Contemporary models of declarative memory state that when initially learned, all novel information is encoded by the hippocampal system before being consolidated or transformed to depend on neocortical structures subserving semantic memory. Based on observations with functional magnetic resonance imaging (fMRI), this thesis presents evidence that novel associations may be directly encoded by the semantic system in humans. While the hippocampus is often involved in information processing at the early stages of learning, the semantic system is seen to encode associative memory traces in the first instance (chapter 2). Furthermore, it is proposed that the hippocampus is not involved in learning when associative information is gradually accumulated across a series of ambiguous events. This is characteristic of cross-situational learning (xSL) which allows for the acquisition of word-object associations (i.e. nouns) during infancy. It is shown that xSL is not well accounted for by a prominent model of contextual learning - the temporal context model (chapter 3). Additionally, fMRI data suggest that neocortical structures rather than components of the hippocampal system are preferentially involved in xSL compared to traditional methods of training (chapter 4). Finally, it is suggested that rapid hippocampal learning mechanisms rely on specialised neuronal-microglial interactions. The administration of a microglial inhibitor (minocycline) was found to modulate hippocampal function and bias its use when other learning systems would have been more advantageous (chapter 5). Collectively, these findings suggest that the hippocampal system is specialised for rapidly encoding information that is explicitly provided, yet may not be recruited when associative information is collated across ambiguous events. At the same time, the neocortical semantic system may be able to learn new information at faster rates than previously thought. As such, it is hypothesised that amnestic patients may be able to acquire some forms of declarative material if presented in an appropriate manner.

Age-related decline in cognitive performance and brain structure can be offset by increased exercise. Little is known, however, about the cognitive and brain structural consequences of long-term high-effort endurance exercise. In a cross-sectional design, we recruited older adults who had been engaging in high-effort endurance exercise over at least twenty years, and compared their cognitive performance and brain structure with a non-sedentary control group similar in age, sex, education, IQ, depression levels, and other lifestyle factors. We hypothesized that long-term high-effort endurance exercise would protect against the age-related decline in memory, attention, and brain structure. Our findings, in contrast to previous studies, indicated that those participating in long-term high-effort endurance exercise, when compared without confounds to non-sedentary control volunteers, showed no differences on measures of speed of processing, executive function, incidental memory, episodic memory, working memory, or visual search. On measures of prospective memory, long-term exercisers performance suggested a self-imposed increase in effort, which did not impact on ability to complete the PM task. In complex attention tasks, they displayed a differential strategy to controls. Structurally, long-term exercisers only displayed higher diffuse axial diffusivity, an index of axonal integrity, than controls, but this did not correlate with any cognitive differences.

Cognitive process accounts of the advantages conferred by diagrams in problem solving and reasoning have typically attempted to explain an idealised user or a reasoning system that has equivalent to practised knowledge of the task with the target representation. The thesis investigates the question of how diagrams support users in the process of solving unpractised problems in the domain of probability. The research question is addressed by the design and analysis of an empirical study and cognitive model. The main experiment required participants (N=8) to solve a set of unpractised probability problems presented by combined text and diagram. Think-aloud and eye-movement protocols together with given solutions were used to infer the content and process of problem interpretation, solution interpretation and task execution strategies employed by participants. The data suggested that the diagram was used to facilitate problem solving in three different ways by: (a) supporting sub-problem identification, (b) supporting prior knowledge of diagrammatic sub-schemes used for interpreting a solution and (c) supporting the process of interpreting and testing the specific meaning of given problem instructions and self-generated solution instructions. These empirical data were used to develop cognitive models of canonical strategies of the three identified phenomena: • Sub-problem identification advantages are accounted for by proposing that the spatial semantics of diagrams coupled with competences of the visual-spatial processing system and opportunities for demonstrative interpretation strategies increase the probability of goal-relevant data being made available to central cognition for further processing. • Framing advantages are accounted for by proposing that represented diagrammatic sub-schemes (e.g. part-whole portions, icon-arrays, 2D containers etc.) facilitate access to existing prior knowledge used to frame, derive, and reason about information analogically within that scheme. • Advantages in instruction interpretation are related to the specificity of diagrams which support the opportunity to demonstratively test and evaluate the referential meaning of an instruction. The cognitive model also investigates and evaluates assumptions about the prior knowledge for solving unpractised probability problems; a representational scheme for addressing the co-ordination of sub-goals; a deictic problem representation to support online processing of environmental information, a meta-cognitive processing scheme to address self-argumentation and intention tracking and visual and spatial competences to address the requirements of diagrammatic reasoning. The implications of the cognitive model are discussed with regard to existing accounts of diagrammatic reasoning, probability problem solving (PPS), and unpractised problem solving.

For people with synaesthesia, a percept or concept (inducer) triggers another experience (concurrent) which is usually in a different modality. The concurrent is automatic, and in the case of certain types of synaesthesia also consistent, however the relationship between the inducer and concurrent is not fully understood and shall be investigated in this thesis from different perspectives. The first is using hypnosis to suggest synaesthesia-like phenomenological experiences to participants, and measuring behavioural responses to see whether they behave in a similar manner to developmental synaesthetes. Results from hypnotic; 1) grapheme-colour (GC) synaesthesia; 2) motion-sound synaesthesia; suggest that phenomenological experiences similar to developmental synaesthesia can be experienced by highly susceptible participants, but is not associated with the same behaviour as developmental synaesthetes. Developmental GC synaesthetes were tested to determine whether a grapheme presented preconsciously binds with the concurrent colour to the extent that it influences behaviour or evokes the phenomenology of colour. Two techniques were used, gaze-contingent substitution (GCS) and continuous flash suppression (CFS). Using GCS, it was shown that although digits can be primed preconsciously, they don't bind with their concurrent colour to influence behaviour. Nevertheless, many synaesthetes still experienced colours though they didn't necessarily match the primed digit. CFS experiments showed that the colour of a grapheme's concurrent, or whether the grapheme is presented in the correct or incorrect colour for that synaesthete, doesn't influence the time for conscious perception of a grapheme, even though colour words presented in the correct colour are perceived faster than those in the wrong colour. Phenomenological differences were compared to the behavioural measures using questionnaires modified using factor analysis (the R-RSPA and R-ISEQ). Overall, inducers must be seen consciously for them to bind with the concurrent, and experiencing the phenomenology of synaesthesia is not sufficient to behave like a synaesthete.

It has recently been found that during recognition memory tests participants' pupils dilate more when they view old items compared to novel items. This thesis sought to replicate this novel ‘‘Pupil Old/New Effect'' (PONE) and to determine its relationship to implicit and explicit mnemonic processes, the veracity of participants' responses, and the analogous Event-Related Potential (ERP) old/new effect. Across 9 experiments, pupil-size was measured with a video-based eye-tracker during a variety of recognition tasks, and, in the case of Experiment 8, with concurrent Electroencephalography (EEG). The main findings of this thesis are that: - the PONE occurs in a standard explicit test of recognition memory but not in “implicit” tests of either perceptual fluency or artificial grammar learning; - the PONE is present even when participants are asked to give false behavioural answers in a malingering task, or are asked not to respond at all; - the PONE is present when attention is divided both at learning and during recognition; - the PONE is accompanied by a posterior ERP old/new effect; - the PONE does not occur when participants are asked to read previously encountered words without making a recognition decision; - the PONE does not occur if participants preload an “old/new” response; - the PONE is not enhanced by repetition during learning. These findings are discussed in the context of current models of recognition memory and other psychophysiological indices of mnemonic processes. It is argued that together these findings suggest that the increase in pupil-size which occurs when participants encounter previously studied items is not under conscious control and may reflect primarily recollective processes associated with recognition memory.

This thesis aims to provide a compelling and distinctive response to the Problem of Consciousness. This is achieved by offering a bipartite analysis of the epistemic gap at the heart of that problem, and by building upon the hypothesis that the apparent problem is symptomatic of our limited conception of the physical. Chapter 1 introduces the problem. The key question is whether phenomenal consciousness is onticly dependent on the physical, or onticly independent of it. There are powerful arguments for the Primitivist view that consciousness is independent of the physical. These arguments rest on the apparent epistemic gap between the physical and the phenomenal. I propose that this apparent gap must be understood as a composite of two deeper conceptual gaps pertaining to the subjective character and qualitative character of consciousness respectively. The ‘–tivity gap' claims that physical states are objective, phenomenal states are subjective and that there is no entailment from the objective to the subjective. The ‘–trinsicality gap' claims that physical properties are extrinsic (structural), that phenomenal qualities are intrinsic (non-structural) and that there is no entailment from the extrinsic to the intrinsic. After refining the case for Primitivism, I consider the compelling reasons for rejecting Primitivism in favour of Physicalism. The challenge posed by the Problem of Consciousness is to resolve this antinomy between Primitivism and Physicalism. In Chapter 2 I consider standard responses to the problem. The failings of these positions lead me to introduce three criteria that an adequate response must satisfy. I reject the view that Primitivism can be salvaged, and hold that a satisfactory response to the problem must protect Physicalism. I reject standard ‘Type-A' responses according to which there is no epistemic gap between the physical and the phenomenal, and argue that a satisfactory response cannot deny the manifest reality of phenomenal consciousness. Finally, I reject ‘Type-B' responses according to which the epistemic gap does not entail ontic distinctness. I hold that if Physicalism is true, the entailment from the physical facts to the phenomenal facts must be knowable a priori for an epistemically ideal subject. Chapter 3 evaluates a non-standard Type-A response to the Problem of Consciousness which promises to satisfy all three criteria. According to Stoljar's Epistemic View (EV), consciousness only seems inexplicable in physical terms because we have a limited conception of the physical. I argue that EV should be supported iff two demanding challenges can be met: the Relevance Condition requires adequate reason to believe that unknown physical properties could address the –tivity gap and the –trinsicality gap. The Integration Condition requires adequate reason to believe that there is a specific blind-spot in our current conception of the physical that is plausibly occupied by properties that perform the requisite explanatory role. To satisfy these conditions, the advocate of EV must make positive claims about the content of our proposed ignorance. In Chapter 4 I argue that EV stands or falls with the plausibility of the Russellian Ignorance Hypothesis (RIH). According to RIH, we have no concepts of the intrinsic properties of physical entities, and those intrinsic properties are integral to the physical explanation of consciousness. I argue that we are indeed conceptually ignorant of intrinsic physical properties. I also argue that RIH meets the Integration Condition, and goes some way to satisfying the Relevance Condition. RIH plausibly undermines the –trinsicality gap by showing that some physical properties are intrinsic, though they are beyond our current conception. The apparent gap is then an illusion resulting from the fact that all known physical properties are extrinsic. RIH fails, however, to address the –tivity gap. I conclude that no version of EV can offer a full response to the Problem of Consciousness. In Chapter 5 I explore an entirely different kind of response to the Problem of Consciousness. Representationalism claims that consciousness is explicable in terms of intentional properties, and that intentional properties are explicable in terms of physical properties. I argue that standard Representationalist proposals are unable to account for the qualitative character of conscious states, and diagnose this failure in terms of the –trinsicality gap. However, the prospects for a Representationalist account of subjective character are more promising. Specifically, Kriegel's Self-Representationalism holds that a mental state is a phenomenal state in virtue of suitably representing itself. I argue that this proposal plausibly addresses the –tivity gap. RIH and Self-Representationalism each deal with one of the two apparent conceptual gaps between the physical and the phenomenal, but not the other. In Chapter 6 I develop a hybrid proposal that combines the best of both positions. The ‘Neo-Russellian Ignorance Hypothesis' (NRIH) claims that a mental state is a phenomenal state at all in virtue of suitably representing itself, and has its qualitative character in virtue of the intrinsic physical properties involved in its implementation. I expand this claim and defend it against a number of potential criticisms. I also explore the relationship between its two components, suggesting that they are each founded on a common epistemic insight. I argue that NRIH successfully addresses the –tivity and –trinsicality gaps and, moreover, that it provides a compelling account of why consciousness appears to be inexplicable in physical terms. I conclude that NRIH offers a powerful response to the Problem of Consciousness that successfully undermines the case for Primitivism. Furthermore, I conclude that NRIH has substantial advantages over competing attempted responses, and offers the best possible way of capitalising on the insights of EV and Representationalism.

This thesis confronts a fundamental shortcoming in cognitive science research: a failure to be explicit about the theory of concepts underlying cognitive science research and a resulting failure to justify that theory philosophically or otherwise. It demonstrates how most contemporary debates over theories of concepts divide over whether concepts are best understood as (mental) representations or as non-representational abilities. It concludes that there can be no single correct ontology, and that both perspectives are logically necessary. It details three critical distinctions that are frequently neglected: between concepts as we possess and employ them non-reflectively, and concepts as we reflect upon them; between the private (subjective) and public (inter-subjective) aspects of concepts; and between concepts as approached from a realist versus anti-realist perspective. Metaphysical starting points fundamentally shape conclusions. The main contribution of this thesis is a pragmatic, meticulously detailed, and distinctive account of concepts in terms of their essential nature, core properties, and context of application. This is done within the framework of Peter Gärdenfors' conceptual spaces theory of concepts, which is offered as a bridging account, best able to tie existing theories together into one framework. A set of extensions to conceptual spaces theory, called the unified conceptual space theory, are offered as a means of pushing Gärdenfors' theory in a more algorithmically amenable and empirically testable direction. The unified conceptual space theory describes how all of an agent's many different conceptual spaces, as described by Gärdenfors, are mapped together into one unified space of spaces, and how an analogous process happens at the societal level. The unified conceptual space theory is put to work offering a distinctive account of the co-emergence of concepts and experience out of a circularly causal process. Finally, an experimental application of the theory is presented, in the form of a simple computer program.

In the universe of problems humans face every day there is subset characterized by a salient dynamic component. The FireChief task (Omodei & Wearing 1995) is a fire-fighting computer simulation that can be characterized as the acquisition of interactive skills involving fast-paced actions cued by external information. This research describes the process followed to create a cognitive model of this complex dynamic task where full experimental control is not available. The cognitive model provides a detailed description of how cognition and perception interplay to produce the interactive skill of fighting the fire. Several artefacts were produced by this effort including a dynamic task fully compatible with ACT-R, a tool for analysing the data, and a cognitive model whose features enable the replication of several aspects of the empirical data. A key finding is that good performance is linked to an effective combination of strategic control with attention to changing task demands, reflecting time and care taken in informing and effecting action. The contributions of this work towards our understanding of complex problem solving are the methodological approach to the creation of the model, the design patterns embedded in the model (which are a reflection of the cognitive demands imposed by the nature of the task) and mainly an explanation of how skill, described in terms of strategy use, is acquired in complex scenarios. This study also provides a deeper understanding of the interactions observed in the Cañas et al. (2005) dataset, including a computational realisation of how cognitive inflexibility occurs.

This research explored strategies for learning programming via worked-examples that promote schema acquisition and transfer. However, learning style is a factor in how much learners are willing to expend serious effort on understanding worked-examples, with active learners tending to be more impatient of them than reflective learners. It was hypothesised that these two learning styles might also interact with learners' cognitive load. The research proposed a worked-example format, called a Paired-method strategy that combines a Structure-emphasising strategy with a Completion strategy. An experiment was conducted to compare the effects of the three worked-examples strategies on cognitive load measures and on learning performance. The experiment also examined the degree to which individual learning style influenced the learning process and performance. Overall, the results of the experiment were inconsistent. In comparing the effects of the three strategies, there were significant differences in reported difficulty and effort during the learning phase, with difficulty but not effort in favour of the Completion strategy. However no significant differences were detected in reported mental effort during the post-tests in the transfer phase. This was also the case for the performance on the post-tests. Concerning efficiency measures, the results revealed significant differences between the three strategy groups in terms of the learning process and task involvement, with the learning process in favour of the Completion strategy. Unexpectedly, no significant differences were observed in learning outcome efficiencies. Despite this, there was a trend in the data that suggested a partial reversal effect for the Completion strategy. Moreover, the results partially replicated earlier findings on the explanation effect. In comparing the effects of the two learning styles, there were no significant differences between active and reflective learners in the three strategy groups on cognitive load measures and on learning performance (nor between reflective learners in the Paired-method strategy and the other strategies). Finally, concerning efficiency measures, there was a significant difference between active learners in the three strategy groups on task involvement. Despite all these, effect sizes ranging from a medium to large suggested that learning styles might have interacted with learners' cognitive load.

In this thesis I develop a novel approach to conscious perception, which I label “radical sensorimotor enactivism” (RSE). In chapter one, I explain how the development of RSE is guided by the tenets of activity and knowledge-how. In chapter two, I outline and explain RSE. Throughout the thesis, I will pit RSE against cognitivist accounts of conscious perception and argue that RSE is to be preferred. In chapters three and four, I highlight two problems facing cognitivist accounts of conscious perception which RSE avoids. I argue that cognitivist accounts of conscious perception face the ‘hard problem of perceptual consciousness', whilst RSE can provide a phenomenologically plausible deflation of this problem. I next explain why cognitivist accounts are incapable of providing a satisfactory explanation of split-brain syndrome. Then, I argue that RSE can provide a parsimonious explanation of this syndrome. Theories predicated on activity and knowledge-how are often rejected for being incapable of accounting for the brain's role in conscious perception. In chapter five, I argue that RSE can account for the brain's role by adopting a non-representational version of predictive processing (PP). Moreover, I argue that the resultant account improves upon cognitivist alternatives. Then, in chapter six, I argue that even representational explanations of PP can be subsumed within RSE by accepting fictionalism about their representational posits. Consequently, I conclude that RSE cannot be objected to for failing to account for the brain's role in conscious perception. Finally, in chapter seven, I discuss ‘non-veridical' experiences. Accounts like RSE are often rejected because it is thought they are incapable of explaining the existence of these phenomena. I explain how the existence of such phenomena is wholly compatible with the truth of RSE. Thus, I conclude that RSE should not be rejected solely on the basis that non-veridical experiences exist.