Richard Bright: Can we begin by you saying something about your background?

Tyler Burge: I was an undergraduate at Wesleyan University, Connecticut, and a graduate student at Princeton University, graduating with a Ph.D. in 1971. I began teaching at UCLA in 1971, and have remained. Since 2015, I have been the Flint Professor of Philosophy at UCLA. I have taught, as visitor, at MIT, Munich University, Bayreuth University, Stanford University, Harvard University, University of Bologna, and University of Zurich. I have written in areas of philosophy that I see as closely connected: philosophy of psychology, philosophy of mind, epistemology, philosophy of language, philosophy of logic, and history of philosophy– principally Descartes, Kant, and Frege.

RB: Have there been any particular influences to your ideas and work?

TB:  The philosophers whom I have worked on in the history of philosophy were certainly influential. Among more recent philosophers, Hilary Putnam (deceased) and Saul Kripke (just deceased) have probably influenced me most. A further source of influence has been the science of perception. There are too many fine scientists here to name. I single out David Marr (deceased) and Wilson Geisler.

RB: What is the focus of your research?

TB: Centrally, philosophy of mind and psychology and epistemology. Philosophy of psychology is philosophy that starts by reflecting on what is known in scientific psychology. I have focused mainly on the psychology of perception. For vision, the science studies how we and other animals see. It tries to discover the laws and causal processes that lead from stimulation of the eyes by light arrays to perceptual states that often represent things in the environment accurately. This area of psychology, psycho-physics, is the first mathematically rigorous, predictive, productive psychological science. With its cousin, sensori-motor psychology – it is far ahead of other areas of psychology on all important measures of the maturity of a science.

I have centered, for some years now, on what I call ‘lower representational mind’. This is the system of capacities grouped around perceiving in various modalities (visual, auditory, tactile, proprioceptual). For humans vision is central. Perceptual capacities guide action and are served by satellite capacities: perceptual memory, perceptual anticipation, perceptual attention, perceptual imagination, perceptual affect (primitive forms of emotion), and perceptual learning. I have reflected on how these capacities interface with the wider environment to provide information about it, on what representational forms the capacities take, and on how they interact with one another. In such reflection, I make central use of the science.

Lower representational mind contrasts with upper representational mind. Upper representational mind consists, principally, in capacities for propositional thought and propositional reasoning. These capacities include not only reasoning, but also propositional-level action-initiation, memory, anticipation, attention, imagination, affect, and learning.

So, the capacities that use and serve perception and thought come in different levels–the perceptual level and the propositional level. Propositional thought is thought that has the same form as a declarative sentence. A thought that Russia has invaded Ukraine is propositional, because the form of the thought is shared with the declarative sentence ‘Russia has invaded Ukraine’. I doubt that propositional thought is always linguistic (carried out in an inner mental language), although it may sometimes be. It just shares a form with declarative sentences. Perceptual-level states have a simpler, non-propositional form. Their form is more like that of a noun-phrase: that spherical body. This form of perception is shared with pictures and maps. The simpler form marks a simpler set of psychological capacities.

All of that is philosophy of psychology. How does it relate to epistemology – roughly, theory of knowledge? Our first knowledge as children, and the knowledge that the empirical sciences depend on for developing sophisticated scientific theories, are perception-based. A child cannot yet know science. But a child can know that that’s a moving object, or an edible object, or that that’s Mama. Such knowledge depends, both psychologically and for having the sort of support that knowledge must have, on perception. Epistemology is fundamentally a normative discipline: it tries to describe epistemically good ways of forming beliefs–ways that are conducive to having true belief and knowledge. But knowing what ways are epistemically good depends on knowing what ways are psychologically available. Thus, doing well-informed normative work in epistemology requires understanding the descriptive psychology of perception.

A lot of work in philosophy of perception tries to reflect on perception without knowing the science of perception, either ignoring it or using it in uninformed, merely decorative ways. A lot of work in epistemology tries to reflect on epistemic norms without knowing the science of perception. Both of these ways of doing philosophy are inevitably limited. They are too compartmentalized to provide deep, informed accounts of their subject matters. I try to do philosophy–both about the nature of perception and about norms for forming good beliefs based on perception – in a way that is informed by relevant science.

RB: Can you say something about your recent book, Perception: First Form of Mind, which focuses mainly on visual perception and vision science? What are its aims and structure?

TB: The aims of the book are three-fold. First, it lays out an account of what perception is, and how it contrasts with non-perceptual sensing and with thought. Second, using techniques from the semantics of language, it explains the representational form of perception–which is decidedly not linguistic. Third, it situates perception within lower representational mind–develops its relations to action-guidance, memory, anticipation, attention, imagination, affect, learning. It characterizes perception as the central capacity in a psychological system that includes all these other capacities. I pursue these three aims through close attention to what is known from the psychophysics of perception, principally visual perception – the type of perception about which by far the most is known.

The contrast between perception and non-perceptual sensing marks the difference between having a mind, in the sense of having distinctively psychological states, and not having one. Plants are sensitive to light and grow to maximize exposure to light. Bacteria, paramecia, snails, clams, and simple worms sense light; they use this sensing to move in ways that tend to benefit them. These types of sensitivity and sensing are non-perceptual. These organisms lack a psychology. Their states and processes are standardly explained, in science, in causal and functional terms that do not imply anything distinctively psychological. Slightly more complex animals–various insects–have capacities that are best explained, and are explained in science, in a different way. The capacities represent the environment accurately or inaccurately. Explanation by reference to states that can be accurate or inaccurate does not occur in botany or in the sciences of the simplest non-plant organisms. (Popular psychological glosses on such explanations do not line up with the actual explanations.) This different way of explaining an animal’s sensory capacities–as being capacities to get things right about the environment – marks different capacities, not shared with the plants, bacteria, paramecia, snails, clams, or simplest worms. Various insects are known to have perceptual states, explained in this different way. The first step in achieving this first aim of the book is explain wherein perceptual capacities differ from non-perceptual sensory capacities. The second step in achieving this first aim is to elaborate differences between perceptual capacities and thought. A capacity for thought is centrally a capacity to reason propositionally. I discuss other capacities that are, almost certainly supra-perceptual, and that are not propositional.

The book’s second aim is to elaborate in some detail the representational structure of perception. This structure constrains ways in which perception can be accurate or inaccurate. This is a relatively technical project. It is like doing a semantics and grammar not for language, but for perception. Broadly, it develops ways in which perception picks out entities that are perceived and characterizes some of their properties. For example, one might perceive (pick out) a body, and simultaneously perceive it as a body and as spherical (thus characterizing it). These two roles – picking-out and characterizing – are fundamental in perception. They are the only two representational capacities in perception, always operating together. They are, however, applied in rich, complex ways. If one sees a body, one sees a lot else besides: its surface, surface parts, texture, color, shape, size, orientation (how it is slated or tilted), distance and direction, spatial relations with other bodies, and so on. Developing how these matters are codified in a single perceptual state is non-trivial. There are literally upwards of billions of pickings-out and characterizations in any normal human visual perceptual state.

The third aim centers in perception’s relation to the other capacities: initiation of action, memory, anticipation, attention, imagination, affect, learning. Traditionally, most of these other capacities have been thought to be higher-level than perception, because they are farther from the “here and now”. Traditionally, it has been common to treat use of memory or imagination as in itself a type of thought, more advanced than perception. By looking closely at what is known about these capacities, and how they relate to perception, I show that they all have a level of representation that is in no way more advanced than perception. Usually, it is less so. This result motivates discovering and explaining wherein they are, with perception, part of a single system, and how this system differs from, yet interacts with, the higher-level capacities for thought, as well as the lower-level sensory capacities.

RB: In the book you use the phrases ‘the iconic nature of perception’ and the ‘mapping of perception’. What do you mean by this?

TB: ‘Iconic’ is a technical term. For present purposes, it applies to representation that relates to what it represents in a certain way. Iconic representations represent partly through natural, not-purely-conventional relations to their subject matters. For example, a road map represents roads through lines on the road-map. The lines bear various natural relations to the roads that they represent. The length of the lines is proportionate to the length of the roads. The shape and directions of the lines on the road-map correspond to the shapes and directions of the roads. The spatial relations among the lines on the road-map correspond to the spatial relations among the roads in the territory. And so on. This way of representing roads is iconic. It contrasts with describing in English the length, shape, directionality, and spatial relations among the roads. The descriptions in English bear a non-iconic relation to the roads, because the words in the descriptions bear no systematic natural (here, spatial) relations to roads. The roadmap represents iconically because of its reliance on natural spatial analogies between its representations and the roads. It is certain that iconic representation in perception, and even in communication among animals, preceded the largely non-iconic representation that shows up in most modern languages.

Perception is more like the road-map than the set of English sentences in the way in which it represents. The evidence for this point is complex. I will not go into detail here. But it consists partly in the fact that there are literally spatial layouts in the brain that correspond to spatial layouts in the environment, and in evidence that these brain maps affect the way in which perceptual states represent. In vision science, perceptual states are modeled as grids or matrices with map-like structure, rather than as a collection of sentences or purely conventional linguistic descriptions.

‘Mapping’ is a mathematical term. It applies to a systematic correspondence or correlation between aspects of a representation or representational state (a roadmap or a perceptual state, for example) and aspects of the environment. For example, there is a top-row, left-most-column position in a grid (or matrix of columns and rows) that might be the structure of a road-map or perceptual state. This position on the grid corresponds to (is “mapped to”) a corresponding a position in the environment. Other positions on the grid also map (correlate with) corresponding positions in the environment. The overall system of correspondences is called a mapping of the grid onto the environment.

A key feature of the book is to connect the idea that perception is iconic with the point, discussed earlier, that it involves those two capacities–picking out and characterizing. I think that perception, road-maps, and certain noun-phrases in language share a representational structure at a very high-level of abstraction. They are all complexes made up of pickings-out and characterizations. Perception and road-maps do this in iconic ways that differ from the ways in which picking-out and characterization occur in (largely) non-iconic languages like English.

RB: One of the chapters is devoted to ‘Perceptual Constancy’, which you describe as the ‘first mark of the representational mind’. Can you say more about this?

TB: A perceptual constancy is a psychological capacity to produce any one of a large range of perceptual states. Each state in the range has something in common with every other state in the range. Each state in the range responds to and represents a common environmental property or object. The states differ in that they derive from different, sometimes dramatically different, types of proximal stimulus. A proximal stimulus is a stimulus that immediately impacts an individual’s sensors.

For example, for human vision the primary proximal stimulus is an array of light just as it slams into the million sensors in each eye’s retina. For example, our visual systems are capable of responding to a given size of a surface, whether the surface is fairly far away or nearby. In far-away position, the proportion of the retinal stimulation by the surface will be small. In nearby position, the proportion will be large. Yet our visual system, and the visual systems of many other animals, automatically scale their representations so as to allow for the stimulus difference and to produce different representational states (differing because they are produced by different stimuli) that have in common an approximately accurate estimate of surface size. Similarly, a visual system can represent a given color (say, white) of a surface both when the surface is bathed in white light and when the surface is bathed in dark blue-ish light. The two conditions produce significantly different proximal stimuli, because the proximal stimuli are combinations of the surface reflectance (roughly, the color) and the illumination (roughly, the character of the light shining on the surface). The proximal stimuli provide all the current information that a perceptual state has to go on. Yet the visual systems of humans and many other creatures, including bees and other insects, can produce the similar responses and approximately accurate perceptual representations of the surface’s color.

There are perceptual constancies not only for size and color, but also for 2-D and 3-D shape, surface texture, ordinal depth (whether one surface is nearer or farther than another), location with respect to the perceiver, orientation (slant and tilt) of shape, speed and path of motion, and so on. Ability to track a given object even as it changes shape and color is another type of perceptual constancy -called ‘object constancy’. Perceptual constancies occur not only in vision. They also occur in hearing, touch, proprioception, and various perceptual modalities that we do not have (for example, perception by electrical impulses in some fish). A barn owl can, in total darkness, auditorily represent the location and movement-trajectory of a mouse from a single sound produced by the mouse.

It is not securely established whether sensory systems for taste and smell produce perceptual constancies, although these capacities can be exquisitely complex. Ants operate primarily from smell, although they show rudimentary perceptual constancies in vision.

Perceptual constancies are a remarkable phenomenon. Although proximal stimuli are the only sources of new information available to perceptual systems, perceptual systems can process very different proximal stimuli to yield responses to and representations of given environmental properties and objects. Studying this formation process in detail–and discovering what causal laws govern it–is the heart of the science. Human perceptual states that are members of a range of possible states in a perceptual constancy can be formed in about 60-70 milliseconds from onset of a proximal stimulus. That is approaching one-twentieth of a second. The simplest animals that are known to have perceptual constancies are various insects. Most animals above the evolutionary level of insects show one or another type of perceptual constancy.

Of course, even humans have many types of non-perceptual sensory capacities: possibly taste and smell, certainly various sensory capacities that affect balance, regulate heartbeat and contraction of blood vessels, and various capacities in perceptual systems that operate before the formation of perception, such as the sensing of light by retinal receptors.

Saying that perceptual constancies are the basic marks of perception sounds circular but it isn’t. There is broad scientific agreement about what cases count as perceptual constancies, and the agreement is not based on any antecedent definition of perception. Scientists know them when they see them. In the book, I discuss various cases in some detail. The discussions are supposed to help us understand what the scientists are going on, when they identify a capacity as a perceptual constancy.

I and many vision scientists think that perceptual constancies are the capacities that mark off perception from the sensitivities of plants and the non-perceptual sensing of organisms like bacteria, snails, and very simple worms. These organisms show no perceptual constancies. Perception is the simplest type of state whose role in causal processes is explained scientifically in representational terms–in terms of states that can be accurate or inaccurate. Representation is (along with consciousness) a mark of mind–or of a psychology. So, perception is the simplest known representational mental or psychological state. So, in marking the difference between perception and non-perceptual sensing, I take perceptual constancies to mark the simplest, evolutionarily first, form of representational mind, or real psychology. Perceptual constancies are the key, known phenomenon that distinguishes the most primitive organisms that have minds from organisms that lack minds.

RB: Can you say something about perception vs conception, particularly in reference to the ideas of Locke and Hume?

TB: Locke and Hume epitomized a tradition of distinguishing perception from conception (really, from thinking) as a matter of degree, not kind. Perception was supposed to be more concrete; conception, more abstract. They developed their opinions from introspection. It does seem that perception of a color shade is more concrete than a conception of color. And it is certainly true that certain relatively abstract matters can be conceptually represented, but not perceptually represented: mathematical structures, quarks, the shape of space-time, evolution, economic recession, political ideologies, and psychological processes, like thinking, perceiving, and so on. (In perception, perception is not perceived or perceptually characterized. Perception focuses “outward”.) Concepts for these and many other subject matters have no counterparts in perception–in capacities to perceptually characterize.

However, the Locke-Hume picture is seriously mistaken. Perception can be very abstract, and thought can be very concrete. For example, our visual system represents not only the most specific shade of color that it is capable of discerning. (Humans can distinguish about two million distinct color shades.) All known visual systems group color shades in ways that are useful for certain tasks. For example, although we see a particular shade of green, our perceptual systems also group that shade with neighboring shades; they treat the neighboring shades equivalently for certain tasks. It does not matter for the edibility of some plant that it be the exact shade of green that it is. Many other shades are equally good signs of edibility. All perceptual systems adapt to such matters. There are many groupings, at various levels of abstraction, depending on task-relevance. In fact, since perceptual systems process color differently from shape, it is, I think, correct to say that they characterize something as colored, whenever they characterize a specific color. The genus color is pretty abstract. Correspondingly, we not only represent the genus color in thought. We can think about the very most specific, discernible shade of a color and reason about it–how it might fit in with the rest of the decor in a room. So, perception can characterize very abstractly, and thought can characterize very concretely.

The difference between perception and thought lies not in degree of abstraction. It lies in the kinds of capacities involved: a difference of kind, not degree. To oversimplify: Perception just picks out and characterizes; thought can do those things, although in a different way. But thought also generalizes, produces reasons, explains, narrates, represents the possible that is not actual. Kant was the first great philosopher to distinguish perception from conception as a deep difference in kind.

Modern science supports Kant. It relies much less on introspection and much more on experiment that probes what happens in a mind. What happens is mostly unconscious and is mostly too fast to be introspected.

RB: Following on from the previous question, can introspection give an insight into the causal sequence of perception? (I’m thinking primarily of Buddhist philosophy and meditative insight practices)

TB: As just noted, introspection provides very limited insight into the causal sequences of perception. Introspection, especially helped with therapy or other training, can yield more insight into the slower processes of consciously available thought. Often, what is popularly taken to be perception is a mix of perception with thinking that is closely associated with perception.

RB: In your view, is there a correspondence between perception and imagination and, if so, what is the correspondence?

TB: Yes, some types of imagination re-use perceptual representations. Dreaming is a type of imagination. Animals that probably lack thought can dream, using perceptual materials. Humans can form re-arrangements of perceptions in imagination. There is experimental work on relations between perception and perceptual imagination – experiments on how humans rotate in their imagination shapes that were initially perceived. Visual artists do that sort of thing sometimes. Imagination also occurs at the level of thought – or in perceptual and conceptual mixtures. The imagination expressed in a novel is largely conceptual.

RB: Finally, are you working on a future book?

TB: The work on knowledge described above centers on ways perception causes belief and makes it epistemically good. Another part of my work, including what I hope will be my next book, centers on epistemology that concerns knowledge further removed from perception. I am working on confronting classical scepticism, associated with Descartes. Classical scepticism questions our knowledge that there is an external world, or our belief that we have hands (even as we seem to be looking at our hands), or our belief that we are not in a matrix on some unnamed planet, being fooled by sinister scientists.

Of course, we do know the things that scepticism challenges. We do know that there is an external world, and so on. It would be seriously neurotic to really doubt such things. For many, including those engaged in science, sceptical doubts seem quaint and silly. I think that this attitude is right, as far as it goes. We do not have to answer sceptical doubts to have knowledge. We can reasonably ignore them. Answering them is not a matter of protecting knowledge from a real threat. It is more part of an inquiry that seeks deeper understanding.

So understood, scepticism invites us to understand better what we already know very securely. Scepticism challenges us to explain how we know these commonplace things and why scepticism gives no reason to doubt them. A key limitation on such explanation is that it not “beg the question” against scepticism. Begging the question would be giving an anti-sceptical reason that assumes, explicitly or implicitly, what scepticism challenges. Avoiding begging the question is very difficult, for scepticism challenges very basic assumptions. The interest of the project lies in understanding reasons well enough to show, without begging the question, that scepticism’s challenges, based on its fantastic storytales, are not good reasons, or are incompatible with good and sufficient reasons. So, the investigation is into the nature of good reasons. This is epistemology that centers on upper representational mind.

Consciousness, Imagination, Neuroscience, Philosophy