Pathfound: Revealing the neural basis of semantic memory and its breakdown in semantic dementia and stroke aphasia

We have lots of knowledge about the world. We know about familiar and famous people. We know what objects are for and how they are used. We know the meanings of words. Collectively, this is called semantic memory. Certain types of brain damage in patients with stoke and dementia can specifically affect semantic memory. These patients? impairments impact significantly on every aspect of their lives. They are often unable to work, have problems with everyday tasks and chores, and cannot maintain their interests and hobbies. It is critically important, therefore, to improve our understanding of how the brain encodes all of this knowledge and how this breaks down after brain damage.
The research proposed here will use five different methods to investigate this topic. We will use simple tests of comprehension and language to reveal the nature and qualities of the patients? problems. We will also take brain images to look at which brain parts are used when we comprehend and also which of the remaining brain parts the patients can still use. We will adopt a new technique which uses a magnetic coil to temporarily suppress activity in a small part of the brain for a short period of time. We can use this method to test which regions in the normal brain contribute to semantic memory. In addition, we will build small mathematical models which, like brains, are built up from a network of simple units (like brain cells). The construction of these models can be made to reflect the organisation of real brains. At the same time, the models can be trained to perform the various tasks we use with the patients. By damaging the models, we can then see if they mimic the patients? impairments. In effect, the models help us to bridge the gap between brain (the network of units) and mind (the behaviour of the model). Finally, we will use a new form of brain imaging that shows us how the different parts of the brain are connected together. By putting all of this information together, we will be able to work out how the brain supports our knowledge of the world and why it breaks down in patients.

Semantic memory represents our knowledge of the meanings of words, objects, people, etc. as well as our general knowledge about the world. This knowledge is central to everything we do (including speaking, reading, recognising and using objects). The two linked, primary goals of our research programme follow from this centrality of semantic memory; they are to improve our understanding of (a) the nature of semantic impairment in various neurological disorders (i.e., clinical research) and (b) the neural basis of semantic memory (basic science). When semantic memory breaks down, patients are left with significant deficits that affect their employment, acts of daily living, social lives and private pursuits. The current literatures - based on neuropsychological studies of patients, functional neuroimaging (with PET or fMRI) and computational modelling - provide inconsistent answers to these two research questions. We will, therefore, use a deliberate conjunction of five methods to resolve these inconsistencies, thereby securing a foundation on which to interpret patients? deficits. They will also give critical insights about the patients? clinical management and better interventions by speech therapists.Our considerable track record on the topic of semantic memory is already based on a combination of detailed neuropsychological studies of patients, functional neuroimaging of normal and dysfunctional semantic processes, and computational modelling. These models provide a formal method by which neuroimaging and behavioural data can be combined: the architecture of the models (combinations of simple mathematical processing units and connections) can be made to reflect normal or impaired neuroanatomy in a way that simulates the behavioural data from healthy participants and patients. To this methodological mixture we will add: (1) improvements to standard fMRI methods to allow functional neuroimaging of the anterior temporal lobes; (2) recently developed methods for MR multi-fibre tractography (both (1) and (2) from our MRC Pathfinder grant); and (3) transcranial magnetic stimulation studies designed to extend and mimic patient performance in normal participants. We will also extend our europsychological studies to encompass additional patient populations. The five methodologies will be used together in order to answer four key questions: (1) What is the organisation/structure of representations within the anterior temporal lobe (the damage associated with semantic dementia)?; (2) What is the role of frontoparietal areas in semantic cognition/control?; (3) What are the individual contributions of temporoparietal and prefrontal regions (the damage producing comprehension impairments in CVA)?; (4) How does the full neural network give rise to semantic cognition?