A neuro-computational account of success and failure in acquiring communication skills

Project Information

NeuroCompSkill

Grant agreement ID: 833694

DOI

10.3030/833694

EC signature date 19 July 2019

Start date 1 August 2019

End date 31 January 2027

Funded under

• EXCELLENT SCIENCE - European Research Council (ERC)

Total cost

€ 2 499 888,00

EU contribution

€ 2 499 888,00

Coordinated by

THE HEBREW UNIVERSITY OF JERUSALEM
Israel

Periodic Reporting for period 4 - NeuroCompSkill (A neuro-computational account of success and failure in acquiring communication skills)

Reporting period: 2024-02-01 to 2025-07-31

Summary of the context and overall objectives of the project

Dyslexia and autism are two developmental difficulties characterized by reduced ability to obtain expertise in two skills that are fundamental in modern society - written and social communication, respectively. The project's aim is to understand the atypicality in their learning mechanisms, from a combined cognitive, computational and neural perspectives and use it to understand the sources of these difficulties. The hypothesis that underlies the project is that the learning patterns in dyslexia and in autism differ in opposite manners from typical learning dynamics. In general, proficiency is acquired by implicit learning of environment's statistics, in particular, that relevant for the trained task. This implicit learning allows us to behave efficiently in complex daily situations – as we do not need to calculate everything from scratch. Our perception and action are largely determined by our previous experience in similar situations. When the environment is stable, our accumulation of information plays a stabilizing role. However, when the environment changes, the most informative history is the recent one. But we cannot always tell which situation we are at. Typical learning takes both time windows into account – very recent events have a large effect on our predictions, and accumulative information also plays a role – allowing both fast updating and stability of our predictions. Among dyslexics we find that the most recent information has a typical impact on perception and motor plans, but earlier information has a reduced effect. Among autistics we find the opposite atypicality – they update predictions slowly, yet persistent exposure allows adequate acquisition of long-term statistics.
These two characteristics can explain their basic difficulty: dyslexics are not fluent readers since they don't fully benefit from the statistics of their native language (with which familiarity is greatest) affords, since they do not efficiently accumulate regularities. Familiarity with these regularities is very important for proficient reading. Such regularities include: syllable frequency, morphology, word frequency, etc. By contrast, people with autism are slow in online updating and integrating of external stimuli into their perceptual priors and motor programs, which poses a particular impediment in social online situations. We now aim to better understand these impediments so that we can structure environments which maximize the efficiently used information.

Work performed from the beginning of the project to the end of the period covered by the report and main results achieved so far

The period was a challenging one – with COVID limiting so many of our experiments. Still, we conducted many studies.
A. We made substantial progress in measuring behavioral difficulties in detecting temporal regularities, in dyslexia and in autism compared to the general population (WP1). We found that, as we hypothesized, people with autism have difficulties in fast updating of their motor plans, as manifested in paced finger tapping, where people with autism are much more variable. Modelling the tapping data (WP3) allowed us to specifically locate the difficulty to slow error correction rather than to generally noisy representation of temporal intervals (Vishne et al., Nat Com, 2021).
B. We conducted an imaging study (WP5) on dyslexia, aimed to measure whether difficulties in detecting non-consecutive regularities are manifested already in sensory cortices. We found that dyslexia is indeed characterized by difficulties in benefitting from structural regularities in simple pitch discrimination, and that these difficulties are manifested in no-adaptation to such regularities, in contrast to good readers’, in auditory cortices including the primary auditory cortex (Gertsovski & Ahissar, 2022).
C. We collected tapping data of typical children ages 5-9 years (WP2), to trach the developmental trajectory of the ability to synchronize. At age 9 they still do not attain adult level performance. Modelling their tapping shows that their variability, is larger than neurotypical adults’ and is similar to that of adults with autism. Yet, it is due to a different mechanism. Unlike in autism, children’s error-correction mechanism is efficient. However, their retention of temporal intervals is noisy. Thus, though descriptively similar to adults with autism, their underlying mechanisms are quite different (Dobner et al, ISCOP 2022).
D. We have recorded EEG activity to assess the dynamics of adaptation in autism (WP4). We used a protocol we developed – presenting fixed inter stimuli interval and different inter-trial intervals in different blocks. Our preliminary results suggest that adaptation lasts longer in ASD, perhaps explaining their inflexibility in modifying predictions.
E. We launched a set of behavioral studies that systematically characterize the dynamics of learning in ASD and in dyslexia in both simple and complex stimuli, in different modalities (WP1): face categorization, motor planning, and perception of temporal intervals. These studies are in progress.

Progress beyond the state of the art and expected potential impact (including the socio-economic impact and the wider societal implications of the project so far)

We plan to
A. Complete a series of EEG studies (WP4) that assess the dynamics of adaptation to auditory stimuli in autism, using a set of newly designed behavioral and passive (watching a silent movie) protocols. These studies will measure different time constants, different complexity of regularities and different levels of noise (masking regularity). The aim is to bind, within a unified framework, a range of seemingly contradicting studies, which used different protocols and obtained mixed results.
B. Extend the age range of children’s tapping data to 13 years old, and to complete tracking the dynamics of their developmental improvement. Integrate data and modelling to clarify whether our preliminary results are valid: whether synchronization difficulties in typical children are explained by additional noise rather than rate of error corrections, as found for adults with autism (WP2).
C. Increase the range of motor programs that we assess behaviorally (WP1). Add a protocol where motor programs need to change online – as target position changes, so that the rate of updating them can be directly evaluated.
D. Measure in the magnet (fMRI) whether ASD is associated with sluggish inputs from the cerebellum, and/or from the basal ganglia, by measuring tapping, temporal estimation and motor updates in adult neurotypical and ASD groups.
E. Compare ability to acquire important expertise in a range of domains (e.g. .face discrimination)– between dyslexia and autism. We predict that people with autism will adequately acquire expertise that relate to long-term statistics, when we ensure that they have attended practice sessions. By contrast, people with dyslexia will have difficulties in acquisition based on item-specific repetition leading to fine discriminations, when given similar training conditions. These predictions are counter-intuitive and unique to our hypotheses.
Finally, we hope to substantially advance the state of the art by integrating all these diverse results into a unified conceptual framework, which specifies the updating mechanisms in the brain, i.e. : when do we update the representations of our perceptual predictions and motor programs; what is its underlying system-level neural circuitry; what are the external triggers (i.e. time or by events); where do these mechanisms differ in autism and in dyslexia.

figure-5.png

figure-3.png

figure-2.png

figure-1.png