The Commentaries

David Booth

Phenomenology is art, not psychological or neural science

Abstract

It is tough to relate visual perception or other achievements to physiological processing in the CNS. The diagramatic, algebraic and verbal pictures of how sights seem to this author do not advance understanding of how we manage to see what is in the world. There are well known conceptual reasons why no such purely introspective approach can be productive.

To see something is an achievement. That is to say, the claim to have performed correctly can be tested. Indeed, we can investigate how that task of visual recognition was successfully carried out. We can try to infer the information-transforming (cognitive) processes mediating the performance by varying what is visible and observing changes in response (i.e. doing psychophysics): this is an example of psychological science.

The physical "engineering" of these processes of seeing can also be studied by varying the optical input, but this time observing what is projected onto the retina and activity in the CNS, from the rods and cones to V1 and beyond. Considerable progress has been made in relating cellular neurophysiology to the psychophysics of elementary features of the visible world. It is not so easy to get psychophysical evidence that distinguishes between a cognitive process being in consciousness and transiently out of consciousness (Booth & Freeman 1993), although it is clear than some visual information processing never enters consciousness. When we can't specify a mental process as conscious, there can't be a theory of the neural basis of that process. Lehar's complaint that neuroscience fails to explain visual consciousness is vacuous.

Furthermore, what we know to be the case through use of our senses is a very different kettle of fish from the contents of consciousness, in the sense of how things seem to us while we discount our beliefs about how they actually are. By definition, how things seem can't be checked against how things are. So the systematisation of expressions of subjective experience is an art-form. Lehar's diagrams, his field equations and his verbal exposition are sophisticated elaborations of the sort of thing that I draw when I wake up and try to sketch the visual imagery that I was experiencing as I woke. His and my graphic, algebraic and verbal efforts cannot be wrong or right: they merely express how it appeared to be.

Lehar says that his visual experience is holistic. I can empathise with that impression. Yet also I have visual experiences that are not holistic. I bet that he does too but chooses to ignore them. Any artist may do that, on the grounds that it would spoil the picture or detract from the story. However, that's aesthetics, not science.

I'm not being positivistic. On the contrary, it is Lehar who commits the empiricists' and rationalists' epistemological fallacy of trying to build public knowledge on the basis of impressions or ideas that seem indubitable because they are private and so can't be wrong - but then neither can they be right. Lehar writes: "These phenomena are so immediately manifest in the subjective experience of perception that they need hardly be tested psychophysically" (page 52 of 66). In words of one or two syllables, "What appears seeming to seem in seeing is so clearly clear that there is no need to test it against success at seeing."

Lehar's paper is built on equivocation in use of the word "perception" between the objective achievement and subjective experience. (The word "conscious" in his title is redundant: experiencing subjectively is the same as being conscious.) Like most philosophers, mathematicians and physicists who expatiate on consciousness, he shows no sign of having considered what was shown, and how it was shown, by any psychological experiment on the perceiver's achievement in a visual task. He also ignores the philosophical advances following the later Wittgenstein's debunking 60 years ago of the pervasive fallacy of supposing that when a patch that is red (in the world that we all live in) is seen as red, this is a 'seeming' in another world (Lyons, 1983). Worse, since these appearances, subjective experiences, conscious qualia or whatever are part of each of us, Lehar (like many) locates them in our heads, or as neurocomputations if we are foolish enough to look for consciousness among the brain cells (Booth, 1978). This is all a big mistake about the grammar of the verb "to seem." When we are viewing something but have reason to doubt that we perceive it correctly, then we may retreat to a claim that it seems to be so. We are not looking at a world inside our minds; we are having problems in seeing the colour of the patch out there.

The grammar of 'seeming as though' or 'seeing as' also shows what the subjective experience is isomorphic to. The syntax of 'as' is the figure of speech known as simile. Subjective visual experience is holistic, at least at times, because the world in which we operate is 'holistic' in its optics; black holes are pretty uncommon in everyday life. Lehar actually says this on page 10 of 66, although he has hidden the point from himself by a tangle of the conceptual mistakes that Wittgenstein (1953) cut through. "The perceptual experience of a triangle cannot be reduced to just three phenomenal values but is observed as a fully reified triangular structure that spans a specific portion of perceived space." Delete the reference to a contrary and all the redundancies and we get, "The perceptual experience of a triangle ... is ... as [sic] ... triangular ..."

Furthermore, a triangle is not a triangle in any world unless it "emerges" "whole," "real," and "invariant." If a Gestalt is taken to be a subjective experience (rather than a perceptual performance), then it is consciousness simply of "seeing the world as it is."

Booth, D.A. (1978). Mind-brain puzzle versus mind-physical world identity. Commentary on R. Puccetti & R.W. Dykes: Sensory cortex and the mind-brain problem. Behavioral and Brain Sciences 3, 348-349.

Booth, D.A., & Freeman, R.P.J. (1993). Discriminative measurement of feature integration in object recognition. Acta Psychologica 84, 1-16. Lyons, W. (1986). The disappearance of introspection. MIT Press, Cambridge MA. Wittgenstein, L. (1953). Philosophical investigations. Blackwell, Oxford.

Birgitta Dresp

Double, double, toil and trouble - fire burn, theory bubble!

Abstract

In this target article we are told that "...the most serious indictment of contemporary neurophysiological theories is that they offer no hint of an explanation for the subjective experience of visual consciousness...". Lehar attacks "good old" Neuron Doctrine by stating that, as a theoretical approach to visual perception, it has reached a dead end because he (Lehar) finds it ..."hard to imagine how...an assembly of independent processors (neurons) could account for the holistic emergent properties of perception identified by Gestalt theory". He then proposes his own doctrine, the Gestalt Bubble Model. The Gestalt Bubble is presented as a computational approach to the perceptual representation of 3-D visual space using a volumetric matrix of dynamic elements, each of which can exist in one of several states: transparent for the representation of void space, opaque-coplanar for the representation of smooth surfaces, opaque-orthogonal for the representation of corners, and opaque-occlusion for the representation of surface edges. The supposed transformation of the physical world outside by a perceptual process taking place inside the brain is defined as the turning on of the appropriate pattern of elements in the volumetric matrix of the model in response to visual input. The Gestalt Bubble thereby replicates the three-dimensionality of visual objects as they are experienced in the subjective percept. The principal merit of this model resides in the fact that it translates some major Gestalt laws of visual perception such as emergence, reification, multistability, and invariance into computational codes.

What the author fails to make clear in his target article is the supposed link between his Gestalt Bubble model and general theories of consciousness. All he does here is demonstrate that modern computer technology produces algorithms that allow us to translate the laws of perceptual organization formulated in Gestalt theory into formal codes within the framework of a computational model. What the model has to do with consciousness, however, remains totally unclear. Neither the fact that we are able to consciously experience and describe 3-D shapes as entities and wholes, nor the fact that we can find laws or codes describing how these emerge perceptually, implies or proves that consciousness is necessary to see and move around in 3-D space. In addition, while Lehar seems to imply that his Gestalt Bubble provides a ready model of what he refers to as visual consciousness, he fails to provide clear definitions of what we are supposed to understand by visual consciousness, phenomenal awareness, subjective perceptual experience, or consciousness in general. In the title of the target article, he uses the term "subjective perceptual consciousness". Does this suggest that there should be an objective perceptual consciousness as well?

Moreover, the author readily assumes the existence of a "visual consciousness" as a particular form of consciousness. This assumption needs to be justified. How would a visual consciousness operate in comparison to an auditory, tactile, or olfactory consciousness, for example? In fact, by using ambiguous terminology in his text (terminological danglers ?), switching readily from one level of explanation to another, the author fails to convince his readers that he knows what he is talking about when he discusses the question of consciousness. Moreover, the fundamental difference between Lehar's "picture-in-the-head" model and the concept of isomorphism from classic Gestalt theory is not discussed in a satisfactory manner. After a lengthy introduction that confronts the reader with odds and ends of numerous general theories of mind and consciousness, the author, all of a sudden, pops up his own version of the Gestalt hypothesis of isomorphism by suggesting that we see the outside world as we do because that is, and has to be, the way the world is represented in the brain. This "picture-in-the-head" view goes far beyond the classic Gestalt concept of isomorphism because it assumes not only a functional, but also a structural correspondence between the visual percept and its brain representation. It is introduced here as the only rightful answer to Koffka's question "why do we see things as we do?" ; the original Gestalt viewpoints (e.g. von Ehrenfels, 1890, Metzger, 1936, Kohler, 1961, among others) on isomorphism are not discussed.

Interestingly, the author seems to have overlooked that his "picture-in-the-head" hypothesis (structural isomorphism) stands or falls on the validity of the assumption that one of the key principles formulated by Gestalt theory, that of the common fate of parts (Ganzbestimmtheit der Teile, Metzger, 1936), reflects the result of a neurophysiological mechanism. In the early sixties, some psychophysicists questioned the neurophysiologcial validity of precisely this principle of perceptual organization. Pritchard (1961) presented figures as stabilized images on the retina and showed that the constituent elements of these figures disappeared from phenomenal awareness one by one, not all at once as the principle of common fate of parts would predict if it reflected the result of a neurosensory mechanism (see also Pritchard, Heron, & Hebb, 1961). In any case, even if the "picture-in-the-head" view could be proven right, Lehar would still have to come up with an explanation of the mechanism(s) by which the picture in the head is made available to consciousness. Also, a rigorous distinction between "awareness", like awareness of the emergent properties of a visual object at a given moment, for example, and "consciousness", like the consciousness of being aware of the emergent properties of a visual object and its significance within a general context, for example, would then have to be made.

Lehar writes that it is of central importance for psychology to address what "all that neural wetware" is supposed to do, and to determine which of the competing hypotheses (presented in the introduction of his target article) "reflects the truth". Who said that science has to bother with metaphors such as "truth" ? As far as I understand it, science is all about facts and measures, collected within a specific context of boring constraints, usually called " conditions ", and therefore inevitably requires a diversity of methods and hypotheses. The concept of " truth " does not appear to be of much use here. Are we not often enough reminded to take care not to get trapped by the metaphors we use to construct hypotheses and explanations ? The overwhelming "Unsumme" (as defined by Metzger, 1936) of bits and pieces of philosophy and phenomenological "brain teasers" we are confronted with in this target article somehow shows how easily we can end up like the Sorcerer's Apprentice in Goethe's poem, who tries all sorts of curses and invokes all sorts of spirits, but is finally unable to take control.

Back to Top

Wlodzislaw Duch

Just bubbles?

Abstract

Lehar misrepresented neuron doctrine and indirect realism. His conclusions on consciousness are unjustified. The Bubble Gestalt perceptual modeling disconnected from neuroscience has no explanatory power.

1. Perception has not evolved for our enjoyment, it serves action, exploration of the world (O'Regan, Nöe 2001) Although the richness of visual perception may partially be an illusion, sensory data should elicit brain states that reflect important features of perceptual organization. Such functional representation would be very useful, facilitating information retrieval from visual and auditory cortex, stored in attractor neural networks after termination of direct sensory inputs (Amit 1994). Persistent brain activity may be responsible for visual imagery, filling in, illusory contours and other such phenomena. This internal representation, being a physical state of the brain, is focused and interpreted by other brain areas, gating it to the working memory and facilitating conscious perception. It is constructed from sparse information obtained from eye fixations between saccades (as is evident in the change blindness experiments, O'Regan, Nöe 2001), and thus may not be so faithful and rich as it seems. Since for many people endowed with visual imagination (individual variance seems to be quite large in this respect) visual experiences are rich and vivid, filling in of missing information must be strong.

2. Construction of the inner perspective is a difficult task. Lehar does not even attempt to enumerate the dimensions required for perceptual modeling that could replace (or at least complement) neural modeling. I have argued myself (Duch 1997) that an intermediate level of cognitive modeling should be useful. It should represent mental events in a way that is closer to our inner perspective, acceptable to psychologist, but also should facilitate reduction, at least in principle, to the neural level. Complex neural systems reveal emergent processes (responsible, as Lehar has noticed, for Gestalt phenomena), requiring a higher level of description characterized by new laws and phenomena. The usual approximation to neural activity misses the perceptual level by going from states of recurrent networks (such as Grossberg's adaptive resonant states, Grossberg 1995), to states of finite automata (cf. Parks et al. 1998 for neural models in psychiatry). A shortcut from neuroscience via neural networks to behavior is satisfactory only to behaviorists. Mind states and mental events may emerge as "a shadow of neurodynamics" in psychological or perceptual spaces (Duch 1997). This is in accord with ideas of Shepard (1987, 1994), who believed that universal laws of psychology may be found in appropriate spaces. Psychological spaces are spanned by subjective dimensions (such as color, shape, and motion), and one may use them to explain subjective perception and to talk about mental events implemented at the neurodynamical level. Therefore I sympathize with Lehar's goal, although details of his proposal are not satisfactory.

3. Trivializing the "neuron doctrine" Lehar writes about neural networks as the "quasi- independent processors", and "an assembly of independent processors". The whole essence of neural networks is in the interaction of their elements, cooperative computational abilities that facilitate their holistic emergent properties. Recurrent neural networks are certainly not "the atomistic feed-forward model of neurocomputation" (Parks et al. 1998). The Neuron Doctrine paradigm has been completely misinterpreted in the target article.

4. The arguments evoked against indirect realism are strange to say the least. Lehar mixes mental and physical levels freely, writing statements like "the world that appears to be external to our head is actually inside our head", and "beyond those perceived surfaces is the inner surface of your true physical skull encompassing all that you perceive". How can physical scull encompass non-physical, inner world? "The world inside the head" is a metaphor, and it does not make much sense to invert it, unless one believes that there is some kind of physical world squeezed inside the scull.

Indirect realism claims that we perceive and comment upon states of our own brain. These states reflect properties of the environment, but interpretation of the spatial structure of the states of visual system has nothing to do with their physical location. There is nothing strange about it, as there is nothing strange in transmission of the voice and images via wires and radio waves. The spatial world inside the head is there in the same sense as panoramic image in the integrated circuit of a computer graphic chip. Subjective reversal of a multistable percept follows the change of neural dynamics. It has to be experienced vividly as an inversion of a perceptual data structure, since visual experiences are a reflection of neural dynamics - how else could changes of visual cortex states be experienced?

5. It is certainly not clear "that the most fundamental principles of neural computation and representation remain to be discovered". Churchland (1984) argued against it already 20 years ago, and since that time computational neuroscience has made a lot of progress. It may very well be that Hebbian learning is the only fundamental principle that is needed and that sufficiently complex models of the brain will be able to simulate its emergent functions.

6. It is quite probable that "our own conscious qualia evolved from those of our animal ancestors". But certainly the "conclusion is that all matter and energy have some kind of primal proto-consciousness" is not inescapable. In fact I am regularly loosing my consciousness in sleep, while anesthetics and damages to the reticular formation lead to coma, obliterating consciousness. Complex organization of matter is not sufficient for consciousness. Instead of looking for conditions necessary for manifestation of consciousness - a fruitful way is to use here a contrastive approach between perception and reception (Taylor 1999) - Lehar goes down the beaten track of thinking about consciousness as some kind of a substance that is present in all matter, although sometimes in watered down form. Conclusion of this line of reasoning is absurd: proto- consciousness of soap bubbles.

Of course since the concept of consciousness is not defined one may try to extend it to all matter, but taking about stomachs being "conscious" leaves no semantic overlap with the word "conscious" applied to a baby, or to a cat. If consciousness is a function, and plays functional role, as Lehar seems to believe ("It seems that conscious experience has a direct functional role"), the inescapable conclusion is rather that not all brains are equal. Language is unique to humans, and even though one can extend the concept of language to some more primitive forms of communication, interaction between internal organs of the body, or message passing between components of a computer system, is not the same "language" as natural languages. The difference between a "field" in agriculture and "field" in physics is comparable to the difference between animal "consciousness", and "consciousness" of a soap bubble due to the physical forces that determine its shape. We should not be deceived by words.

Back to Top

Charles R. Fox, O.D., Ph.D., F.A.A.O.

Back to Top

Stephen Grossberg

Linking Visual Cortex to Visual Perception: An alternative to the Gestalt Bubble.

Abstract

Lehar's lively discussion builds on a critique of neural models of vision that is incorrect in its general and specific claims. He espouses a Gestalt perceptual appoach, rather than one consistent with the "objective neurophysiological state of the visual system" (p. 1). Contemporary vision models realize his perceptual goals and also quantitatively explain neurophysiological and anatomical data.

Lehar describes a "serious crisis" (p. 1), "an impasse" and a "theoretical dead end" (p.2) in contemporary models of vision and advances as a possible alternative his Gestalt Bubble approach "which is unlike any algorithm devised by man" (p. 2). He also claims that "Gestalt aspects of perception have been largely ignored" (p. 2) by neural models of vision and then goes on to describe presumed dichotomies between equally desperate attempts to understand how the brain sees. Lehar particularly comments about modeling work by my colleagues and myself, noting that "the most serious limitiation of Grossberg's approach...is that, curiously, Grossberg...did not extend their goal to...three-dimensional spatial perception [and] no longer advocated explicit spatial filling-in" (p. 11). He also says it is "impossible for Grossberg's model to represent transparency..." (p. 12). These general and specific claims unfortunately do not accurately represent the published literature about neural vision models. Lehar seems motivated to trash neural vision models because his own model makes no contact with neurophysiological and anatomical data about vision.

In reality, there is an emerging neural theory of 3D vision and figure-ground perception, called FACADE theory, for the multiplexed Form-And-Color-And-DEpth representations that the theory attempts to explain (Grossberg, 1987, 1994, 1997). Lehar refers to my 1994 article in summarizing the deficiencies of our models. However, this article explains many 3D figure-ground, grouping, and filling-in percepts, including transparency, and uses an explicit surface filling-in process. Later work from our group developed these qualitative proposals into quantitative simulations of many 3D percepts, including 3D percepts of daVinci stereopsis, figure-ground separation, texture segregation, brightness perception, and transparency (Grossberg and McLoughlin, 1997; Grossberg and Kelly, 1999; Grossberg and Pessoa, 1998, Kelly and Grossberg, 2000; McLoughlin and Grossberg, 1998).

These studies laid the foundation for a breakthrough in understanding how some of these processes are organized within identified laminar circuits of cortical areas V1 and V2, notably processes of cortical development, learning, attention, and grouping, including Gestalt grouping properties (Grossberg, 1999a; Grossberg, Mingolla, and Ross, 1997; Grossberg and Raizada, 2000; Grossberg and Seitz, 2003; Grossberg and Williamson, 2001; Raizada and Grossberg, 2001, 2003; Ross, Grossberg, and Mingolla, 2000).

This LAMINART model has been joined with the FACADE model to develop a 3D LAMINART model that quantitatively simulates many perceptual data about stereopsis and 3D planar surface perception, and to functionally explain anatomical and neurophysiological cell properties in cortical layers 1, 2/3A, 3B, 4, 5, and 6 of areas V1 and V2 (Grossberg and Howe, 2003; Howe and Grossberg, 2001), and uses 3D figure- ground and filling-in concepts to do so. More recently, the 3D LAMINART model has been generalized to explain how 3D percepts of slanted and curved surfaces and of 2D images are formed, and clarified how 3D grouping and filling-in can occur over multiple depths (Grossberg and Swaminathan, 2003; Swaminathan and Grossberg, 2001). This work includes explanations of how identified cortical cells in cortical areas V1 and V2 develop to enable these representations to form, how 3D Necker cube representations rival bistably through time, how slant aftereffects occur, and how 3D neon color spreading of curved surfaces occurs even at depths which contain no explicit bottom-up inputs. All these studies are consistent with the grouping interpolation properties that Kellman et al (1996) have reported (p. 51), and the 3D grouping properties summarized in Figure 16 that Lehar seems to think cannot yet be neurally explained.

These modeling articles show that many of the perceptual goals of Lehar's Gestalt Bubble model are well-handled by neural models that also provide a detailed account of how the visual cortex generates these perceptual effects. In summary, we do not need analogies like soap bubble (p. 28), or rod-and-rail (p. 32), or different local states to represent opaque or transparent surface properties (p. 35), as Lehar proposes (p. 28). The brain has discovered a much more interesting solution to these problems, which links its ability to develop and learn from the world with its ability to see it.

Lehar makes many other claims that are not supportable by present theoretical knowledge. He claims that "we cannot imagine how contemporary concepts of neurocomputation...can account for the properties of perception as observed in visual consciousness [including] hallucinations" (p. 9). Actually, current neural models offer an explicit account of schizophrenic hallucinations (Grossberg, 2000) as manifestations of a breakdown in the normal processes of learning, expectation, attention, and consciousness (Grossberg, 1999b).

Contrary to Lehar's claims on pp. 43-45, recent neural models clarify how the brain learns spatial representations of azimuth, elevation, and vergence (see Figure 14) for purposes of, say, eye and arm movement control (Greve, Grossberg, Guenther, and Bullock, 1993; Guenther, Bullock, Greve, and Grossberg, 1994). Lehar defends "the adaptive value of a neural representation of the external world that could break free of the tissue of the sensory or cortical surface..." (p. 46). Instead, What stream representations of visual percepts should be distinguished from Where stream representations of spatial location, a distinction made manifest by various clinical patients.

Lehar reduces neural models of vision to capacities of computers to include navigation as another area where models cannot penetrate (p. 49). Actually, neural models quantitatively simulate the recorded dynamics of MST cortical cells and the psychophysical reports of navigating humans (Grossberg, Mingolla, and Pack, 1999), contradicting Lehar's claim that "the picture of visual processing revealed by the phenomenological approach is radically different from the picture revealed by neurophysiological studies" (p. 48). In fact, a few known properties of cortical neurons, when interacting together, can generate emergent properties of human navigation.

Back to Top

Keith Gunderson

Steven Lehar's Gestalt Bubble Model of Visual Experience: The embodied percipient, emergent holism, and the ultimate question of consciousness.

Abstract

Aspects of an example of simulated shared subjectivity can be used both to support Steven Lehar's remarks on embodied percipients, and triangulate in a novel way the so-called "hard problem" of consciousness which Lehar wishes to "sidestep" but which, given other contentions of his regarding emergent holism, raises questions about whether he has been able or willing to do so.

Steven Lehar's Gestalt Bubble Model (GBM) is said to emphasize the often ignored fact "...that our percept of the world includes a percept of our own body within that world, and it remains at the center of perceived space even as we move about in the external world." (6.4) I offer here a friendly, if folksy, example of a simulation of shared 1^st person subjectivity designed to reinforce Lehar's brief but interesting claims concerning the prominence of the embodied percipient in visual perception, but one which leads to other questions regarding his analysis. I have labeled the example elsewhere and with variations The Cinematic Solution to the Other Minds Problem and invoked it earlier against B.F. Skinner's view of subjective privacy and scientific inquiry, also objected to by Lehar for his own reasons.(Gunderson 1971,1984) Suppose a film director wishes to treat us to the subjective perceptual experiences of another person, say Batman, as he gazes on the traffic far below from some window perch. How is this best done? Not, to be sure, by simply showing us the whole superhero perched on the ledge with the traffic moving by on the street below. This would not be anything like being privy to Batman's subjective perceptual experience. It would only amount to our own visual experience coming to include Batman. Instead what is characteristically done is that Batman's filmed body (or at least the better part of it) is somehow (gradually or suddenly) subtracted from the screen in such a manner that we become insinuated into roughly whatever space and orientation Batman's body occupies, and are thereby made party to the visual field (sense of height, traffic passing below, etc.) that we can asssume would be Batman's from that perspective. We cannot literally, of course, occupy (even cinematically) exactly the same space that Batman does - a prerequisite to having his visual experience - but the tricks of the art permit us to enjoy a simulation of such an occupancy. It is the sleight-of-camera with respect to our seemingly ubiquitous embodied presence in visual perception which carry with it tactics for conjuring a sense of the usual "subjectivity barrier" between us and another percipient being breached. And here it occurs in a florid phenomenological manner, obviously different from the "relational information" which can cross that barrier as described by Lehar (5.1). Notice too, that a "pre-set" feature of the whole typical movie experience involves the darkened theatre, and no focused sense of our own body being either present in the audience, or as an inclusion in the screen action. The effect is that where we are not assuming specifically Batman's perspective, we are assuming one belonging to no one in particular, or rather one "belonging" to anyone in the vicinity, as it were.

So the possibility of the cinematic simulation of shared subjectivity seems to presuppose the inclusion of an embodied percipient in our visual perceptions along lines suggested by Lehar. But the apparent friendliness of the example has a complicated provocative side as well. For if what it takes to create the illusion is the clever collapsing of our perspective (or someone else's) into another's, then the epistemic-ontic primacy of the 1st person point of view becomes obvious, and the "hard problem" of consciousness can be rephrased with respect to it this way: there is no analogous thought experiment which would render subjectivity or a point of view (one's own or another's) as being somehow manifest in any set of neurophysiological processes to begin with, such that another consciousness might appear as somehow insinuated into it. But there should be if consciousness is to be modelled (displayed, illustrated ) within any 3^rd person physicalistic conceptualization. This rather flat and crude sounding point is not, I think, irrelevant or naively realistic. In a nutshell, that there can be no cinematic type simulation of a solution to the mind-body problem parallel to the other minds one, can be seen to stem from our inability to cling to our sense of experiencing a point of view while being in some neurophysiological locus (however represented).For Lehar the salient residual problem(s) is this: although the contents of all our subjective visual experience for the GBM are subsumed under the subjective, we lack any vivid demonstration of how the having of a 1st person point of view itself which is a prerequisiite to their being any such phenomenal contents, lies within that experience. Simply specifying underlying neurophysiological conditions for consciousness takes us nowhere we haven't already unsatisfactorily been. That there is, and how there is, any locus at all for our perceptions remains unexplained within any micro or macro frame of reference. We think, of course, that the locus of our locus of perceptions lies in some way within the embodied. But to be apprised of all this does not thereby help us to see how any subjective perspective occurs in the first place, or why it is uniquely ours! (Nagel 1965). The problem of explaining it arises independently of whatever type of metaphysical substance the perceiver is believed to be embodied in, even as part of a panpsychic or panexperientialist scheme such as Chalmer's (as in 6.5). And it can be reiterated with respect to any type of substance of any kind of complexity so far we can tell.

Back to Top

Julian Hochberg

Backdrop, Flat and Prop: The stage for active perceptual inquiry.

Abstract

Lehar's revival of phenomenology, and his all-encompassing bubble model, are ambitious and stimulating. I offer an illustrated caution about phenomenology, a more fractured alternative to his bubble model, and two lines of phenomena that may disqualify his isomorphism. I think a perceptual-inquiry model can contend

Steven Lehar's ambitious bubble metaphor is highly stimulating, assuming a unified phenomenal visual world that explains and predicts our perceptual experience. Herewith, a cautionary reminder about phenomenology as such; an alternative to Lehar's specific enclosing bubble model; and two lines of phenomena that Lehar ignores but that are difficult to reconcile with the particular isomorphism he espouses. Phenomenology should indeed guide psychophysics and neurophysiology. But phenomenology is certainly not incontestable. For example, Lehar cites the CIE as a description of phenomenological color space. The Helmholtzean dogma, that the experience of yellow consists of red plus green experiences, lurked within mainstream sensory physiology until after WWII (and was often attributed to the CIE); following Hering instead, Hurvich & Jameson's phenomenologically-guided opponency-oriented psychophysics and model (1955) explained to neurophysiologists what their microelectrodes later revealed, thus changing our view of neurophysiology and liberating our relevant phenomenology. (In fact, Jameson & Hurvich showed later(1967) that the CIE is no phenomenological summary -- two very different colors come out at the same point on the graph.) Phenomenology must be both consulted and contested. Accordingly, a different model follows.

Lehar's tackling of encompassing space is an important step, but other phenomenological details might support a different, less wholistic model - a stage or set, not a bubble: Several quite different aspects of our visual ecology afford distance information. Their zones of efficacy, as in Figure 1A (after Cutting & Vishton, 1995), surely are important for any account of our encompassing visual world. Assume that the furthest zones form an essentially equidistant region like the backdrop on the stage in Figure 1B. Railroad tracks visible in those zones appear to converge. In nearer zones, the depth information effectively specifies the tracks as parallel, and holds the backdrop in its place upstage.

Figure 1. Onstage and backdrop scenery. A.The strength of the major depth cues with egocentric distance, adapted from Cutting and Vishton, 1995 (with permission). To the eye as actor, the backdrop usually lies between 10 and 50 feet upstage. B. The experienced stage in which visual inquiry proceeds. The viewer's normal actions provide no distance information beyond the plane labeled "backdrop;" and they can readily generate and therefore incorporate information about the downstage prop. The curves of Figure 1A account for, but are not salient in the experience of B. C. Attention extends the stage. When the inquiring eye visits a scene, its boundaries are remembered as further out than they were (see Intraub, 1997); this is not merely memory, since such Boundary Extension (BE) is a function of where the viewer plans to look.(Intraub et al, 2001).

This implies discontinuities (e.g., between backdrop and stage) that are not firmly fixed, since where the viewer attends, and with what intentions, affects what information is recovered and used (cf. Figure 2B, C). Figure 1A can therefore serve only as a conditional account; and as Figure 1C implies, the phenomenal layout itself varies somewhat with the viewer's perceptual intentions. In this model, therefore, distance to the end of the internal world is not a continuous variable, nor continuously defined. Why are not the discontinuities spontaneously evident? Next, evidence of such overlooked discontinuity.

Figure 2A seems to reverse as a whole, and has been offered as one example of how a minimum principle (including Lehar's version), leads to perceiving an entire 3D structure (Kopfermann, 1930, Hochberg & MacAlister, 1953). But Figure 2B shows that, when tested, perfectly possible objects display the same dependence on what the viewer attends as was previously shown by the Penrose & Penrose (1958) impossible figures. Perceptual consequences (Hochberg, 1998; In Press) like the effects of rotation described in Figure 2B, and the surface-lightness effect in Figure 2C, attest that these are perceptual phenomena. They also share some aspects of Lehar's isomorphism. (And the absence of any salient break between the different spatial zones of the environment in Figures 1A and 1B, and in the apparently-continuous bubble that Lehar describes, merely parallels what happens within objects.)

Figure 2. Some shapes isomorphism must take. A. The reversible Necker Cube. Sometimes offered as an example of how a minimum principle (or something like it, in Lehar's version), leads to perceiving an entire 3D structure (Kopfermann, 1930, Hocherg & MacAlister, 1953). (B) The partly-reversible Killer Cube. When attended at (a), the present cube appears of definite and nonreversible 3D structure; when attended at (b), it soon starts reversing, though the same Gestalt remains in view (though off attentional center). The reversals are attested by their perceptual consequences: When rotated clockwise around its vertical axis, the perceived motion is clockwise when (a) is attended; when (b) is attended and when it appears nearest the viewer, motion appears counterclockwise. Such perceptual consequences help validate one's otherwise unsupported phenomenology, as in the next figure. (Hochberg & Peterson, 1987; Hochberg, In Press) (C). Adelson's Impossible Staircase. With no discernible discontinuity, the right and left sides here are incompatible as 3D structures; showing that they are actually seen that way, note that the same print density appears of higher reflectance (lighter paint job) at (b) than at (a). (after Adelson, 2001, with permission); see text. D. Do configuration-based organizational factors first provide figure- ground segregation which thereby offers a shape to be recognized? Not so you can tell: See text. (Peterson and Gibson, 1993; see Peterson, 1994).

Such phenomena raise difficulties for any wholistic proposed isomorphism powered by the physical relationships as perceived. Gestaltist visions of isomorphism were of course concerned mostly with flat shapes, not 3D structures (see Hochberg, 1998). The fact that Peterson and her colleagues (see Figure 2D) have shown that meaningful (denotative) shapes preempt figural status when in their familiar orientations (Figure 2Dc,d) but not when the physically identical configurations are inverted (Figure 2Da,b) makes it hard to even imagine what an appropriate formulation of isomorphism would be like. A phenomenology centered on query-directed TOTE-like machinery might be easier to manage (cf. Hochberg, In Press; 1970; O'Regan & Nöe, 2001).

Back to Top

Donald D. Hoffman

Back to Top

Donald Laming

Psychological relativity

Abstract

'Psychological relativity' means that 'an observation is a relationship between the observer and the event observed'. It implies a profound distinction between 'the internal first-person as opposed to the external third-person perspective'. That distinction, followed through, turns Lehar's discourse inside-out. This commentary elaborates the notion of 'Psychological relativity'; shows that while there is already a natural science of perceptual report, there cannot also be a science of perception per se; and draws out some implications for our understanding of phenomenal consciousness.

Lehar is lacking an essential idea. Physicists have it-'relativity'-but Lehar does not. Lehar chances to mention (§1) 'the internal first-person as opposed to the external third- person perspective', but fails to realise how that distinction impacts on his discourse. If the implications of that distinction be followed through, the entire body of problems addressed is turned inside-out.

The overriding principle that Lehar is lacking is

an observation is a relationship between the observer and the event observed

and thereby depends on the observer as well as the event. So, two observers in motion relative to each other make different determinations of the velocity of a third object (Galilean relativity). Figure 2 sketches the set-up for Thouless' (1931a,b) phenomenal regression to real size. The observer has a different view of the experiment to the experimenter.

Figure 1. The different views from four houses on a housing estate. (Reproduced with permission from Understanding human motivation by D. Laming, p. ??. (c) 2003, Donald Laming; published by Blackwell Publishing)

Figure 1 presents an analogy. Looking out from my window, I can see three other houses, separated from me by a road and a green sward. If there is a car in the road, my neighbour and I can readily agree that it is red. By agreeing on a suitable instrument for measurement, we can agree the colour of the car to whatever precision we desire. That arena outside our houses (camera view) is part of the public domain within which experiments can be conducted. But my neighbour and I cannot see into each other's houses. If I telephone my neighbour, I can only describe my interior furnishings by reference to what my neighbour will have seen elsewhere. The scope of experimental procedure can be extended to internal experience only by projecting that experience into the public domain. I might describe my curtains as scarlet, or carmine, or cerise -but my neighbour might think of a different colour referent to the one that I have in mind, and 'seeing red' will then mean slightly different things to the two of us.

I can invite my neighbour into my house to see for himself; but I cannot give him direct access to my visual experience. One might suppose that my internal visual experience could be measured, like the colour of the car in the road. But experimental psychologists have been trying to measure internal sensations for 150 years and have so far progressed nowhere (Laming, 1997).

Some part of our visual experiences can be shared with others; the remainder is private. The Gestalt properties surveyed in §5 and §7 belong to that private part, which is why Gestalt psychology has not proceeded beyond verbal description. There is a boundary between experiences that can be shared and experiences that are essentially private. It is determined by what, within my field of view, my neighbour can also see (see Fig. 1). That is, the boundary is determined within my neighbour's field of view and is not to be found within my own visual experience. My own experience, by itself, contains no distinction between that which lies in camera view and that which is private. The junction is seamless. It is only too easy to confound subjective experience with objective observation; this is what Lehar has done.

It follows that there cannot be a natural science of perception. There is a science of perceptual report, a tradition that goes back to Fechner (1860). But perceptual reports cannot be taken at their face value (here the Gestalt psychologists erred), but must be evaluated by experiment. Lehar is aware of this (§5.2), but asserts that perceptual experience is isomorphic to the neural substrate and thereby denies this distinction.

Lehar's stance is that " the world of conscious experience is accessible to scientific enquiry after all, both internally through introspection and externally through neurophysiological recording." He envisages an isomorphism between perceptual experience as described by the observer and the observations of the natural scientist. Thouless' (1931a,b) experiment on phenomenal regression to real size (Fig. 2) shows why such an isomorphism is not found in nature.

Figure 2. Experimental set-up for the measurement of phenomenal regression to real size. (Adapted with permission from Understanding human motivation by D. Laming, p. ??. (c) 2003, Donald Laming; published by Blackwell Publishing)

The observer's task is to select a disc set normal to the line of sight at distance a to match the angular size of the larger disc at distance b. While people do choose a smaller disc from the alternatives at a, they systematically choose one too large to match (phenomenal regression to real size). Imagine that a neurophysiologist making observations at the neural level of description relevant to understanding how and why this error of judgment occurs. If the observer's perceptions stand in the same relation to the neural substrate as the neurophysiological observations, then there has to be an internal 'observer' looking at internal processes with the same objectivity as the neurophysiologist. The fact that Lehar has a mathematical model to replace the neurophysiological observations does not alter this requirement. This observer is represented by the 'thinks bubble' in Figure 2. Philosophers will immediately identify this internal observer as Ryle's (1949) 'Ghost in the machine' (which is why the 'thinks bubble' is decisively crossed out).

I next ask whether the hypothetical neurophysiologist can also observe the neural substrate of this 'ghost'. If so, the relationship of the ghost to the neural substrate is structurally different to that of the neurophysiologist; otherwise the 'ghost' is pure mind- stuff. In fact, verbal descriptions of what is perceived are produced by the same system as that which does the perceiving, and the relationship of 'observer' (if that term may still be used) to the neural substrate that is supposedly 'observed' is essentially different to that of a third-party neurophysiologist. Several conclusions follow.

There need not be any useful isomorphism between neural process and perceptual experience.

Modelling perceptual experience is not an alternative to understanding the neural process.

There cannot be a natural science of perception, distinct from the study of perceptual report.

The idea of psychological relativity also impacts on consciousness (§6). Since it is impossible to access any other person's subjective experience, it is not possible to observe any other person's consciousness. Even if the hypothetical neurophysiologist were to observe and record a substrate in the brain that subserved consciousness, there is no way in which the observations could be identified as such. However much one explores the brain, all that one finds is brain function. Phenomenal consciousness is simply the quality of subjective experience.

Back to Top

Dan Lloyd

Double Trouble for Gestalt Bubbles

Abstract

The "Gestalt bubble" model of Lehar (BBS target article) is not supported by the evidence offered. The author invalidly concludes that spatial properties in experience entail an explicit volumetric spatial representation in the brain. The article also exaggerates the extent to which phenomenology reveals a completely three-dimensional scene in perception.

The real world is a place of many properties; so also is its presentation as a phenomenal world in the conscious brain. One way for a brain state to present in experience a worldly property P is to duplicate P itself. Like a painter striving for perfect mimesis, an embodied consciousness might use patches of red in the head to represent a red apple. Or, according to Lehar, a brain might use spatial properties to represent external spatial reality:

The central message of Gestalt theory is that the primary function of perceptual processing is the generation of a miniature, virtual reality replica of the external world inside our head, and that the world we see around us is not the real external world but is exactly the miniature internal replica. (Lehar, Conclusion)

Lehar's article makes the case for the internal replica, or "Gestalt bubble," and then develops a model of how three-dimensional spatial modeling could occur in something like a neural medium. In this commentary, I suggest that the evidence in support of the Gestalt bubble is in double trouble. It is both conceptually and phenomenologically flawed.

The coffee in the cup at my elbow is (to me) hot, brown, of a certain weight and size, and in a specific location. We cannot conclude, however, that the state of my brain that is my consciousness of the coffee replicates any of these properties itself. Yet this is an inference Lehar seems to make repreatedly in the target article. For example:

The fact that the world around us appears as a volumetric spatial structure is direct and concrete evidence for a spatial representation in the brain. (5.2)

This is a non sequitor, as can be seen by substituting "colored" for "spatial" in the passage. A slightly more elaborate argument is no less fallacious:

The volumetric structure of visual consciousness and perceptual invariance to rotation, translation, and scale offer direct and concrete evidence for an explicit volumetric spatial representation in the brain, which is at least functionally isomorphic with the corresponding spatial experience. (5.1)

Lehar is right that functional isomorphism between phenomenal experience and its implementation is required to avoid "nomological danglers," but once again "explicit volumetric spatial representation" is in no way entailed - for "rotation, translation, and scale" substitute "hue, saturation, and brightness," and the fallacy will be apparent. Nor does Lehar's claim that phenomenal spatiality preserves the relational structure of spatial objects entail an internal replica, since (once again) a three-dimensional relational structure defines "color space" without in the least implying that the color solid appears somewhere in our brain. Functional isomorphism, meanwhile, is readily preserved between spatial objects/scenes and their representations without invoking replicas. For example, the World Wide Web is well stocked with virtual worlds that preserve functional isomorphism with spatial scenes, each of them encoded is some non-spatial computational idiom like VRML, etc.

In sum, the conceptual arguments in the target article do not support the author's main conclusion. Nonetheless, the brain does have properties, and some of its properties do determine the contents of conscious experience. Lehar's arguments do not establish that the brain must use space to represent space. Does phenomenality license any inferences at all about the neural medium? There are two ways to approach this question, beginning either with contingent generalities about perception, or with its essential structures. The first approach begins with features of phenomenality (as revealed by perceptual psychology, including the Gestalt demonstrations of our perceptual capacities). The second analysis isolates essential or necessary structures of phenomenality. The second approach accords with classical phenomenology, as exemplified in the works of Husserl (e.g. Husserl 1974). In either case, the hope is that the analysis of phenomena will constrain the search for computational architectures sufficient to generate some or all of the features of phenomenality.

On neither approach is there compelling reason to posit the spatial virtual world proposed by Lehar. I don't doubt that I live in a spatial world, but in my visual field, i.e., what I see before me right now, conveys far less spatial information than Lehar's Gestalt bubble encodes. At the focus of attention I'm aware of surfaces, distance from my eyes, and edges, but outside of focal attention I experience only a very indefinite spatiality, which seems to me to be inconsistent with the continuously present three-dimensional models constructed in the Gestalt bubble. The supposition that my experience specifies a full 360-degree diorama in my head arises from the "just-in-time" availability of spatial information with every attentional focus. The information is there when and where I need it, and experience presents an ordered sequence of focally attended presentations rather than a single wrap-around replica of the spatial world. This seems to be phenomenologically "given," but it is also amply confirmed in psychological studies of "inattentional blindness" (Mack and Rock, 1998) and "change blindness" (Simons 2000). (Section 8.8 briefly acknowledges the effect of successive gaze fixations in different directions, suggesting that parts of the replica fade while outside of the visual field. This suggests either that the replica has an absolute spatial orientation and does not turn with the head, or, if the replica does turn with the eyes, only a small focal part of it has the spatial detail Lehar describes.)

This disagreement can be made more rigorous, and more properly phenomenological. One essential property of the phenomenal world is expressed in our ability to distinguish properties by location. That is, I can be aware of a red circle and a green square at the same time, without confusing the pairings of colors and shapes. Austen Clark refers to the problem posed by this pervasive perceptual ability as the "Many Properties" problem, and he argues that it can be solved only by coding places along with other perceptual properties (Clark 2000). So "red" and "circle" must be assigned a location, and "green" and "square" a second location. Experience, of course, solves the Many Properties problem easily, and arguably it is essential to the very concept of phenomenality that consciousness solve it. This argument so far provides support for Lehar's position, but immediately raises the question, how many spatial dimensions are required? Lehar advocates three, Clark suggests two, but the argument necessitates just one, a linear dimension along which one point is tagged "red" and "circular" and another "green" and "square." The basic dimension, then, would be temporal, and experience an orderly ensemble of phenomenal leaps and bounds, a time line. Spatiality emerges from trajectories encoded in proprioception, that orient each momentary percept to those before and after. This proposal conforms well with classical phenomenology (Husserl 1966; Husserl 1974), and in other work I present evidence for its implementation in the brain (Lloyd 2002; Lloyd in press). This alternative cannot be defended here, but it does suggest that the Gestalt bubble is not entailed by phenomenology.

Back to Top

Riccardo Luccio

Isomorphism and representationalism

abstract

Lehar tries to build a computational theory that succeeds in offering the same computational model for both the phenomenal experience and the visual processing. However, the vision that Lehar has about isomorphism in Gestalttheorie as representational is not adequate. The main limit of Lehar`s model derives from this misunderstanding of the relation between phenomenal and physiological levels.

The Gestalt psychology was fundamentally misunderstood in the United States (but it too had some responsibilities: see Kanizsa, 1995). After the Second World War it had a meager destiny, cultivated only marginally in Germany and in America, more intensively in peripheral countries like Italy or Japan. However, mainly in the last few decades, some concepts of the Gestalt psychology appear frequently in the psychological debate, as praegnanz, isomorphism, minimum principle, and so forth, and it demonstrates the inability of cognitive psychology to buy some highly significant aspects of our way to pick up the reality that is around us. Lehar`s paper doesn't confine itself to stress the importance of some classic Gestaltist ideas, taken in isolation, as other scholars in the past have done, in a never completely successful attempt to integrate part of the Gestaltheorie inside the cognitive psychology. Instead, Lehar tries to build a computational theory that succeeds in offering the same computational model to both the phenomenal experience and the visual processing.

This highly interesting attempt however deserves some comments. In my opinion, the vision of Gestalttheorie that Lehar has is not fully adequate, and this has some consequence on his theorizing. The point on which I disagree almost completely with Lehar is the following. He claims that there is a central philosophical issue that underlies discussions of phenomenal experience as seen for example in the distinction between the Gestaltist and the Gibsonian view of perception. The world we see around us is the real world itself, or merely an internal perceptual copy of that world generated by neural processes in our brain? In other words, this is the question of direct realism, also known as naive realism, as opposed to indirect realism, or representationalism. I note parenthetically that, though Gibson (1966, 1979) called himself a naïve realist, this was only a provocation. The theory of direct perception is neither naïve nor realistic. As Michaels and Carello (1981, p. 90) clearly put it, "the test of the veridicality of perception concerns the mutual compatibility of the action of the actor/perceiver with the affordances of the situation". Here we are very far from the veridicality requested by genuine naïve realism.

More important is the picture that Leahr offers to us of Gestalt psychology. It is well known that in Gestalttheorie there was a strong Spinozian attitude. For instance, Wertheimer (greatly impressed by Spinoza's Ethica from childhood on: see Luchins and Luchins, 1982) remained all the life in this orientation. So, we can speak in terms of indifferentism about the problem of representation. in general, Gestaltist isomorphism has to be considered a variant of psychophysical parallelism (see Boring, 1942, 1950, mainly Ch. 13; for a recent survey of this issue, see Luchins and Luchins, 1999). But almost the same could be said about almost all other Gestalt psychologists. Lehar quotes extensively Koehler. But Koehler too never said that "the world we see around us... (is) ... generated by neural processes in our brain". Köhler, indeed, was in some instance a little ambiguous on this topic (for instance, Köhler, 1969). But he was absolutely clear when he had to address directly the mind-body problem. He conceived the Gestalt position as a variant of parallelism (Köhler, 1960, pp: 20-21), and said: "The thesis of isomorphism as introduced by the Gestalt psychologists modifies the parallelists' view by saying that the structural characteristics of brain processes and of related phenomenal events are likely to be the same" (italics added).

Lehar (p. 16), quoting Köhler (1969), insists that the isomorphism required by Gestalt theory is not a strict structural isomorphism, but merely a functional isomorphism". But Köhler always spoke of structural isomorphism. He was very clear in stating (f. i. Köhler, 1940, Chs. 2 and 3) that the processes that run in our brain do not have any necessary correlate in our phenomenal experience. what is structurally identical is their interaction with what happens in bordering areas of the brain and the interaction that there is in the phenomenal field; their dynamics, and the dynamics of the phenomenal field.

The structural identity between phenomenal world and physiological processes doesn't imply any causal relationship between the two levels. It means only that we are made up of one and only one matter. The physical laws that rule the matter lead to structurally identical outcomes, when we consider the phenomenal level as well as the physiological one. In this sense, the Gestalt psychology is neither representationalist, nor antirepresentationalist: it is indeed indifferentist.

The main limit of Lehar`s model derives in my opinion from this misunderstanding. His computational model, as I can assess it, works perfectly for a world, which is organised in terms of soap bubbles (Koffka's metaphor, 1935, used too by Attneave, 1982). A soap bubbles world is in Gestalt terms a world in which the forces of the perceptual field tend to dispose themselves as to make an outcome that is maximally good, pregnant in the sense of ausgezeichnet. In Lehar's model, this happens at phenomenological as well as at neurophysiological level. The fact is that, as I believe to have demonstrated with Kanizsa (Kanizsa and Luccio, 1986, 1990), doesn't exist a tendency of this kind in perception. These tendencies are instead well present in thinking, in memory, in all that Kanizsa (1979, Ch. 1) called "secondary processes", to distinguish them from primary processes of perception. But they are beyond the scope for which the concept of isomorphism is here interesting.

Back to Top

William A. MacKay

The Unified Electrical Field

Abstract

The electrophysiological perspective presents an electrical field that is continuous throughout the body, with an intense focus of dynamically structured patterns at the cephalic end. That there is indeed an isomorphic mapping between the detailed holistic patterns in this field and perception (at some level) seems certain. Temporal binding, however, may be a greater challenge than spatial binding.

The independent processor model of individual neurons has given rise to the widespread impression, echoed by Lehar, that neurophysiology fails to deliver a unified basis for the holistic properties of perception. If there is any 'illusion' it is not in the unity of perceptual awareness, but in the portrayal of physical separation by techniques such as extracellular recording and fMRI. Overlooked is the axis of continuous activity stretching from the spinal cord to the cerebrum. The tonic activity in the brainstem activating systems (cholinergic, serotonergic and noradrenergic) plus the histaminergic activating system of the hypothalamus, is responsible for our state of (un)consciousness (Pace- Schott & Hobson 2002). All sensory and motor activity feeds into this axis and influences the general distribution of activity. Also, the activating systems can directly trigger synchronization of activity within the cerebral cortex (Munk et al. 1996).

Furthermore, it is extremely doubtful that action potentials are of much significance in the direct link to perception. They are far too fleeting. It is the more sustained membrane potentials that are likely to correlate the best. Discrete neuronal activity in the brain, however isolated it may appear, is simply a local distortion in an unbroken continuum of electrical flux. All cells produce membrane potentials, even if static, such that an electrical field encompasses the entire body. The 'panexperientialism' view would also suggest that perceptual awareness is linked to something like an electrical field. This is the only obvious property that is shared by both the atom and organism, and is increasingly elaborated as one ascends to the organism. One might postulate that the higher the degree of complexity in the electrical field, the higher the level of consciousness experienced. Using fMRI it can be seen that the same cortical areas are active whether a stimulus is perceived or not. The difference in the case of perception is that the level of activation is greater (Moutoussis & Zeki 2002). This could mean that either more neurons are depolarized within the given area, or the same synapses are active but at a higher frequency, or both.

Neurons and their attendant glial cells manipulate membrane potentials like no other part of the body. This is their 'game'. Many attributes of neuronal electrical activity extend the range of information coding. Not a single one of them is the essence of conscious perception, but collectively they can raise (or lower) the level of consciousness. Spike synchrony is unquestionably relevant. For example, Riehle et al. (2000) have shown that unit pairs in motor cortex synchronize activity to a very significant degree exactly at the moment of an expected signal. However synchrony is not essential for 'binding'. In area MT, Thiele & Stoner (2003) recorded from pairs of units, one preferring the direction of motion of one visual grating, and the other preferring another grating direction. The units did not usually synchronize activity when the gratings were perceived as moving together in a coherent plaid. Synchrony elicited by coherent plaids was the same as for non- coherent ones. Again it is probably not spiking activity per se that is ultimately important, but the associated changes in membrane potential and possibly phenomena such as depolarization fields manifested in superficial layers of cortex (Roland 2002).

The various states of Lehar's Gestalt Bubble model can easily be construed as hypothetical neuronal feature detectors. One could not ask for a better set of discriminators of planar properties in depth, and I suspect that something very similar lurks somewhere in the association areas between V1 and inferotemporal cortex. The transformation from a 2-D image on the retina to a 3-D percept would follow a process as outlined by Lehar when the stimulus is an everyday, familiar experience with established expectations. For any unfamiliar object, whether presented to the eye or hand, exploratory movement is requisite to clarify ambiguities. Here Lehar is correct to emphasize the translation/rotation invariance of the perception, divorced from the motion of the explorer. The target is perceived as it relates to its environment external to the viewer. This is the essence of the great transformation from egocentric (parietal cortex) to allocentric representation (presumably in hippocampus or prefrontal cortex). The constancy of the percept over time as another data sample is added with each exploratory movement is also rightly highlighted.

It is essential that perception integrate over time as well as space. Even within one sampling episode, different sensory attributes such as color, and motion, are processed at slightly different times although they are perceived as a unity. Thus Zeki & Bartels (1998) postulate the existence of multiple 'micro-consciousnesses' in the brain which are asynchronous with one another. This raises the problem of how they are integrated. A simple possibility is that everything processed within a finite window is integrated, just as two colors flashed within less than 40 ms are blended together. But it cannot be that simple because haptic exploration of an object can continue for hundreds of ms.

Figure-ground designation also involves time constraints. Neurons in inferotemporal cortex which are selective for shape, maintain that shape preference when light-dark contrast is reversed (negative image) but not when a figure-ground reversal is made. Just as the perception of shape depends on whether a visual region is assigned to an object or background, so the visual analysis of form depends on whether a region is perceived as figure or ground (Rubin 2001). One cannot relegate the problem of resolving border- ownership of edges to earlier stages in the visual stream. It occurs quickly, within 10-25 ms of response onset and really requires feedback from higher cortical areas. Thus it is an instantaneous, holistic decision of the entire visual system, presumably selecting the most probable choice.

Back to Top

Slobodan Markovic

The Soap bubble: phenomenal state or perceptual system dynamics?

Abstract

The Gestalt bubble model describes a subjective phenomenal experience (what is seen), without taking into account the extra-phenomenal constraints of perceptual experience (why it is seen as it is). If it intends to be an explanatory model then it has to include either stimulus or neural constraints, or both.

While presenting the theoretical background of his approach, Lehar attempts to keep a critical equidistance toward both indirect and direct realism. However, instead of a radically new approach, he offers a combination of some constructivist and some Gibsonian premises. On the one hand, like many constructivists (e. g. Gregory, 1971; Hochberg, 1978; Marr, 1980; Rock, 1983), Lehar adopts a representational paradigm which defines perception as a subjective conscious description or as an internal virtual copy of the external world. On the other hand, inconsistent with the constructivists' perspective and more similar to the views of proponents of direct realism (e. g. Gibson, 1979; Shaw & Bransford, 1977; Show & Turvey, 1981), Lehar does not postulate any mediating mechanisms that process the representations within a perceptual system.

Moreover, Lehar's exact position concerning the question of direct perception of distal objects is not quite clear. At one point he explicitly claims that "the internal perceptual representation encodes properties of distal objects rather than of a proximal stimulus". At another point he states that "the direct realist view is incredible because it suggests that we can have the experience of objects out in the world directly, beyond the sensory surface, as if bypassing the chain of sensory processing". Why would the thesis that distal objects are mapping onto the phenomenological domain without neural intervention be incredible and mysterious, while the idea about the projection of internal representation onto the external perceptual world not be incredible and mysterious? How is it possible that perception is partially indirect (representational), and partially direct (distally oriented)?

In his criticism of neurophysiologic modeling, Lehar rejects not only the classical Neuron doctrine, but also some recent holistic approaches (cf. Crick & Koch, 1990; Eckhorn et al., 1988; Singer, 1999). For Lehar, hence, atomism is not the greatest shortcoming of neural models, but rather the problem of neuro-phenomenal decoding. That is, how can a fully spatial (topographical) perceptual description be created from spatially less constrained (topological) or even from completely abstract, symbolic and non-spatial neural representation? I find that this epistemological question is a natural consequence of a hidden ontological dualism: how does one domain of reality (consciousness) know how to read and understand the codes coming from the other (neural) domain.

To paraphrase Koffka (1935), the ultimate task for perceptual science is to answer why things look as they do. In the case of Lehar's theory this question might be formulated as the following: why is the phenomenal volumetric space such as it is? Why is it non- linear in a particular way? Implicitly, he proposes that this is an intrinsic property of phenomenal space which is not in a causal relationship with any other domain of reality. My opinion is that without the precise specification of the extra-phenomenological aspects of perception, such as the stimulus and neural domains, it is difficult to answer the question related to why the percept looks as it does. For instance, imagine the difficulty in explaining the path shape and velocity of the planet Earth's motion without taking into account the mass and motion of other cosmic objects (moon, sun, other planets, and so on). A description of the Earth's motion is not an explanation of its motion.

Even Gestalt psychologists, who widely utilized the phenomenological method, did not create pure phenomenological explanations of perception. For instance, Koffka (1935) used the soap bubble metaphor not to describe some phenomenal bubble-like experience, but to point out some basic principles of perceptual (neural) system functioning. Attneave (1982) also used the metaphor "soap bubble system" to describe the economy of perceptual system behavior. Like the soap bubble which tries to enclose the largest volume within the smallest surface, the perceptual system tends to reduce the global spending of energy (entropy, minimum tendency) while at the same time striving to increase its effective use (dynamics, maximum tendency) (cf. Köhler, 1920, 1927; see also Hatfield & Epstein, 1985; Markovic & Gvozdenovic, 2001).

Back to Top

Niall McLoughlin

Bursting the Bubble: Do we need true Gestalt isomorphism?

Abstract

Lehar proposes an interesting theory of visual perception based on an explicit three- dimensional representation of the world existing in the observer's head. However, if we apply Occams Razor to this proposal it's possible to contemplate far simpler representations of the world. Such representations have the advantage that they agree with findings in modern neuroscience.

Lehar proposes to model visual perception using his subjective visual experience as his source of data. He proposes a perceptual modelling approach since "conventional concepts of neural processing offer no explanation for the holistic global aspects of perception identified by Gestalt theory". This allows him to conveniently ignore current research in visual neuroscience while concentrating on the central issues of the representation of the visual field and of our subjective visual experiences. As he correctly points out the world we see and experience surrounding us exists only as nerve impulses within our head. Lehar proposes that since our subjective experience of the world is that of a high-resolution three-dimensional volume, and since this representation must exist in our heads, it must therefore be some form of a high-resolution three-dimensional structure. However this does not necessarily follow. For example on a computer system it's possible to generate a sparse representation of the world into which it is placed so that the computer could interact with objects in the world in a meaningful manner. Objects could be represented at tokens at such and such x,y, and z location and so forth. There would be no explicit representation of empty space within this sparse representation. Who's to say what the subjective experience of the computer might be?

There is no doubt that my subjective experience of the world is that of a three- dimensional solid environment that I perceive in equal detail in all directions. Yet as visual scientists and practiced observers we know that this is patently not the case. Each of our eyes responds to incoming photons in a non-uniform manner and this non- uniformity is father exaggerated in the cortex. The over-representation of the fovea is magnified between the retina and cortex and the multiple inter-connected cortical regions amplify this distinction even further. Most naive observers are surprised to discover that they have a fovea and amazed that they have a blind spot in each eye. How do we fool ourselves?

The very fact that we are genuinely fooled (until we make careful observations) calls into question the use of subjective experience as the basis for theories of visual perception. Furthermore, while the Neuron Doctrine is indeed the foundation for most modern neuroscience research I refute the notion that this doctrine implies purely feedforward models of neurocomputation. Certainly recent findings in both neuroanatomy (e.g. Bosking et al., 1997; Angelucci et al. 2002) and neurophysiology (Kapadia et al. 2000; Levitt & Lund 1997) emphasise the roles played by feedback and lateral connections in visual processing. Likewise a number of popular modern computational theories make use of feedforward, feedback and lateral connections (for example Grossberg 1994). If a Gestalt Bubble Model subserves perception then why do we have so many visual areas each containing a retinotopic map of visual space?

Is there any evidence for gestalt-like processes at work neurophysiologically? Recent electrophysiological recordings from as early at the lateral geniculate and V1 have found interactions well outside the classical receptive field (for example Blakemore & Tobin 1972; Felisberti & Derrington, 2001; Jones et al., 2000, 2001; Kapadia et al., 2000; Levitt & Lund 1997; Stettler et al. 2002; Solomon et al., 2002). While the source of these interactions (whether they are mediated by feedback, or lateral connections) remains to be elucidated it's clear that many aspects of grouping, completion and emergence may well arise from such non-local interactions. In addition, recent neurophysiological studies in the primate (e.g. Livingstone & Hubel, 1988) suggest that different aspects of a visual scene are represented primarily in different visual streams and areas. While there is some disagreement as to the amount of segregation of function numerous neuropsychological studies in humans back up the suggestion that multiple representations exist for different attributes and/or functional roles. One such patient, studied by Humphrey and Goodale (1998) suffered from visual form agnosia (Farah, 1990). She was unable to discriminate between visual forms let alone recognize her friends and family yet her color vision was close to normal and she could recognise shapes when placed in her hands. Such case studies suggest that the brain encodes the external world using multiple representations, each one perhaps subserving a different role or task, rather than a single isomorphistic one.

What Lehar seems to have forgotten is that the high-resolution representation is generated only when we pay attention to the input and focus our eyes on the object or texture under inspection. We need not represent even our immediate environment in high resolution unless we need to interact directly with it. Why waste time and space representing the world in vivid detail when we only interact with a small part of it at any one time? Surely our central representations should be goal directed. We can always direct our vision to different locations in a scene to find out what's there and given that most useful scenes are dynamic why waste effort representing space in high-resolution when it's constantly changing? O'Regan (1992) argued along a similar line when he suggested that "seeing constitutes an active process of probing the environment as though it were a continuously available external memory"(p. 484). He suggests that seeing does not involve the reification of a three-dimensional spatial representation of the external world in the observer's head but rather depends on ones ability to interrogate the environment through directed eye movements. It may well be that we have a fuzzy three-dimensional representation of the external world in our heads that we use to help direct eye- movements but I remain to be convinced that we would need or want anything more complex. If we need the detail we look.

Back to Top

Axel Randrup

Relations between three-dimensional,volumetric experiences and neural processes: Limitations of materialism

Abstract

Certain features of perception, the quale red for instance and other qualia, must be regarded as additions to the materialist neurophysiological picture of perception. The perception of three-dimensional, volumetric objects can also be seen as qualitative additions to the neurophysiological processes in the brain, possibly without additions to the information content.

In the history of science and philosophy the world has been regarded as material, mental (idealist philosophy), or dualist (both material and mental). Like many people today Lehar has chosen the materialist view, and he attempts to avoid dualism by assuming the mind-brain identitiy position ("consciousness is a physical process taking place in the physical brain"). Still he writes, that there remains a subjective quality (or quale), to the experience of red for example, which is not in any way identical to any physical variable in the brain. I think, this must mean, that the experience of qualia adds something to the assumed material world, and that Lehar therefore does not stay consistently within the materialist frame of reference. Lehar also writes (section 2.3), that sense data, or the raw material of conscious experience, are the only thing, we can know to actually exist, and that all else, including the entire physical world, is informed conjecture based on that experience. To me this statement appears as a departure from materialism; actually it is close to the idealist view.

I now suggest, that also the perceptual experience of three-dimensional, volumetric objects and of empty space is something, that "subjective conscious experience" adds to the assumed material electro-chemical processes in the brain, possibly without changing the information content, a qualitatively different representation. Lehar thinks, that the gap between the materialist descriptions of neurophysiology and the phenomenological descriptions of Gestalt features of perception may be due to the present "embryonic" state of neurophysiology, but I regard this as a promissory belief rather than an explanation.

Analogously (and staying within the materialist frame of reference) I believe, that a computer can produce a three-dimensional, volumetric figure, namely, if it is connected with a device that can construct that figure. The figure will then be another representation of the information content which inside the computer is represented by electrical processes. Of course a human person can also construct a three-dimensional figure with his hands or describe it in words and drawings, as Lehar does. In this case it is the connection with the body, particularly with the muscles and the hands, that enables the brain to make these constructions and descriptions from its information content.

I think, that materialism has served science well within a rather large domain, but with studies of cognition such as Lehar's we move into a domain, where materialism reveals significant shortcomings. I find, that such shortcomings appear in Lehar's work. Thus, on his materialist background Lehar rejects direct (naive) realism which suggests, that we can have experience of objects out in the world directly, as if bypassing the chain of sensory processing. Provided that the materialist background is retained, I agree with this rejection. But if we apply an idealist world view, our perceptions are of course experienced directly, and based on these perceptions we form concepts such as the concepts of a "material" object, a "material" world, and perceptual models such as Lehar's Gestalt Bubble model. I see these concepts and models as mental constructs representing features of the perceptual reality such as quantitative features and three-dimensional Gestalt features. These constructs are of course also experienced directly, and they can be made unambigous and precise. Here I agree with Lehar who thinks, that perceptual models remain "safely on the subjective side of the mind/brain barrier", and writes about "objective phenomenology" leading to "perceptual modeling" (section 4). It is, when we accord the "material" concepts a special existence of their own, principipally different from the existence of conscious experiences, i.e., when we move to materialism, that we run into the troubles with direct realism.

Lehar finds troubles with indirect realism as well but eventually accepts this view on the premise, that the world, we see around us, is not the real external world, but a miniature virtual-reality replica, an internal data structure within our physical brain. I think, that this view gives only an incomplete, imprecise conception of the "external world" including our "physical brain". This incompleteness and imprecision is shared with other philosophies assuming indirect realism such as "hypothetical realism" (Löw 1984; Randrup submitted; Wuketits 1984), "common sense realism" (Ruse 1986), and Kant's concept of "the things in themselves" versus "the things for us". According to Kant's philosophy we know actually nothing about the things in themselves, except that they are supposed to exist. I think, that this uncertainty or renunciation of knowledge compares unfavorably with the precision of the "material" concepts based directly on perceptual data in the idealist world view.

It is also a shortcoming of materialism in relation to the study of cognition, that it is difficult to consistently avoid dualism, as it appears from Lehar's statement about qualia quoted above. And if dualism is admitted, it is hard to see, how conscious experiences can be generated by material processes in the brain, as Lehar thinks, they are (section 2.4). In the alternative idealist view of the world it is not so hard to see, conversely, how "material" concepts are generated by the mind; the history of science shows, how such concepts have been created (e.g. quanta, superstrings) or deleted (impetus, phlogiston, the ether) following the advent of new perceptual (observational) experiences.The special material type of exstence is not a part of the idealist philosophy. For a more extensive discussion of the mind-matter and mind-brain problems in relation to cognition, reference is made to a few recent publications (Knight 2001; Randrup 1997; 2002).

Back to Top

Antti Revonsuo

Consciousness as Phenomenal Ether?

Abstract

The Gestalt Bubble model of visual consciousness is a courageous attempt to take the first-person's perspective as primary in the study of consciousness. I have developed similar ideas as the Virtual Reality Metaphor of consciousness (Revonsuo 1995, 2000). I can thus only agree with Lehar about the general shape of a proper research startegy for the study of consciousness. As to the metaphysical basis of the research program I have however several reservations about panexperientialism.

I agree with Lehar on several points, but disagree about the ultimate metaphysical nature of consciousness. I shall first describe points of agreement, and then proceed to a criticism of panexperientialism.First, any research programme on consciousness should start by taking the explanandum seriously, constructing a systematic description of it. This is Lehar's "objective phenomenology". In the context of the biological sciences, this is the initial, descriptive stage of inquiry. All branches of biology have begun with the descriptive stage, and the study of consciousness should be no exception.

Second, in the study of consciousness the top-down approach should be of at least as much importance as the bottom-up approach. Once we have a detailed description of the structure, organization and dynamics of a higher level of organization (in this case subjective phenomenology), it will impose significant constraints on the possible lower- level (neural) mechanisms that could account for the higher level features. The lower- level mechanism must be capable of supporting exactly the kind of structure, organization and dynamics as is found at the higher level of phenomenology, otherwise the proposed mechanism is not a plausible candidate to explain the phenomenon. The bottom-up strategy is important too, but it should be combined with the top-down strategy. Otherwise bottom-up approaches may lead to either the elimination of consciousness (because it is so difficult to see how single-neuron activity could add up to holistic features of consciousness), or to the search for the mere neural correlates of consciousness (rather than the directly underlying constitutive mechanisms that explain the phenomenon), because the signals that are collected from the brain usually originate nowhere near the higher levels of organization where consciousness itself resides (Revonsuo 2001).

Third, indirect realism as a theory of perception seems to be the only alternative that can give a plausible explanation of dreams and other hallucinations. Dream experiences show that the brain in REM sleep can bring about a fully convincing simulation of the perceptual world, and a simulated self embodied inside this virtual world. Dreams are temporally progressing "being-in-the-world" -experiences generated inside the brain. During dreaming, phenomenal consciousness is causally isolated from the stimulus environment, from the concurrent state of the physiological body, and from behavioral output systems. As I have argued in my previous BBS commentaries on Pessoa et al. (1999) and O'Regan & Noë (2001), their theories of visual consciousness cannot account for our vivid visual experiences in dreams.

Although I thus largely agree with Lehar as to what the proper approach to the study of consciousness should be, there is one core issue on which we seem to have differing views. His fundamental metaphysical commitment is panpsychism (or panexperientialism), according to which (a simple form of) consciousness is a fundamental property of physical matter. According to this view there is no radical discontinuity between any physical systems as to the possession of consciousness; it is just a matter of degree. Everything is more or less conscious; simple physical systems to a lesser degree, the human brain perhaps to the highest possible degree. This smooth continuum of consciousness across all physical entities is supposed to have the following explanatory strengths: (1) Consciousness is a fundamental property of physical matter and therefore need not be explained in terms of (nonconscious) physical matter, (2) There is no radical conscious/nonconscious dichotomy to be found anywhere in the natural order (e.g. in phylogeny or ontogeny).

This approach raises some severe problems. There are clear, well-demonstrated dichotomies between the presence and the absence of the state of consciousness (caused by anesthesia, epileptic seizures, fainting, coma) and between the presence and absence of particular contents of consciousness even though the stimuli are implicitly processed (as in blindsight or neglect). Any theory of consciousness should be able to explain these radical subjective differences between the presence and absence of consciousness. The panexperientialist is however forced to say that these are not really cases where the presence and total absence of consciousness in the brain could be strictly contrasted. The contrast is only between primitive and more sophisticated forms of consciousness. According to the panexperientialist the primitive form may be something so simple that we would hardly recognize it as consciousness at all. Thus, what we thought was the total absence of experience is actually the presence of a primitive form of experience; we just cannot recognise it as experience.

Unfortunately this move will not help us to understand the radical contrast between the presence and absence of conscious experience in the above cases. Regarding everything as conscious (to some degree) does not remove the radical conscious/nonconscious contrasts. In fact it leads to a position as difficult (but the exact opposite) as the eliminativist position defended by Dennett. If we take either the panexperientialist position that phenomenal consciousness is everywhere in the world, or the eliminative position that it is nowhere, we are no closer to explaining the radical empirical differences that we want to understand.

Furthermore, panexperientialism smacks of a misuse of the concept of experience. It is difficult to see why the postulated "primitive form" of consciousness - which we might not even recognize as experience - should be placed in the same category with our vivid phenomenal experiences. There seems to be no clear idea what "proto-consciousness" could be, whether it exists at all, or how the claims for its existence could be empirically tested or theoretically modelled, and how exactly the primitive form of consciousness relates to our ordinary vivid phenomenal consciousness.

Thus I do not regard panexperientialism as an advisable metaphysical commitment for a research program on consciousness. I would rather postulate that the sphere of subjective experience is a higher level of biological organization in the brain. Phenomenal experience only exists at that level, and in those creatures whose brains can realize that level. Otherwise the physical universe is devoid of phenomenal consciousness. When we totally lose consciousness, as we do during anesthesia, for example, our brain is temporarily incapable of supporting the phenomenal level of organization. The radical difference between the presence and the absence of phenomenal experience is to be described and explained in terms of biological levels of organization in the brain. Physical matter at lower levels of organization perhaps may be said to contain the potentiality of being conscious, but only in the weak sense in which all physical matter contains the potentiality to be alive. The mere potentiality does not make simple physical systems (say, carbon atoms or diamonds) alive, and it would be a waste of time to study the microphysical structure of diamonds in order to understand the biology of living systems. In a similar vein, I fear that the assumption that all physical systems (diamonds, toothbrushes, bacteria and so on) are conscious (or "proto-conscious") is going to be a useless, untestable hypothesis for the science of consciousness.

Back to Top

Victor Rosenthal & Yves-Marie Visetti

Gestalt bubble and the genesis of space

Abstract

Lehar (rightly) insists on the volumetric character of our experience of space. He claims that three-dimensional space stems from the functional 3D topology of the brain. But his 'Gestalt Bubble' model of volumetric space bears an intrinsically static structure - a kind of theatre, or 'diorama', bound to the visual modality. We call attention to the ambivalence of Gestalt legacy, question the status and precise import of Lehar's model, and the phenomenology that motivates it.

Lehar should be applauded for making strong case for the fundamental character of volume and depth in our experience of space. The originality of his proposal resides inter alia in the radical claim that three-dimensional experience of space stems from the functional 3D topology sustained by the human brain (not to be naively equated with brain topography). He posits that subjective spatial experience criterially requires a 3D topological substratum - a device lacking 3D topological-dynamical structure could never account for volumetric experience of space. In other words, the only viable option for a functionalist indifferent to brain physiology is 3D topological-dynamical functionalism.

Lehar depicts his model as an outgrowth of Gestalt tradition. Indeed one can easily recognize two essential features of Gestalt theory: its phenomenological approach to subjective experience, and the postulate of psychophysical isomorphism. Phenomenological space, its emergence, and its scientific explanation as a brain process, are, according to Lehar, grounded in pre-given, continuous and coherent topology, specifically a 3D functional topology.

Lehar may not be aware that the way Gestalt psychologists treated space was in reality quite equivocal. Although they were in principle cognizant of the fundamental status of volumetric space, they granted it low priority in their scientific agenda, and tended 'provisionally' to treat space as a series of transparent/opaque surfaces, if not as ambient ground against which to set a figure. On the other hand, it is true that Köhler's theory of psychophysical isomorphism explicitly referred to 3D functional brain topology to construe not only 3D geometrical static structures but also 2D structures evolving in time (see Koffka, 1935). The theory combined empirical and phenomenological constraints with speculative brain physics (e.g. the theory of cortical fields) so as to represent both brain process and phenomenological experience in a single dynamical scheme (see Rosenthal & Visetti, 2003).

Several attempts have been made to model Gestalt principles of perception in accordance with neurophysiology, and in particular with the doctrine of neural coding (e.g. of perceptual micro-features). For instance, models of neural fields or neural repertoires feature a 2D functional topology which corresponds to a topographic 2D arrangement of units in primary areas (e.g. retinotopy) (e.g. Hoffman, 1989; Koenderinck, 1990, see Petitot, 1999, for a review). Less discrete models, unconstrained by brain physiology, developed in the context of image processing and sometimes resorted to fairly complex mathematics, but maintained set to a bi-dimensionality of their input (retinal or pictorial, see Morel & Solimini, 1995). The very idea of 3D functional topology was hardly taken into consideration in the few attempts to account for depth (e.g. Grossberg's), which thus had to resort to hosts of specialized coding units: a patently implausible solution, as Lehar rightly noted.

The solution advocated by Lehar is original and certainly deserves attention. He defines a 3D topological milieu where any local element can be in one of four states (corresponding to local surface elements). Each individual element (or point in a perceptual matrix) exerts a field influence on adjacent elements for them to take on a similar state (or to be prevented from this by inhibition). Reciprocal determination between surface elements is assumed to generate equilibrium in which the relevant features are stabilized. The input to the model is an image set in frontal plane (much like a retinal image). The output (actually the first step in 'geometrization' of space) is a distribution of geometrical surface micro-features in a 3D space. Although Lehar does not mention this issue, one can readily deduce that unit formation or individuation is assumed to take place in this 3D visual matrix. The originality of this proposal should be highlighted: while the majority of rival models first individuate 2D units (from 2D image input), then categorize them as faces of 3D units, Lehar sets his 3D structure ab initio, and whatever is to populate this 3D distribution of geometrical micro-features supposedly comes next.

It isn't clear, however, which scientific question Lehar has set out to answer. He doesn't seem to attempt another perspectival reconstruction of the visual field, for his model, in contrast to its alleged purely phenomenological motivation, builds on a physicalist metaphor. Although Lehar dismisses neurophysiological concerns, the analogy between his model and neural net models jumps to the eye: traditional 'neurons' with their receptor fields are replaced by elements or points in perceptual matrix, and neural connections are supplanted by fields of influence. Moreover, Lehar alludes to the possibility that the model take a discrete or granular form (see Figure 7A). Why then, does he hammer so loudly his physicalist credo? It seems that Lehar believes that the process by which space is constituted necessarily sheds light on the way we perceive space. Then, why doesn't he try to motivate his model genetically? Clearly, Lehar needs to tell us the rules of the scientific game he plays more explicitly (does he want to model the constitution of space from a purely phenomenological viewpoint, or does he attempt a free mathematical reconstruction of subjective experience?).

Lehar could have mentioned that during the past century other theorists put forth elaborate proposals concerning the constitution of space experience (e.g. Husserl, 1997; Poincaré, 2001; Gibson, 1950). Instead of sticking to neurophysiology, they referred to the structure of the organism or the lived body. These were strongly dynamic, sensori- motor 'models' of constitution of phenomenological space, which assumed a multi-modal origin of volumetric space and explicitly related its dimensionality to repertoires of self- generated movements. Although none of these 'models' can be regarded as fully effective, they account for the ontogenesis of space in a dynamic fashion, and, for a variety of phenomena of adaptation (e.g. to distorting goggles). We suggest that considering the dynamics of genetic, multi-modal and sensori-motor character of the constitution of space is as important in modeling perceived space as neurophysiology and the kind of static geometry on which Lehar elaborates. What comes along with such dynamics is the constitutive relationship between external and bodily space. Lehar appears to be aware that perception of space involves one's own body, but instead of taking this as a constitutive relation he treats the body as just another object in space.

Back to Top

Helen Ross

Back to Top

James A. Schirillo

Spatial Phenomenology Requires Potential Illumination

Abstract

Collapsing three-dimensional space into two violates Lehar's "volumetric mapping" constraint and can cause the visual system to construct illusory transparent regions to replace voxels that would have contained illumination. This may underlie why color constancy is worse in two-dimensions, and argues to Lehar revise his phenomenal spatial model by putting "potential illumination" in empty space.

Lehar's phenomenological description of space neglects the fact that empty space is actually full of illumination. For example, if a cast shadow crosses half of this page and you move your finger from a word under shadow to one under full illumination you are not surprised when your finger crosses the shadow, even though your finger is closer in depth than the page. This is because every voxel between your eye and the page contains some amount of light. It is unfortunate that Lehar overlooks this fact since he correctly asserts that depth information is volumetric, while current neurological models fail to "...represent transparency, with multiple depth values at every single (x,y) location, or to represent the experience of empty space between the observer and a visible object" (page 12). These same models also ignore that every voxel of "empty" space contains light of some intensity and chromaticity.

This confusion probably results from naively accepting the popular notion that humans care only about the location and qualities of objects, making the perception of illumination irrelevant. This assumption is so prevalent that much of color research is devoted to determining how the visual system "discounts the illuminant". However, a viable solution to the Gestalt problem of color constancy will only emerge with a more complete description and understanding of how we subjectively experience illumination. Ironically, Lehar's aspiration to describe the subjective experience of spatial vision in terms comparable to those of color vision reveal that current color vision research is also in peril. That is, he claims that color phenomena are reducible to hue, intensity and saturation because that is how the brain represents them physiologically (page 7). Yet models of hue, intensity and saturation cannot be the "primitives of raw conscious experience" (page 13-14), in that these qualities remain invariant as illumination changes across space.

This confound is apparent when Lehar discusses Figure 1 as containing "...explicit volumes, bounded by colored surfaces, embedded in a spatial void" (page 15), where "...every point can encode either the experience of transparency, or the experience of a perceived color at that location" (page 21). His more accurate intuition is that there are also intermediate states between transparent and opaque "...to account for the perception of semi-transparent surfaces" (page 35). I suggest Lehar consider filling these semi- transparent voxels with "potential illumination" "at multiple depth values at every single (x,y) location". This would also strengthen his second and third conclusions that "volumes of empty space are perceived with the same geometrical fidelity as volumes of solid matter" and that "multiple transparent surfaces can be perceived simultaneously" (page 55). Having semi-transparent voxels contain "potential illumination" is a more parsimonious description of the void between your eyes and this page. You can actualize the "potential illumination" of these voxels by placing your finger in front of any shadow cast on the page. More accurately, Lehar's phenomenological model allows only the plane of voxels directly in front of a given surface to contain cast shadows (i.e., less illumination), since the voxels that compose the surface must be the color of the opaque surface itself (page 36).

Note that this concept is not merely peripatetic (Aristotle, 1976), or an ether explanation, in that we are always subjectively aware of the illuminant. For example, by looking from your illuminated reading room into a dark hallway, your subjective experience is not only that the hallway walls are under less illumination but also that the space itself contains less light. In this way, "potential illumination" can also address why color constancy differs in two- verses three-dimensional scenes. For example, Gilchrist (1977) had observer's look through a pinhole into a room containing a doorway into a second room. The near room was dimly illuminated and the far room was highly illuminated. Attached to the doorframe were several papers, arranged so that a mid-gray paper appeared either adjacent to the doorframe, or (with its corners removed) on the far room's back wall. The lightness of the paper shifted in the direction of lightness constancy depending on whether it appeared on the doorframe or on the far wall. Schirillo et al. (1990) generated equivalent stimuli in two-dimensions using a CRT and stereoscope, yet their replication produced only a fraction of Gilchrist's constancy. I hypothesize this occurred because stereoscopic space does not contain the actual voxels of either high (e.g., near room) or low (e.g., far room) illumination. In essence, Schirillo et al. failed to preserve Lehar's necessary condition of "volumetric mapping" (Figure 1D, page 16).

Back to Top

Peter Ulric Tse

Back to Top

Max Velmans

Back to Top

Edmond Wright

Back to Top