Search Results

This chapter examines the components of the impaired neural networks that might underlie the presentation of autistic symptoms. Topics covered include the neural network models of autism, central nervous system ontology, candidate regions in autism, trouble at the cellular level, trouble with neurotransmitters, trouble with circuitry in autism, and trouble with myelination of neural networks. It is shown that the autistic syndromes may be considered as one extensive set of impaired final common circuits presenting with dysfunctional information processing of behavior and cognition in very young children. Deficits in pragmatics, linguistic abilities, mindreading, executive functions, episodic memory, self-awareness, central coherence, and affective processing have been documented. These deficits are caused by many disease entities whose shared symptoms likely occur owing to malfunction of certain distributed neural networks.

This chapter proposes a normative neural model of visual processing under the assumption that visual spiking neurons optimally detect the presence of independent objects in dynamic visual scenes. Rather than by their receptive fields, these neurons are better described by their causal field, that is, by what they predict for the sensory input. As a result, visual layers can discover the set of objects that probably explains the input from the previous layer, solving two problems at once: integrating the sensory input optimally to detect visual objects efficiently, that is, performing optimal combination of visual cues, and resolving ambiguities between similar objects, that is, performing explaining away. This model provides a fresh view of sensory processing at the level of spiking neurons and small circuits as a form of redundancy reduction in time and space.

This chapter presents models and mechanisms of masking according to five distinguishing characteristics: (i) models based on spatiotemporal response sequences; (ii) models adopting some version of an overtake hypothesis; (iii) models based on two separate neural processes or channel activations; (iv) models relying on stimulus or object substitution; and (v) models based on emergent properties of distributed neural networks. The models discussed in detail include spatiotemporal sequence models (Kahneman's impossible motion model, Matin's three-neuron model, and Burr's spatiotemporal receptive field model), two-process models (Ganz's interactive trace decay and random encoding time model, Reeves' temporal integration and segregation model, and Navon and Purcell's integration and interruption model), neural network models (Bridgeman's Hartline–Ratliff inhibitory network, Weisstein's Rashevsky–Landahl two-factor neural network, perceptual retouch model, and Boundary Contour System (BCS) model), and object substitution models.

This chapter concerns the computational modeling of one of Les Cohen's most important discoveries in infant information processing—the developmental shift from learning about visual stimulus features to learning about correlations between these features. It describes the theoretical origins of this work, and reviews the relevant psychology experiments and the several attempts to simulate it with artificial neural networks, before presenting a new neural network model. The modeling suggests a somewhat different explanation than originally proposed, based on depth of learning rather than qualitative shifts in learning strategies. Computational models of this shift may be equally relevant to several other documented developmental shifts from learning about stimulus elements to learning about relations between those elements.

This chapter focuses on a heuristic neural systems model of motivated behaviour. This model provides hypotheses for mechanisms underlying changes in behaviour across development and psychopathology. The fractal triadic model (FTM) posits that goal-directed behaviour results from the interaction among three nodes of behavioural control. These three functional nodes are centred on the amygdala, striatum, and medial prefrontal cortex, which contribute to avoidance, approach, and modulation, respectively. They feed two distinct neural circuits: one that is modulated primarily by appetitive stimuli and serves approach behaviour, and one that is modulated primarily by aversive stimuli and serves avoidance behaviour. The behavioural output results from the integration of the information that is processed by these two neural circuits and is submitted to the control of the supervisory node. Such organization of three functional nodes subserving two neural circuits relies on the well-described structural and functional heterogeneity of these nodes. In addition, asynchrony in the maturational trajectories not only among the nodes, but also among the subunits of these nodes, is the central principle that underlies the typical behavioural changes seen in adolescence. Functional neuroimaging research is beginning to examine ontogenic changes in neural responses to reward-related processes that can further inform this heuristic model. The chapter addresses the major points mentioned above and ends with selected questions proposed as priority for future research.

This chapter gives a full account of the theoretical development and experimental investigations of the hypothesis that the primary visual cortex (V1) creates a bottom-up saliency map to guide visual attention exogenously. The chapter details the background motivations, theoretical formulation, and experimental tests of the hypothesis, as well as a neural circuit model of the primary visual cortex for the underlying neural mechanisms. The hypothesis links two bodies of data: one is of physiological data on intracortical interactions in V1 and the consequent contextual influences in V1 neural responses; the other is of behavioral data on attention capture, visual search, and visual segmentation. In light of the saliency map in V1, the chapter additionally discusses the roles of the extrastriate visual cortices, contrasts the roles of the central and peripheral visual fields, and reflects on the dissociation between attention capture and perceptual awareness.

This book concentrates on data-driven modeling, i.e., the use of fairly standardized modeling methods to replicate the behavior of neural systems at different levels of detail. The chapters are structured in intuitive order and try to cover all aspects of neural modeling, from molecules to networks. Each addresses the equations needed to simulate these models, sources of data for the model parameters, approaches to validate the models, and a short review of relevant models. This chapter provides an overview of those that follow.

This chapter examines the structural characteristics of a cognit, that is, the morphological feature of the cognitive networks of the cerebral cortex. Knowledge about oneself and the environment is the subject matter of cognitive science and the matter of all cognitive functions: perception, attention, memory, intelligence, and language. For more than a century, neurologists and experimental psychologists have endeavored to assign one or another aspect of cognition, beginning historically with language, to one or another sector of the cerebral cortex. Largely because it derives from neural inferences, connectionism has become the most plausible model of the organization of knowledge in the cerebral cortex. All connectionist models and neural network models assume the distribution of knowledge in assemblies of units, neurons, or nodes that constitute and represent the component elements of knowledge. This chapter also looks at different categories of knowledge, cortical modularity, cortical hierarchy of perceptual networks and executive networks, and hierarchical representation in association cortex.

Many accounts linking music and language to the brain represent the brain as a network of boxes, each of which has an active role in providing some resource, but once we move away from the auditory periphery there are very few models that offer finer-grain explanations of the underlying circuitry that supports these resources and their interaction. This chapter thus offers a bridge to future research by presenting a tutorial on a number of models that link brain regions to the underlying networks of neurons in the brain, paying special attention to processes which support the organization of events in time, though emphasizing more the timing or ordering of events than the organization of sequential order within a hierarchical framework. Our tour of models of the individual brain is complemented by a brief discussion of the role of brains in social interactions. The integration of cerebral activity is charted with that in other brain regions, such as cerebellum, hippocampus, and basal ganglia. The implications for future studies linking music and language to the brain are discussed which offer increased understanding of the detailed circuitry that supports these linkages. Particular emphasis is given to the fact that the brain is a learning machine continually reshaped by experience. Published in the Strungmann Forum Reports Series.

Central to the present theory is the making of connections, guided by emotion. Chapter 7 examined making connections in a close-to-experience sense, and discussed connections in the sense of similarity, condensation, and metaphor. At the level of the brain, making connections involves connections between neurons and assemblies of neurons in the brain—especially in the cerebral cortex. This chapter discusses the same event—the dream—from a different standpoint. While the last chapter took a “top-down” approach, this chapter takes a “bottom-up” approach, examining the underlying biology, and also an “in between” approach in discussing neural net modeling.

Tinnitus research is making tremendous progress in both the understanding of mechanisms and in developing tinnitus management. Yet, we still do not know why only 2/3rd of people with hearing loss develop tinnitus, or why tinnitus occurs in people without hearing loss. We also do not know the reason why both loudness and annoyance of tinnitus are enhanced by stress, whereas the two are typically only weakly correlated. Most importantly we do not know how findings in animal models and those in humans can be integrated. An attractive common ground could be to study cortical neural networks in animals as well as humans using either drug treatment, or local electrical or magnetic stimulation. The ultimate cure for tinnitus might be to restore the hearing loss by regeneration of cochlear hair cells, but this is still far away and even if successful may create its own problems.

This chapter starts with a brief review of what has become a generic framework for modelling the hippocampal and neocortical roles in long-term memory for personal experience, following from the work of David Marr (1971). It considers some of the ways in which this model has been developed over the years in terms of general theoretical issues concerning memory, such as time-courses, capacities, representations, and interference. It attempts to relate these issues to questions regarding the nature of episodic memory as compared with other forms of memory. It discusses a neural-level model of the medial temporal and parietal processes involved in the retrieval of the spatial context of an event.

This chapter aims to contribute to the ongoing debate about the neural correlates of consciousness from the viewpoint of a particular experimental approach: the study of distributed neuronal processing and of dynamic interactions which implement specific bindings in neural network architectures. The now classic notion of binding and the search for potential binding mechanisms has received increasing attention during the past decade. Having been introduced first in psychological discourse, the issue of binding has now advanced into the focus of research in other disciplines within cognitive science in areas such as neural network modeling, philosophy of mind, and cognitive neuroscience.

In the final chapter of Volume 1 of Neuroconstructivism, the most exciting implications, both theoretical and practical, of the neuroconstructivist approach were laid out. Among these were several lessons for how hands on research could be carried out within a neuroconstructivist framework. One such lesson was to incorporate more computational modelling within research programs exploring the origins of cognition. Meanwhile, this present volume has focused on examples of how computational and neural network modelling can be used as tools to explore factors interact to shape representations. It will become apparent that these are still early days for the neuroconstructivist approach to brain and cognitive development.

This chapter reviews recent tests of a neural net model of imprinting. The timing of imprinting, the features that most readily trigger learning, and the motor systems that are linked to representations stored as a result of learning are all specific to the functional context of forming a social attachment to one or both parents. Tune plays a different role in classical or instrumental conditioning from that which it does in perceptual learning. The order in which different events are experienced may matter a lot when one event causes the other. However, the order does not matter at all when the experiences are different views of the same object. Some behavioural and physiological evidence from studies of imprinting in chicks suggests that these two broad functions are served by different subprocesses but that the subprocesses are, nevertheless, in touch with each other.

This chapter focuses on the contribution and goals of functional neuroimaging. It presents a detailed discussion on the goals of functional neuroimaging, including neuroanatomical localization of cognitive processes, testing current theories of cognition, and neural models. The chapter reviews the work of functional neuroimaging in cognitive science and looks into its progress, and, finally, addresses questions related to functional neuroimaging and cognitive neuroimaging studies.

This chapter brings up to date the summary of results obtained in studies of localization of sound in space and of discriminations of changes in frequency, pattern, and duration of tones. A comparison is presented of the findings for the cat and the monkey, and related clinical studies on man are noted. Finally, an attempt is made to construct a simple, neural model that will explain the differences between discriminations that can and cannot be made after ablation of auditory areas of the cerebral cortex. Auditory discriminations that can be made after bilateral ablations of the medial geniculate projection areas of the cortex differ from those that cannot be made, in that for the former the positive stimulus causes new neural units in the auditory afferent system to be excited. In discriminations that cannot be made after the cortical ablations, the same neural units are excited by both negative and positive stimuli.

This chapter presents a self-organizing map (SOM) neural network model of tonality based on experimentally quantified tonal hierarchies. A toroidal representation of key distances is recovered in which keys are located near their neighbours on the circle of fifths, and both parallel and relative major/minor key pairs are proximal. The map is used to represent dynamic changes in the sense of keys as cues to keys becoming more or less clear and modulations occurring. Two models, one using tone distributions and the other using tone transitions, are proposed for key-finding. The tone transition model takes both pitch and temporal distance between tones into account. Both models produce results highly comparable to those of musically trained listeners, who performed a probe tone task for ten nine-chord sequences. A distributed mapping of tonality is used to visualize activation patterns that change over time. The location and spread of this activation pattern is similar for experimental results and the key-finding model. In general, experimental studies suggest that the sense of tonality undergoes dynamic and subtle changes when a listener hears music.

This chapter reviews the efforts of theorists in developing and testing formal (i.e. not purely verbal) models of cognitive control during task switching. First, it provides an overview of the architecture of extant models of task switching. To ease exposition, the models are grouped into three classes: mathematical models, computational models, and neural network models. The second section covers key empirical/theoretical concepts in the task switching literature, discussing how the models explain them. In particular, it focuses on each model's explanation of task sets and the switch cost. The third section highlights areas that the models do not cover or that require more elaboration in future modeling work. The chapter concludes that, despite the impressive progress made on the development of theories of task switching, there are important and rather neglected aspects that need to be addressed in future models. One important direction is the formalization of inhibitory processes during task switching.