TOPOLOGICAL VERSUS DYNAMICAL ROBUSTNESS IN A LEXICAL NETWORK

Abstract

Semantic memory is the cognitive system devoted to the storage of conceptual knowledge. Empirical semantic networks constructed from adult generated free word association data represent a good map for this system. Concepts, represented as words, link each other with a certain weight representing the strength of their relation. Everyday experience shows that word search and retrieval processes, which can be assimilated to traversals on a semantic network, provide fluent and coherent speech, i.e. are efficient and robust. Brain pathologies, such as Alzheimer's disease or schizophrenia, severely damage neural structures and associated capacities. Thus, semantic networks must also undergo disruption, but the question is how long cognitive processes, which depend on the underlying structures, can operate before collapsing. Interestingly, we find that degradation of the original structure has a dramatic impact on the topology of semantic network, whereas the dynamics upon it evidence much higher resilience. We define this problem in the framework of percolation theory.