ retrospect confidence ratings (confidence rating assigned to past decisions) discriminate between correct and incorrect responses better than those of monolinguals. Importantly, this increase in discriminative power is due to lower confidence of bilinguals on incorrect responses, while confidence ratings on correct responses do not differ between bilinguals and monolinguals.
3. The mechanism that underlies this metacognitive enhancement is enhanced error-detection and error-monitoring in bilinguals compared to monolinguals.
4. The data on whether metacognitive enhancement is transferrable across operational domains is inconsistent. We have observed electrophysiological evidence of the transfer (in terms of error-related negativity, error-related positivity and other correlates), yet this did not surface at the behavioural level. In other words, bilinguals exhibit electrophysiological evidence of enhanced error-detection and error-monitoring mechanisms in non-language domain compared to monolinguals, yet retrospect confidence ratings on correct and incorrect responses did not differ between bilingual and monolinguals populations. Neither did we observe differences between populations in terms of metacognitive sensitivity/metacognitive efficiency in non-language domain. Further studies are needed to properly address the limits of transferability of metacognitive enhancement across domains.
5. The experiment aimed to understand whether the neural underpinning of metacognitive processing is domain-general or domain-specific has not been conducted (see full report).
Main results:
1. In language domain, bilinguals indeed outperform monolinguals in metacognitive efficiency.
2. Bilinguals better estimate the likelihood of making an error in language tasks compared to monolinguals. This leads to lower confidence in their response when the probability of making an error is high. Therefore, bilinguals’ retrospect confidence ratings (confidence rating assigned to past decisions) discriminate between correct and incorrect responses better than those of monolinguals. Importantly, this increase in discriminative power is due to lower confidence of bilinguals on incorrect responses, while confidence ratings on correct responses do not differ between bilinguals and monolinguals.
3. The mechanism that underlies this metacognitive enhancement is enhanced error-detection and error-monitoring in bilinguals compared to monolinguals.
4. The data on whether metacognitive enhancement is transferrable across operational domains is inconsistent. We have observed electrophysiological evidence of the transfer (in terms of error-related negativity, error-related positivity and other correlates), yet this did not surface at the behavioural level. In other words, bilinguals exhibit electrophysiological evidence of enhanced error-detection and error-monitoring mechanisms in non-language domain compared to monolinguals, yet retrospect confidence ratings on correct and incorrect responses did not differ between bilingual and monolinguals populations. Neither did we observe differences between populations in terms of metacognitive sensitivity/metacognitive efficiency in non-language domain. Further studies are needed to properly address the limits of transferability of metacognitive enhancement across domains.
5. The experiment aimed to understand whether the neural underpinning of metacognitive processing is domain-general or domain-specific has not been conducted (see full report).
So far, the effect of individual experience on metacognitive efficiency was only studied in low-level perceptual domain and memory. We explored this effect in high-level cognitive functions and established the effect of bilingualism (enriched experience with language structures) on artificial grammar learning (i.e. extraction of regularities from environmental input and using these regularities for making decisions in regard to classification of the future environmental input).
We found that bilinguals indeed outperform monolinguals in metacognitive processing in language domain. Enriched cognitive experience in language domain allows bilinguals to better estimate the probability of giving a wrong response when decision-making process engages those mechanisms that are also activated while processing natural speech. Bilinguals are more sensitive to those cases when they are likely to make a mistake, and thus they report that they are not confident in the decisions on such cases. This effect can potentially be transferred to non-language domain, and this is where the project might have societal implications. We observed the evidence of the transfer at the electrophysiological level, but not at behavioural level. In those cases, when this transfer happens at the behavioural level, decision making and evaluation of decisions might differ between bilinguals and monolinguals, thus leading to potentially different decisions at the group level in bilingual and monolingual populations. However, within the framework of this project (24 months), we could not attend to this question.
We found that bilinguals indeed outperform monolinguals in metacognitive processing in language domain. Enriched cognitive experience in language domain allows bilinguals to better estimate the probability of giving a wrong response when decision-making process engages those mechanisms that are also activated while processing natural speech. Bilinguals are more sensitive to those cases when they are likely to make a mistake, and thus they report that they are not confident in the decisions on such cases. This effect can potentially be transferred to non-language domain, and this is where the project might have societal implications. We observed the evidence of the transfer at the electrophysiological level, but not at behavioural level. In those cases, when this transfer happens at the behavioural level, decision making and evaluation of decisions might differ between bilinguals and monolinguals, thus leading to potentially different decisions at the group level in bilingual and monolingual populations. However, within the framework of this project (24 months), we could not attend to this question.