Structuring spatial knowledge through domain-general, non-spatial learning mechanisms (OutOfSpace)

Is space the main organizer of our mental reality? The answer to this question is apparently "yes".

According to recent views, nearly any type of knowledge would be organized through low-dimensional geometries relying on the same computations that are at play in the navigation of the physical space, as attested by the involvement of the hippocampal-entorhinal region in high-level cognition.

From an evolutionary standpoint, spatial processing mechanisms might have thus developed from originally mapping the navigable environment to representing cognitive spaces.

Moving beyond this spatiocentric view of the human mind, OutOfSpace will test the fascinating yet apparently counterintuitive hypothesis that non-spatial associative learning mechanisms are active (if not the main) ingredients in structuring spatial representations.

To pursue this aim, OutOfSpace will employ cognitively plausible computational models (i.e., distributional semantic models) based on non-spatial associative learning mechanisms to extract latent knowledge from natural language; this non-spatial information will be then used to predict a variety of spatial representations, using a pioneering interdisciplinary approach that combines computational, behavioural, eye tracking, fMRI, TMS-EEG and intracranial EEG work in both sighted and blind individuals.

OutOfSpace will thus make a breakthrough on two fronts: (i) probing whether language - a non-spatial learning environment - can encode and recode spatial knowledge without the need for a dedicated spatial memory system; (ii) attesting the strict interplay between spatial and non-spatial learning mechanisms in structuring mental representations.

More generally, these findings will open the venue to developing a more comprehensive, empirically-based cognitive neuroscience framework for processing and representing spatial information.