Three Senorimotor Accounts of Perception

Tom Roberts

Abstract

Introduction
The central claim of a sensorimotor theory of perception is that we perceive certain properties of objects in virtue of understanding the ways in which the appearances of those objects change with our movements. Perception, on a model of this sort, is an active and exploratory process rather than a passive reception of sensory stimuli - by moving around our environment and keeping track of the changes in the way that it appears to us, we are capable of perceiving its constant, invariant features.

Appearances are objective, albeit relational, properties of objects. From any given spatial location, there is a determinate way that an object appears, a side or facet that it presents to an observer at that location. The three versions of the sensorimotor theory that I will discuss differ over their interpretation of the access that perceivers have to these appearances, and therefore over what it is that is tracked or understood to change with movement. As a result, they offer contrasting accounts of which properties are perceived as a consequence of sensorimotor knowledge.

Version 1
The first account, which emerges from the work of Alva Noe and J. Kevin O'Regan, construes our access to appearances as a subpersonal matter. In actively encountering the environment, our perceptual systems undergo changing patterns of stimuli (for example, shifting images on the retina). We perceive real properties of shape and colour (etc) in virtue of understanding the ways in which these patterns of stimulation change as we move our bodies. This understanding could, it is claimed, be embodied in a recurrent neural network - the task is a form of pattern recognition. This version is offered as a solution to the Explanatory Gap - the phenomenal qualities of experience are said to be explained in terms of subpersonal neural activity.

Problem: It is unclear how such a mechanism could provide the agent with contentful, qualitative experience. That is, why tracking patterns of stimuli should feel like anything at all.

Version 2
Access to appearances is a personal-level, perceptual relation. Experiences have two levels of content: the first specifies the apparent properties of objects, which are dependent upon the location of the observer and change with his movements; the second, which is the result of understanding these changes, specifies the real, invariant properties of objects. For example, a circular object presents a series of elliptical apparent shapes to a moving observer. These are perceived by the agent, and thus form the first level of content of his visual experience. In understanding the ways in which the ellipses change with his movement, the agent sees the real circular shape of the object. This forms the second level of perceptual content; the object is seen as both elliptical and circular. The proposal builds apparent colours and apparent sizes into the first level of visual content in addition.

Problem: This account is phenomenologically inaccurate. It is not the case that circular objects look elliptical - or even elliptical and circular - they simply look circular. It is not the case that nearby objects look larger than those which are further away, they look closer. If this is true, then the putative second level of real shape and size content does not emerge from sensorimotor mastery.

Version 3
Access to appearances is again a personal-level, perceptual relation and experiences have two levels of content. However, the apparent properties proposed by version 2 are not present. The first level of content specifies, instead, those parts of the object that are immediately perceivable from the agent's current location. For example, the facing side of a tomato. The agent's movements bring new parts of the object into view, and it is in understanding these changes that the agent comes to see the 3-dimensional shape of the object, the second level of perceptual content. We see only this side of the tomato, but we see the tomato as being solid and spherical. Likewise objects that are partially occluded - we see only those parts that are immediately accessible, but we also see them as extending behind the surfaces by which they are hidden.

Benefits: Phenomenologically appealing; room for explanation of illusion / hallucination; compatible with m any philosophies of perception. This is the version that I endorse; while not an attempt to close the Explanatory Gap, this account provides an explanation of our perception of objects as extended and 3-dimensional in spite of our immediate, 'sensational' access only to their facing sides.

Brain Correlates of Conscious Perceptions

Marie L. Smith, Philippe G. Schyns andFrederic Gosselin

Abstract

A fundamental problem in vision is to understand which visual information correlates with the perception of a stimulus and how the brain extracts and aggregates this information to arrive at a conscious perception.  We examined this problem in a case study involving the perceptual reversal of an ambiguous image (here, Dali's painting the Slave Market with the Disappearing Bust of Voltaire ).  We applied Bubbles in 3D space (2D image space x 5 spatial scales) to five observers disambiguated the image.  We determined the visual information correlated with each perception from the observers' responses ("nuns" vs. "Voltaire" vs. "don't know").  Simultaneously, we recorded the observers' EEG at 512 Hz with a 64 electro-cap and determined the visual information correlated with modulations of amplitude in the theta, alpha, beta and gamma bandwidths of oscillatory EEG activity.

Brain correlates of conscious perceptions are likely to integrate the scale information that determines the "the nuns" vs. "Voltaire" behavioral responses.  To find the time points of highest scale integration, we first derived the sensitivity of oscillatory brain activity to the scale information associated with behavioral responses, independently for each EEG band.  We computed these sensitivity curves from –200 to +800 ms around stimulus onset, in 2 ms independent time steps. The sensitivity curves (one for each spatial scale considered) effectively project EEG activity into the space of the scale information associated with perception. To find the time points of highest information integration, we computed phase-locking factors between all possible pairs of sensitivity curves and averaged them.

            With this new approach, we were able to isolate the time points at which brain signals are mostly correlated with an integration of the spatial scale information underlying the perception of each observer.  This suggests a new approach to find the brain correlates of conscious subjective perceptions.

Neural Correlates of Consciousness and the Unity of Experience

Julian Kilverstein

Abstract

Consider the familiar phenomenological point that what we experience at any given moment occurs as a part of a unified field.  I am currently undergoing an experience of my computer screen as I type these words, but I seem to experience more than just my computer screen at this moment.  The papers and books, the other tables and chairs in my office also form a part of what I am currently experiencing as does the taste of the coffee I have just drunk; the feeling of the chair supporting my back; the pressure the table makes on my arms as I type these words etc.  All of these diverse phenomena combine to form the experience I am currently enjoying.  Should we think of the experience I am currently enjoying as composed of various parts which somehow coalesce to give me a unified experience?  Should we instead think of conscious experience as a unified field that cannot be decomposed into parts, but which is constantly being modified by sensory input from the external world?  Which is the correct understanding of the unity of experience?

I propose to examine this question by considering the different accounts of the neural correlates of conscious experience which each entails.  Searle (2000) has called the first conception of the unity of experience, the building-block conception and the second conception of experience, the unified field conception.  The building block conception entails the kind of definition of neural correlates of consciousness (henceforth NCC's) we find in Chalmers (2000). Chalmers suggests that NCC's should be thought of as "minimal neural representational systems" whose contents map one-to-one onto the contents of conscious experience.  I will argue that there is good reason not to expect a perfect match in content of this kind between neural representational systems and the contents of conscious experience.  But the building block conception of experience entails the account of NCC's we find in Chalmers (2000), or so I shall argue.  So whatever reasons we find for rejecting Chalmers' conception of NCC's might reasonably be thought to likewise cast doubt on the building block conception of conscious experience. 

One reason for scepticism about Chalmers' characterisation of NCC's lies in the phenomenological point that our experience is of a figure-on-a-ground located in an egocentric space.  This egocentric space is defined by the possibilities for movement which an animal's surrounding environment affords.  To experience the possibilities for movement that the local environment affords, an animal's experience must represent the locations of things in relation to itself.  It is far from clear how a neural representational system could have this kind of content.  I agree with No‘ and Thompson (2004) that it is difficult to see how 'anything other than a whole perceiver as an intentional agent could be the bearer of this sort of content.' (op cit p.91) 

I will finish up by arguing that the unified field conception of experience can readily account for this kind of representational content.  I will suggest that consciousness on this model is best understood as non-conceptual self-consciousness of the kind Bermudez (1998) describes.  To the extent that the unified field conception of experience can account for something that the building block conception cannot this provides indirect support for the former view over the latter.

 

Bibliography

Bermœdez, J. 1998 The Paradox of Self-Consciousness (Cambridge, MA: MIT Press)

Chalmers, D. 2000: 'What is a Neural Correlate of Consciousness?' in T. Metzinger (Ed.) Neural Correlates of Consciousness: Empirical and Conceptual Questions (Cambridge, MA: MIT Press)

No‘, A, and Thompson, E. 2004: 'Sorting out The Neural Basis of Consciousness' in Journal of Consciousness Studies 11: 87-98

Searle, J. 2000: 'Consciousness' in Annual Review of Neuroscience 23: 557-78

Mental Imagery and Synaesthesia

Mary Jane Spiller & Ashok S Jansari

Abstract

Synaesthesia, the condition in which someone has the conscious experience of dual perceptions, has received an increase in interest over the past decade.  Research has found that some individuals have a consistent synaesthetic experience in which seeing a letter or number elicits the perceptual experience of colour (Mattingley et al, 2001), and that there is differing neural activation between synaesthetes and non-synaesthetes when a number or letter is viewed (e.g. Paulesu et al, 1995).  Research has also found that for some synaesthetes the inducer does not need to be externally presented in order for the synaesthesia to occur (Dixon et al, 2000; Jansari et al, 2005).  After an overview of the recent research into synaesthesia, the current paper outlines an empirical study that further explores this finding.  The aim of this study was to find objective evidence that an internal visual mental image of an inducer can elicit a synaesthetic experience, as many subjective reports have suggested.  As visual perception and visual mental imagery have been found to use similar neural mechanisms (Kosslyn, 2005), a synaesthetic experience from a mental image might not be surprising.  The results are discussed in relation to previous synaesthesia literature and previous mental imagery literature, and to what these findings can tell us about the interface between mental imagery and synaesthesia.

The Geometry of Visual Experience

- Reconsidering Luneburg's Theory of Binocular Vision

Phillip Meadows

Abstract

In this paper I consider Luneburg's empirical theory that the geometry of binocular vision is hyperbolic. I first argue that the theory is intended to provide a description of visual sensations, and so is a competitor to recent claims by Gideon Yafe, James Van Cleve and Gordon Belot that the geometry is spherical. Next I criticise two philosophical objections to Luneburg's theory: one proposed by Richard Angell and one by Robert French. The former concerns the theory's reliance upon the judgments of the experimental subjects and upon physical measurements. The latter offers a competing explanation of Luneburg's results in terms of the size constancy phenomenon. I argue that neither of these objections is satisfactory and I offer alternative reasons for believing Luneburg's arguments to be inconclusive.

Attention Effect of High and Low Valence Visual Stimuli: An FMRI Analysis

Lynda Shaw, Michael Wright and Justin O'Brien

Abstract

Background

Implicit processing of positive and negative emotional visual stimuli was investigated using a foveal/peripheral paradigm.  The visual system has limited processing capacity, and multiple objects in the visual field compete for neural representation.  This competition can be biased by selective attention.   An exception is the automatic processing of certain emotional stimuli, which by pass this mechanism.  The purpose of this experiment was to measure the interaction of foveal (voluntarily attended) and peripheral (not voluntarily attended) stimuli as a function of the emotional valence of the stimuli.  Using functional magnetic resonance imaging (fMRI), we tested the effects of pictures of animals, faces, scenes and inanimate objects calibrated for both high valence (HV) (negative emotion), and low valence (LV) (positive emotion) on activations in the fusiform gyrus, parahippocampus, amygdala, insula and anterior cingulate.

Method

 

Sixteen participants viewed stimuli passively in the fMRI scanner.  The stimuli were 24 HV and LV pictures from four categories (animals, faces, scenes and inanimate objects) and 24 neutral pictures (IAPS, Center for the Study of Emotion and Attention 2001).  Using a block design, stimuli were presented in three different conditions: (1) "large-field", whereby single pictures occupy a rectangle 16x10 deg; (2) "small-field" conditions showing single pictures (3.75x2.5 deg); (3) "combined" conditions consisting of small-field pictures superimposed centrally on large-field pictures of opposite valence and equivalent category.

Results and Conclusions

The important contrasts for testing the experimental hypotheses are those that show the activation produced by condition (3) versus condition (2) as a control condition. Thus by matching the valence and category of the small-field stimulus in (3) and (2), we show by subtraction the effect of the unattended surround stimulus. The fusiform gyrus, parahippocampus and insula were bilaterally activated by an unattended HV or LV surround, and some of these effects were category-specific.  However, the fusiform gyrus was not shown to be a face specific area, nor did the parahippocampus only react to scenes alone, thus supporting the theories that these areas are not category selective regions. The RH was dominant in the insula for both unattended LV and HV stimuli, providing evidence this area is not specific to unattended negative emotions.  The left amygdala was activated when viewing unattended LV stimuli across all four categories and the exact opposite occurred when presented with unattended HV stimuli.  This is a significant indication that the left and right amygdala subserve different functions in emotional processing.  Bilateral activations in the anterior cingulate were evident for all four categories for unattended HV, but the left anterior cingulate was only activated when processing unattended LV scenes. This is consistent with the reported involvement of anterior cingulate in tasks requiring divided attention or response competition, but suggests that attention is fully captured by the central stimulus in the case where the centre stimulus is HV and the surround is LV.  Overall, the right hemisphere (RH) was dominant for unattended LV stimuli, and bilateral activations were found for all four categories for unattended HV stimuli. This result is a surprise as evidence to date suggests that unattended LV stimuli are more likely to be processed than unattended HV stimuli.  The results provide evidence that all five target areas can discriminate stimuli based on valence and categories for the processing of unattended emotional stimuli.   

Objects, Properties, and Two Kinds of Binding

Jose Luis Bermudez

Abstract

This paper discusses some of the implicit philosophical background behind current discussion of the psychology of object perception. It has two parts. In the first part I discuss and criticize the particular conception of what an object is, and hence correlatively of what it is to perceive an object, that underlies current research into the so-called binding problem. The default assumption in play is that objects are nothing more than bundles of features and therefore that perceiving an object is essentially a matter of perceiving a bundle of features co-instantiated in a particular place. On this conception, therefore, the binding problem is the problem of explaining how it comes about that the various different features of an object, many of them processed separately in different areas of the brain, come to be bound to each other in a way that allows them to be experienced as co-instaniated in a particular place. I will suggest, however, that this is not the best way of understanding objects and what it is to perceive them. Perceiving a bundle of co-instantiated features is a necessary but not a
sufficient condition of perceiving objects. On the alternative account of the metaphysics of object perception that I will sketch out and defend, we should understand objects as bundles of features that obey certain high-level physical principles.

In the second part of the paper I show how this account of the metaphysics of object perception to
work in explaining current research into object perception in infancy. The basic idea here is that we should understand object perception in terms of perceptual sensitivity to certain canonical high-level physical principles that define what it is for a bundle of features to count as an object. I use this basic idea to explain how infant perception differs from adult perception as well as putting it forward as a way of understanding infant naive physics without attributing to them some form of theoretical understanding of the relevant principles.