Liste des mémoires du Cogmaster année 2019-2020

AUBRAIN, Kevin

Association between ventro-lateral prefrontal cortex activity and pupil size using local-global paradigm in monkeys

sous la direction de Ghislaine Dehaene Lambertz

Pupil size have been shown to be associated with a broad variety of cognitive function, including attention. The neural activity of the dorsal stream of attention, especially the one in the frontal eye field has been shown to correlate with the pupil size. However, no direct link between the neural activity of the ventral stream of attention and the pupil size have been established yet. Here our aim is, by using a visual local-global paradigm and recording of pupil size and neural activity in the ventro-lateral prefrontal cortex of two monkeys, to see if there is any link that can be drawn between pupil dynamics and the neural activity in the ventral stream of attention. These findings could potentially be useful to assess the wealthiness of the ventral stream of attention in subjects with disorders of consciousness or attention troubles.

34 p., 2020

Spécialité : Neurosciences

Cote pour l’emprunt : Master 1969

___________________

BACQUET, Maxime

Influence of preterm birth on Superior Temporal Sulcus:
Modification of the depth profile in the Superior Temporal
Asymmetrical Pit measured on MRI at equivalent-age-of-term in a 184 children cohort born from 24 to 41 weeks of postmenstruation age

sous la direction de Ghislaine Dehaene Lambertz

Superior temporal Sulcus is one of the main language hub in the brain. Its asymmetry was found to be almost constant in human but absent in other primate, and therefore hypothesized to be a landmark of language. To investigate the influence of age-of-birth on STS depth, we analyzed anatomical MRI from 184 newborns. Our study included 92 children born preterm from 24 to 37GW and a control-group of 92 term matched for gender and age-at-scan, made available by the developing Human Connectome Project (Release 2). We applied the same method as used for the definition of the Superior Temporal Asymmetrical Pit, with complementary analyses on STS maximum depth to compare it with STS Mean depth. We replicated previous findings showing a shorter STAP area in newborns than in adults, limited to its posterior half. We found the STAP to be significantly longer in preterms than in control terms. Also we found that age-of-birth significantly impact STS mean depth in the STAP area, in both side, but not the asymmetry. Finally we rejected the maximum depth as a potential proxy in the study of STS depth.

28 p., 2020

Spécialité : Neurosciences

Cote pour l’emprunt : Master 1970

____________________

BRACON, Lucie

Étude de la spécialisation progressive des capacités d’inhibition froide et chaude au cours du développement à travers l’activité cérébrale (électroencéphalographie, EEG)

sous la direction d’Emilie Salvia

La présente étude s’intéresse à la spécialisation progressive des capacités d’inhibition chaude (i.e. en contexte émotionnel) et froide (i.e. en contexte neutre) au cours du développement à l’aide de l’électroencéphalographie (EEG). Pour cela, 31 adultes, 10 adolescents et 12 enfants ont été recrutés pour participer à cette étude. Les participants ont effectué une tâche de Stroop froid et une tâche de Stroop chaud. Dans le Stroop froid, les participants doivent identifier la couleur de l’encre d’un mot de couleur tout en ignorant le sens de ce mot. Dans le Stroop chaud, ils doivent indiquer l’émotion exprimée par le visage tout en ignorant le mot écrit sous ce visage. Chaque tâche est constituée d’items congruents (i.e. sans conflit ; par exemple le mot « rouge » écrit en rouge et le mot « colère » écrit sous le visage exprimant de la colère pour le Stroop froid et chaud respectivement) et d’items incongruents (i.e. avec conflit ; par exemple le mot « rouge » écrit en bleu et le mot « joie » écrit sous le visage exprimant de la colère). Dans ce mémoire, compte tenu de la taille de l’échantillon adolescents (n = 5 après pré-traitement des données EEG), seuls les résultats adultes et enfants ont été analysés et présentés. Les principaux résultats comportementaux montrent un effet d’interférence (score d’interférence (SI) = TRs aux items incongruents – TRs aux items congruents) plus fort dans la tâche de Stroop froid que dans celle de Stroop chaud aussi bien chez les adultes que chez les enfants. De plus, le SI au Stroop froid et le SI au Stoop chaud ne sont pas corrélés chez les adultes et les enfants. Les principales analyses en potentiels évoqués mettent en évidence, chez l’adulte, des corrélats neurophysiologiques classiques de l’inhibition (i.e. ondes N2 et N450), avec des amplitudes plus fortes aux items incongruents qu’aux items congruents pour ces ondes. De plus, des différences d’amplitudes et de latences sont retrouvées tout au long de la réponse électrophysiologique selon que le conflit soit de nature émotionnel ou non-émotionnel, pouvant suggérer des mécanismes dissociables. Chez l’enfant, les résultats préliminaires montrent des différences au niveau des étapes précoces de la réponse électrophysiologique, à savoir au niveau des traitements perceptifs, selon que le conflit est de nature émotionnel ou non-émotionnel, pouvant ainsi suggérer des mécanismes dissociables. En revanche, au niveau des étapes plus tardives de la réponse, impliquées dans le contrôle cognitif, les résultats ne semblent pas montrer de différences entre les traitements des deux types de conflit, suggérant, peut-être, des mécanismes communs.

36 p., 2020

Spécialité : Psychologie

Cote pour l’emprunt : Master 1971

____________________

CARILLO, Pablo

Caractérisation computationnelle des effets motivationnels de la kétamine chez le sujet sain

sous la direction de Fabien Vinckier

La kétamine est un antagoniste non-compétitif du récepteur glutamatergique NMDA. Employée initialement en anesthésie, de nombreuses publications ont récemment souligné ses effets antidépresseurs rapides et sa supériorité par rapport aux traitements habituellement utilisés. Les mécanismes cognitifs sous-tendant ces effets restent cependant peu connus. Dans ce travail, nous cherchons à caractériser les effets de la kétamine sur la motivation chez le sujet sain. Notre stratégie consiste à combiner une batterie de tâches à une approche de modélisation computationnelle pour disséquer la motivation en processus cognitifs plus élémentaires tels que la sensibilité à la récompense, à la perte ou à l’effort. Afin d’étudier les effets de la kétamine, nous avons utilisé un dessin expérimental en cross-over, ketamine vs. placebo, chez le sujet sain. Nos résultats préliminaires, obtenus sur un groupe de 21 sujets, indiquent une augmentation de la sensibilité à la récompense à 24h de la perfusion de kétamine. Des analyses exploratoires complémentaires suggèrent également une modification de la sensibilité à l’effort sans effet consistant au travers des sujets. A notre connaissance, il s’agit des premiers résultats permettant de mieux caractériser les effets cognitifs de la kétamine sur la motivation et obtenus à 24h de l’administration, soit un délai compatible avec le pic d’efficacité antidépressive de cette molécule. La poursuite du recrutement sera nécessaire pour en préciser les résultats. Par la suite, un paradigme identique pourrait être utilisé en clinique, chez des patients souffrant de dépression résistante, pour explorer les mécanismes d’actions psychopharmacologiques de cette molécule.

38 p., 2020

Spécialité : Neurosciences

Cote pour l’emprunt : Master 1972

____________________

CHUNG, Victor

Physiological synchrony and emotion

sous la direction de Julie Grèzes et Elisabeth Pacherie

In social contexts, individuals interacting with each other sometimes show similar patterns of autonomic activity. This phenomenon has been labelled physiological synchrony and analysed in a wide range of populations and experimental contexts. Previous research in psychophysiology typically assumes that physiological synchrony results from idiosyncratic social interaction between the participants, whereas other studies emphasize the role of contextual or individual factors such as participants’ emotions. Yet, evidence remains scarce regarding the relation between emotion and physiological synchrony. This experimental study investigates the possibility that physiological synchrony may occur in response to an emotional stimulus and in the absence of any social interaction. Activity of the corrugator and the zygomatic, as well as electrodermal activity, were recorded in eighteen participants separately attending to an emotional movie. Physiological synchrony was measured as phase coherence (ISPC) between participant’s time series of autonomic activity. Results suggest that ISPC in electrodermal activity increased in response to some of the movie highlights identified a priori by the film director as emotion-eliciting events. In addition, ISPC in electrodermal activity during the movie positively correlated with the similarity in participants’ self-reports of valence experienced during the movie. These findings suggest that physiological synchrony can occur in the absence of social interaction, which calls for a rigid control of participants’ emotions in research on synchrony. Finally, these results provide support for the idea that physiological synchrony could reflect the broader phenomenon of synchrony in emotions, although further research is needed to specify the relation between these two phenomena.

x-37 p., 2020

Spécialité : Psychologie-biologie

Cote pour l’emprunt : Master 1973

____________________

DABADIE, Lucien

Religious morality and time-preferences in Latin Christianity (3rd -14th Century)

sous la direction de Nicolas Baumard

By applying textual analysis to almost 900 theological works written from the 3rd to the 14th century C.E., the current study seeks to test two central predictions of the cognitive branch of Life History Theory, namely, that affluence, for which population density is chosen as a proxy, is positively correlated (1) with an increase of morality-in this case, religious morality-and (2) with delayed time-preferences. Simple word frequencies, bigram frequencies (controlling for frequency of isolated words) and Mutual Information scores are computed to measure variations of intensity in several religious representations: basic cues of morality such as the system of vices or modes of qualification of the divinity; punishment and reward in the afterlife, interpreted as a marker of both heightened moral content and stronger future-orientations; and finally the idea of gradual spiritual ascent, interpreted as a sign of increasingly delayed time-preferences. The overwhelming majority of linear mixed models employed to analyze the data fail to reveal any significant relationship between the dependent variables and population density; and interpretation of the rare positive significant coefficients is made difficult by the lack of similar findings on almost equivalent metrics. Moreover, further statistical tools introduced to deal with assumptions of linear mixed models being violated only confirm, on the whole, the initial results. The most promising findings consist in a large spike during the 11th and 12th centuries of metrics pertaining to spiritual ascent, closely in line with our predictions. Although the relationship loses significance once the frequency of isolated terms in the bigrams is controlled for, the results remain intriguing and deserve further investigation.

114 p., 2020

Spécialité : Cognition sociale

Cote pour l’emprunt : Master 1974

____________________

DAILEY, Megan

Allophonic Cues in L2 Processing

sous la direction de Sharon Peperkamp

The perception of speech sounds is shaped by a listener’s native language phonology (Best, 1994). Second language (L2) learners, guided by the perceptual tendencies acquired early in development, rely on different word processing strategies than native speakers in L2 word recognition (Shea & Curtin, 2010; Ito & Strange, 2009; Kondaurova & Francis, 2008). To better understand how language learners process L2 speech, it is important to uncover the extent of their knowledge of L2 phonological contrasts. The present study looks at L1 French speakers’ representation and processing of nasal vowel allophony in English. Native English speakers use the allophonic distribution of nasal vowels in English, such that they only precede nasal consonants (i.e. man – [m n], but mad – [m.d]), to predict upcoming nasal segments and facilitate word recognition (Lahiri & Marslen-Wilson, 1991). However, in French, nasal vowels are in phonemic contrast with oral vowels. In order for French listeners to use vowel nasality as a cue to an upcoming nasal segment in English, they will have had to learn the association between nasal vowels and consonants in their L2. In the present study, we implemented a gating task with nasal-oral minimal pairs (such as bunny – [b ni] and butter – [b    ]) to determine whether French participants and a control population of English participants could predictively isolate words with nasal vowels from their oral vowel competitors. A series of linear regression model analyses revealed that French learners of English successfully isolated nasal stimuli from oral stimuli, and vice versa, before the onset of the following disambiguating segment. The results demonstrate that L2 learners can learn novel L2 allophonic contrasts when the relevant contrast is phonemic in their first language. Furthermore, L2 learners use their knowledge of allophonic contrasts to aid in L2 word recognition.

25 p., 2020

Spécialité : Langage

Cote pour l’emprunt : Master 1975

____________________

DE DAMPIERRE, Charles

Quantifier et expliquer la productivité culturelle dans l’Histoire

sous la direction de Nicolas Baumard

Nous présentons une base de données  Cultura 1.0  composée de 62 927 biographies d’artistes et de scientifiques du monde entier extraites de Wikidata et de 13 Wikipédia diff´erentes entre l’année 3000 avant JC et l’année 1850. Combinée aux informations disponibles sur la démographie et l’économie de chaque pays, cette nouvelle base de données nous a permis de créer une mesure de la productivité culturelle pour chaque individu et pour chaque pays, avec une résolution temporelle de l’ordre de la décennie. Dans un premier temps, nous avons vérifié que notre indice répliquait l’état de la littérature en histoire culturelle, notamment à travers les périodes d’âges d’or bien connues comme le Ve siècle Grec, le XVIe siècle Espagnol et le XVIIe siècle Hollandais. Nous avons ensuite montré que notre indice permettait d’explorer avec plus de précision des questions anciennes en histoire culturelle comme la dynamique de divergence entre l’Europe et l’Asie, ou entre l’Europe catholique et l’Europe protestante. Enfin, nous avons testé l’hypothèse, issue des sciences cognitives, que la productivité culturelle d’un pays, ou autrement dit sa créativité et son degré d’innovation, dépendent non pas de la taille de sa population ou de la masse totale de richesses disponible (Produit National Brut), mais de degré de développement économique de l’ensemble du pays (PNB par habitant).

102 p., 2020

Spécialité : Cognition sociale

Cote pour l’emprunt : Master 1976

____________________

DE LA CHAPELLE, Aurélien

Integration of External Stimuli during Wake and Sleep

sous la direction de Delphine Oudiette

It is unclear how we switch from an idea to an other when our mind is unconstrained from conscious goals and stimuli : what drives the shift towards a specific idea during our spontaneous thoughts ? Such thoughts thoughts are known to arise regularly during the day (for ex. during mind wandering) and the night (for ex. during dreaming), but the specific mechanisms that guide our thoughts in this unconstrained state are unknown. We investigate the Spreading Activation Theory that suggests thoughts flow towards semantically related concepts, and we hypothesize that some external stimuli may get integrated in our spontaneous thoughts. Here we used a modified Targeted Memory Reactivation design (TMR-like), which is a paradigm where a learning-related cueing is presented during sleep. Unlike this, we presented auditory verbal stimuli during an induced state of mind wandering, and we collected mentation reports after these stimuli. Each verbal stimulus was also associated with a second word in order to guide the thought flow. The semantic relations between these two words were investigated in each participant, who had to find semantic associations between the two words : they had to find 1, 2 or 3 words that related one to the other in order to semantically link the two target words. The words they reported in this task were categorized according to their semantic distance to the verbal stimulus.These semantically related words were then searched for in the mentation reports obtained during mind wandering depending on their relatedness to the presented cue and their semantic distance from the stimulus word. We found cue-related words to be more often integrated in mentation reports (1,18% of words integrated) than cue-unrelated words (0,47%). The cued word had a higher integration rate, but not its paired word or their semantically related words. Typical TMR effects observed during sleep such as improved memory or creativity were not replicated during mind wandering.

60 p., 2020

Spécialité : Neurosciences

Cote pour l’emprunt : Master 1977

____________________

DERRIEN, Diane

Dissociations between perception and awareness in brain-damaged patients with visual field defects

sous la direction de Sylvie Chokron

We designed a protocol of letter detection and discrimination in the framework of signal detection theory, sensitive to the phenomena of blindsight and blindsense, to assess objective and subjective visual performance, in patients with homonymous hemianopia resulting from a stroke in their visual cortex. Importantly, our task is designed to also measure visual metacognitive capacities using the meta-d’ index. We jointly assess the presence of anosognosia and the occurrence of hallucinations in these patients using the Q3H questionnaire designed for patients with visual field defects. We thus provide a complete profile of the dissociations between perception and awareness for post-stroke patients with a blind hemifield. In our group of three patients, we find no evidence of neither blindsight nor blindsense. Moreover, metacognitive performance in these patients is not different to that of 5 control participants. We compare our protocol with a previous methodological design, discuss each of their advantages and propose ideas for a combination of the two which would provide the most sensitive measures in these patients. Finally, as the behaviour of one patient suggests a remarkable metacognitive bias, we propose complementary measures to be performed in the future in this patient population.

51 p., 2020

Spécialité : Neurosciences

Cote pour l’emprunt : Master 1978

____________________

DI FOLCO, Cécile

Early environmental and cognitive predictors of literacy

Sous la direction de Franck Ramus

Learning to read is crucial for the development of individuals and their insertion in society. Reading is acquired throughout schooling, but its foundations are laid long before instruction begins. Identifying which early cognitive, socio-emotional and environmental characteristics underlie reading acquisition is essential to guarantee the mastery of this skill by all.
We study the role of a large set of such predictors in the EDEN cohort, following 297 children from age 5.5 (assessment of cognitive and socio-emotional variable) to 11.5 (assessment of reading comprehension, accuracy, speed and spelling). The role of environmental predictors was further investigated in a mediation analysis, allowing to disentangle its direct effects on late literacy from indirect effects through early cognitive and socio-emotional variables.
When literacy outcomes were directly regressed on language predictors at 5.5 years, phonological processing and short-term memory predicted reading accuracy when short-term memory and conceptual knowledge predicted spelling. A second model explored the relative roles of latent variables of language, visuo-spatial and fine motor skills; only language predicted literacy. Adding socio-emotional variables in this model, we found a negative effect of hyperactivity/inattention at 5.5 on reading accuracy. Directly regressing literacy outcomes on environmental variables, parental education and cognitive stimulation predicted reading accuracy, and parental education further predicted reading comprehension. Effects of parental education on literacy at 11.5 seem to be part direct and part mediated by early language skills, when effects of home stimulation were mostly mediated by early cognitive skills. Our final model respectively explains 41.5%, 30.6%, 29% and 22.7% of variance of reading comprehension, accuracy, spelling and reading speed.
These results shed lights on how a wide set of factors drive the development of literacy in children, and the mechanism of action of environmental variables .

94 p., 2020

Spécialité : Langage

Cote pour l’emprunt : Master 1979

____________________

FASSAYA, Judicaël L.

Haut quotient intellectuel et risques psychopathologiques et somatiques

sous la direction de Franck Ramus

Contexte : L’intelligence englobe la capacité d’apprentissage, à raisonner et à résoudre des problèmes d’une personne et se veut être un des principaux prédicteurs de sa réussite scolaire, professionnelle, de sa santé ainsi que de son l’espérance de vie. Néanmoins, la relation entre d’importantes capacités cognitives et un éventail de troubles mentaux et physiologiques demeure à ce jour débattue. Alors que certains défendent l’hypothèse qu’une plus grande intelligence caractérisée par un Haut Quotient Intellectuel (H.Q.I) joue un rôle protecteur vis-à-vis de certains troubles mentaux et somatiques, d’autres suggèrent qu’il s’agirait à l’opposé d’un facteur de risque pour ces mêmes pathologies, ou bien encore que ses effets variraient suivant le phénotype étudié. Cependant, de nombreuses études rapportant une plus grande prévalence de troubles chez les H.Q.I comportent également certaines lacunes méthodologiques rendant difficile la généralisation de leurs conclusions. Parmi ces défauts retrouve-t-on notamment l’absence de groupe contrôle, les variations dans la définition du groupe H.Q.I, ainis que le recours à des échantillons de faible effectif.
Objectif : La présente étude vise donc à étudier si la prévalence de troubles mentaux et somatiques diffère chez les H.Q.I par rapport à la norme tout en tenant compte des limites précedemment évoquées en tirant parti des données épidémiologiques offertes par la large cohorte U.K Biobank (N   150 000).
Méthodes : Des groupes hommes et femmes H.Q.I comprenant les participants admettant un score d’intelligence fluide du U.K Biobank ≥ 2 écarts-types (E.T) par rapport à la moyenne et Moyen Quotient Intellectuel (M.Q.I) compris entre 2 E.T autour de cette dernière ont premièrement été constitués. Dans un second temps, les différences de prévalence de troubles mentaux et somatiques particuliers dont dépressifs, anxieux, de l’addiction et asthmatiques ont été analysés grâce au test du Chi. et des régressions logistiques et linéaires multiples.

73 p., 2020

Spécialité : Psychologie

Cote pour l’emprunt : Master 1980

____________________

FOUCAULT, Cécrid

An artificial neural network-based framework to reveal efficient brain algorithms and representations for adaptive learning in complex structured environments

sous la direction de Florent Meyniel

Learning to predict future events based on past observations from the environment is a fundamental problem that the brain needs to solve. In real-life environments that are uncertain, changing, and structurally complex, the brain exhibits sophisticated behaviors for solving this problem. Bayesian models explain such behaviors as optimal for specific environments but present two major limitations: they fail to provide a general model architecture due to their dependence on built-in assumptions about the environment, and they do not provide biologically plausible or efficient algorithms for their computation. In this thesis, we propose a new modeling framework that adopts explicit requirements for overcoming these limitations and leverages the use of artificial neural networks. In three wellstudied environments, we asked networks to predict the next observation in a sequence, let them learn from their own prediction error, and examined four questions: 1) What minimal complexity allows the networks to perform quasi-optimally? 2) Do they exhibit adaptive learning, which is optimal in uncertain and changing environments? 3) Do they rely on confidence-weighted learning, which underlies adaptive learning in Bayesian models? 4) Do they develop an internal model of the environment’s complex hidden structure? We found that with only eleven gated recurrent units, the networks consistently achieve more than 97% of optimal performance and exhibit adaptive learning at short and long timescales. Analysis and perturbation of internal representations revealed that the networks discover novel confidence-weighted learning algorithms and develop probabilistic representations of the environment’s hidden transition structure. These results demonstrate the power of our framework to identify a general, biologically plausible and efficient architecture for quasi-optimal probabilistic learning, and reveal what algorithms and representations the brain can develop in a given environment.

52 p., 2020

Spécialité : Neurosciences

Cote pour l’emprunt : 1981

____________________

GAUDFERNAU

 Source reconstruction from EEG data to analyse neural dynamics during social interaction with a virtual avatar in autism spectrum disorder

sous la direction de Guillaume Dumas

Background: Exploring neural networks dynamics during social interaction could help identify biomarkers of Autism Spectrum Disorder (ASD), a neurodevelopmental disease characterized by social dysfunction. Functional connectivity has been widely investigated in autism, but studies have yet to include the cerebellum, a structure believed to play a major role in social cognition. Here, we proposed to investigate communication between cortex and cerebellum during social interaction. To our knowledge, this is the first study to investigate neural networks dynamics involving the cerebellum in patients with ASD during an interaction task

Methods: We combined the Human Dynamic Clamp (HDC), a novel paradigm for studying realistic social behavior, with high-resolution electroencephalography (EEG). Using source reconstruction, we investigated neural dynamics between the cerebellum and key cortical areas in ASD and typically developed (TD) participants.

Results: Power and connectivity analyses in the theta band revealed elevation of power in the cerebellum and elevation of cerebro-cerebellar connectivity during interaction compared to resting state in both groups, though the effect was only significant in the ASD group. They also revealed higher cortical and cerebellar activation in the ASD group compared to the TD, along with connectivity alterations between cerebellum and sensorimotor motor cortex and between cerebellum and parietal areas. However, these inter-group effects were not significant.

Discussion: We provided a pipeline to analyse neural dynamics at source level from EEG data. To gain statistical power, our analyses will have to be re-run with a bigger sample and improved artifacts detection. To our knowledge, this is the first EEG-based study to show increased theta power and connectivity involving the cerebellum during an interaction task in ASD participants, which is consistent with the known involvement of cerebellum in motor and social functions.

68 p., 2020

Spécialité : Neurosciences

Cote pour l’emprunt : Master 1982

____________________

GIORDANO, Jonathan

Anxiety-Depression Symptom Dimension Related to Deficits in Reinforcement Learning

sous la direction de Claire Gillan et Stefano Palminteri

Depression and anxiety are the most common mental disorders worldwide. In comparison with healthy individuals, depressed and anxious ones display increased sensitivity to negative events and bad news. Indeed, previous research has shown that healthy controls maintain a positivity bias in which they welcome good news, but refuse to accept, on a non-abstract level, bad news. In contrast, depressed and anxious individuals do not display a positivity bias. This can be conceptualized as a deficit in learning, which can be modeled within a reinforcement learning (RL) framework. One parameter within RL models which could explain these deficits are learning rates (LR), a parameter which weights the impact a learning experience has on an individual. LRs can be valence dependent, meaning different LRs can exist for better-than-expected outcomes (i.e. “good news”) and worse-than-expected outcomes (i.e. “bad news”). If LRs are responsible for the deficits in learning experienced by depressed or anxious individuals, then this link can be identified by using a cognitive task, modeling an individual’s performance by their behavioral data on said task, and associating model parameters, in this case LRs, with clinical scales measuring these disorders.
This thesis aimed to do just that by examining the data collected by Gillan et al., 2016. This dataset consists of behavioral data from 1,961 participants who completed the ‘2-Step task’ and self-report psychiatric questionnaires composed from nine different clinical scales. In keeping with the methodology used by Gillan et al., this study approached mental illness with a transdiagnostic, dimensional approach. That is, treating mental health as something with normal variation among the general population, and non-discrete boundaries between healthy controls and patients. In addition to this methodology, multiple RL models were examined in order to determine which model best captures the data, and does so parsimoniously. This includes a RL model with a single LR which served as a base comparison for models with LRs for better-than-expected outcomes (reward LR) and worse-than-expected outcomes (omission LR). Based upon the LR distribution among subjects, and associations with clinical scales, this study found that reward and omission LR did capture different information. Unexpectedly, reduced reward LR (i.e. reduced learning from better-than-expected outcomes) was significantly associated with increased severity in Obsessive Compulsive Disorder (OCD) and eating disorders. A supplementary analysis was also carried out to examine the effects of removing the stickiness parameter in the RL model, a variable which captures an individual’s tendency to select a choice in the task given they had selected it in the previous trial.

64 p., 2020

Spécialité : Neurosciences

Cote pour l’emprunt : Master 1983

____________________

LAMARQUE, Loia

Measuring and explaining the evolution of genderbiased attitudes in Early Modern France

sous la direction de Nicolas Baumard

Numerous qualitative works suggest that between 1550 and 1900, women’s condition improved in France in the conjugal, economic, political and intellectual spheres. While many theories compete to explain variations in gender inequality, we propose that these variations can be attributed to a psychological shift induced by a change in the environment. In this study, we attempt on the one hand to measure the evolution of patriarchy in French theater (N= 1034, CE= 1550-1900) via the importance of women on stage and their cantonment to traditional role topics (child, family, marriage). On the other hand, we test a psychological hypothesis derived from Trivers-Willard’s one, namely that a decrease in male reproductive variance is associated with a decrease in son-biased parental investment, thus decreasing gender inequality. Yet, male reproductive variance is affected by the environment, since a wealthier environment shifts individuals’ strategy from mating effort to parenting effort. We therefore test the hypothesis that economic development precedes a decrease in gender inequality in theater plays. If our results do not demonstrate a significant decrease in patriarchy in Early Modern period, they partially validate the Trivers-Willard hypothesis, since time-lag analysis reveals that an increase in GDPpc precedes a decrease in patriarchy for two of our four indicators. Further research should consider, in addition to economic development, the distribution of wealth, a key variable in the explanation of patriarchy proposed by evolutionary psychology theories.

42 p., 2020

Spécialité : Cognition sociale

Cote pour l’emprunt : Master 1984

____________________

LENORMAND, Diane

Neural activation patterns of episodic memory in children and adolescents: an fMRI study

sous la direction de Francis Eustache, Dr Lucie Hertz-Pannier, Dr Bérengère Guillery-Girard

Episodic memory is is the conscient memory of personal past events encoded with their spatial and temporal context (Tulving, 2002). It comprises objective elements, such as contextual details (“Who gave me the information?”), and subjective elements (when the subject feels like “reliving” the event). Despite the importance of episodic memory for human beings, little is known about the neural networks sustaining its objective and subjective components in children and adolescents. In this study, an fMRI dataset with subjects aged 8 to 15 years old, based on a subsequent memory effect paradigm, was analysed to try to shed some light on this still largely unexplored issue. After the pre-processing of the data, three different contrasts were implemented: Correct Recognition minus Forgotten items for general memory, Remember minus Forgotten items for the subjective part of episodic memory and Context minus Forgotten items for the objective part. The behavioural analyses explored the linear and polynomial correlations of performance with age. The observations made based on the results computed by SPM12 showed a greater activation for all three contrasts in the middle temporal lobe, which is consistent with previous studies, while unexpectedly no activation was found in the prefrontal cortex. The age did not correlate with the behavioural data or with the functional activation. Other contrasts were explored to further the analyses: the three contrasts with scrambled images as control showed greater activation in the occipital regions, hinting at a recruiting of the visual areas rather than memory. All these observations and experiments bring new light to the subjective and objective parts of episodic memory in children and adolescents, showing that the middle temporal lobe is involved in both. Further analyses, such as functional connectivity, could help precise this aspect.

30 p., 2020

Spécialité : Neurosciences

Cote pour l’emprunt : Master 1985

____________________

MORENO, Sarah

The neuro-computational dynamics of dysfunctional credit assignment in Obsessive Compulsive Disorder

sous la direction de Philippe Domenech

Obsessive Compulsive Disorder (OCD) is a complex neuropsychiatric disorder for which pathological doubt is a core clinical feature that might account for the emergence of obsessive thoughts and compulsions (Mataix-Cols, 2008, O’Connor et al., 2009). The pathophysiology of the disorder remains largely unknown (Pauls et al., 2014) but past research (Chamberlain et al., 2008, Gunaydin et al., 2016, Burguière et al., 2013, Ahmari et al., 2013) has highlighted functional and structural alterations in the lateral orbitofrontal cortex (LOFC); a region whose role in credit assignment – i.e., the process of assigning an outcome to the correct action, has been repeatedly argued (Noonan et al., 2017, Noonan et al., 2010, Walton et al., 2011). Building on this, Weiss*, Rondot* et al.’s study (in prep.,) characterized a selective impairment in credit assignment involving the LOFC in OCD patients and, since research suggests that credit assignment in the LOFC relies on belief state cognitive maps that are shaped by dopaminergic signals from the SNpc (Gershman & Uchida, 2019, Zhang et al., 2017), we believed that impaired credit assignment in OCD critically depended on the interaction between the SNpc and the LOFC. Thus, we characterized the subcortical dynamics of inference in OCD patients by recording local field potentials in the SNpc and the STN. We observed that conflicting evidence was encoded in the beta frequency band of the SNpc, while consistent evidence encoding was limited to the ventral STN. Although more research is necessary, we argue that this difference could be explained by the prediction error associated with cues that conflict with OCD patients’ strong prior beliefs in credit assignment.

2020

Spécialité : Neurosciences

Cote pour l’emprunt : Master 1986

____________________

MORLAAS, Orphée

When reality catches up: how the COVID-19 pandemic lockdown reduced optimism bias in belief updating

sous la direction de Liane Schmidt et Philippe Fossati

The optimism bias is a robust cognitive bias that is thought to be caused by an asymmetry in belief updating: people take more into account information that is favourable them, compared to unfavourable information. This in turn influences prospective risk perception, leading to risk underestimation, which partly determines behaviour. In the COVID-19 pandemic context, accurate risk perception is key to ensure compliance with preventive behaviour guidelines. Determining the effects of the COVID-19 lockdown on belief updating biases has not been achieved, and has a two-fold importance because: i) prospective risk beliefs relate to preventive behaviour, and ii) belief updating mechanisms are thought to be at the root of the human optimistic bias, which appears to be beneficial for both mental and physical health. To date, we do not know how the pervasive and ubiquitous threat that represents this crisis – most salient in France during the lockdown – has affected specific cognitive abilities, and important cognitive biases. To answer this question, we assessed belief updating within one group of participants before the very emergence of the pandemic in France (October/November 2019), and during the first three weeks of the lockdown. We also tested performance on the task in a second group of participants, recruited during the lockdown, who completed the task online three times within a one-week period. We found that before the pandemic, there was an asymmetry in belief updating: participants updated their beliefs significantly more when they received favourable information (i.e. good news), in line with previous findings reporting on optimistically biased belief updating. During the lockdown, no such bias was observed, in any group, and at any timepoint: belief updating became symmetrical, and the bias vanished. This restored symmetry in belief updating during the lockdown suggests that humans have the ability to adapt their biases to the threat context.

68 p., 2020

Spécialité : Neurosciences

Cote pour l’emprunt : Master 1987

____________________

NOBLET, Jonas

Résolution de schémas winograd par exploitation de la sélection verbale

sous la direction de Pascal Amsili

La résolution du pronom est le processus par lequel un pronom anaphorique est mis en relation avec son antécédent. Ce processus est complexe, et fait intervenir une grande variété d’éléments lexicaux et syntaxiques. Sur la base de problèmes de résolution anaphorique spécifiques – les schémas de Winograd – un défi a été conçu pour mettre à l’épreuve les systèmes d’intelligence artificielle. Un schéma de Winograd est une paire d’énoncés qui divergent par un ou deux mots, et qui contiennent un pronom ambigu à résoudre. Ces schémas sont conçus pour nécessiter des connaissances du monde pour être résolus. Les schémas de Winograd sont une tâche qui teste à la fois les capacités de raisonnement et la compréhension du langage naturel. Nous proposons dans cette étude, une méthode de résolution basée sur la restriction de sélection verbale. La sélection est le processus par lequel un prédicat filtre et catégorise sémantiquement les entités qu’il prend en argument. Nous nous appuyons sur des statistiques construites sur des corpus de données linguistiques pour modéliser ce phénomène et résoudre les schémas. Le succès partiel de cette méthode met en relief l’importance de la sémantique verbale dans l’interprétation du langage.

53 p., 2020

Spécialité : Langage

Cote pour l’emprunt : Master 1988

____________________

PAGLIARI, Marine

Modulation of the sense of agency in human-machine interactions. Evidence from metacognitive reports: Control-used and Control-felt

sous la direction de Valérian Chambon et Bruno Berberian

Recent technological advances have largely improved safety in many areas. However, the increasing presence of automation between operators and automated systems tends to distance operators from the results of action, leading them to leave the control loop. To counteract this “undesirable effect” of automation, we looked for practical solutions, and we were interested in the theoretical framework of agency. In a series of two experiments, we investigated the influence of automation, difficulty and a possible design solution (explicability) on explicit and implicit measures of the sense of agency through a cognitive psychology experiment with a randomized setup. Our first experiment focused on an implicit measure of the sense of agency, the intentional binding (IB). We were interested in the influence of automation and found that automation increased IB. Then, we found that adding supplementary information about what the system will do next, i.e. system explicability, also impacts the sense of agency by increasing IB. Finally, and more surprisingly, we found an increased IB with difficulty, something that we did not predict. By analyzing more carefully these results, we found that this observation was true only in forced trials. These findings suggest that sense of agency is increased only in situations where participants faced increased difficulty and where they selected a choice that the system had made for them (forced-choice trials). To better understand these results, we planned to realize a second experiment by introducing explicit measures of SoA. Unfortunately, we were not able to complete the project due to the national health circumstances. To conclude, our most interesting result concerns the increased sense of agency with explicability, demonstrating the importance of providing additional information about the system’s intentions in human-computer interactions.

67 p., 2020

Spécialité : Interface Homme-Machine

Cote pour l’emprunt : Master 1989

____________________

PENAUD, Sylvain

Incarnation et mémoire épisodique : Etude en réalité virtuelle

sous la direction de Pascale Piolino

Les études sur la conscience de soi se sont tournées vers l’étude des bases corporelles de la conscience et des mécanismes qui permettent à un individu d’être incarné dans un corps et de se constituer en tant que sujet de l’expérience consciente. Au-delà de son importance dans notre expérience immédiate du monde, le soi joue aussi un rôle important dans l’émergence d’une conscience étendue dans le temps. Ainsi, plusieurs études récentes suggèrent que le sentiment d’incarnation pourrait avoir un effet sur différentes dimensions de la mémoire épisodique, néanmoins à ce jour, ces effets restent mal connus. Dans cette expérience, nous avons réalisé une étude immersive en réalité virtuelle afin de mieux comprendre le rôle du sentiment d’incarnation au regard des différentes composantes de la mémoire épisodique. Nous avons successivement immergé des participants dans une série de villes virtuelles enrichies d’événements saillants selon trois niveaux d’immersion. Dans une condition d’immersion faible, la navigation des participants était entièrement contrôlée à distance par l’expérimentateur. Dans la condition d’immersion moyenne, la navigation était contrôlée par l’expérimentateur mais les participants devaient agir comme s’ils contrôlaient la navigation en marchant sur place et en levant un bras pour tourner. Enfin, dans la condition d’immersion forte, les participants avaient un contrôle total sur la navigation et étaient incarnés dans un avatar virtuel reproduisant leurs mouvements en simultané. Les participants étaient ensuite soumis à une série de tests de rappel et de reconnaissance des événements (éléments factuels et contextuels associés) perçus dans les villes virtuelles. Nos résultats montrent un bénéfice de la condition d’immersion forte sur les performances de reconnaissance factuelle et sur le sentiment de recollection associé. Nous suggérons que le sentiment d’incarnation permet de lier l’information encodée à des représentations de soi en mémoire à long terme renforçant l’association des éléments

58 p., 2020

Spécialité : Psychologie

Cote pour l’emprunt : Master 1990

____________________

RINKE, Patricia

Cognitive flexibility in social and non-social contexts in a reinforcement learning task

sous la direction de Mehdi Khamassi et Ouriel Grynszpan

The present study seeks to analyse common cognitive flexibility mechanisms for social and non-social contexts in a reinforcement learning task, in addition to processes that may have developed specifically for social cognition. This is an essential question when studying learning impairments of people with Autism Spectrum Disorder (ASD) – the long-term goal of the present project – since those common mechanisms might be altered in ASD. In a pilot experiment, it was examined whether cued learning yields different results when the cue is provided by a social agent or a non-social agent (Grynszpan et al., 2018). Model-free and model-based data analyses showed no significant difference between social and non-social cues.

The objective of the internship was two-fold: (1) to further analyse the data of the pilot experiment; (2) to contribute to the development and test of a new experimental protocol with an improved social framing and added cost to request advice. We predicted that typical adults show common mechanisms of cognitive flexibility in learning when they are free to request information from a social vs. a non-social cue. Indeed, our results suggest that subjects learn equally well in both contexts, and that they show differences in performance between the conditions “congruent”, “incongruent”, and “no info”. Moreover, subjects show signs of automaticity when they learnt the rule and slow down to initiate exploration after a rule change, independently of the type of cue. Finally, we can neither confirm nor fully deny that subjects’ decision time is longer when analysing the reliability of a social vs. a non-social cue, and recommend replicating the study for a definite answer.
On the basis of this study and the model-free data analyses, a computational model will be fitted to the data of the new experiment after the internship. This model could later help design interactive agents in serious games for pupils with autism, and maybe even support diagnosis one day.

81 p., 2020

Spécialité : Intelligence artificielle

Cote pour l’emprunt : Master 1991

____________________

TAL, Roy

The role of prioritized memory access and hippocampal replay through the lens of reinforcement learning

sous la direction de Mehdi Khamassi

How do we use information about our past experiences to inform decisions that we make in the present, or to better prepare for situations in the future? These types of personal experiences are stored as memories in the hippocampus, which is a brain region that has been shown to replay events in certain situations, potentially for planning, learning and memory consolidation. The question can thus be specified in terms of replay: when, where and in what order should memories be accessed to best inform our decisions? There are several replay strategies that can improve the efficiency of an RL agent, though at the expense of a computational cost for planning. We thus investigated these behavioral outcomes for different simulated prioritized replay strategies. The first model was based on the Expected Value of a Backup (EVB; Mattar and Daw, 2018), which considers which memories a rational agent ought to replay and is computed as the product of a ‘gain’ and a ‘need’ term. The former considers how much the agent could gain by replaying each event, whereas the latter measures how much it need be replayed given its immediate relevance. The next two models isolate each of these terms (i.e., the ‘Gain-only’ and ‘Neediii only’ models). Prioritized sweeping was also explored, which ranks experiences by how ‘surprising’ they were. Finally, two control models were included (i.e., no replay and random replay). After subjecting these models to extreme environments, it was apparent that the EVB and Gain-only models were most optimal but came at a high computational cost and low flexibility to changes in reward. In contrast, prioritized sweeping, which had a negligible computational cost, achieved very similar performance to these models in small environments but not in larger ones. Hence, a tradeoff emerged between complex but optimal prioritization strategies versus heuristic-based approaches, which are more biologically plausible.

64 p. + appendice, 2020

Spécialité : Intelligence artificielle

Cote pour l’emprunt : Master 1992

____________________

WAGELMANS, Anna

Egocentric and allocentric representations in mental time travel

sous la direction de Virginie Van Wassenhove

Mental time travel, or the projection of oneself in time (Suddendorf and Corballis, 1997; Tulving, 2002), allows us to imagine ourselves in the past or in the future, where we can re-live past events or imagine future ones. Similarly to the way humans can mentally project themselves in space towards a destination, imagining themselves in another location, mental time travel allows one to project themselves towards a specific event in time. This ability to navigate time requires not only access to memories, but also that events be ordered on a cognitive map that may be conceptualized as a “mental time line” (Arzy et al., 2009). The self holds a particular place on this timeline, as it can serve as the reference for the mapping of events on the timeline, in which case the representation of events will be characterized as egocentric. On the other hand, in an allocentric representation, events are mapped on a reference frame independent from the self. Just as in spatial navigation, these two types of representation interact to allow the navigation of the self through time (Buzsáki and Moser, 2013). However, the way the distance of self-projection is computed remains a question in research. Recent works (Gauthier and van Wassenhove, 2016a,b; Gauthier et al., 2019) have provided evidence that the ability to project oneself through time may rely on mental imagery. As such, distance effects specific to mental imagery may be expected in mental time travel. This thesis aims to explore such distance effects. Unfortunately, due to the CoVid situtation, sufficient data could not be collected to conclude on this question. Though the results may be inconclusive, this work proposes different perspectives on the state-of-the-art and how to further research in mental time travel.

35 p., 2020

Spécialité : Neurosciences

Cote pour l’emprunt : Master 1993

____________________