Laboratoire de Sciences Cognitives et Psycholinguistique
29 rue d'Ulm
75005 Paris, FRANCE
I am a CNRS senior research scientist and adjunct professor at the Laboratoire de Sciences Cognitives et Psycholinguistique, Institute of Cognitive Studies, Ecole Normale Supérieure in Paris, within which I head the "Cognitive development and pathology" team.
My research bears on the development of language and social cognition in children, its disorders (developmental dyslexia, specific language impairment, autism), its cognitive and neural bases and its genetic and environmental determinants.
Raphaël MIZZI obtained a tenure position as "Maitre de conférences" at Aix-Marseille University.
Congratulations Raphaël !
Page under construction
The general theme that interests me and unifies the different branches of my research is that of the brain and cognitive development of the child. Within this theme I am particularly interested in the acquisition of language, as well as the development of other high-level cognitive functions. The questions that lie at the heart of my research are: What are the cognitive predispositions that underlie the development of the child? How are these predispositions implemented in the infant's brain? How does the human genome build a brain with such predispositions? How do genetic and environmental influences interact to constrain the development of cognitive functions?
I approached language acquisition from the pathological side by studying developmental dyslexia, which I see as a neuropsychological model of phonological acquisition. I naturally went on to study specific language impairment. I have also begun to investigate the neural and the genetic basis of dyslexia.
More recently I broadened my interests to social cognition, as well as to other neurodevelopmental disorders (including autism and schizophrenia) using epidemiological tools insofar as they can inform us on the genetic and neural basis of cognitive development.
1. Exploring the phonological deficit in developmental dyslexia and specific language impairment
A long-standing project aims to better understand the nature of the phonological deficit in dyslexia. This project was supported by funding from the Dysbrain project (ANR), which explicitly includes a fairly extensive cognitive investigation component in the auditory and phonological field.
In general, our data support the idea that phonological representations of dyslexic people are not fundamentally degraded, but that their access to certain demanding cognitive tasks is compromised. With my collaborators Trevor Agus and Daniel Pressnitzer, I conducted a study showing that dyslexic adults memorize sounds as well as controls, contrary to the predictions of a number of hypotheses about hearing and phonological deficits in dyslexia (Agus, Carrión Castillo, Pressnitzer, & Ramus, 2014).
My research on phonological processing in dyslexia is paralleled by similar work on children with specific language impairment, in collaboration with Heather van der Lely, Chloe Marshall and Stuart Rosen (UCL) and with funding from the ESRC.
2. Genetic, neurological and cognitive characterisation of developmental language disorders : the Genedys project
Together with geneticist Thomas Bourgeron at Pasteur Institute and many collaborators all over France, I set up the Genedys project, which explores the genetic bases of developmental language disorders (dyslexia and Specific Language Impairment, SLI), in relation to both cognitive and neuroanatomical phenotypes. It is based on the participation of dyslexic, SLI and control children, aged 8 to 12, and their family members (in the case of multiplex families). Children undergo a complete behavioural test battery (around 3 hours of tests covering all the main aspects of general cognitive functioning, oral and written language), donate either a blood or a saliva sample for DNA extraction, and a subset of them undergo a neuroanatomical MRI scan, in order to define a neural phenotype for those disorders.
The Genedys project also became part of the European project Neurodys, led by Gerd Schulte-Körne in Munich, and including many other collaborators. Within Neurodys, around 1500 dyslexic and 1500 control children from 9 countries are behaviourally tested on a common test battery (a subset of the Genedys battery) and donate a DNA sample.
In this latest study, based on hierarchical logistic regression analyzes, we confirm that phonological awareness and fast naming are the two main predictors of dyslexia. On the other hand, we show that these two variables have about the same relative weight in the prediction at all levels of orthographic complexity. On the other hand, their predictive power (and that of the complete model) increases with orthographic complexity. Thus, in spelling languages such as Finnish and Hungarian, a standard deviation deficit in phonological awareness multiplies the risk of dyslexia by 2, but in the most complex spelling languages such as English and French it multiplies it by 4. In the spelling languages, other (perhaps visual) factors are obviously at work, which have not been measured in this study.
The global analysis of cognitive data shows the quasi-universality of the phonological deficit among dyslexic children. On the other hand, we do not observe any real case of visual stress and a limited impact of the visual-attention span reduction. In addition, visuo-attentional disorders always coexist with a phonological deficit, unlike the independence reported by Bosse, Tainturier and Valdois (2007). Overall our results challenge these alternative theories of dyslexia (Saksida et al., 2016).
In addition, we contributed French genetic data to an international consortium (from but not limited to Neurodys), to the first genome-wide analysis of dyslexia (Giallusi et al., 2019).
Genetic data were also partially analyzed but did not replicate known associations (Becker et al., 2014).A later collaboration with Anne-Lise Giraud and Katia Lehongre attempts to investigate cortical oscillations supporting auditory sampling of speech in dyslexia.
These efforts were continued in the Dysbrain project, which gathered very high resolution images at the 7 Tesla MRI, and MEG responses to auditory stimuli.
Cognitive aspects of developmental dyslexia
Cortical oscillations in developmental dyslexia
Neuroanatomy of developmental dyslexia
Genetic basis of developmental dyslexia
3. Dysbrain Project : Neuroanatomy of dyslexia
The work of analyzing neuroanatomical T1 images of dyslexic children and witnesses has been a very important part of the work done by my team in recent years, involving Irene Altarelli (PhD student, 2009-2013), Katarzyna Jednorog (postdoctoral 2010-2011, collaborator in Warsaw since), Jingjing Zhao (postdoctoral fellow, 2012-2014, collaborator in Xi'an), Lou Scotto di Covella (PhD student, 2013-2017) under my supervision and with the help of collaborators (Ghislaine Dehaene, Michel Thiebaut from Schotten). These analyzes focus on the 64 brains scanned during the Genedys project, combining occasionally with additional images acquired by G. Dehaene-Lambertz as well as international collaborators (K. Jednorog, S. Heim).
Methods were the folllowing: voxel-based morphometry (VBM), surface based morphometry (SBM), morphometry of targeted regions, in particular the temporal planum and the Heschl gyrus; morphometry of the furrows; and multi-parametric classification algorithms.
We conclude from our analyzes that the abnormal functional development of the visual word-form area in dyslexic children is associated with a smaller thickness of cortex, and that this can not be attributed exclusively to the least experience of these children in reading (Altarelli et al., 2013).
Our results are more generally consistent with the idea that the brain disturbances underlying dyslexia are probably partly different between boys and girls, joining earlier conjectures (Humphreys, Kaufmann, & Galaburda, 1990, Ramus, 2006) as well as more recent studies (Evans, Flowers, Napoliello, & Eden, 2013). We develop this point of view in a book chapter (Ramus et al., 2017).
The initial analysis of VBM on the French data proving to be unconvincing, we have changed scale, thanks to the work of our ex-postdoctoral fellow K. Jednorog who gathered the data of several countries: 84 French children (from Genedys as well as a study by G. Dehaene-Lambertz), as well as 71 German and 81 Polish. We suggest that the previously published results of VBM on dyslexia are probably false positives. We hope that this demonstration will help cognitive neuroscientists to raise their methodological standards, notably by increasing the numbers, and by systematically carrying out replication studies, as is now the case in genetics in genome-whole studies (Ramus et al., 2018).
Among the other ways to analyze brain anatomy on the basis of T1 images, our doctoral student Lou Scotto di Covella performed a morphometric analysis of grooves (Scotto di Covella, submitted), while K. Jednorog conducted an analysis, multiparametric classification being written (Płoński et al., 2017). These two analyzes were conducted on 236 French, Polish and German brains.
Finally, we have also begun to exploit the dissemination data, thanks to our post-doctoral fellow Jingjing Zhao and our collaborators Michel Thiebaut of Schotten and Jessica Dubois. We have shown that two bundles appear to have a different pattern of symmetry in dyslexic children: the lower occipito-frontal bundle, which has a less pronounced asymmetry to the left in the dyslexic child than in the control, and the second bundle segment superior longitudinal, which has a more pronounced asymmetry to the right in the dyslexic child. We now plan to continue with a thorough analysis of the corpus callosum in dyslexia (Zhao et al. 2016 ; Lou et al. 2019).
4. Cognitif development and litteracy acquisition in french and chinese infants (CROSSLINGDYS)
My colleagues at Beijing Normal University (BNU) (PI: Hua Shu) followed more than 300 children from the age of 3 to 14, collecting many interesting measurements over the years, especially in connection with learning to read. The data from this cohort can make it possible to ask the following questions in particular: 1) What are the early (cognitive, environmental) predictors of later acquisitions in the field of oral and written language? 2) What are the cerebral bases of these acquisitions? 3) What are the early predictors and cerebral bases of dyslexia? 4) Comparing with our data acquired in alphabetical languages, to what extent do the answers to these questions depend on the language?
BNU doctoral student Mengmeng Su spent a year in my team (2013-2014) thanks to funding from the China Scholarship Council, notably to train in the analysis of diffusion imaging (an MRI was passed at age 14 in 79 of these children), and to carry out various longitudinal analyzes. Hua Shu and I have obtained PRC funding between the CNRS and the National Natural Science Foundation of China, which will allow us to continue to exchange and collaborate in the next 3 years.
In a first study (Su et al., 2017), we analyzed the predictive power of early family factors and initial cognitive abilities measured at 3-5 years of age on reading skills assessed in great detail at the age of 11 years. We found on the one hand a prediction of 20 to 34% of the variance, substantial over such a long time, on the other hand that the nature of the predictors varies according to the capacity of reading measured (character recognition, fluence of reading, text comprehension, spelling). In addition, mediation analyzes show that the effect of family factors is entirely mediated by the measured early cognitive abilities.
The following two studies exploit the MRIs obtained with these children, including diffusion imaging. The first shows that the connectivity (fractional anisotropy) of the left arched beam (long and posterior segments) is associated with different vocabulary development trajectories from 3 to 11 years (Su et al., 2017). The second shows that Chinese dyslexic children differ from controls by the connectivity of left arched (long segment) and lower longitudinal bundles (Su et al., 2018). In addition, arc beam connectivity is correlated with phonological capabilities, as in the alphabetic languages, while the lower left longitudinal beam correlates with morphological capabilities, which probably reflects a specificity of morphographic languages (Su et al., 2018).
5. Cerebral bases of voice perception in dyslexic adults
The hypothesis we are testing is that the endogenous cortical rhythms of dyslexic people may be less differentiated between the left and right auditory cortex, or that the left auditory cortex may be less receptive to gamma band oscillations than matched controls. Alternatively, dyslexics could have quite normal cortical rhythms, which would be consistent with our results obtained by psycholinguistic methods (Ramus & Szenkovits, 2008) (In collaboration with Anne-Lise Giraud and Katia Lehongre).
After a first MEG study (Lehongre, Ramus, Villiermet, Schwarz, & Giraud, 2011), we conducted a simultaneous fMRI / EEG study (Lehongre, Morillon, Giraud, & Ramus, 2013), which confirms the anomaly left auditory cortex responses at 25-35 Hz, while showing that brain responses to other frequency bands relevant to speech (1-3 Hz and 4-7 Hz) are normal, challenging Goswami's competing theory (2011).
These results are particularly remarkable for their strength and their adequacy with the predictions of Giraud et al. (2007). Nevertheless, their precise interpretation remains uncertain, in particular the alleged cognitive consequences, which will have to be the subject of targeted investigations (Giraud & Ramus, 2013). It was the focus of a portion of the Dysbrain project that was completed in December 2015, and will continue to be analyzed in 2017-2020 with funding from the Agir Foundation for Hearing.
6. Social cognition and disabilities
With Emmanuel Dupoux and Pierre Jacob, we carried out, thanks to ANR funding for 2009-2013, the Socodev project on different aspects of the development of social cognition.
We have studied how different states of an agent can be intuitively extracted from simple visual scenes and represented by an observer focusing on four types of representations: (1) the representation of the animacy character (the difference between inanimate objects and self-propelled animated agents); (2) the representation of the goals of an agent; (3) the representation of the beliefs of an agent and (4) the representation of the social attributes of an agent.
Our general method has been to use nonverbal stimuli (silent animations) and to contrast explicit verbal tasks with more implicit nonverbal measures, such as eye fixations. We studied these abilities in adults, children, and their deficits in autism spectrum disorder (ASD) and in schizophrenia. This work was carried out by two successive psychiatric PhD students: Baudouin Forgeot d'Arc for the initial development of the battery of tests and its application to autism, then Paul Roux for later developments and application to schizophrenia. The data collected is very rich and is analyzed and published progressively.
6.1 Representation of the animacy character
This is the most basic component of the system: the identification of agents as such, as opposed to inert objects. We are trying to get more implicit measures of animated character detection, notably using eye-tracking.
Our animation measures of triangles Frith and Happé were validated on 17 participants (Roux, Passerieux, & Ramus, 2013) and conducted in patients with schizophrenia, which we have shown that their eye followings are normal, even though verbal descriptions of these stimuli are impoverished (Roux, Smith, Passerieux & Ramus, 2017). These findings are consistent with the idea of a normal social perception in schizophrenia and a deficit at a more explicit and metacognitive or language level. This protocol is nevertheless limited by its heterogeneity, which has led us to prefer more controlled stimuli described below.
6.2 Representation of the goals of an agent
In order to go beyond the limits of triangles animations, Paul Roux has developed a new experiment based on the protocol for detecting the tracking of an agent (circle) by another of Gao et al. (2009) (tested on 40 healthy adults). The manipulated kinematic parameters satisfactorily modulate the subjects' ability to detect stalking. In addition, we have shown that simultaneous eye-tracking measurements predict the behavioral response to a large extent, allowing for a totally implicit measure of tracking detection.
This protocol was then adapted and used on a group of 29 schizophrenic patients and 29 matched controls. The results showed a lower sensitivity to detection of stalking in patients. Eye tracking analyzes revealed that this lower sensitivity was manifested both in implicit treatment and in explicit cognitive processing.
Finally, an important element in the detection of the goals of an agent is to decode the direction of his gaze to determine the object of his attention. The question of whether this ability to follow the eye is deficient or not in autism remains debated. This protocol was passed by 33 patients with ASD (Autism Spectrum Disorder) and 38 matched controls in age and IQ. Our results suggest a slight deficit in monitoring the direction of gaze in patients (Forgeot d'Arc et al., 2017). Nevertheless, they leave open the question of the direction of causality between this deficit and the more general disturbances of social functioning in autism.
6.3 Representation of believes
During his thesis, Baudouin Forgeot d'Arc created a new experimental paradigm designed to test the mentalizing abilities, while avoiding the many pitfalls and limitations of the classic tasks of false belief (Sally-Ann type).
These are animated films made with Flash software, featuring scenarios of the wrong type. The idea is that the eyes of the subjects will betray their surprise when the scenario will lead to an unexpected fall; the prediction of the fall involving the mentalizing abilities in the mentalist scenarios, but not mechanistic ones.
In addition to a validation study of this paradigm (Forgeot d'Arc & Ramus, 2011), we used it with the same group of schizophrenic patients as before. We found that patients had lower belief-granting performance than controls, in line with an already important literature. In addition, by analyzing the fixations on different areas of interest of the scene (object that is moved or modified, face of the agent who sees or does not change), we showed that patients looked less at the face of the agent (but not the object), and that this difference in access to information made it possible to explain most of the difference in the explicit task (Roux, et al., 2014). Our results therefore suggest that either eye-tracking difficulties in schizophrenic patients or their lower interest in faces has an impact on their ability to attribute mental states to complex social situations.
6.4 Representation of the social attributes
Beyond the mental states of an agent, human beings commonly attribute to others (rightly or wrongly, but most often spontaneously) different social attributes (kind, sympathetic, intelligent, etc.). It is interesting to try to understand at the same time the mechanisms leading to these social judgments, and their possible deficiency in certain populations. For this purpose, B. Forgeot d'Arc has developed a task in which the participants see pairs of faces, and must decide which one looks the most "nice". Some of the stimuli are photographs of natural or synthetic faces in which the facial features associated with kindness are manipulated, presented in pairs differing in these facial features in a subtle or important way.
The results indicate that autistic patients have judgments less consistent with the reference population in the photographs, but not on the synthetic faces, regardless of the level of difficulty (pairs differing little or much) (33 autistic individuals and 38 controls). This suggests that either the difficulties of autistic people in social judgments are not due to perceptual deficits, or they are very subtle perceptual deficits concerning indices present in natural but non-synthetic faces (Forgeot d'Arc et al., 2016).
7. Genetic and environmental basis of cognitive development
As I say in my introduction, the study of factors determining individual differences in cognitive development requires an epidemiological approach. With this in mind, the EDEN project is currently my main database. I also participate with my Chinese colleagues in similar analyzes of the data of a population of Chinese children.
The EDEN project (http://eden.vjf.inserm.fr/index.php/fr/accueil) is an Inserm cohort of 2000 children followed since gestation in all aspects of their health. They include a lot of data on the environment of children, as well as questionnaires filled regularly by parents. It also includes a cognitive component, including a 2-year vocabulary measurement (McArthur-Bates questionnaire completed by parents), a cognitive test battery at 3 and then 5 years, and questionnaires on behavior and psychological disorders (Strength and Difficulties Questionnaire: SDQ).
This cohort will allow us to ask a whole series of new questions on the developmental trajectories of different cognitive abilities and learning, and the interactions between genetic and environmental factors influencing them. Hugo Peyre, doctoral student and then associate researcher in my team and child psychiatrist trained in epidemiology, worked on the cognitive data acquired so far.
7.1 Development and stability of language skills between 2 and 3 years
The first analysis focused on the development of language skills between 2 and 3 years and on the stability over time of identified risk profiles at 2 years. The results show that language delays are still unstable between 2 and 3 years, in both directions (recovery of delay, and delay occurring later). Beyond the general factors influencing the development of language (sex, prenatal alcohol, breastfeeding, primiparity, gestational term, parental education, maternal stimulation), changes between 2 and 3 years are predicted particularly by the consumption of alcohol during pregnancy. pregnancy, breastfeeding, parental education and the frequency of maternal stimulation (Peyre et al., 2014)(Peyre et al., 2019).
7.2 Association between language difficulties and attention deficits
Is there a cause-and-effect relationship between the two, and in this case, in which direction? We found that 3-year language skills predict some of the variance in inattention symptoms at 5 years (estimated by the SDQ), but not vice versa. Our results are therefore consistent with theories that postulate a role of language in the child's behavioral regulation and executive control (Peyre et al., 2016a).
7.3 What are the environmental factors of cognitive development that produce global effects on all cognitive functions?
Which of these are the ones that produce possibly specific or more important effects on certain functions? We found that while most factors have general effects, some have greater (if not exclusive) effects on verbal abilities. This is the case, for example, in the level of parental education, breastfeeding, but also the number of older siblings (negative effect). Our results make it possible to begin to distinguish the factors that induce positive correlations between the cognitive functions contributing to the factor g of general intelligence, and those which on the contrary induce a differentiation of cognitive functions (Peyre et al., 2016b).
7.4 Are gifted children more anxious?
Based on the IQ and behavioral symptom (SDQ) data of more than 1,000 5-year-old EDEN children, we found that the 23 that are more than 2 standard deviations above the mean (IQ> 130, "gifted" children) do not actually have more behavioral symptoms than others, challenging popular belief (Peyre et al., 2016c). It will be interesting to repeat this analysis in early adolescence when the data at 11 years old were collected, some theories assuming that it is only on arrival at the college that the "maladjustment" of gifted children to the system school is fully revealed.
7.5 Is there a relationship between growth and IQ?
We have investigated the extent to which the child's major developmental stages, as assessed by relatively brief clinical examinations at 4, 8, 12 or 24 months, predict a child's IQ measured at 5-6. years. We found that developmental milestones at 24 months (including those for language) can predict about 20% of IQ variance at 5-6 years. On the other hand, measurements at earlier ages have almost no predictive power (Peyre et al., 2017).