The general goal of this subproject will be to select well-defined cognitive tasks, already partially studied by cognitive neuroscience (perception action; motivation, decision and reward; learning and memory; space, time and numbers; from sensory processing to multimodal perception; capabilities characteristic of the human brain) to apply standardised stimulation protocols (localisers) and to dissect associated patterns of brain activation and response dynamics. The observed patterns of activation and dynamics will make it possible to identify a) the brain regions involved in the task, b) the likely circuitry connecting these brain regions, and c) principles of information processing within and between these brain regions. We refer to this information collectively as the cognitive architecture for the task. Cognitive architectures will provide the constraints for the cognitive brain models developed by Subproject 4. Models of cognitive architectures will span scales ranging from high-level conceptual models to more explicit models with individual simplified neurons.

Successful models will be implemented into neuromorphic computing systems using the Neuromorphic Computing Platform (Subproject 9) and tested for their cognitive capabilities using the Neurorobotics Platform (Subproject 10). Combined with behavioural performance, cognitive architectures will guide reconstruction of multi-scale models where data is missing and will provide benchmarks for the validation of biologically detailed brain models produced by the Brain Simulation Platform. The objective for the ramp-up phase will be to i) identify cognitive architectures for a subset of identified cognitive functions; ii) develop localiser protocols for these functions; iii) gather, or if necessary acquire, unique theory-constraining data to be provided to modellers as benchmarks; iv) work with Subproject 4 to translate these cognitive architectures into cognitive models, specifically focusing on two theoretical models: a cognitive architecture for spatial navigation and a cognitive architecture for visual action recognition. The data generated within Subproject 3 will be complemented by data generated by projects responding to a competitive call - Cognitive Architectures.

What People are Saying

  • Our goal within the Human Brain Project is to describe cognitive architectures that allow the human brain to perform certain cognitive functions such as reading, language, face recognition, social cognition. What we call a cognitive architecture is a set of brain areas, their internal codes, and their internal connections that allow the brain to perform a specific function.

    Prof. Stanislas Dehaene, INSERM-CEA Cognitive Neuroimaging Unit
    Commissariat a l'energie atomique et aux energies alternatives, leader of the Cognitive Architectures subproject