_{This project explores metacognitive monitoring: the ability of individuals to track decisions, cognitive states, and behavior in uncertain situations.}

_{In earlier studies, I showed that Basque-Spanish bilinguals outperform monolinguals in metacognitive efficiency in tasks that engage cognitive processes activated when we hear natural speech. I suggested that bilinguals estimate the likelihood of making an error in language tasks better than monolinguals, i.e., they have enhanced error-monitoring skills in language tasks. However, bilinguals do not represent a homogeneous population. They differ in age of acquisition, relative proficiency in their languages, switching behavior, and also in typological distance between languages in their inventory. Basque and Spanish are typologically different languages.}

_{I hypothesize the variety of typologically different structures presents additional cognitive challenges and increases speech processing load, enhancing metacognition. I will test that the modulatory effect of bilingualism should be more modest in those bilingual populations where languages are typologically closer and the variety of language structures to be processed is narrower. I predict that exposure to more diverse phonological cues will enhance monitoring phonological tasks of any type, irrespective of phonological properties in bilinguals' native languages.}

_{I will also explore whether this metacognitive enhancement will be transferred from phonological to syntactic tasks, even if bilinguals' languages have typologically different phonology but similar syntax. Finally, I will explore if metacognitive enhancement in language tasks is transferred to non-linguistic behavior. In contemporary society, where bilingualism is a norm rather than exception, it is important to be aware to what extent metacognitive enhancement due to individual linguistic experience is transferred to non-language behavior and influences decision making both at individual and group levels.}

_{More information:}

_{https://www.uc.pt/en/spgi/strategic-areas-unit/erc-uc/typometaling/}

_{https://erc.easme-web.eu/?p=101041475}

_{https://cordis.europa.eu/project/id/101041475}

_{The writing system – the transparency of orthography in alphabet-based systems and differences between logographic and phonetic-based systems – can affect the efficiency of inferential word learning when words are introduced visually. It can also shape how people self-evaluate their learning success (we refer to such type of self-evaluation as metacognitive monitoring of word learning).}

_{Writing systems differ in the size of the unit that is coded by one symbol (a discrete semantic concept in logographic systems, a syllable in syllabic systems, and phonemes in alphabetic systems). Alphabetic systems also differ in terms of syllabic complexity and orthographic depth, both of these factors affect the sound-to-symbol mapping. In languages with transparent orthography, the sound-to-symbol correspondence in explicit, while in languages with opaque orthography, the relations between symbols and sounds are not consistent. Syllabic complexity reflects the phonotactic rules, or possible constraints on how many consonants can make a single syllable, and depending on the orthographic depth, the syllabic complexity may differ across visual and auditory modalities (in spelling, the words may have many consonants, which are not pronounced or heard in spoken speech). Hence, alphabetic languages can be syllabically complex and transparent vs. opaque (German vs. Danish, correspondingly) and syllabically simple and transparent vs. opaque (Spanish vs. French, correspondingly).}

_{In the proposed project, we will first explore how experience with different writing systems modulates metacognition during language tasks that involve versus those do not involve the processing of written information. To this end, we will administer an inferential word learning task in the auditory and visual modalities to native speakers of the following alphabetic languages: German, Danish, Spanish and French (to explore the effect of orthographic depth in syllabically complex and simple languages). Further, we will administer the task to Chinese, Arabic, Japanese and Hindi speakers, to explore the effect of the unit size that is coded by each symbol. Chinese has a purely logographic writing system, Japanese speakers are familiar both with logographic (kanji) and syllabic (kana) writing, Hindi speakers use syllabic script, and Arabic speakers use abjad script, which is a transition between purely syllabic and purely alphabetic systems, and each symbol codes a consonant, while vowels are sometimes represented by adding diacritics to the consonantal symbols, and sometimes omitted.}

_{Participants will learn the pseudowords that are phonetically and phonotactically possible in their native languages, and we will test how well they have learnt new words in the visual and auditory modality, and how precisely they can monitor and evaluate the learning process. Preliminary work already showed that the writing system affects efficiency and metacognition of word learning in the visual modality, but not in the auditory modality, however we do not know which aspects of the writing system (unit size, orthographic depth or syllabic complexity) affect metacognition of word learning in the native language, and how these aspects interact during the learning process.}

_{Second, we will test how writing system of a target language affects word learning by native Portuguese speakers who learn a foreign language with different writing systems. The observed differences between populations using different writing systems and between perceptual modalities may potentially modulate the effectiveness of vocabulary acquisition activities during foreign language learning.}

_{The third objective of the proposed project will be aimed at developing audio-visual aids to help word learning by Portuguese speakers depending on the writing system in their target language. This is the foundation for applied research in educational psychology, aimed to design optimal methods for vocabulary building.}

_{The fourth objective is to explore the causal relations between metacognition and learning outcome. Earlier studies that showed the effect of metacognition on learning are purely correlational. In the proposed project, we will use non-invasive high-definition tDCS to modulate activity in the neural substrates that underpin metacognitive processing. We will explore if the changes in metacognitive sensitivity, induced by tDSC during the experiment, will also lead to changes in the visual and auditory word learning in a foreign language.}

_{Multiple studies showed the effect of language on cognition, behaviour and decision-making. We believe that the proximate mechanism for this effect is metacognition. The need to monitor different cognitive mechanisms engaged by diverse language structures affects metacognition, which, in turn, influences non-verbal behaviour and decisions in non-language domain. Importantly, experience with multiple languages may influence metalinguistic awareness, which is closely linked to grammaticality (the individual ability to test whether an utterance can be generated by internalized grammar) and acceptability (judgment of whether an expression is correct or wrong, which is affected not only by grammaticality, but also by socio-cultural factors). Metalinguistic awareness can influence attitude to speakers of other languages, immigrant speakers of national language, regional accents and dialects, and through that modulate tolerance to different lifestyles, values, ideas.}

_{The proposed project will contribute to explanation why people in different communities make different even if they have the same information. We will reveal the social engineering “tricks” – how social attitudes and behaviours can be orchestrated in pre-planned directions via language manipulation. We will develop strategies to make individual decisions resilient to such manipulations by engaging general metacognition and metalinguistic awareness.}

_{Ability to generate and process recursive structures and patterns is considered to be fundamental to language faculty. However, there is a lot of disagreement whether recursion is specific to language faculty or it is domain-general, whether it evolved specifically for reasons other than language or it evolved to make communicative code more efficient once the rudiments of language faculty emerged. The proposed underlying neural networks and a whole range of electrophysiological correlates specific to processing recursive structures are inconsistent across studies. One of the disagreement causes may be related to terminology. Often, recursive grammars are equated with hierarchical grammars, at least conceptually, while hierarchical grammars are opposed to sequential grammars.}

_{We will argue that some hierarchical grammars are not recursive, and some sequential grammars are hierarchical. Hence, to understand the underlying neural mechanisms for processing recursive grammars, we need to clearly separate three types of grammatical structures: (1) hierarchical and recursive; (2) hierarchical and non-recursive; (3) non-hierarchical, in which only sequential regularities preserve. The main objective of the proposed study is to single out the neural networks that underlie processing of recursion and processing of hierarchy in linguistic sequences.}

_{The role of attention in statistical learning is highly debated. In a series of experiments, Toro et al. (2005) showed that attention diverted from a sensory input to a different cover task disrupts segmentation: participants could not differentiate between discrete units from continuous stream and foils presented during the recognition test after the familiarization. Fernandes et al. (2010), on the other hand, showed that attentional load degrades the efficiency of using statistical information, however, the recognition performance during the test was nevertheless above chance. Besides, it was shown that attentional load differentially affects recognition of frequent words (degraded in high attentional load condition) and infrequent words (on the contrary, enhanced in high attentional load condition) embedded into the familiarization sensory input.}

_{Turk-Browne et al. (2005) that when participants, during exposure to two interleaved stimuli set, performed a cover task during familiarization to only one of the sets, the learning of regularities occurred only for the other set (when participants’ attention was not diverted to a cover task). However, when the authors tried to replicate their results (Musz et al., 2015), they also observed learning on an unattended set. Daikoku and Yumoto (2017) suggested that attentional and unattentional statistical learning may have partially distinct and partially overlapping neural bases, thus leading to mixed results across different experiments.}

_{Besides, some cross-experimental differences can be explained by differences in the nature of the material: statistical learning on in unattended condition seems to be more efficient on non-linguistic material, while in attended condition efficiency is higher on linguistic material. Batterink and Pallier (2019) suggested that increased attentional demand actually does not disrupt the process of segmenting continuous sensory input into discrete units based on SL mechanism. They showed that the neural entrainment at the frequency of whole units, which compose the familiarization continuous input, is not modulated by whether participants pay attention to the segmentation task, or whether their attention is distracted by a cover task. Statistical regularities can interact with attention during the familiarization, and as statistical regularities are acquired in the course of familiarization, they capture attention thus reducing the extent to which attention is diverted to the cover task (Alamia & Zenon, 2016). Batterink and Pallier (2019) suggested that it is the memory-based offline performance that is severely affected by diverted attention, not online segmentation. That is, people can use statistical information for extracting the discrete units from continuous input. However, committing these units to memory and subsequent recognition is modulated by attentional load.}

_{In the proposed project, we decided to focus on the following two questions:}

_{1) what is impeded by high attentional load: online segmentation (i.e., extracting statistical regularities and discrete units from continuous stream), or offline recognition of the extracted constituents, or both? We will test whether segmentation happens at all in unattended condition.}

_{2) can people learn and later recognize the discrete units during statistical learning without conscious access to the knowledge? Potentially, conscious access to the knowledge of the extracted units from unattended input is blocked, rather than memorization process during online learning.}

_{The project studies how metacognition in language tasks is affected by individual linguistic experience, by structural differences between native languages, or by structural differences between languages in bilinguals’ inventory.}

_{I predict that phonological differences enhance metacognitive processing on prosodic tasks (e.g., rhythm discrimination), morpho-syntactic differences improve tracking performance on grammar learning (e.g., statistical learning or grammaticality judgment), and vocabulary differences hone metacognitive monitoring in inferential word learning. Bilinguals whose languages do not exhibit typological differences in phonology or syntax do not exhibit enhanced metacognition in language tasks, when all other bilingualism-related factors (age of acquisition, language dominance, etc) or individual differences (e.g., in IQ, in verbal fluency, in socio-economic parameters) are controlled for. Additionally, I will try to separate the effect of socio-cultural differences on metqcognition from the effect of language-elicited differences, by comparing metacognition in language tasks in monolingual speakers from culturally difference societies (e.g., monolingual Spanish speakers from Mesoamerica and Europe).}

_{Metacognition is the ability to evaluate past decisions (meta-monitoring) and estimate the degree of uncertainty when a new decision has to be made (meta-control). Meta-control and meta-monitoring are two separate sets of processes relying on distinct neural networks and evolved for different purposes. Monitoring processes track past behavior by estimating retrospective confidence associated with made decisions. Control processes guide future behavior by evaluating current cognitive states, available evidence about the environment and consequences of past decision.}

_{Efficient metacognition is related to accurate estimation of likelihood of having made an error on past decisions, and with accurate evaluation of uncertainty associated with future decisions. This can lead to assigning higher confidence to those decisions which are more likely to be correct and lower confidence to those decisions which are more likely to be wrong. Metacognitive monitoring skills in a particular domain and activities can be modulated by individual experience in this domain and activities. For example, bilinguals who have more enriched experience with linguistic structures and cues than monolinguals also exhibit enhanced metacognition language tasks. However, we know neither the exact factors which lead to metacognitive enhancement nor the limits of transferability of metacognitive enhancement across tasks and domains.}

_{I will address these issues in two work packages. First, I will explore how metacognition in language tasks is influenced by Age of Acquisition and by relative proficiency in both languages on metacognition. Second, I will explore how metacognition is related to decision-making strategies in the language-irrelevant tasks. The investigation will be done on Danish-German and Frisian bilingual populations. Due to variable and nuanced language landscape in southern Denmark and Northern Schleswig-Holstein, it is possible to test different aspects of the hypothesis.}

_{Error-related negativity (ERN) and error-related positivity (Pe) are event-related potentials that are associated with incorrect responses in perception experiments or in decision-making tasks. They are often considered as electrophysiological correlates of error identification and represent autoregulation strategy of the brain because they occur on error trials without external feedback. However, the origin of these electrophysiological correlates is not yet clear. Moreover, although negative deflection curve in the ERP waveform tends to occur on error trials, it has also been observed on correct responses when the participant was not certain whether his response was correct or not. Therefore, we can advocate three competing hypotheses related to the origin of the ERN/Pe:}

_{1. ERN/Pe emerge when the error is detected; - in this case, electrophysiological correlates will be elicited on error trials only;}

_{2. ERN/Pe emerge when the uncertainty in the environment is higher, leading to increased probability of error (i.e., these are correlates of detecting the cases when the error is more likely, but not detecting the error per se) - in this case, electrophysiological correlates will be elicited on all trials in the condition when unambiguous answer is not possible;}

_{3. ERN/Pe emerge when the uncertainty in one’s own mental states (detecting the cases when the error is more likely because of the perception failure or lack of knowledge/experience) - in this case, electrophysiological correlates will be elicited on all trials when the participant assigns lower confidence to the response/decision and when the task is more difficult, although unambiguous answer is possible.}

_{We will test these three competing hypotheses, which will also inform us whether ERN/Pe can be used as an electrophysiological marker of meta-monitoring efficiency (hypotheses One and Two) or not (hypothesis Three).}

_{Whether and how metacognition is altered in individuals diagnosed with the Autistic Spectrum Disorder (ASD), and in which domains metacognition is most susceptible to alterations in the ASD population, is a hotly debated topic. Here, we measured metacognitive efficiency in ASD individuals from two different geographic regions (Portugal and Spain) and in matched neurotypically developing (TD) control samples from the same regions in decision-making and perception tasks that engage difference cognitive processes (metacognitive mechanisms monitor execution of these processes).}

_{Additionally, we administered gaming tasks (e.g., a trading game, a strategy game) to participants. The rules that allow profit maximization (in trading game) or faster extinguishing of fire (in a strategy game) change in the middle of the game, which required that players modify their strategy to keep the profit at maximum or faster achievement of strategy objective (e.g., fire extinguishing). Measuring how fast people can learn the rules (learning efficiency) and re-learn the rules (cognitive flexibility) will tell us whether metacognition (and which aspects of metacognition, in which tasks) is related to learning and cognitive flexibility in autistic and typically-developing populations.}

_{Metacognition – the ability to monitor one’s cognitive performance and evaluate one’s own decisions – is associated with cognitive flexibility, or a mental ability to accommodate new information during decision-making. Higher metacognition is positively correlated with the ability to incorporate new information to decrease uncertainty.}

_{Bilinguals exhibit enhanced metacognition compared to monolinguals, which leads to lower confidence ratings assigned by bilinguals to error trials. This led us to the hypothesis that bilingual advantage in metacognition might result in differences in the perception and evaluation of unverifiable information by bilinguals and monolinguals, and in different confidence ratings assigned to such evaluations. We hypothesize that bilinguals, compared to monolinguals, will be more critical towards information that is easily proved as false but will be more liberal with taking into account unverifiable information. Such differences in information processing are modulated by metacognition in language tasks. Language processing (speech perception and production, language comprehension) is affected by psychosis, and thus can influence metacognition in language tasks in individuals diagnosed with schizophrenia. This, in turn, may also affect the perception of true, false and unverifiable information in psychotic patients, explaining disorganised thinking, persistent beliefs, and delusions.}

_{We will measure metacognition in semantic and syntactic language tasks in individuals diagnosed with schizophrenia and neurotypically developing bilinguals and monolinguals, and then ask people to evaluate the validity of verifiable (e.g.,”Berlin is further to the north than Madrid”) and unverifiable (e.g.,”Scientists found cure for a rare tropical disease”) statements. We expect that psychosis states will affect metacognition in language tasks, and lead to higher endorsement of unverifiable statements. We also expect that bilinguals have better metacognition and endorse unverifiable statement at a higher rate yet they will express their evaluation with lower confidence than monolinguals. Verifiable statements, by contrast, will be more rigorously evaluated by bilinguals (showing lower tolerance to false information and higher tolerance to unverifiable information).}