Highway to (verbal) memory: Neurocomputational consequences of specifically human connectivity in perisylvian cortex.

Schomers, M.R.; Garagnani, M. and Pulvermüller, F.. 2015. 'Highway to (verbal) memory: Neurocomputational consequences of specifically human connectivity in perisylvian cortex.'. In: 23rd Annual Meeting of the Cognitive Neuroscience Society (CNS 2016). New York, United States. [Conference or Workshop Item]

No full text available

Abstract or Description

Rich long-distance connectivity in the fronto-temporal perisylvian language areas by way of the dorsal arcuate fasciculus (AF) sets apart humans from other primates and appears to be crucial for verbal memory and thus word learning abilities in humans. But how come that a stronger AF entails better working memory and word learning? To investigate this, we used a neurophysiologically plausible computational model implementing Hebbian learning mechanisms (cf. Garagnani et al. 2008) to simulate major regions and relevant neuroanatomical connections of superior-temporal and inferior-frontal cortex. We compared models with links documented in macaques (monkey model, MM) vs. those plus additional ones of the AF recently reported in humans specifically (human model, HM). The models were presented with spoken ‘words’ coded as concurrent neural activity patterns in auditory and motor cortex. 24 randomly initiated networks (12 of each type), were trained with 14 ‘words’ each and evaluated. Compared with the MM, HM models developed larger circuits with especially high circuit cell densities in the higher-association areas. Crucially, long-lasting memory activity was only seen in the HM circuits, whereas MM circuits lost their activation rapidly, thus giving little evidence of verbal memory processes. These functional differences did not depend on the amount of learning or other potentially relevant parameters. In summary, we show that the specific anatomical features of human cortex, especially the stronger connectivity within central perisylvian areas implicated by the human AF, contribute to better learning of spoken word forms and may explain the specifically human ability of verbal working memory.

Item Type:

Conference or Workshop Item (Poster)

Departments, Centres and Research Units:

Computing

Dates:

DateEvent
30 November 2015Accepted

Event Location:

New York, United States

Item ID:

24103

Date Deposited:

10 Sep 2018 08:58

Last Modified:

10 Sep 2018 08:58

URI:

https://research.gold.ac.uk/id/eprint/24103

Edit Record Edit Record (login required)