现有法国科学院和巴黎大学神经科学和认知研究中心所长 Florian Waszak 教授与中 国家留学基金管理委员会(CSC)正在联合招神经科学和认知方向的博士研究生. 阅读下文后有兴趣者可在线投递所需材料.
在线投递简要操作步骤和研究计划全文可通过下列链接下载
https://www.dropbox.com/scl/fi/a3qb9uchb21vozzlicwd1/online_procedure_CSC_2022.doc?dl=0&rlkey=nsk80tl6miqtbldntetqmgpes#opennewwindow
https://www.dropbox.com/s/g83kjobvo7j1rma/projet%20dynamics%20and%20time%20scales.pdf?dl=0#opennewwindow
如果不能下载可与 yangqing165@hotmail.fr 联系
相关博士培养或专业问题可与 Florian Waszak 教授直接联系 f.waszak@gmx.net
Sujet: Human action control: Time-course of sensory effect prediction and time-scales of sensory effect learning
Codirection:
Nombre de mois: 48 mois
Ecole Doctorale: ED 158 - Cerveau, Cognition, Comportement
Unité de recherche et équipe:
Integrative Neuroscience and Cognition Center (INCC) https://incc-paris.fr
The INCC brings together research groups that address the question of how the brain generates its integrative functions, i.e., how it creates the mind and behavior. It gathers researchers who work on all three main targets of neuroscientific research: healthy humans (both adults and infants), patients, (with perceptual, motor, cognitive, psychiatric or neurologic impairments) and animal models. The research topics of the INCC are centered on three axes:
1- Perception, Cognition & Behavior
Almost all teams of the INCC are involved in research on how humans perceive, understand and interact with the environment. The teams that are most active on this axis are the Vision and the Spatial Orientation teams. The Vision team, investigates perceptual processing and how it interacts with other cognitive functions such as action, attention and decision making. It will also study how the cognitive system takes action decisions, choses appropriate actions and automatizes behaviour. The Spatial Orientation team will complement this research by investigating low-level motor functions like multisensory integration, gaze and posture control as well as, not surprisingly, spatial orientation.
2- Development & Plasticity
The teams that are most active on this axis are the Speech & Cognition and the Perception, Action, Cognition (PAC). The Speech & Cognition team studies the acquisition of spoken language. It also studies cognitive functions like the representations of space and quantities, as well as the interaction between language and cognition during development. The Perception, Action, Cognition team investigates the development and plasticity of perception-action and cognition in typical and atypical populations. A big asset of these teams, which makes their research entirely unique, is their collaboration with Paris hospitals (Port Royal, Necker, Bichat) which grants them access to newborns, sometimes only a few hours after birth. In addition, through the Vision and Cognition Unit (Fondation Rothschild, Chokron) the PAC group can run studies in adult and children brain-damaged subjects.
3- Neurophysiology & Neurological Diseases
Several INCC teams study the neural and glial basis of integrative brain functions as well as their cellular and molecular underpinnings, partially with a particular focus on pathological conditions. The teams most active on this axis are the Cerebral Dynamics, Plasticity, Learning team, the Glia-Glia and Glia-Neuron Interactions team, and the Pathophysiology of Psychiatric Disorders team. These teams investigate amongst others the role of astrocytes and SGC signaling in sensory processing, dynamics and information processing in visual cortex, functions and computations of the cerebellum and the basal ganglia. On the pathological side, they study notably Parkinson’s disease, autism, schizophrenia and depression with the goal of unraveling the synaptic and neural processes that are involved in the diseases.
The project will take place in the Vision Group of the INCC. Research conducted by the Vision Group is aimed at better understanding the mechanisms underlying perception, attention, consciousness and the links between perception and action. Our interests include the properties of visual attention and of spatial maps, visual perception during and across eye and head movements and visual motion perception. We also perform research on hearing and touch, especially their interactions with vision. We deploy multiple techniques ranging from behavioral methods such as psychophysics and eye tracking to brain imaging techniques such as functional magnetic resonance imaging (fMRI), electro/magnetoencephalography (EEG/MEG) and transcranial magnetic stimulation (TMS). We use decoding methods to better understand brain processes and for designing the online control of robotic upper limbs.
Coordonnées de l’équipe:
Integrative Neuroscience and Cognition Center
CNRS & Université de Paris
45 rue des Sts. Pères
75006 Paris
Secteur: Sciences de la vie / Life Sciences
Langue attendue: Anglais
Niveau de langue attendu: C1
Description du sujet:
Human action control: Time-course of sensory effect prediction and time-scales of sensory effect learning
Our ability to interact with our environment seems limitless. We can learn to use keyboards, to play violin, to drive a car, and so on. We do all of this so swiftly, it is almost as if our body can become one with its surroundings.Almost? Recent decades gave rise to a school of thought which postulates that, in our brain, perceiving and acting is a unified process. The most straightforward representatives of this ’unified’ view are the ideomotor theory and the common coding principle (Prinz, 1997; Hommel et al., 2001; Waszak, Cardoso-Leite, & Hughes, 2012; Shin et al., 2010). The ideomotor theory proposes that to act purposefully presupposes knowledge about action-effect relationships (cf., Lotze, 1852). Once acquired, this knowledge can be used to select and perform an action by anticipating or internally activating their perceptual consequences (e.g., Greenwald, 1970). The common coding theory goes a step further. It claims that actions are coded in terms of the effects they evoke in the environment (e.g., Hommel et al., 2001; Hommel, 2009, 2015). Consequently, perceiving an action effect involves the same representation as performing the associated action and, conversely, performing an action involves the same representation as perceiving its associated effect (Prinz, 1997, 2015).
Two other approaches contain this idea to a certain extent. The first is predictive coding (Rao & Ballard, 1999; Friston, 2009, 2018, 2019). Within this framework motor control has been described as active inference, where predictions of sensory signals directly elicit motor actions that fulfil the predictions (e.g., Brown et al., 2011; Palmer et al., 2019). The second is computational models of action control (Wolpert, Ghahramani, & Jordan, 1995; Friston, 2011; Wolpert & Ghahramani, 2000; McNamee & Wolpert, 2018). Most similar to the ideomotor concept, many computational accounts incorporate ’inverse’ models that provide the motor command which will result in a desired end state (e.g., a particular sensory state) given a particular current state (Wolpert et al., 1995). In all these frameworks action is conceivedof as perceptual states. Not present perceptual states, but future perceptual states.
According to prediction-based theories of action control a voluntary movement should originate in the representation of the anticipated effect. Desantis et al. (2014) investigated this notion. Participants’ task was to discriminate the direction of a random dot kinetogram (RDK) triggered by the participants’ action. We used signal detection theory to assess d’ (perceptual sensitivity) and c (response bias) (Green & Swets, 1966) in congruent trials where the participant’s key press produced the predicted RDK and incongruent trials where it did not. Importantly, Desantis et al. (2014) presented the effect stimulus not only after but also before movement onset. We found d’ to differ between congruent and incongruent conditions from about 230 ms before movement onset on (see also Dignath et al., 2020).This research shows that perceptual representations are active during motor preparation. However, the paradigm they used bears the disadvantage that it assesses effect anticipation only at a processing stage where it already affects perception. Moreover, the paradigm does not allow us to assess continuously how the action effect is predicted in the course of the preparation of the action. The current project explores the dynamics of action2effect anticipation using Steady-State Visual-Evoked Potentials (SSVEP) to avoid this shortcoming (see Axis 1 below).
The second axis of the current project is based on the fact that humans live in a volatile environment in which action-effect relationships are subject to changes. Moreover, these changes occur at different time-scales, from very transient to very long-lasting. To adapt to these changes optimally, effect predictive processes should use multiple operators, each adjusting on a different time-scale. Long-lasting changes in action-effect relationships should result in long-lasting learning in an operator working on a long-term time scale, whereas transient changes should result in transient learning in an operator working on a short-term time scale. For example, steering a car necessitates a particular set of actions. It is useful to learn strong, long-lasting representations of the action-effect relationships of turning the wheel. However, driving on a very wet street or a muddy ground changes the action-effect relationships considerably. As a matter of course, the action system needs to adapt to these changes quickly, but without weakening the previous representations as the new action-effect relationships are only transient. This would call for multiple operators working on different time-scales. A similar mechanism has recently been demonstrated in perceptual learning (Bao & Engel, 2012). Axis 2 investigates this issue. This research will show whether and, if so, how the action system uses multiple time-scale adaptation to warrant the plasticity and, at the same time, stability of human voluntary action control.
Axis 1: Time-course of effect prediction traced by means of ssVEP
Research aims. To develop a continuous measure of effect anticipation we propose to employ a steady-state visually evoked potentials (ssVEPs) paradigm. Participants will be presented with not only one stimulus every trial, but they will receive continuous bursts of visual stimuli. This stimulation will yield a resonance response in the visual cortex at the frequency of the stimulation (e.g., Mueller et al., 1998). The analysis will then extract the time course of power in the frequency envelope centered on the stimulation frequency that is generated in the visual cortex. The aim here is to investigate the time course of VEP attenuation during movement.
Stimulus, task, design. Acquisition phase: Participants will learn two ideomotor representations: one action will trigger presentation of a vertically oriented Gabor stimulus in the upper visual hemisphere, the other keypress a horizontally oriented Gabor stimulus presented in the lower visual hemisphere. See Figure 1 in attached document.
Test phase: After the acquisition phase, participants will be continuously presented with a vertical and a horizontal Gabor stimulus in the upper/lower visual hemisphere. The two stimuliwill be presented at two different flicker frequencies (about 12 vs. 19 Hz) to evoke two different ssVEP responses in the visual cortex. Participants will be instructed to press a button at a rate of about once every 5 seconds. The analysis will investigate the time course of the modulation in power at the stimulation frequencies as a function of the action.
Hypotheses. Sensory attenuation (here assessed as attenuation of the power of the flicker frequency) will depend on the identity of the action: one key press will attenuate power of one flicker frequency, while the other key press will attenuate power of the other frequency. This will demonstrate that performing an action results in the anticipation of the ensuing sensory effect. More importantly, sensory attenuation will begin prior to the action. The continuous stimulation will allow us to draw a very detailed picture of the beginning/end of this attenuation(and, therefore, of the internal effect anticipation). We also intend to correlate beta power as an index of motor preparation with sensory attenuation. The prediction is that modulations in beta power will be positively correlated with the time course of ssVEP modulation. The experiment will reveal the precise dynamics of effect anticipation.
Axis 2: Plasticity of action-effect relationships represented on multiple time-scales
Axis 2 investigates prediction error brain responses as a function of long-lasting action-effect relationships interleaved with transient changes in these relationships to show whether action-effect representations are learned on multiple time-scales (cf., Bao & Engel, 2012).
Stimuli, procedure, design. Stimuli are two sinusoidal tones (high- and low-pitched). A total of 4 blocks, each comprising 1 long, 1 short, and 1 test phase, are presented. Long, short, and test phase comprise 322, 82, and 102 trials, respectively. In all phases participants are instructed to press randomly one of two keys with right/left index fingers to generate a tone. Importantly, the mapping between the tones and the left/right key presses differs between the three phases to mimic transient changes in action-effect relationships in an otherwise stable action-effect context. In the long phase, the mapping between the tones and the left/right key presses is 80% consistent. That is, 80% of the left key presses would generate tone A and 20% of the left key presses would generate tone B, whereas 80% of the right key presses would generate tone B and 20% of the right key presses would generate tone A. The subsequent short phase is identical to the long phase except that the mapping between the tones and the left/right key presses is reversed. In the following test phase the mapping between the tones and the left/right key presses is random. That is, a left/right key press is equally likely to generate tone A/B. Figure 2 in the attached document illustrates the design.
ERP analysis. In all EEG experiments ERP analysis will be conducted with a temporal principal component analysis which provides data-driven ERP components, avoiding subjective selection of time points and electrodes (Kayser et al., 2003; Hsu et al., 2013). This temporal decomposition provides a set of components reflecting the contribution of each time point on certain temporal components. Components corresponding to the N1, N2, and P3 will be identified. Their component scores serve as inputs for repeated measures ANOVA.
Hypothesis. We expect to find that, in the long and short phases, brain responses will show a complementary pattern: in the long phase tone B will generate a prediction error response after a left key press and tone A after a right key press (because a left key press predicts tone A, whereas a right key press predicts tone B); in the short phase this pattern will be reversed, indicating the system’s capacity to quickly learn the new action-effect relationships.
Importantly, the neutral environment test phase in which the tones are randomly produced by the two key presses will reveal the recovery of the prediction error response of the long phase. These results will suggest that action-effect prediction is controlled by at least two mechanisms, with initial action effect learning in the long phase affecting a longer-term mechanism and relearning in the short phase affecting a shorter-term mechanism in the opposite direction. In the following test phase the short-term effects rapidly decays, revealing again the initially learned longer-term representations. As pointed out by Bao and Engel (2012) in the context of perceptual adaptation a single controlling mechanism cannot account for the observed recovery of effects.
Compétences requises:
- strong background in cognitive neurosciences
- programming (matlab, python)
- knowledge in EEG recording and analysis
- good writing and presentation skills
Références bibliographiques:
see attached document
转自：“科研doge”微信公众号
如有侵权，请联系本站删除！
