Network Level Representation of Conceptual Content

Our ability to store knowledge and represent the world within our minds has spanned multiple disciplines (philosophy, psychology, neuroscience). Currently, theories of human conceptual knowledge suggest that human representation of the world is widely distributed across the brain. Regions involved in sensory/motor simulation as well as amodal systems contribute to our flexible ability to manipulate knowledge. A detailed account of how the overall human semantic system works at a network level is still lacking. To begin our investigation into how knowledge is distributed across brain networks, we will first consider a specific kind of knowledge - person related knowledge. Chapter 2 will look at the behavioural indicators of person-knowledge organisation. We will ask participants to judge explicit/subjective similarity between different person-knowledge domains: social, physical, biographical, episodic and nominal knowledge. This will allow us to investigate whether these processes are independent or related to each other. We will then compare these judgements to implicit similarity measures to see whether correlated patterns of responses or reaction are informative about cognitive similarity. Chapter 3 will look at how the brains core/extended system for face perception coordinates across the aforementioned person-knowledge domains. We will investigate the representational similarity of different person-knowledge domains in individual regions, and crucially - across the network as a whole. This will allow us to address whether cognitions are localised in individual regions or distributed across the whole network. Chapter 4 will investigate the stability of network organisation when going across modalities. Extended system for face perception has been shown to be recruited during familiar name reading. We will ask whether network-level patterns of activation during person-knowledge remain stable across input modalities. Chapter 5 will generalize the network-level approach to investigate broader semantic categories. We will interrogate how diverse regions activated during semantic processing, interact during processing of naturally occurring conceptual categories. We will use a corpus derived semantic distance model and compare it to individual region activity to that of the network overall. We will ask whether information about conceptual distance between categories is contained within individual regions or arises as a product of coordinated effort across the network. Combined, evidence presented in this thesis speak to the distributed nature of cognitive representation. Different kinds of person-knowledge and object categories are highly linked and rely on overlapping neural substrates. We demonstrate that instead of being specialised for particular tasks, brain areas involved in meaning extraction tend to be involved in most kinds of conceptual processing. Individually regions have slight cognitive tunings and can be geared towards specific cognitions. Differences in person- knowledge and object categories emerge as a product of the coordinated interplay between multiple brain regions.