Search Results

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 explains how the examination of arithmetical concepts can lead to knowledge. It argues that in order for an examination of our concepts to supply us with knowledge of an independent reality, it must be that those concepts are appropriately sensitive to the nature of that reality, or what is called here grounded. The core idea here — and indeed the core idea of this book — is that grounded concepts are like trustworthy on-board maps of the independent world. The chapter suggests that it is through the normal functioning of our senses that our arithmetical concepts come to be grounded, and argues that if this is so, then the ultimate source of our arithmetical knowledge is empirical, though that knowledge is still a priori by many standard definitions.

This chapter consists in development of some of the key ideas of the previous chapter. It begins by showing how an arithmetical epistemology along the lines suggested in the previous chapter sits well with a structuralist conception of arithmetic and arithmetical concepts (although it is also compatible with other views). It cautions against a simplistic understanding of the envisaged grounding relationship between concepts and sensory input, arguing that the proposed account allows us to say that our arithmetical beliefs count as knowledge by the lights of Chapter 3. It shows that the account is consistent with realism as characterized in Chapter 1. The chapter spends some time discussing the crucial notion of unconceptualized sensory input, and also offers some comments on what is called here ‘ungrounded’ and ‘unfitting’ concepts.

This chapter begins with a brief discussion of tactile-motor-kinaesthetic perceiving. It argues that through use of the tactile-motor-kinaesthetic sensory input and imagery that we learn the boundaries of self and not-self, and the geography of our own bodies and the three-dimensionality of things and of spaces between them. It then introduces the ‘Feeling Once, Feeling Twice Phenomenon’ that is manifested when you place your hands on a surface and what you feel with your hands does not feel back; then, as you move your hands to come into contact with one another, what you feel does (even quite sensuously so) feel back. What feels back, and what is felt as continuous (for instance, ends of the hair) with that, forms the geography of your body and its limits against what is not your body, namely, what does not feel back.

The 21st century brought about a shift in the types of models used to explain the processing of sensory information by the cerebral cortex. Until now, these models have overwhelmingly emphasized the feedforward direction for the transfer of sensory information, and it is only recently that a more balanced view of the cortical network has emerged. This chapter summarizes this evolution and focuses on the question of the role of feedback connections in the processing of sensory information. It uses the visual system as a model since most of what is known concerning the processing of sensory inputs by the brain comes from studying that system.

The theme of this chapter is polysensory integration. Here, the chapter concentrates on the role of the superior colliculus (SC) of the cat in the integration of information from different senses. This chapter shows that there is a high degree of receptive field alignment in polysensory neurons in the SC, which is crucial for multisensory integration. The essence of multisensory integration in the SC is that changes in the responsiveness of individual neurons in the SC to multiple sensory inputs are more than the sum of the changes in responsiveness to the individual stimuli. Multisensory integration in the SC is critically dependent upon input from a restricted part of the cortex. Finally, the last part of this chapter shows how in cats there is a maturation of polysensory characteristics of SC neurons over the first months of life, with sometimes abrupt changes in individual neurons.

This chapter reviews the evidence for a general process of sensory filtering that attenuates self-generated tactile sensation. It shows that tactile attenuation specifically affects self-generated sensory input and leaves externally generated sensations in the same part of the body unchanged. The findings of the review reveal that the level of attenuation can be reduced by introducing a spatial separation between the active effector and the body part in which the touch is felt, which suggests that tactile attenuation is modulated by the degree to which the context of the action is consistent with self-generation.

The ability to change cortical connections and receptive field properties is a well-established characteristic of the cortex early in development. After a short period of plasticity extending from birth to about six months of age, the connections from thalamus to cortex become fixed. Current evidence, however, shows that receptive field characteristics and cortical topography, even in primary sensory cortex and even in the adult, are surprisingly dynamic. These properties are subject to several influences—the context within which a feature is presented, long-term changes in sensory input, and attention toward stimulus attributes. The dynamic nature of cortical function has profound implications for understanding the mechanisms underlying many aspects of visual perception: the unification of an object's component contours into a single percept, separation of a figure from its background, perceptual constancies, the storage of visual information by neuronal ensembles, and the role of top-down processes in perception.

This chapter pursues some subtle aspects of the mind’s fragility and its indirect relation to the world, which arise from the way in which prediction error minimization must proceed. In particular, the chapter looks at challenges to our seemingly robust background beliefs about the nature of our own body. This section appeals to out of body experiences, and to research on full body illusions. The idea here is that there is a surprisingly wide range of hypotheses that the mind can entertain, many of which seem relatively impervious to prior learning. Having established this very general outlook on us as fundamentally fragile prediction error minimizing machines, the chapter then explores some challenging notions about how we relate to the world in perception; these issues relate to the fundamental situation for the brain, namely that it must infer and act solely on the basis of its expected sensory input.

The brain is exposed to continuous sensory input from the environment (and the body). How does the brain encode such continuous sensory input and translate it into neural activity, e.g., stimulus-induced activity? Results from cellular recordings show that single neurons and a population of neurons represent the stimulus in a rather sparse way so that many stimuli are represented in one neuron’s (or one population of neurons’) activity. This amounts to a many-to-one relationship between stimuli and neurons entailing sparse coding. As such, sparse coding must be distinguished from other coding strategies like dense and local coding that propose a one-to-many and one-to-one relationship between stimuli and neurons. How is such sparse coding possible? The neurons’ (and population of neurons’) activity seems to encode the statistical frequency distribution of stimuli across their different discrete points in physical time and space; that is, their natural statistics. However, this is possible only when presupposing that differences between the stimuli’s different discrete points in physical space and time are encoded into neural activity. In other words, spatial and temporal differences (between the different discrete points in physical time and space) must be encoded into neural activity in order for sparse coding as a many-to-one relationship between stimuli and neurons to be possible.

This chapter introduces the general approach followed in the rest of the book—the starting point for explaining the human mind is the simple mental processes that we share with animals: the perceptual guidance of actions. This mental process is described with the help of the concept of “pragmatic representations”—perceptual representations that represent those features of objects that are relevant for the performance of an action. This approach is distinguished from the two most influential contemporary ways of thinking about the human mind: computationalism/propositionalism (according to which beliefs, desires or other propositional attitudes mediate between sensory input and motor output), and anti-representationalism/enactivism (according to which nothing mediates between sensory input and motor output—they form one intertwined dynamic process). According to the picture explored in this book, there are some special kind of representation, “pragmatic representations”, that (sometimes directly) mediate between sensory input and motor output.

An internal representation of the human body is crucial for a wide range of activities such as planning action, registering the location of sensory input and making judgments about whether one could fit into a particularly appealing item of clothing. Given the central role of body representations in these and many other behaviors it is not surprising that the issue was addressed by a number of investigators in the early part of the twentieth century (e.g., Pick, 1922; Head & Holmes, 1911–1912). In this chapter, Coslett first reviews evidence supporting the existence of multiple discrete but interacting representations of the human body. In the second section of the chapter, he elaborates on the “body schema”, a representation of particular relevance for sensory-motor processing as it mediates between perception and action. As will become clear, the accounts developed are heavily influenced by the work of neurologists and psychiatrists from the 19^th and early part of the 20^th centuries–the original cognitive neuroscientists.

Traditionally, human focal epilepsy has been thought to arise from an area of cortical damage, and the models of focal epilepsy used in research have been based on this concept. However, although the concept holds true for the adult brain, focal epilepsy is more common in children who have no evidence of such a lesion. This chapter gives a description and analysis of two types of focal epileptogenesis that are unique to the developing brain and for which there is no historical, clinical, or laboratory evidence of a structural brain lesion. In the first type, the focal epileptogenesis has its origin in a genetically determined cellular defect. In the second, the focal epileptic process is a result either of deprivation or chronic distortion of sensory input during a critical period of brain development.

There are some tough problems in comprehending the control of head movements. The head-neck system is multijointed and the posture and the movement of the head can be controlled by distinct pairs of muscles that may subserve the same functions or help to perform a particular task. There seems to be considerable redundancy. The behavioral degrees of freedom are few, yet simple movements such as rotating the head may result from the contraction of many muscles acting in a coordinated manner manifesting the necessity for some constraints. Another problem is that different tasks may need to be performed and the organization of the sensory inputs and the motor outputs must be appropriate for a particular task, such as controlling gaze or posture or both at the same time.

This chapter outlines selected recent developments in psychophysics that may be relevant to problems of sensory organization. It is assumed in this discussion that psychophysics is concerned primarily with “terminal” activities, in that specified sensory inputs are related to specified behavioral outputs, without direct observation of the complexity of intermediate neural events. Psychophysical data are thus necessary, but not sufficient, for a complete model of sensory organization. From this point of view, psychophysics is examined to suggest crude alternative models or broad strategies of action for approaches to the sensory organization of the human observer. All references to the nervous system in this chapter refer to a conceptual nervous system that serves between the terminal activities of stimulus and response.

This chapter explores the idea that scene representation is a spatio-centric representation that incorporates multiple sources of information: sensory input and several sources of top-down information. It discusses behavioral, neuroimaging, and neuropsychological research on boundary extension that supports a spatio-centric alternative to the traditional description of scene representation as a visual representation. Boundary extension provides a novel window onto the nature of scene representation because the remembered extended region has no corresponding sensory correlate. This alternative view bears a relation to theories about memory and future planning.

This chapter describes the dynamic causal modeling (DCM) equations, demonstrates how the ensuing model is inverted using Bayesian techniques, and reports the use of Bayesian priors to derive better magnetoencephalography/electoencephalography (EEG) models. It discusses the current DCM algorithms and some promising future developments, and explores the EEG data acquired under a mismatch negativity paradigm. The three plausible models defined under a given architecture and dynamics are examined. The chapter shows that evoked responses, due to bilateral sensory input (e.g., visually or auditory), could be analyzed using DCMs with symmetry priors.

This chapter argues against conventional accounts of implicit learning. It proposes that implicit learning designates an adaptive mode in which people's behaviour is sensitive to the structural features of a previously presented situation, without this adaptation being due to the internal exploitation of knowledge about these features. The chapter puts forward a Subjective Unit Account of implicit learning, according to which people process complex material by parsing it into small and disjunctive units. It further argues that implicit learning generates conscious perceptual units, and that these are directly responsible for the improvement in performance observed in typical implicit learning studies. The chapter suggests that, with training, conscious perceptual units become increasingly independent of the sensory input and, hence, form the explicit representations tapped by conventional tests of explicit knowledge.

The primary focus of this study is multi-sensory integration and its motor control and behavioral consequences. The series of experiments involve moving a spatial visual stimulus provided by a roughly hemispherical “rotating dome” filling the subject's field of view. It is discussed in this chapter how elementary reflexes that act to maintain an animal's head in the upright position depend upon numerous sensory inputs acting independently or in a group. This study demonstrates that the presence of lateral forces on the shoulders is enough to inhibit visually induced motion and the associated postural reactions of the neck and trunk.

This chapter argues that the experience of authorship strongly depends on the capacity to form predictions about the outcome of an action. It reviews cognitive psychology and neuroscience research indicating that subjective agency experience depends on the integrity of outcome predictions generated by internal forward models in the brain. Outcome predictions can be used for filtering sensory input through a continuous comparison with the actual sensory feedback. Current computational models explaining agency in terms of forward modeling and predictive coding are discussed. Moreover, scientific work has demonstrated that dysfunctional outcome prediction may contribute to aberrant agency experience in schizophrenia and obsessive-compulsive disorder. The chapter concludes with a discussion about the role and nature of internal predictions underlying functional and dysfunctional experience of agency—critically distinguishing the psychological and neural mechanisms operating at different phenomenological levels of agency.