A neurologically plausible implementation of a meaning-based knowledge

The aim of this thesis is to develop an artificial neutral network model that specifically addresses the problems of achieving both human neurological and psychological plausibility. Many models of the brain have been developed that address either just the psychological or the neurological properties of the brain, but few address both. By combining together a model of the psychological and neurological aspects of the brain, an hybrid model may be produced that captures both aspects of the human brain, and may also provide further evidence supporting each component model. The problem is tackled by looking to the psychologist for the cognitive properties of the brain and a good model capturing these properties, and looking to the neurologist for information on neural physiology, anatomy and possible organizational principles. This thesis tries to combine the prepositional network model of meaning-based knowledge representation with Gerald Edelman's Selective Neuronal Group theory of neuronal organization using a plausible neuronal model of the cerebral cortex. This thesis presents the background material from which these models are drawn and which is used to test and validate the resulting hybrid model. A mathematical description of the model is presented along with the reasons for choosing the basic models. A computer simulator of the model is then developed to test the model and the ideas. These tests are designed to approximate simple learning paradigms (i.e. simple conditioning, paired-associate and serial learning) and to see if the model exhibits any psychological properties in common with the human brain (e.g. associative priming and interference effects). These results are discussed before a critical look at the model is made, conclusions of the work are drawn and future work suggested