Stages dans un laboratoire du DEC



Responsable des stages : Yves Boubenec (yves.boubenec at ens.fr)

Tous les élèves et étudiants de l’ENS, quel que soit leur département d’affiliation, et quelle que soit leur année de promotion, peuvent effectuer un stage dans un des laboratoires du DEC, pour une durée variable.

Un certain nombre des projets de rentrée du DEC peuvent donner lieu à stage plus long (qui n’implique par nécessairement une participation au projet de rentrée correspondant). Les normaliens intéressés peuvent directement contacter les responsables d’un projet de rentrée, qu’ils aient ou non participé à un tel projet au cours du mois de rentrée.

Tout normalien intéressé par un stage de rentrée peut également prendre conseil auprès de Brent Strickland (brent.strickland at ens.fr).



LISTE DES STAGES PROPOSÉS PAR DES MEMBRES DU DEC
(Last update: october 2016):


Institut Jean Nicod (IJN) / Laboratoire de Sciences Cognitives et Psycholinguistique (LSCP) / Laboratore des Systèmes Perceptifs (LSP) / NeuroPsychologie Interventionnelle (NPI)



Experimental epistemology: Craftsmen’s and farmers’ intuitions about reliability and knowledge

Laboratoire: Institut Jean Nicod
Responsables: Martin Fortier (Institut Jean Nicod) & Edouard Machery (University of Pittsburgh)
Emails: martin.fortier@ens.fr & machery@pitt.edu
Durée: at least 3 months
Langue du stage: French & English


Description:
Epistemologists have put several definitions of knowledge forward. One of the most influential definitions (notably advocated by Alvin Goldman) identifies knowledge with reliably produced true beliefs. According to reliabilism, then, if a process reliably produces true beliefs, the products of this process are justified, and they count as knowledge if they are also true. Consequently, knowledge does not require internal justification. Experimental philosophers aim at studying lay people’s philosophical intuitions. In this project, we are interested in collecting data examining intuitions lay people have as to what counts as knowledge and especially as to whether knowledge requires internal justification. We have already collected data from American participants (through Mechanical Turk). The internship will consist in adapting the tasks used previously and in running the experiment with French craftsmen and farmers. Our working hypothesis is that these specific participants may have somewhat different intuitions about the connection between reliable processes and knowledge. A very good level in French is required as the experiment will be run with French-speaking participants.

Etats de connaissance et choix du mode

Laboratoire : Institut Jean Nicod (CNRS/ENS)
Responsable : Alda Mari (IJN, CNRS/ENS)
Email : alda.mari29 at gmail.com
Durée : 3 mois minimum
Langue du stage: Français ou Anglais


Dans le cadre du projet: “(Non-)veridicality and subjectivity” en collaboration avec l’Université de Chicago.
Descrption: Ce stage théorique s’adresse aux étudiants intéressés par la modélisation logique des données linguistiques empiriques. Le domaine empirique étudié est celui du `mode’ (indicatif - subjonctif). Il sera proposé d’analyser les attitudes mentales, et de dégager les types de connaissance qui justifient l’emploi du subjonctif à travers les langues. Une modélisation en termes de mondes possibles est également attendue (après une etude guidée du cadre théorique).

Le français sera la langue étudiée par défaut, mais si l’étudiant(e) a une autre langue maternelle marquant la distinction indicatif/subjonctif, elle/il sera encouragé(e) à travailler sur cette langue, en apportant des données nouvelles. Ce stage est de trois mois minimum. Il s’adresse aussi bien à des étudiants modélisateurs désireux de se rapprocher des sciences du langage, qu’à des profiles plus littéraires, désireux d’apprendre à maîtriser des outils de formalisation. La/le stagière participera à une collaboration avec des chercheurs de l’Université de Chicago.

Epistemic states and mood choice

The internship targets students interested in modelling empirical linguistic data. Specifically, it will focus on mood shift (indicative-subjunctive) under epistemic verbs like `believe' or `know'. The theoretical goal is of spelling out the types of mental states that enhance the use of the indicative vs. subjunctive under these verbs, across languages. A modelization in terms of possible worlds is expected.

French will be the default target language (although the language of the internship itself can be English). If the student is native of another language in which the indicative-subjunctive shift is key to establishing a typology of epistemic attitudes, s/he will be encouraged to use data from that language, broadening the empirical landscape under investigation. The internship will last at least three month. Both students with a mathematical/computational background leaning towards human sciences and students with a literary background interested in formal models are welcome. The intern will contribute to an ongoing project with the university of Chicago on modality and subjectivity and will have the chance to interact with researchers from this university.



La perception des agents et patients (c’est à dire « qui à fait quoi à qui »)

Laboratoire : Institut Jean Nicod (CNRS/ENS)
Project supervisors : Brent Strickland & Tiziana Zalla
Contact : stricklandbrent at gmail.com or tiziana.zalla at gmail.com
Durée : 3 mois minimum
Project language: French and / or English


Une des leçons les plus importantes des vingt dernières années de l’étude de la perception humaine est que le système visuel est capable non seulement de se représenter des éléments de « bas niveau » comme la luminosité, le contraste, ou la couleur, mais peut aussi se représenter de manière automatique les catégories qui ont une importance sémantique telles que la distinction entre les objets solides et les substances, ou le genre biologique. Dans ce projet, on se demandera jusqu'à ou peut aller cette capacité sophistiquée. Notamment on veut savoir si le système visuel peut automatiquement distinguer un agent (c’est à dire l’acteur qui a été la cause ou l’initiateur d’un évènement) d’un patient (c’est à dire l’acteur qui a subit les effets de l’évènement), et si oui à quelle vitesse.

Pendant ce stage, on demandera au stagiaire de s’initier à “Blender,” un logiciel pour créer des vidéos animées et photo-réalistes. Le stagiaire aura pour but de créer et réaliser une expérience sur le thème de la perception des agents et patients au cours de l'année. Celle-ci comparera la performance d’ adultes en bonne santé (dans la détection des agents/patients) et celle de personnes autistes

Pré-requis : compétences minimales en programmation.



Langage et cognition : la corrélation entre la distinction grammaticale massif/comptable et la distinction cognitive matière/objet

Laboratoire : Institut Jean Nicod (CNRS/ENS)
Encadrants : David Nicolas & Brent Strickland
Email : dnicolas@gmx.net, stricklandbrent@gmail.com
Durée : 3 mois minimum
Langue du stage: Français ou Anglais


Dans de nombreuses langues, les noms communs sont divisés en deux sous-classes grammaticales, les noms massifs (comme “eau”, “bois”, “mobilier”, “sagesse”) et les noms comptables (comme “lac”, “chat”, “meuble”, “idée”). Cette distinction semble en partie basée sur une distinction cognitive entre matières et objets (Soja et al 1991). Mais c’est une corrélation très imparfaite, avec de nombreuses exceptions. Quelle est donc la nature exacte du lien entre la distinction langagière et la distinction cognitive ? Dans ce projet, nous commencerons par caractériser les prototypes mentaux associés à différents types de noms communs en examinant les images Google qui leurs correspondent. Nous verrons alors si les différents types de propotypes ainsi identifiés correspondent à différents types de noms, dont les emplois grammaticaux peuvent être dit plus ou moins comptables, ou plus ou moins massifs.



Measuring neural response to external stimuli during sleep

Laboratoire : Laboratoire de Sciences Cognitives et Psycholinguistique (LSCP - CNRS/ENS/EHESS)
Project supervisor : Sid Kouider
Contact : sid.kouider at ens.fr
Durée : 3 mois minimum
Project language: French and / or English


A fundamental issue in cognitive neuroscience concerns the extent to which the brain continues to respond to external stimuli during sleep. Falling asleep leads to a loss of sensory awareness and the inability to interact with the environment. Yet, while this was traditionally thought as a consequence of the brain shutting down to external inputs, it is now acknowledged that environmental stimuli continue to be processed, at least to some extent, during sleep. For instance, auditory stimuli with a relevant meaning (e.g. own names, own baby's cry, fire alarm) are more likely to lead to awakening. More compelling are recent studies from our lab showing that the sleeping brain, despites the absence of consciousness and overt behavioural response, continues to analyses and even prepare for acting on relevant stimuli. These studies rely on high-density electroencephalography (EEG), with signatures of semantic processing and motor preparation (N400, Lateralized readiness potentials on premotor cortex) to probe neural responses during sleep. In this new project, we will test whether the sleeping brain, despite any overt behavioural responses, can ‘focus attention’ towards the most relevant/meaningful auditory stream in a noisy multi-stream environment.



Developing a self-reflecting mind

Laboratoire : Laboratoire de Sciences Cognitives et Psycholinguistique (LSCP - CNRS/ENS/EHESS)
Project supervisor : Sid Kouider
Contact : Sid Kouider, sid.kouider at ens.fr
Durée : 3 mois minimum
Project language : French and / or English



Developmental cognitive neuroscience constitutes a booming new field in which researchers are probing the functional architecture of the infant brain. Babies can’t talk and tell us what they think about, but now thanks to new non-invasive methods such as high-density electroencephalography (EEG), it becomes possible to probe the electrophysiological signatures of major cognitive functions (perception, attention, memory, inferences, language understanding, etc). These electrophysiological signatures are called event-related potentials and correspond to specific patterns in the EEG signal. Here, we will attempt to answer the two fundamental issues of whether infants have a capacity for metacognition (do they know they know) and whether they experience self-consciousness (do they feel themselves as a unitary entity). Examining these self-reflection mechanisms, through implicit behaviours and EEG brain responses, will address the issue of whether humans in the initial state have a primitive self, or are actually unconscious about their own person.



How does the input shape babies' language acquisition ?

Laboratoire : Laboratoire de Sciences Cognitives et Psycholinguistique (LSCP - CNRS/ENS/EHESS)
Project supervisor : Christina Bergmann, LSCP (post-doctoral researcher 2014-2017)
Contact : chbergma at gmail.com
Website : sites.google.com/site/chbergma
Project language : French and / or English


Background
Research with American babies indicates that they know what words mean from 6 months on, that is long before they start producing words themselves [1, 2]! At the same time, babies differ a lot, even at the same age (try wordband.stanford.edu for an impressive, interactive example). Previous work has tied even slight differences in how clearly parents speak with how well an infant can discriminate two sounds [3]. It may seem surprising to you, but in fact a great deal of experimental, computational, and theoretical work suggests that how many people talk matters for word learning. Everybody sounds a bit differently and it can be tricky to recognize the same word when two very different people say it (as I explained in this blog post: [4]). At the same time, you may imagine that babies who hear lots of speaker also get a chance to identify the key aspects of how the word should sound -- so hearing many talkers could be helpful. At this point, we simply do not know which of the two, being distracted by differences or identifying important features of words and sounds across talkers, impacts language development.

Aim
This internship takes place within a larger research agenda on how the number of speakers a child is exposed to affects early language acquisition. For more information, see the project website.

Your master internship will contribute to answering this question, by helping with study design, experimentation, and data analysis.

Further details
We are interested in babies aged between 6 and 12 months, just as they are beginning to learn words. Since we want to understand the relationship between number of speakers and word comprehension, we measure both.

We measure the number of speakers in the environment in through both a questionnaire that the parent fills in, and via a small recording device that the child wears for two whole days. Finally, we measure babies’ word recognition with an eye-tracker in the lab. This project would allow you to get hands-on experience with three very different methodologies (questionnaires, daylong recordings, and behavioral experiments) -- in addition to working in a very successful babylab.

References
[1] Tincoff, R., & Jusczyk, P. W. (1999). Some beginnings of word comprehension in 6-month-olds. Psychological Science, 10(2), 172-175. [2] Bergelson, E., & Swingley, D. (2012). At 6–9 months, human infants know the meanings of many common nouns. Proceedings of the National Academy of Sciences, 109(9), 3253-3258. [3] Cristia, A. (2011). Fine-grained variation in caregivers’ /s/ predicts their infants’ /s/ category. The Journal of the Acoustical Society of America, 129(5), 3271-3280. [4] Bergmann, C. (2015). Why should different voices matter? Blog post 04.05.2015, available at http://bootphon.blogspot.fr/2015/05/why-should-different-voices-matter.html



Reverse engineering infant language development

Laboratoire : Laboratoire de Sciences Cognitives et Psycholinguistique
Responsable : Emmanuel Dupoux
Webpage : http://www.syntheticlearner.net
http://www.lscp.net/persons/dupoux/bootphon/page_4.html
Email : syntheticlearner@gmail.com
Durée : typically 5-6 months; part time possible; background in maths or programming or linguistics preferable
Langue du stage : Français ou Anglais


Infants learn their first language effortlessly, without explicit supervision, while being immersed into a complex and noisy environment. Infants do not seem to follow a hierarchical order (sounds, then words, then sentences) as adults would do, but rather, they start learning all of these linguistic levels in parallel. The aim of our project is to decipher this puzzling learning process by applying a 'reverse engineering ' approach, i.e., by constructing an artificial language learner that mimicks the learning stages of the infant. We use engineering and applied maths techniques (automatic speech recognition, signal processing, machine learning) on large corpora of child-adult verbal interactions in several languages. We develop psychologically plausible (unsupervized) and biologically plausible (bio-inspired) algorithms which can discover linguistic categories (words, syllables, phonemes, features). The validity of these algorithms are then tested in infants or newborns using behavioral techniques (eye tracking) or noninvasive brain imagery (Near InfraRed Spectroscopy, EEGs).



Other internships at LSCP
Webpage : http://www.lscp.net/stages.php?lang=fr




Imaging learning- and attention-induced changes in brain activity

Laboratoire : Neuro Plateform/Audition team, Laboratoire des Systèmes Perceptifs, IEC
Responsable : Yves Boubenec
Email : boubenec@ens.fr
Durée : 1 semester, starting anytime
Langue du stage: Français ou Anglais


Perceptual decision-making in detection and discrimination tasks results from a complex and subtle integration of sensory and contextual information. How task structure is learned and encoded in the brain is still an open question. In particular, stable large-scale recordings of brain activity appears to be technically challenging. As such, a complete understanding of the changes elicited in neural sensory maps during the course of learning is missing. Previous studies suggest that enduring receptive field plasticity in the adult auditory cortex may be shaped by task-specific top-down feedbacks [1,2], interacting with bottom-up sensory inputs and reinforcement-based neuromodulator release. However, the time-course of these neural correlates still remains to be investigated, as it could provide crucial insight on how they precede or follow increases in performance. In addition, how such changes occur across the whole auditory pathway (from midbrain to high-order auditory areas), and how gating modulates task-irrelevant sensory structures remains unclear.

We aim at tackling these two questions using functional UltraSound (fUS [3]) imaging in awake, behaving ferrets. This newly developed technique, based on Doppler effect to measure changes in blood flow, allows us to image brain activity on a large scale (1x2cm) at a high spatial (100µm) and temporal (down to the cardiac cycle) resolution over several days. Our experimental setup thus provides a perfect framework for imaging modifications of tonotopic maps and stimulus processing during learning, or even more dynamically across changing brain states. The goal of this internship is to acquire long-term spatio-temporal patterns of brain activation in multiple brain areas (from subcortical sensory structures to possibly frontal cortex) while ferrets are learning behavioral tasks, and then to identify possible changes in evoked response properties and functional correlation. This involves a combination of experimental skills (behavioral training, ultrasound recordings) coupled with data analysis.

[1] D. B. Polley, Perceptual Learning Directs Auditory Cortical Map Reorganization through Top-Down Influences, Journal of Neuroscience, vol. 26, no. 18, pp. 4970–4982, 2006.
[2] J. Fritz, S. Shamma, M. Elhilali, and D. Klein, Rapid task-related plasticity of spectrotemporal receptive fields in primary auditory cortex, Nature neuroscience, vol. 6, pp. 1216–1223, nov 2003.
[3] Osmanski, B.-F., Pezet, S., Ricobaraza, A., Lenkei, Z., & Tanter, M. (2014). Functional ultrasound imaging of intrinsic connectivity in the living rat brain with high spatiotemporal resolution. Nat Commun, 5, 5023.



Visual feature extraction from natural scenes

Laboratoire : Vision team, Laboratoire des Systèmes Perceptifs, IEC
Responsable : Peter Neri
Email : neri.peter@gmail.com
Durée : minimum 3 months
Langue du stage : Anglais


Visual processing of simple image elements (such as lines and edges) does not happen inside a cognitive vacuum: it may differ when those simple elements are embedded within natural scenes that look more like what we see every day, as opposed to th e featureless backgrounds that are normally used in the laboratory. We know a good amount about the mechanisms that support vision under the latter conditions (i.e. involving a simple stimulus with no natural meaningful content), but we know virtually nothing about how those mechanisms may change and/or be augmented/replaced by new mechanisms under conditions that are closer to natural vision (i.e. when the image starts making sense and contains recognizable objects). Your project would attempt to understand this transition from elementary vision to natural vision. Below is an example of a relevant publication that looked at this question: Neri, P. (2014). Semantic control of feature extraction from natural scenes. Journal of Neuroscience, 34, 2374-2388. Prior experience with computer programming (C++, Matlab, Python) is highly desirable.



Le rôle du striatum dans le traitement des ambiguïtés syntaxiques

Laboratoire : Laboratoire NPI (NeuroPsychologie Interventionnelle) au DEC (Ens, 29 rue d’Ulm, Paris) et à l’Hôpital Henri Mondor (Créteil)
Responsables : Maria Giavazzi et Charlotte Jacquemot
Email : maria.giavazzi@ens.fr / charlotte.jacquemot@ens.f
Durée : 3 mois minimum
Langue du stage: Français ou Anglais


Les modèles anatomo-fonctionels de traitement du langage impliquent principalement deux structures corticales: l’aire de Broca et l'aire de Wernicke. Cependant, les études de patients cérébrolésés- menées au sein de notre équipe aussi bien que par d'autres chercheurs - ont pu mettre en évidence que les structures sous-corticales, le striatum plus particulièrement, jouent un rôle important dans certaines étapes du traitement du langage, notamment pendant la résolution d’ambiguïtés (Sambin et al., 2012; Kotz et al., 2003; Mestres-Misse, 2012). Le but du stage est celui de développer et de tester une tâche comportementale afin d'étudier le rôle du striatum dans la résolution d’un certain type d'ambiguïtés qui sont les ambiguïtés syntactiques (e.g. La petite ferme...la porte, ou bien en anglais The horse race past the barn fell). Pour évaluer le rôle du striatum dans le traitement des ambiguïtés syntaxiques, nous comparerons les performances de patients avec une lésion du striatum (les patients atteints de la Maladie de Huntington) et de sujets sains. Des corrélations entre le taux d'atrophie du striatum chez les patients et leur performance linguistique seront également étudiées avec la technique de la Voxel Based Morphometry (VBM).