I. Introduction

I suggest that many of the difficulties in mental imagery accounts come from overlooking or misconceiving some fundamental features of mental images. One overlooked feature is that mental images have important similarities to public, visible images. One misconception is that mental images are pure visual data or unidentified "pictures." In fact, an analysis of imagery experiences shows that images are "objects" that have intentional depth. Many images are more like topics, exemplars, or designated symbols than pictures. Another overlooked feature of mental images is that, according to traditional psychology, they fall into a set of fairly well-defined types. These types may be categorized according to a variety of established psychological parameters, including apparent location in space, degree of conscious control, and type of conceptual content. The most important feature of mental images is that they involve conscious states. As a consequence, their description depends upon how we understand consciousness and its objects. In this chapter I attempt to see how these and other features may be brought together in order to form a descriptive account of the nature of mental images.

I argue that the common sense view, that mental images are indeed real and have meaning, turns out to be substantially correct once it is amended by philosophic distinctions. I follow many insights brought to bear on the nature of mental images by traditional psychology and philosophers such as Aristotle, Brentano, Moore, Ducasse and Jackson. Some metaphysical issues are left unresolved, but I attempt to find what is involved in formulating the various alternatives at hand.

II. Preliminaries to the Study of Mental Imagery

A. The Visible Realm: Things, Images, and Symbols

Much of the confusion in the history of mental images is due to failure to make clear the similarities and differences between the contents of, and the human activities appropriate to, two realms: the visible realm and the mental realm. Imagists have made a close comparison between these two realms. They find images, objects, and percepts in both. Descriptivists find no images or image-like objects in common to the two realms. They reserve the concepts of perception and its derivatives for use in the visible realm.

Visible things are obviously among the things we encounter and interact with in the real world. If we want to understand how mental life is manifested in publicly-observable ways, we need to examine the kinds of visible things there are, their properties, how they may affect us, and how we may use them.

Visible, that is, publicly observable, things include a wide class of phenomena. First, this class includes solid physical objects, such as cars and stones. These are unified physical objects with determinate properties such as size, apparent color, shape, texture, and location in space. They are also visually inspectable objects, for although we see them as unified objects, we can also identify visually separate parts, or features, of these objects. Second, the class of visible things includes physical objects that do not have a determinate color, shape, size, or texture, and may only have an approximate or inexact location in space (e.g,. transparent objects, mirrors, clouds). Such objects are also visually inspectable, if less precisely so. A cloud, for example, may have an area of greater translucency on the left part, a deep shadow on the bottom, and so on. Finally, the class of visible things also includes items that have no actual location in space, but only appear to have a location in space, such as rainbows or holographic projections. Visible things, then, are most broadly identified as things of which we can form unified percepts and that have at least an apparent location in space. They may be either determinate or indeterminate with regard to other visibly-inspectable properties, such as color, shape, size, number of parts, or visibly distinguishable regions.

Publicly observable, visible images are a sub-class of visible things, and share some of the properties of visible things. Virtually any visible thing may be deemed to be an image, or visual likeness, of something. We need not limit images to photographs or drawings. A paper cutout, a cloud, or even the random discolorations of an old bit of cloth are equally capable of being interpreted as a visible image, or likeness of something. A visible image need not be any more or any less determinate in its physical properties than a visible thing. Like visible things, public images can also be inspected. They may have distinct visible parts or features to the same degree that any visible thing has these. Finally, like other visible things, images can obviously affect us. Our reaction to seeing the image of a person, place, or thing, can be as powerful as actually encountering the person, place, or thing.

Visible images differ from mere visible things not by any specific physical characteristic, but entirely by the fact that some things are deemed to have enough visible similarity to other things to be accepted as images of other things. Although we sometimes speak of seeing an image as a way of implying that what we see is either vague or not really a physical object, but only an illusion, this is not the primary dictionary meaning of "image." To be an image in the primary sense, and, so far as we shall be concerned in this discussion, the proper philosophic sense, is to be an image of something. Hence, rainbows are not images; they are not likenesses of anything else.

Visual similarity cannot be defined by any strict set of requirements. The reason for this is that it is in part driven by context. Both a life-size photograph of Clinton and a quick sketch of his profile on a slip of paper may be counted as images of Clinton. The photograph is likely to be universally recognized as an image. The quick sketch, while it might not be recognized as an image of Clinton if it is mixed with the trash in an alley, will almost surely be accepted as an image of Clinton if it is hanging on a wall in the White House. While some degree of visual similarity remains a requirement for something to be called a proper image of something else, verisimilitude becomes secondary to the requirement that there is some context in which an observable thing is understood as a visual reference to some other thing.

The human ability to shift from seeing something as merely a visible thing to seeing it as an image, is a circumstance woven into the fabric of the visible realm. It is due to the existence of intelligent, conscious agents capable of intentional shifts. When a slip of paper is moved from an alley to the White House, and in the process becomes an image, the physical properties of the image do not change. Rather, we have changed; we have shifted our inner perspective on the identical physical item. Another example: consider a visible vertical line.

This might be considered just a dark mark on a white background, a visible object only -- an image of nothing at all. But it might, if one were to creatively reflect on it, be considered a reasonable likeness of something: say, a fence post in the snow, seen at a great distance.

Apparently, the only thing changed when a visible thing becomes an image is the subject's intentional state -- how the subject understands or creates a meaning for something it considers. Such changes occur within individual minds, but images need not remain private. Public images exist as part of our real world -- part of the visible realm -- not just as part of private mental worlds. The world of the shared conventions of human agents, whether linguistic or visual, is both a public reality and part of individual mental realms. Public images and the intentional shifts that create them are just as much a part of shared reality as language elements and linguistic conventions. We communicate through public images, we react to them, we use them, understand them, and interpret them, just as we do linguistic signs. The conventions that apply to our interactions with and communication through images evidently differ from those of language -- arguing how and why this is true is a philosophic difficulty -- but this should not block us from accepting the fact that public images play a significant part in our mental life.

Intentional shifts also enter the account we must give for a related phenomenon that is part of the visible realm: the existence of symbols. Suppose now that the vertical mark above is to be understood as a symbol of a fencepost. Again, nothing changes in the thing; only our attitude toward it -- what the visual appearance is understood as -- changes. This is an entirely new intentional attitude because now visual similarity is irrelevant. I might as well have designated this mark as a symbol of a man, a car, or even something that has no visible appearance, such as the number one. Our apprehension of what a symbol means does not require any visual resemblance.

I understand the possibility of the shifts from visible thing to image and from image to symbol, as well as the reverse of these same shifts, as a datum of the human world.

thing <-----> image <-----> symbol

That these shifts exist within the context of the visible realm, that their existence and use is accepted as part of the common-sense view of the world, I see as unquestionable. Why and how they exist, the philosophic problems, come after this starting point.

We need to consider one such philosophic problem before closing this review of things in the visible realm. This is the problem of how to describe the manner of the existence of public images. I shall adopt the position that images are best described as relative beings. The reason for this is straightforward. Public images are real, but are relative beings because they owe their existence qua images to the existence of other things. This is true of anything (e.g., also symbols) that are representations of other things. Existing as a representation can be signified formulaically.⁽¹⁾ When we say

X represents Y to Z,

and X is an image, Y the object it represents, and Z a conscious agent, we express the fact that representation is a triadic relation, and X is in relation to both Y and Z. If one of the terms is removed from the relation we are no longer speaking of the same thing. This itself is not an insignificant metaphysical point, for some philosophers apparently believe that representation is only dyadic. They assert

X represents Y

can be an unadorned fact about the world -- a fact not requiring the existence of conscious agents (Z). Dretske's philosophy of the flow of information in the natural world hinges on this interpretation of what it means for one thing to represent another. Other views about what it means to be a representation are sometimes implied or stated in philosophic accounts. Among them:

X causes Y,

X is correlated with Y, or

X functions as if it were Y.

It is often argued that one of these relations is sufficient to characterize how physical states in the brain are "representations." This is a point we shall review when we discuss contemporary models of human psychology. In the meantime, I shall assume the only correct way, and, not coincidentally, the way most compatible with common sense, to understand representation is to insist that it is inextricably bound up with human intentionality.

In summary, the visible realm includes images that serve as representations of other things by way of visual resemblance. These images, like visible other things, are visually inspectable and may figure in causal explanations of human behavior. Public images are relative beings that can be transformed to symbols, and when this is done visual similarity ceases to be a factor in our relation to it.

B. Mental Images

Let anyone who doubts the existence of private, non-visible, mental images attempt to explain the existence and function of images in the visible realm without assuming that individual minds make for themselves representations of objects that are absent from perception. Even strict descriptivists do not deny there are such private representations. What they do deny is that these representations are distinctively visual, or that they are visual in any important sense that would make them much like public images. This false transference from the visible to the invisible realm (mental) realm can be characterized by two related faults which descriptivists justly level against the imagists.

The first fault is to speak of the mind as if it were entirely passive when mental images are present. We do not come upon mental images in surprise fashion, as we come upon images in the visible realm. Imagists speak as if mental images were objects planted in the dark recesses of the mind, suddenly and accidentally illuminated by the light of consciousness. Such passivity, descriptivists point out, does not reflect what actually happens. The contents of the mind are brought about by the intent to recall or to be engaged in constructive thought about something.

The second fault is to describe mental images as if they were items that can be visually inspected in the manner of an external stimulus. Imagists sometimes speak as if, when images appear on the inner stage of consciousness, we take some time to reflect on their properties -- how they "look," how "vivacious" they are, whether there is a "feeling" of belief associated with them, and so on, and then decide what kind of images they are.

Descriptivists are right in arguing this simply cannot be the case. Something more must be active in the mind if such presentations are to have any meaning or any effect, for raw images can neither have meaning nor be causally effective until they are interpreted. By saying, for example, that I have a memory image, I do not ordinarily mean to imply that I inwardly see an unprocessed visual presentation which, upon scrutiny, I have decided must be a memory. No such inner visual scrutiny occurs. The imagery of active, waking life seldom has the ambiguity of dreams or hallucinations. We do not, except rarely, wonder "what kind of image is this?" The image and a consciousness of its type forms a unified experience. In the vast majority of cases we automatically accept images as being of a certain type. Otherwise, we would be continuously lost in speculation about the contents of our own consciousness.

Despite these problems for imagism, imagists surely are correct in identifying visual experiences as a unique form of experience. Suppose someone were to make the following proposal:

Let us replace all the portraits and sketches of Clinton in this gallery with descriptions. In place of each image, we will place a descriptive paragraph, containing the information that was in the image. People can still visit here and read the descriptions. Since the information is the same, everything will be just as it was, and nothing will be lost.

The absurdity of such a proposal is evident when we speak of visible images. Surely, the "information" in the replacement descriptions is not the same. "Information" is not even the correct word to describe what was replaced, since visible images do not have a definitive informational content in and of themselves. The imagist perspective leaves open the notion of "visual presence," in all its ambiguity and possibility, as a factor in mental life, both public and private. One wonders why, when descriptivists offer essentially the same replacement proposal described above with respect to mental images, it is even given a moment's credibility.⁽²⁾

Imagism is also correct in insisting that no matter how "conceptual" or "verbal" cognition is thought to be, it does not change the fact that we have object-like presentations that, while lacking physical properties, induce in us sensations similar to those we experience in seeing. If imagists described the mind as too passive, passivity cannot be eliminated altogether. Mental images do resemble visible images in that they can stimulate the mind -- not, perhaps, precisely in the manner of a pure stimulus -- but in a way that produces similar effects. The mind is not entirely active; it is also passive in the sense that internal representations are presented to it.

A compromise position is called for. The similarity/dissimilarity between visible and mental images is not an all or nothing proposition. The mind is both active and passive with regard to internal representations, and neither linguistic nor visual artifacts alone constitute its contents.

1. A Definition of Mental Imagery

Richardson's revitalized traditional psychology provides a useful starting point for a philosophic definition of mental imagery. In his book, Richardson defined mental images as follows:

Mental imagery refers to (1) all those quasi-sensory or quasi-perceptual experiences of which (2) we are self-consciously aware, and which (3) exist for us in the absence of those stimulus conditions that are known to produce their genuine sensory or perceptual counterparts, and which (4) may be expected to have different consequences from their sensory or perceptual counterparts. (Alan Richardson, 1969, pp. 2-3).

Writing again on the topic of the definition of mental imagery in 1983, Richardson defended this definition against the advances of cognitive science. The only concession that Richardson made was that the fourth criterion for mental imagery -- different consequences than an original stimulation -- should be slightly more hedged. Although we might generally expect different consequences in the case of mental imagery, it is not always the case. Directed imagination, for example, has been shown to produce effects that are identical to those experienced in the case of external stimulation. This minor point aside, Richardson has consistently maintained his view that this constitutes the definition of mental imagery that is appropriate to psychology.

Clearly, the definition encapsulates some important features of traditional and common sense imagism we have already touched on. Under this definition, mental images are distinctively conscious experiences, can have causal efficacy, and have a distinctive phenomenological quality understood by analogy to vision.

The definition entails methodological prescriptions for the study of mental images in empirical psychology. It implies that since experiences are the subject matter, subject reports of these experiences are the primary data. One of the principal tasks of psychology, in Richardson's view, is to find correlations between various types of imagery and other psychological variables. Imagery could be correlated, for instance, with memory tasks, sports performance, psychotherapeutic exercises, mathematical problem-solving abilities, and numerous other variables of interest. In such studies, overt behavior is not excluded from study as an effect of mental imagery, but is not to be confused with mental imagery itself. Obviously, then, behaviorist or behaviorist-inspired approaches to mental imagery will simply have little or nothing to say about mental imagery under the rubric of Richardson's definition. Another methodological prescription is implied in that normal sensory stimuli should be absent when we are studying mental images. This corresponds to the imagist intuition (e.g., Aquinas's) that images explain behavior in the absence of certain stimuli.

Richardson's definition is only a starting place for attempting to unify the imagist and descriptivist views under a new philosophic paradigm, and in the following sections I attempt to expand upon it.

2. Four-Way Causation

There is a common-sense view of causation that is entirely unencumbered by the conundrum of metaphysical dualism. This view is woven deeply into our ordinary language of causal explanations. Ordinarily, it goes without saying that physical events cause other physical events. Equally, we unhesitatingly accept that mental events can cause other mental events. Mentally entertaining premises can lead to forming conclusions; anticipation of a coming event can lead to a reverie on previous events, and so on. Common sense also recognizes that physical events can have mental effects. Drugs, disease, bodily injury, and brain injury all have mental effects. These physical causes may result in temporary hallucinations, depression, loss of memory, and any other of a number of mental effects. Finally, the common man is quick to acknowledge that mental acts have direct physical results. The most obvious case is the production of physical movements of the body by the impulse of the will, but more subtle cases are also commonly-recognized, such as the influence of mental attitudes, sometimes manifested in the production of mental images, on healing the body. The pre-theoretical, common-sense view of causation, then, recognizes all four forms of possible interaction between the mental and the physical.

A contemporary common sense view may also suppose that four-way causation is possible because mental and physical events are somehow related or correlated (albeit in some unspecified way) in the brain. When the common man is pressed for a deeper explanation of this correlation, dualism may begin to creep into the explanation. But I assume that in contemporary times many people will seek to avoid dualism by holding that mental events always have corresponding physical events in the brain, even if they are not sure about the details of the relation (e.g., whether the reverse, that physical events always produce mental events, is true, and so on). Given this, it seems rather easy to explain the presence of mental images in terms of brain functions. One need only assume mental images and their effects are possible because the brain is capable of recording the visual appearances of objects or events, and then of playing them back via physical changes in the brain that restimulate the mind in a manner similar to the original experience.

This physical/mental correlational explanation of mental imagery, however meager and insufficient it may appear to be in light of the demands of a reductive science, is a substantial part of traditional psychology and its contemporary offshoots. In 1890, James suggested that since sensory (peripheral) neurons sent signals in toward the center of the brain for additional processing, it must also be possible that centrally produced stimulation could be transmitted back out to the sensory organs (James in Beakley, p. 188). He also noted that some subjects had reported "negative after-images of objects they had been imagining with the mind's eye," thereby establishing a possible link between his suggestion and subject reports (1892/1961, p. 178). It turns out that James's reverse flow explanation is now supported by many studies in psychology that correlate mental events with measurable physiological changes.

Finke (1984) designed a well-controlled experiment to prove that imagined colors produce optical illusions, previously thought only capable of production under exposure to actual stimuli. Finke also showed that imagined, rather than actual, physical movements could create measurable "training" effects in hand/eye coordination. Physical training by means of imagined performances is now used by athletes all over the world and has been proven to be effective. In a series of experiments in which subjects were asked to think of one color while looking at another, Kunzendorf (1990) demonstrated that some subjects were able to induce, at will, a measurable pattern of electrochemical responses on their retinas, corresponding to what was normally induced only by a colored stimulus. Some subjects were even able to produce the "red" response at will, while simultaneously observing a green stimulus, thus completely overriding the natural, physically-induced, response pattern. Kunzendorf's studies demonstrated continuity in the so-called "higher" senses with other previously studied, and very well-known, reverse-flow phenomena, such as conscious control of the heart and other autonomic functions.

Brain studies also show a close association between the physical visual centers and the mental process of using visual imagination. When mental images are reported, radioisotope studies show that the visual centers of the brain are activated. Kosslyn has shown that the same area of the brain is activated when the capital letter "A" is imagined as is activated when the letter is actually seen. Even the size of the brain area activated varies directly with the size of the letter imagined, just as it does when the letter is seen (Kosslyn, 1994, pp. 17-19).

Additional evidence of the association of visual centers with memory comes from brain damage studies. If the visual centers are destroyed, the ability to form descriptions based on visual memories is impaired. One famous case reports of a man who lost vision on the left side of the visual field (Kosslyn, 1983, p. 70). When asked to describe the village square that he knew very well since childhood, he could only describe the shops and objects on the right-hand side of the square. The descriptive knowledge was not lost, however, because the man could recall the other side of the square by imagining himself standing in a different location. This case argues that the physical generation of images in the appropriate hemisphere is prior to and necessary for description of certain memories.

There is, then, a common-sense view, supported by empirical psychology, that explains mental imagery by allowing physical and mental states to be causally connected in both directions (as depicted below). Mental states of conscious visual contents can be the result of physical inputs to the organs of visual presentation or they can be the result of mental inputs to these same organs, as Aristotle's view suggests. A physical state (physical₁ in the diagram) can produce some mental state (mental₂), this mental state can produce another mental state (mental₃), and this mental state can produce still another physical state (physical₄).

Brentano (1973/1874, p. 6) thought this way of describing causation was to be accepted as part of the normal circumstances for doing psychology. Like chemists and physicists working together by explaining different parts of the same process, psychologists and neurologists need not come into conflict. Any series of mental events, traced back in time, can be found to have a physiological correlate, and vice versa for physiological events. The neurophysiologist will have the job of explaining immediate physical causes of sensation, and the psychologist will have the job of finding "the first mental phenomena which are aroused by a physical stimulus" (Brentano, 1973/1874, p.7).

The philosophical upshot of the correlational view is that it leaves open the issue of the priority of causes. Neither physical nor mental causes need to be given absolute priority in describing and explaining the phenomena of interest. This, again, corresponds with our direct experience, our ordinary means of explanation, and the common-sense knowledge we have that we are not fully in control of all our conscious mental experiences. When physical causes are thought to predominate (e.g., in experiencing an after-image), we appeal to neurophysiological states as the predominating influences; when we are thinking, we feel that mental causes predominate.

It is, of course, precisely this ad hoc scheme of assigning influences and causes that reductionists object to. But the correlational view has some advantages. The correlational view has metaphysical plausibility because it implicitly adopts metaphysical neutral monism. It implies that the states studied in psychology are not ultimately either just mental or just physical. Again, this is (roughly) like a traditional Aristotelian view in that the changes studied are of a single substance, the psyche, that incorporates both material and non-material (formal and final) causes. We need not assume that one type of change is always prior in time to another. It is possible that mental states and their physical correlates spring into existence simultaneously without being determined by any prior states in any sense that we might presently be familiar with according to physical laws. Wittgenstein (Zettel, 1967) suggested this as a possibility: we can think of the brain as moving from a beginning state to a final state as a natural process, without the intervening states being determined by causal law (para. 613). Wittgenstein realized that thinking of brain processes in this way has little precedent. But he replies, "If this upsets our concept of causality then it is high time it was upset" (para. 610).

I believe the correlational view has some plausibility, and it will be useful in describing mental images. Initially, then, I shall adopt it.

3. Subject and Object

In a famous passage, Brentano attempts to define mental phenomena in terms of intentionality:

Every mental phenomenon is characterized by what the Scholastics of the Middle Ages called the intentional (or mental) inexistence of an object, and what we might call, though not wholly unambiguously, reference to a content, direction toward an object (which is not to be understood here as meaning a thing), or immanent objectivity. (Brentano, 1874/1978, p. 88)

This passage has often been misinterpreted. The word "inexistence" in this context does not refer to the ability of the mind to conceive of objects that do not exist. That is a capacity of the mind, but not the definition of the mental, and not the topic Brentano is addressing here. The capacity in question is, as Brentano states, reference to a content or direction toward an object. The "in" of "inexistence" means that the presentation is in the mind; presentations capable of reference to non-present objects only exist in the mind.

In what sense mental images may be said to be objects, and in what sense they may be said to exist "in" the mind are topics we take up latter in this chapter. As we observed in chapter 2, some philosophers regard images as non-intentional because they (or, more precisely, their raw appearances) have no intrinsic content. It is now clear, however, that this describes mental images as if they existed apart from the lived-through experiences of the subject -- as if they were public, rather than mental images. Public images have "content" (when a subject interprets them) in two senses: (1) the visual content, or the visible features that one may notice, inspect, or be aware of, and (2) the intentional content, or the content derived from what the image is understood to be an image of. To attend to (1) misses the point of what it is to be an image. We do not inspect the properties of the ink to determine what the sketch of Clinton is to connote. "Clinton," not "this sketch," is the intentional content of the image. The same is true of mental images; they have both visual and intentional content. It is, of course, possible to conceptually or theoretically strip away the consciousness of the raw appearances of mental images but (with a few exceptions) that is not the way we experience them. They have an intentional depth or content they are directed toward that is not actually separable from their being.

The meaning of introspecting mental objects needs to be clarified, since imagists have been accused of holding the characteristics of mental images to be given infallibly through introspection. Brentano, while he insisted on the general validity of inner consciousness as a form of understanding, did not even think introspection was possible. Introspection in the sense of inner observation (beobachtung) is not possible, because there are no separate inner objects that retain their properties independent of the acts of the subject. Brentano, like Wundt, insisted that we could not directly observe inner "objects" because the act of having an object was inseparable from the subject's attitude toward it. The only valid sense of introspection was "inner perception" (from wahrnemung, containing the roots for "true" and "to take"). I understand introspection to be absolutely certain only in the tautologous sense implied by Brentano's distinction: when we have a presentation, the only thing certain is that we have it. Everything else about it -- its causal history, and sometimes even its type⁽³⁾ - is subject to occasional error. The "properties" one does claim for images, then, belong not to objects per se, but are ways of characterizing our experiences and their general presumed causes and purposes.

4. Conscious Mental Images

If having mental images implies that we are conscious of them, we need to consider two other problems. First, what of the common-sense intuition that we have mental images even when we are not conscious of them? We do, of course, as both the accounts of Aristotle and Locke show, tend to refer to mental images in unconscious memory as if they were stored away in some physical space. If a visible image is removed from its hanging place on the wall of a museum and placed in storage where no one sees it, it remains an image. Is there a meaningful analog to this when a mental image disappears from consciousness?

An image disappearing from consciousness is occasioned by the deactivation of visual centers (among others) in the brain. What is left is the ability of that subject to (possibly) regenerate some similar state of activation of the visual centers. This ability is, presumably, underwritten by some physical state of affairs in the brain. These physical states, whatever they are, have the ability to serve as intermediating instruments in the recreation of a conscious presentation, but while they are in this state of latency they themselves neither bear a likeness to anything nor serve as a symbol for anything to an agent. Hence, they are not images. If mental images were like actual pictures, it would be as if pictures disintegrated into a collection of molecules once they disappeared from consciousness. Only some additional cause could reorganize the molecules to make them into an image again. In the case of mental images, this additional cause is fortunately at hand, in the form of the intentions and purposes of the subject.⁽⁴⁾

As Locke said, unconscious images are nowhere; they have only potential being when not in consciousness. Mental images require that we acknowledge the being of potential being. Only by accepting being in this sense can we say we have unconscious mental images. Had Aristotle been more scrupulous in applying his own principles, he would have made this point more clearly and not referred to them as "images."

A second problem for the view that having mental images implies we are conscious of them is the opposing view or intuition that mental images exist as subconscious or near conscious concomitants of ordinary conscious perception. In this way, subconscious mental images could be operating causes in mental life.

Three possibilities in defense of subconscious mental images emerge. First, there is a meaning of the term "mental image" by which we understand a complex of ideas about some subject matter, as in "he has a mental image of the French people." There may be many actions attributable to having an image of something in this sense, but this use does not imply any particular visual presentation that uniquely defines such an image. It implies, rather, a set of prejudices or concepts as motivations for behavior. We shall not be concerned with an analysis of mental images consistent with this meaning.

A second temptation, very much a live possibility both within the scope of common-sense and contemporary psychology, is to assume that unconscious mental images are causally operative in perception. One such way is the so-called "template matching" view of perception. It holds that what we see generates an image that is then compared to other stored images of previous perceptions. The input image is then identified according to its similarity to one or more of the stored, unconscious images. Thus, stored (unconscious) mental images account for all visual recognition. Some imagists appear to hold this view, and it is often held to be quite consistent with nominalism and an empiricist epistemology. It has also been revived in the form of visual recognition programs for computers.

There are numerous difficulties with the template matching theory, and I will just briefly indicate them here. First, let us consider what simple template matching would involve. One problem is the number of templates needed to accomplish even simple recognition tasks. There are a large class of items that are almost infinitely variable in appearance, such as the human hand. If there were inner pictures for each possible appearance of a hand (as a fist, with one finger pointing, fingers together, fingers curved, etc., and from each possible point of view) the resulting picture file would contain hundreds or perhaps even thousands of stored images. Supplying adequate visual templates for other visual concepts is even more difficult to imagine. How many images are needed to encompass the possible visual appearances of "dog" for example? While there may be some visual types, such as "box," that do not have significant appearance variability, the idea that each and every case of visual recognition could be supported by matching (or even approximately matching) visual templates, ultimately implies a nearly infinite supply of separate images that must be sorted and then scanned each time they are to be used. While this might be done by a huge computational system, the theory lacks plausibility on this point.

One can significantly improve the viability of the theory by proposing that it is not the sheer appearances that are used, but some model of the inner structure of an object plus its possible outer appearance. Thus, a hand or a dog would be modeled by means of a presumed inner structure of bones and joints that can move in certain permissible ways. This would reduce the number of templates, or "prototypes" as they are sometimes called in this version of the theory, to as few as one in the case of "hand" and perhaps only a dozen or so in the case of "dog."

The problem with the improved proposal is that it is no longer the visual qualities alone that are instrumental in making the comparisons. The possible configurations of visual appearances must be constrained by rules governing the possible configurations of the inner structure. This is no longer a pictorial theory per se. Rather, it is a combination theory. If computationally instantiated, it would require both "analog" or "visual" data and propositional rules about their relationships.

Either a pure picture or a picture/propositional theory must also face the following objections. Matching proposals, as H. H. Price has pointed out, are rather like strong imagism. In taking the matching of two templates to be sufficient for causing our judgments, it is rather like saying that visual images are our concepts. It inverts the relationship (in Price's view) between concept and exemplar: it is the concept that enables us to recognize an image, not vice versa. As to what concepts are, Price argues, we have no direct access. It may be that there are locked away in the human soul some image-like "schemata," as Kant termed them, that are instrumental in elementary processes of recognition. Presently, however, we can only speculate about their nature.

As unconscious operations, it is difficult to see by what criterion can say templates, prototypes, or schemata must be "images," if the way in which they could "resemble" or "look like" anything we perceive is unspecified. To insist that the inner representations are literally images also invites the homunculus objection. What intelligent agent, operating unconsciously, visually (rather than by other means) compares the incoming image with the store of templates or prototypes? I take up a question directly related to this in chapter 4, in the discussion of Kosslyn's theory. His theory seems to claim that there are unconscious images that are "inspected" in the visual sense. I attempt to show that his computational theory alone cannot encompass any form of visual inspection. The question of existence of biological recognitional modules in the brain, however, is another matter. These might exist, but how inner, unconscious "inspection" actually occurs here (if it can be said to occur at all), is still a mystery.

More might be said about this issue. For the purposes of my study, which concentrates on the concept of the conscious image, I believe the initial difficulties of such theories outweigh advantages of pursuing these issues further. Consequently, I shall not be concerned with the notion of causally-operative unconscious mental images.

The final case that philosophers and psychologists have sometimes used to argue for the causal effectiveness of unconscious mental images is the case of the ambiguous figure. The duck/rabbit figure, the face/vase, the old woman/young woman figure and many others have been cited as examples of mental imagery determining what is perceived. In looking at the duck/rabbit figure, for example, one is tempted to say that the reason I now see it as a duck, and a second later see it as a rabbit, is that I store unconscious mental images of these that temporarily determine what I actually see. Finding what I see with my eyes to correspond equally well to either of these inner representations, I shift back and forth between recognizing the figure as one then the other. These unconscious operations involving mental images so profoundly influence my perception that it literally changes, seemingly of its own accord.

There is, in my view, a convincing counter argument that retains the definition of mental images as conscious affairs, distinct from perception (I am indebted to Richardson, 1969, pp. 131-135 for this argument). First, the appeal to unconscious mental imagery really adds nothing to our understanding of how we interpret an ambiguous figure. It is simply an ad hoc theory that gives a suggestive name to processes of which we are not directly aware.

Second, closer examination reveals that interpretation of ambiguous figures actually involves some conscious mental work. Ambiguous figures involve a time-delay between seeing and perception. Perceiving the ambiguity in such figures is not a case of ordinary, automatic perception, but a case of conceptual and visual problem solving. If we immediately see, without effort, an ambiguous figure as one thing, it is just a perception, and we give it no more thought. If we fail to have an interpretation of an ambiguous figure or are told "this is an ambiguous figure," conscious work begins. We ask ourselves, "what is this?" and then consciously explore interpretive strategies to find an answer. These conscious strategies involve visual and conceptual posits. We consciously posit that some visible feature represents, say, an ear, and we then attempt to see the figure in a new way, given this posit. If the strategy is successful, a percept is formed. The percept may present itself so powerfully that we cannot "see" the figure as being anything other than, say, a rabbit. The temporary overcoming of the raw visual information by the mental image is consistent with the reverse-flow phenomenon discussed above.

The entire interpretive process involves two distinct stages, one in which conscious mental imagery is involved in resolving a pre-perceptual problem, and one in which a percept appears automatically as being a representation of something specific. When the percept stage is reached, the mental imagery stage is over and is no longer involved.

To conclude, then, we shall not be concerned with unconscious or near-conscious causally-effective images. In describing mental images, we shall be concerned with consciousness and its contents.

C. Descriptive Metaphysics and Objectives of the Study

Prior accounts have often described the nature of mental images in a way that makes them conform to the needs of a pre-supposed metaphysical system. This need takes a general form in that it starts from a definition of the unit of perception and ends in a description of the units of thought in terminology directly derived from perception. To coin a phrase: the perceptron (unit of perception) is transformed into the mentron (unit of cognition). For the imagist camp, the perceptron had to be intuitively obvious, directly derived from conscious experience, and never lose its fundamental characteristic of being an image through its various incarnations as memory, imagination, and thought. Berkeley, for example, claims that in order to attain clarity and truth, he will direct his mind to objects, "divested of words," so that he will understand the contents of the mind through "attentive perception of what passes in my own understanding" (Berkeley, 1710/1965, p. 59). Descriptivists also wanted to reveal the nature of the mentron, and believed they found it not in images, but in language or linguistic analogues, Still, from a general standpoint, the descriptivist theory is identical to the imagist theory because it also retains the continuity of transformation from perceptron to mentron. In the descriptivist case, the continuity is kept intact by construing perception as descriptive.

I shall attempt a bottom up, rather than a top down, approach to the topic of mental imagery. This attempt is aimed primarily at a description of what mental images are, a definition and a list of characteristics that circumscribe mental imagery, rather than finding a convenient place for them once a metaphysics is already established. The lessons of history and many recent critiques of the very idea of a neutral or purely "descriptive" metaphysics may make such an attempt seem absurd on the face of it. I do not claim any description to be neutral, only that I try to avoid presuppositions that drive other accounts. I allow the possibilities, for example that there is no perceptron, at least as others have apparently understood it, and that percepts are indefinite in form. I also allow there may be (as a consequence of percept variability) no universal psychology, and that no physical causal chain (in the conventional sense of 17^th century physics) can be invoked as an ultimate explanation of mental imagery events.

The strategy I shall follow is simple and consists of four objectives:

1. Recovery of lost information;

2. Creation of a philosophic inventory of imagery types;

3. Establishment of an alternate understanding of what the explanatory problems are for mental images;

4. Indication of ways to avoid technical mistakes in empirical psychology.

The first objective, to restore lost information and lost perspectives to the topic of mental imagery, consists in recovering the insights of Wundt, Fechner, Galton, James, Brentano, and Richardson and incorporating them with contemporary discoveries in the neurophysiology of imagery.

Traditional accounts of mental imagery include a good deal of reasoning, speculation, and experiments with imagery types (see Holt, 1964). The reason for this is that, as these psychologists correctly understood, there is no way to isolate the vast complex of all the possible psychic and physical changes associated with the general term "mental image" and make this an object of study. Because there is always some varying mix of psychic and neurophysiological components in mental imagery, the field must be broken up somehow, even if arbitrarily, in order to design empirical investigations that are directed to a fairly isolated range of phenomena. Thus, the second objective in my study is to redirect attention to imagery types, and improve upon the insights of traditional psychology. In this way, we can distinguish stimulus-like properties from conceptual properties, and, instead of giving a univocal answer to the question of the source of image content, we can separate the cases, using imagery types, and give a nuanced description of how visual content may be induced, generated, and provisionally bound to conceptual content.

The third, or metaphysical, objective is to clarify the description of mental images to such an extent that the problem of explaining them is clearer, even if it remains indecisive how to proceed. While a complete reductive explanation of consciousness, and therefore of mental imagery, is presently lacking, we do have causal explanations of mental imagery in terms of the neuro-physiological states correlated with imagery. This metaphysical approach places the primary burden on description. It is a topic neutral approach in that it does not seek to specify precisely which components of imagery experiences are due more to "mental" or more to "physical" factors. This is, I believe, also consistent with some aspects of current folk psychology. We cannot always readily separate the physical versus the mental components or causes in our discussions of why people enjoy smoking, or running, or contemplating. The same is true of mental imagery, but there are fairly distinct variations according to types, in the so-called mental/physical mix. I have called this a "correlational view," but this an epistemic rather than a metaphysical separation and is not to be understood to imply there are separate metaphysical substances causing changes in each other. The "explanation" I give of mental images is quite limited, but by comparison to the other presently popular accounts discussed in chapters 4 and 5, is actually less speculative and stays to the topic (intentionality and conscious forms of awareness) instead of reverting to explanations in terms of mechanisms driven by principles of 17^th century physics.

The definition of the problem of explanation has implications for the fourth objective, avoiding mistakes in empirical psychology. These implications come to light in chapter 5. Several important contemporary investigations measure phenomena that simultaneously involve perceptual problem solving and short term memory images. The ambiguity in these experiments might have been avoided, given a better understanding of the nature of mental imagery and its various types. Experiments in a more traditional mold, I show, are more fruitful and true to the phenomenon to be investigated.

III. The Range of Mental Imagery

........[text of this section omitted, except for summary]

G. Summary: The Nature of Mental Images

Mental images exist on a continuum from those in which automatic, physiological factors predominate (after-images, perhaps eidetic images), to those in which interests in responses to the environment predominate (memory and perhaps dreams), to those in which self-directed creative impulses are set in the context of the possible and the actual (imagination and thought images). Each of these states represents a deepening of conceptual components in the imagery experience and may be associated with a corresponding decrease in importance of visual content. Having memory, imagination, and thought images is each characterized by intentional shifts with a distinctive character, independent of the purely visual components, and dependent upon the distinction between object of presentation and symbol. Memory, imagination, and thought images can provide a mediated awareness, replete with multi-modal capabilities, of a non-present environment -- a space that was, or might be, experienced. These forms of mental images are not pure visual presentations, but are provisionally bound to conceptual content. What we mean, in these cases, when we say that we are having mental images is not that we are hallucinating or dreaming aimlessly, but that our mind is taking something under consideration, and part of that process involves a quasi-visual experience in and through which a particular content is present to us. Mental images of memory, imagination, and thought are more like mental topics than they are pictures in the mind. The human mind is (perhaps) unique in being able to release a pictorial presentation from its content, to consider it as something new, or even strange, as just a visual presentation with no history or purpose. This ability is the foundation of radical skepticism (for which we gave reasons to presume false), but also the foundation of thought, which lifts the "given" and pronounces whatever it wishes a symbol for the absent.

IV. Explaining Mental Imagery

We have completed the first two objectives of this study, the restoration of lost (traditionalist) information and the creation of a descriptive, intentional analysis of imagery by types. It remains to complete the third objective, to clarify the options for an explanation of mental images. In this section, I am primarily concerned with laying out the options, but in the process I argue for my own view, that descriptive metaphysics is at present the most viable way we have of examining the nature of mental images.

A. Images and the Mind/Body Problem

As we know (chapter 2), one of the most well-known passages in the history of mental images is Aristotle's comparison of the formation of a memory image to the impression of a seal made in soft wax. Less well-known is Aristotle's introduction, in the immediately following passages, of the problem of presence and absence. Aristotle asks:

[W]hen one remembers, is it this impressed affection that he remembers, or is it the objective thing from which this was derived? (On Memory and Reminiscence, 450b13)

Bear in mind that we have already established that (as Aristotle also argued) memory presentations cannot be recognized as such unless they are known to be representations of something absent. Then Aristotle asks:

how is it possible that, though perceiving directly only the impression, we remember that absent thing which we do not perceive? (On Memory and Reminiscence, 450b15)

In other words, how is consciousness of the absent possible at all? Aristotle attempts to answer this question by using an example. Aristotle points out that when we contemplate a painting we can contemplate it as something absolute, just as it is itself, or as a representation of something else simply by a change in point of view: "one changes his point of view, and regards it [the presentation] as relative to something else" (On Memory and Reminiscence, 451a8). So, he concludes, if a presentation is "actualized in consciousness," and if "the soul perceives it qua related to something else, then...the experience involved in this contemplation of it is different" than if it were just a mere presentation (On Memory and Reminiscence, 450b25-451a3).

This explanation or, more precisely, observation sets out a path we have already explored at the beginning of this chapter. We followed Aristotle's implied suggestion that the answer lies in understanding how images are used in the public realm. It must now be admitted that this was a bit of misdirection. This does not explain the inner realm, it just moves the problem of inner representation out into the public realm and asks us to accept what goes on there as proof of the non-mysteriousness of the inner realm. We are told that images can represent the objects they do by a change in point of view. No doubt. In the public realm, the agreed-upon meanings of images might be accounted for by some complicated system of verbal and other non-vocal cues sustained by implicit agreements among various observers. In the private realm, no similar hypothesis suggests itself. My mental image of a man holding his hand inside his shirt can suddenly become a representation of Napoleon. My mental presentation of Tim can suddenly become a representation of Tom, his identical twin. But how is this to be accounted for? What, precisely, is a change in a mental point of view?

We said these changes in point of view were intentional changes -- again, as if that were a sufficient explanation, and as if descriptive metaphysics were an unambiguous project. We need to define intentionality more precisely and locate it within a broader system of terms. In defining intentionality Brentano states "no physical phenomenon exhibits anything like it" (1973/1874, p. 89). This apparently means that intentionality is utterly without possibility of explanation in terms of physical entities and the causal laws presumed to govern them. This understanding of intentional phenomena, when combined with what the history of mental images accepts as the origin and function of mental images, is the foundation for the most intractable philosophic problem about mental images. The other old saws of the history (role of imagery in thought, homunculus problem, etc.) are by comparison far less problematic.

The nature of the problem is indicated in the following set of propositions:

1. Mental images exhibit intentionality because they can represent something absent from present perception. The content of images is defined by what they represent.

2. The origin of mental images is in the physical workings of the perceptual equipment of the body.

3. The mental realm is defined by intentionality, reference to a content, or the ability to understand an (inner) presentation as being of something absent. Nothing in the physical realm appears to exhibit intentionality.

We now have a set of philosophically problematic theses. Each thesis has intuitive appeal, yet finding a harmonious interpretation of them that yields a satisfactory explanation of human mental powers proves to be enormously difficult.

The main problem, historically, is that theses 2 and 3 are thought to conflict. How can a physiological, materialist-based explanation of the origin of mental images be reconciled with the mental capacity to take a given presentation as representing something absent? Physical things are either present or not. Physical things do not (at least in any obvious way) indicate or intrinsically represent anything which is absent. How can a physical residuum from sensory processes become transformed into or serve as the basis for a mental phenomenon that represents something absent? Mental and physical modes of describing events and processes seem to be utterly different. It is therefore a problem when we are asked to refer to physical processes in order to explain mental processes.

What solutions present themselves to this trilemma? Let us review some options, beginning with the present account.

B. Accepting the Mystery

The solution I find the most persuasive is to accept all three theses as basically true. The apparent conflict between theses 2 and 3 can be ameliorated, if not resolved, if we accept four-way causation. Mental images can be caused by physical events, but that does not mean that they must somehow also remain physical events. So we do have an explanation -- a causal explanation -- of how they are brought into being. It is the mode of being of mental images that is in question not their causal history. Because mental things are a new order of being, we do not have a reductive explanation, i.e., an explanation that would -- to put it paradoxically -- explain how mental images, despite being mental, are nonetheless physical. What many philosophers and psychologists want to insist on is that we must have not only a causal explanation, but a reductive explanation in order to make headway on the mind/body problem. If we release this restriction, it is possible to further analyze the being of mental images without being distracted by the demands of reductionism.

There are three general reasons in favor of releasing restrictions on the types of explanations and the types of causality that are appropriate to entertain in conjunction with the mind/body problem. These derive from a priori, practical, and scientific considerations. I shall indicate briefly what each of these is.

Hume argues that we lack an a priori reason for supposing that causal connections cannot bridge the gap between the mental and the physical. It is only a prejudicial supposition that mental and physical events are too dissimilar to be causally connected that prohibits us from accepting the possibility of their causal connection. Hume defended this position eloquently in the following passage:

We need only reflect on what has been prov'd at large, that we are never sensible of any connexion betwixt causes and effects, and that 'tis only by our experience of their constant conjunction, we can arrive at any knowledge of this relation. Now as all objects, which are not contrary, are susceptible of a constant conjunction, and as no real objects are contrary; I have inferr'd from these principles, that to consider the matter a priori, any thing may produce any thing; and that we shall never discover a reason, why any object may or may not be the cause of any other, however great, or however little the resemblance may be betwixt them. (Hume, 1739/1978, p. 247)

And, as far as experience teaches us, Hume continues, we perceive a constant conjunction between thought and motion all the time: "every one may perceive, that the different dispositions of his body change his thoughts and sentiments" (Hume, 1739/1978, p. 248). Therefore, Hume concludes, "matter and motion may be regarded as the causes of thought as far as we have any notion of that relation" (Hume, 1739/1978, p. 250).

One of the long-standing premises of traditional biology and psychology is that these sciences have a different starting place than sciences such as chemistry or physics. The most compelling biological and psychological explanations often take the form of referring to an organism's inner desires, needs, beliefs, and other properties that have no apparent form of practical reduction into terms of equal explanatory value. While biological or psychological terms might, according to some accounts of physical causality, be reduced to an elaborate history of physical micro-events, there is still an enormous practical value of retaining vocabularies that pick out the biological and psychological factors of our direct acquaintance and immediate concern.

One of the assumptions that seems to be accepted along with the notion that reductive explanations are, or will be, "final" is that physics itself is (or will be) the final terminus of our knowledge of forms of causality and the ultimate entities that make up the world. If this supposition rests on a faith that contemporary physics has itself resolved what the ultimate principles of the universe are likely to be, then it is simply mistaken. The contemporary controversies over the nature of wave/particle duality and the meaning of Bell's theorem demonstrate that contemporary physics is in no position to state its own principles unambiguously, much less stipulate the ultimate principles of any science whatsoever. The supposition, often implied or stated as a "principle of reductive explanations," that in the final analysis the principles of mental causation or the fact of human consciousness will not be included in physics is far from secure. I state this point, without any elaborate defense, merely as a reminder that there are many possible points of view about what scientific reductionism amounts to.

We lack, then, any compelling reason to say that because we cannot (presently) provide a reductive explanation of mental phenomena, we have no explanation at all. Let us now see how we can use the principles of causal history and description of mental phenomena to elucidate mental images, the homunculus problem, and the epistemic claims of psychology.

We find that it is impossible to reduce mental imagery to either a pictorial or a descriptivist phenomenon. We find it impossible to maintain the dichotomies that form the basis of the differences between the imagist and descriptivist perspectives. We find it impossible to completely separate the conceptual and pictorial, the active and the passive, and the being-present-to and being-present-for in having mental images. I suggest that the inability to maintain these dichotomies is due to the nature of mental images themselves. There are ineffable, vision-like, pictorial features to mental imagery experiences, but these are never without the effects of the subject's intentional awareness toward these experiences, be it the subtle awareness that what is presented is not an external stimulus, or a more intrusive conceptual frame that what is presented is logically possible but false to the actual world.

Because mental images retain these ineffable vision-like features they cannot be reduced to mere description. Since a description never captures the "object" of visual consciousness, one refers to the mental image as an object, in order to identify the state of consciousness we mean to designate. Mental images are not physical objects, but we adopt an object language in speaking of them analogically as objects of inner vision.

The necessity of this analogical language resolves the homunculus problem. When taken literally, the object language requires an inner homunculus to make mental images inwardly seen. Yet we can no more explain in strictly physical terms or strictly descriptive terms inner seeing than outer seeing. Neither kind of experience has any explanation whatsoever in light of a complete description of the physiological processes and the physical laws thought to be governing them. In vision, we can only trace the progress of the electro-chemical changes inwardly from the eye to the cortex so far, and then at some point announce "and then we see" -- more precisely, we see that we see. We see not in the manner of sheer insect-like receptiveness to light stimuli, but in the manner of self-conscious awareness. The physical transforms into the mental. The same is true when we trace the development of mental images from some inner impulse outward toward the visual cortex and even to the retina. The only "reductive" or "physical" explanation we have of the similarity of these experiences is quite general. Apparently, some of the same visual centers of the brain and eye are activated in both seeing and imaging. This general physiological explanation shows the so-called homunculus is just ourselves becoming conscious of visual presentations using the same physiological equipment that we do in seeing. The homunculus, then, disappears when we stay within the realm of general physiological explanation, but the sheer fact of consciousness through seeing and inner awareness remains as mysterious as ever if one attempts to reduce this explanation to more primitive terms.

The epistemic claims possible for a philosophic psychology, under this conception of mental images, are modest. There is no claim to be made as to the ultimate explanation for conscious mental images, unless there were to be an ultimate explanation of consciousness itself. Other than to say that consciousness and its contents are phenomena we directly encounter, and that these are evidently the result of biological processes in living beings capable of inner representations of the environment, we have no deeper, or more reductive, explanation. Our "explanation" takes the form acknowledging a brute fact -- a description of the world. We agree with Aristotle and simply observe that the mental act of supplying a reference to an image is something we can do. Any further details we can provide represent attempts to understand imagery states as states of which we are self-consciously aware, on their own terms. We use "mental" terms, such as "understand in a new way," "supply a content to," "make a mental reference to," and so on, to explain what it is we do. If this view is correct, as I believe it is, then description of this sort will be all we have. This description uses the irreducible terms drawn from experiences and refers to the "objects" with which we are directly acquainted.

If we accept this form of descriptive metaphysics combined with bio-physical causes of mental states, then our work is essentially over. By this I mean that we can, of course, elaborate our description, expanding our inventory of mental acts and mental dispositions, but we have now accepted or established the realm, i.e., the mental realm, in which our further descriptions will be sought. Whatever we find in this way is unlikely to be either any more or any less mysterious than the already established fact of human consciousness.

Further investigations of mental imagery can also continue in practical psychology and physiology. Richardson's program of discovering the subjective and objective conditions of imagery experiences remains intact, as does the goal, originally envisioned by traditional psychologists like Fechner and Wundt, of finding all the physiological concomitants of mental states.

Finding and interpreting mental/physical correlations defines the prospects and limits for traditional philosophic psychology. Wundt's attitude toward the limits of psychological knowledge is exemplary (see Foundational Period: 1860-1901, chapter 1). Wundt adopted what was essentially explanatory dualism while maintaining metaphysical monism. For Wundt, the separation of mental and physical explanations had to be accepted as a given in psychology. This did not mean there were two separate metaphysical substances in which mental and physical events occurred. Rather, Wundt thought of the series of mental and physical changes as two parallel streams of events. Any series of mental events had physical circumstances that coordinated with it, and vice versa, but each series of causes retained its own integrity as a viable object of study. In the final analysis, however, the physical accompaniment of psychic changes were of secondary importance, because Wundt considered the subject matter of psychology proper to be the conscious contents of the human mind. The primary aim of this science is solving practical problems for the benefit of human beings; it is not an attempt to empirically investigate how consciousness of the absent is possible in physical terms.

This type of philosophic psychology supposes a metaphysics in which mental images have being because one accepts potential being, relative being, and real relations among subject, object, and signs indicating objects. In this universe, we can account for mental images essentially by describing both the physiological and mental changes that occur when subjects have mental images. In this universe, too, my accepting either a visible thing or a mental presentation as an image or a symbol is a new state of the universe, and it makes no necessary connection with any particular physical states in the body (though, it is presumed, there is some, unspecified, state corresponding).

C. Mental Images as Mental Objects

1. Review of the Problem

In chapter 1, I stated that the intuitive, common sense view of mental images is that they are kinds of mental objects. I pointed out that many images seem to exhibit continuity over time. Memory images, for example, seem to be recalled over and over again, and imagination images, even though they may be transitory, one-of-a-kind images, seem to be stable while we are having them. Some mental images, for example, can be mentally "manipulated," in a manner analogous to physical manipulation, by rotating or moving them in an imagined space. Finally, I pointed out that mental images might be regarded as objects in the sense that they are like "materials" or "constituents" in the overall economy of mental life. We seem to use or employ images as fundamental constituents in memory, imagining, and thinking activities, and we are subject or witness to them in the form of dreams, hallucinations, eidetic images, and after-images.

Despite these considerations, I argued in chapter 1 that it was necessary to put quotes around the word "object" because even elementary considerations show that mental images do not fit our concept of ordinary physical objects in the external world. Mental images do not literally have size, shape, or color. Other difficulties for accepting mental images as objects were indicated in the review of history in chapter 2. This showed that object language results in a tendency to conceive of mental images as if they could be literally seen by an inner eye, and this requires a homunculus. Descriptivists attempted to remove the homunculus by showing how linguistic elements and concepts could fulfil the role attributed to images. The descriptivist correctives, I argued, went too far, resulting in loss of a correct description of the phenomenon to be examined. The review of history was therefore inconclusive, and the proper sense for the "object" designation of images was not found.

In this chapter, I have attempted to consolidate common sense with science and philosophic psychology by accepting a definition of mental imagery as referring to quasi-sensory or quasi-perceptual experiences. This definition also accepts the fact that these experiences may be significantly like perceptual encounters with actual objects and that these mental encounters may have consequences similar to their sensory or perceptual counterparts. The sense of "objects" implied by this definition also stands in need of elaboration.

Finally, I have also referred to Brentano's definition of intentionality. Again, there is mention of objects and their "inexistence" in the subject. The "in" of the intentional inexistence of the object needs to be clarified since it is not obvious how there can be objects either "in" or "existing" in consciousness.

In short, the philosophic task remains of attempting to sort out what the scare quotes around the term "object" actually amount to. We need to clarify how such objects can be said to be "in" the mind, in what manner, if any, they may be said to exist, and how, or whether, they obtain the properties they seem to have.

I shall begin by briefly describing two competing views about the nature of mental objects. The two views are the adverbial account developed by Ducasse and Chisolm and the act/object account defended by Moore and Jackson.

2. Ducasse's Adverbial Theory

Ducasse states his theory as one to be contrasted with G. E. Moore's "object" or sense data theory of conscious awareness. Moore pointed out that in the case of sensing something blue, it is possible to distinguish the consciousness of sensing from the sensation itself. The sensation is of blue, rather than of, say, green, so there is a distinct, identifiable sensum or object that we are aware of that differs from other objects. At the same time, there is an element that remains the same in both sensations -- consciousness -- so it is the objects, not consciousness, that differentiates the cases. Normally, we think of "blue" and our awareness of it as being unified or at least concurrently existing, but, as Moore pointed out, there is no logical objection to the possibility that "blue" and our consciousness of it could exist independently. This establishes the logical possibility of an unsensed sensation or sensa that has independent existence. Moore introduced this possibility in order to argue against idealism. By arguing against the idea that consciousness of sensory qualities necessarily implied the existence of something "in the mind," rather than something "objective" and possibly capable of independent existence, Moore hoped to weaken one of the central arguments in support of idealism. This argument is that the being of sensible objects consists entirely in their being perceived; as Berkeley put it, esse est percepi. Moore's argument is that the proposition that the esse of anything consists in its being perceived does not even hold true in those cases where consciousness and its objects seem most firmly united, i.e., in the case of the apprehension of sensory qualities such as blue.

Ducasse argued against Moore not in order to defend idealism, but in order to show that esse est percepi is true for a certain class of conscious processes. In Ducasse's view, this does not result in idealism, because he distinguishes between the content of conscious processes and the object of conscious processes. In the case of consciousness of blue, blue is the content of consciousness, not the object of consciousness. For this content, it is true that esse est percepi, for the content exists, although it has no independent existence apart from the

process of which it is a part, and this is the only mode of existence it can have. The object of consciousness, on the other hand, must be an object external to consciousness, having an independent existence.

How are we to understand the mode of existence of the content of consciousness? Ducasse claims that this is a mode of existence of "things" that we readily accept as part of the real world. Consider a waltz. Waltzes are things that can exist, but a waltz has no existence apart from the act of people dancing it. What we find in case we are looking for instances of waltzes, and what we come to know, if we seek to know walzes, are not some independently existing things, but some recognizable parts of a particular kind of process. Similarly, in our consciousness of blue, we become aware not of something independently existing but something that is part of the conscious process itself: "what is known by the knowing process is then its own determinate nature on a given occasion" (Ducasse, 1951, p.260). To say we have consciousness of blue, then, is ultimately a way of characterizing a particular mode of a conscious process. As a waltz is a mode of dancing, consciousness of blue is a particular mode of "being aware of." It is not the case, then, that there is any object that literally is blue when we see blue. "Blue" is not something independently existing in the world but only something brought about by, and inseparable from, conscious activity. This activity takes on the mode of awareness of its own specific form of activity we call sensing blue.

In light of these considerations, Ducasse claims that the essence of the activity of seeing blue would be most perspicuously formulated by eliminating the noun "blue" altogether. The process should be described as "seeing bluely" -- a process verb, adverbially modified to indicate its particular kind. Hence, "blue" is not something objective at all; it is a species of conscious awareness, not an object of awareness.

3. Chisolm's Support for the Adverbial Theory

In support of the adverbial theory Chisolm (1966) argued that the object theory utilizes various forms of invalid inferences and unnecessarily multiplies the entities necessary to explain perception. One error is to conclude from the premise

(1) something appears F

that

(2) something presents an appearance that is F.

This error is easily seen by substitution. From, "this ship appears water-tight," we cannot conclude that "this ship presents an appearance that is water-tight" (Chisolm in Rosenthal, p. 382). This, Chisolm would say, is the kind of error Moore commits when he attributes "blue" to a sense datum or an appearance in consciousness. From the fact that there is an appearance of blue, we cannot conclude that the appearance itself is blue. Reasoning in that vein will end in infinitely multiplying entities. For if any perceived object A with property X generates another entity (an appearance) B, having property X in virtue of which A is said to have X, then B must also generate an appearance entity C, in virtue of which it may be said to have property X, and so on. In order to eliminate these unnecessary entities, Chisolm concludes, following Ducasse, that it is necessary to avoid attributing properties to appearances.

Chisolm's argument is an important one to bear in mind in determining the issue of the qualities of mental objects. For the purposes of the discussion below, I shall accept Chisolm's argument that to attribute properties to appearances themselves is an error. I shall refer to this as "the sense datum error." This error is closely related to another one that we must also consider. To say that from "I see a ghost" it follows that there exists a ghost that I see, is also an error. All of the mental images we report might be likened to such an example. I shall refer to this as "the existence error." We should note that each of these errors might be corrected by a suitable interpretation placed on terms such as "appearance," "see," or "exists," but that without such embellishments they remain logical errors.

4. Moore's Objection and the After-Image

It is important to note that what we have said so far about the adverbial theory is only indirectly related to our investigation. Ducasse and Chisolm concentrated on an analysis of perception as an adverbially defined mode of consciousness. In one fundamental sense, it may even be said that Moore and Ducasse are bound to agree on the essence of perception as it applies to our investigation. Both accounts endorse versions of common sense realism: independently existing physical objects are the objects that we know. Whether the means by which we know them are designated "contents" or "objects" of consciousness is immaterial as long as it is recognized that some form of conscious mediation is involved in the process. When a physical object is present, the nature of this conscious mediation may be debated; whether it is most accurately designated a mode of consciousness or whether it involves consciousness and a sense datum is a problem, but it is not precisely the same problem as the one we face in the case of mental images. Ex hypothesi, for the purposes of our investigation, there are no conscious mental images involved in perception.⁽⁵⁾ We are concerned with the relation of consciousness to its "objects" when consciousness is not directed toward an external stimulus.

Now, both Moore and Ducasse did address the problem of one sort of mental image we have mentioned: after-images. Moore thought the ability to see blue in the absence of any physical object supported his position, since the after-image is a form of direct presence of a perceptible quality in consciousness. Since the question of the ultimate relation between consciousness and the physical object was now ancillary, the after-image case supported even better, he thought, the claim that consciousness is necessarily of non-physical objects (Moore in Ducasse, pp. 282-283). Ducasse responded that after-images are not simple objects (of any kind) at all. Rather, they are complexes of qualities. That is, an after-image is not a single intuition of a single color quality but always appears with the additional qualities of size, shape, and location. While Ducasse allowed that there was a sense in which the after-image constituted a whole, and might even be designated a "sensa" or "sense datum," he held that (1) it was nevertheless composed of distinguishable elements, and that (2) in order to determine the status of the whole, one need only determine the status of the parts because "the existential status of the sense datum...can be no other than that of its elements" (p. 284). The elements, in turn, were most aptly described in terms of modes of consciousness since they literally were qualities and did not themselves bear qualities (as Moore's sensa were reputed to bear). In order to formulate the experience of having an after-image according to the adverbial format, Ducasse suggested that we should most properly speak of seeing an after-image, not just bluely, but, for example, bluely, squarely, and largely.

We now have two accounts, both of which seem to be partly right. Ducasse's account seems to be right from the standpoint that in the absence of a physical stimulus, certain forms of conscious awareness should be thought of as modes of consciousness itself. Moore's account seems to be accurate in pointing to the phenomenological fact that we encounter after-images as if they were separate objects and not parts of consciousness. At this point we will do well to consider another aspect of the object theory.

5. Jackson's Object Theory

Jackson (1976) argues that the adverbial theory is not successful in eliminating mental objects because its proposed linguistic recastings of the nature of our experiences fail to do justice to the nature of these experiences, specifically sensations and after-images. While Ducasse has argued that his own theory correctly represented the common sense view and attributed the failure of our linguistic conventions to conform to this to "linguistic inertia" (p. 278), Jackson thinks our linguistic conventions correctly indicate the true state of affairs. By convention, we do in fact refer to many sensations and mental images as singular terms identifying something existing at a particular time and (sometimes) in a particular place. (This corresponds with the assessment of the common sense view I have maintained since chapter 1.)

Jackson claims that mental objects can be eliminated only if we can find a correct, non-question begging analysis of a statement such as

(1) Bob has a pain.

A non-objective account needs to show that "pain" (or some other named sensation) does not appear in a way that logically implies its objective status. But it is immediately obvious that there is not, for example, such a thing as someone simply having "a pain." Bob's having a pain implies that he has a pain somewhere, and this gives an objective status to the pain. The adverbial approach, in order to capture this, must resort to awkward constructions such "Bob senses painly-legily." Similarly, the adverbial approach cannot successfully recast the meaning of what it is to be conscious of an after-image that has more than one property. For example, if (following Ducasse) the suggested formulation for sensing a blue, square after-image is

(2) I sense bluely and squarely,

then this cannot logically differentiate two different cases:

(3) I sense one blue thing and one square thing,

and

(4) I sense one thing that is both blue and square.

According to Jackson's critique, then, Ducasse's proposed solution to identify the locations in the visual field as modes of consciousness simply cannot work. Ducasse's formulation "I sense bluely, squarely" cannot distinguish between sensing an indefinite patch of blue and a square from sensing a blue square. Nor will separating the visual locations of such color and shape intuitions do the trick. Sensing "leftly bluely" and "rightly squarely" would indeed remove ambiguity, but it also implicitly introduces a new form of object identification into the description of the phenomenon. The locations in the visual field are themselves species of objects. We have a sensation "here" or "there" in the visual field, thereby logically implying an object-like status to the segments of visual space themselves. These spatial object areas, designated "left" and "right," then remain as static (non-adverbial) givens even if the mental objects "in" them can be designated as "modes" rather than objects.

We see, then, that there are substantial reasons for accepting that a linguistic account also terminates in the need to accept the object status of mental images.

6. The Need for a Metaphysical Account

Let us review and access these accounts. Ducasse's account seems physically correct, even if metaphysically ambiguous. Moore's account seems phenomenologically correct. Jackson's arguments are reminders of the fact that mental images are linguistically identified as objects, even if only minimally as objectively-given aspects of the visual field.

If both Moore's phenomenological and Jackson's linguistic accounts of mental images grant them an object status, is there a metaphysical account that conforms to this?⁽⁶⁾ In what sense, if any, can we grant object status to mental images? Chisolm's arguments are reminders of the fact that in any object account we need to address the sense datum and existence errors. The analysis we have so far, then, does nothing to settle the issue of the metaphysical object status of mental images.

7. Desiderata for an Object Theory

An object theory of mental images based on our investigation should retain, as far as possible, (1) the characteristics of mental images that have been found central to the revised common sense approach advocated in this chapter and (2) the most valuable insights into the nature of mental images introduced by discussion of the act/object and adverbial accounts. I propose that a metaphysical account of mental images should address and, if possible, decide upon the issues related to the following characteristics.

(1) Object Awareness. Mental images are experienced by us as encounters with presented properties. These properties often cohere, if only temporarily, in a way similar to the coherence of properties given with external objects.

(2) Mode Awareness. We are aware of most mental images (mental images used or created in normal waking states) in a mode that identifies them as experiences distinct from ordinary perception. We are aware of them as being like a sensation or perception of an ordinary object. Imagery experiences, in other words, occur with an accompanying awareness of the context, one that identifies our experiences as being of a certain type or in a certain mode of conscious awareness.

(3) Self-Conscious Awareness. When we are conscious of a mental image (in normal waking states) we are not only conscious of the image, we are self aware, or aware that we are aware of the image.

(4) Relative Being. Mental images are, or can be, understood to be representations of visually similar visible objects. Like publicly visible objects, their being qua image is derivative of the existence of conscious agents who understand the appearance properties of mental images as relative rather than absolute.

(5) Causal Relevance. The common sense view is that mental images bear information in memory and act as tools of imagination and thought. In this sense, they act as mediating elements in a series of causally connected events (e.g., the derivation of a theorem, or the shaping of a statue). In other cases, mental images play a more immediate role and may be identified as either causes or effects themselves. Dreams or after-images, for example, may be either the results or causes of other mental or physical events. Barring strong evidence to the contrary, mental images should be accepted as fitting into the normal causal chain of events in our everyday life.

(6) Inexistence. There should be some identifiable sense in which mental images can be said to "in" the subjects that experience them. This would help give some intelligible meaning to the commonly-accepted idea that statements like "she has a mental image" can be a true statement correctly reflecting the implied linguistic analysis that refers to an inner object.

Enough has been said on the way in which we are aware of the various types of imagery (the mode awareness in which mental images occur) and the various properties we may say "inhere" in the mental images belonging to these types. In the sections that follow, I shall be primarily concerned with the last four desiderata.

8. Self Awareness and Object Awareness

When we are conscious of a mental image we need not be conscious of it alone. We can also be aware that we are aware of the mental image, and this, as Moore says, is one of the aspects we mean to capture when we say that something is a mental, as opposed to a physical, fact. Moore says we are aware of the object of awareness and aware of that awareness in exactly the same way (Moore, 1903, p.449). This analysis cannot be entirely right if it is meant as a full explanation of how awareness is related to objects in consciousness. Aristotle's treatment of this issue indicates one of the problems. Aristotle asked by what sense could we be aware that we are aware of a colored object (On the Soul, 425b10-427a15). Aristotle felt there were only two options. First, we could say that there are two senses aware of the same object. The first sense would be aware of the object alone and the second sense would be aware of the object plus the awareness of that object (as indicated in figure).

Figure 1. Two Models of Consciousness and its Objects. Moore's model (left) and Aristotle's model (right).

This appears to be Moore's solution. The problem with this solution is that the second sense does not yet specifically achieve self awareness. It only posits a higher-order awareness that bears a relation to another awareness. This higher-order awareness in turn needs to be aware of itself, but if this is accounted for by a still higher-order awareness, we have the beginning of an infinite regress.

Aristotle thought the only solution was to accept it as a brute fact that there is a single sense that is aware of itself and aware of its object in the first place. Whether or not Aristotle's pre-Cartesian conception of self-conscious awareness can be likened to our contemporary view is controversial (see Modrak, 1987), but this need not concern us here. Our concern is the general form of the relation between subject and object.

Including self-awareness impacts how we understand the way in which the object is "given" in consciousness. It means, among other things, that we are not aware of mental objects simpliciter, e.g., as a fly might be aware of a moving object. As many imagists have tried to point out, our self-awareness carries with it some nascent knowledge of the individual's own cognitive resources that may be applied to an image. One is aware that other images are "present in power" (as Hume observed), that a rotated image might have a different interpretation, that an image might be of something impossible in the actual world, and so on.

When we refer to the fact that self-awareness is involved in imagery consciousness, this also introduces a possible way to understand how the object can be in the subject. We may be tempted to say, as in Ducasse's theory, that what we are aware of in consciousness is not an independent object but a mode of consciousness itself. It is not, after all, the subject that is "colored," or "square," so it must be some part or sub-process within the subject that has these properties. But if we attribute to some part or sub-process these properties we shall commit the sense datum error. It would be more appropriate to say, then, not that the parts or sub-processes have properties but that they are properties. Still, this does not seem correct, because the question reasserts itself: properties of what? Not the subject -- that has been excluded -- and to say "of an object" just starts the problem again.

The above considerations lead us to question the nature of the independence of the subject and the object once more. The fundamental intuition that consciousness of mental images is due to internal processes in the subject is a strong one, and it may be that one could still conceive of the inner object as a part or sub-process of consciousness. Aristotle took up a similar consideration when he asked how the subject and the object could be separate and yet "one." Aristotle reasoned that the object and the sensing of it are "one" in the sense that they come to be simultaneously. That is, the object, qua something perceived, is only potentially perceptible until it becomes actually the percept of a sentient being. Likewise, a sentient being (or a power of sense) is only potentially conscious of something until it senses something. The two potentialities become a single actuality in the moment of perception. Yet, Aristotle claimed, "the distinction between their being remains" (On the Soul, 425b25) and, he added, "they differ in their modes of being" (On the Soul, 426a15-20). We shall follow Aristotle's suggestion, and turn next to examining the mode of being of the inner object in the hope that this will clarify its status as an object.

9. The Mode of Being of the Inner Object

According to our desiderata, we want to capture the sense in which images are relative beings. At the same time, we want to keep, if possible, some intelligible notion of the inexistence of the object. The first question, then, is to consider how a relative being could be in a subject.

Recall that just as in the case of perception of a visible image, we can, when attending to a mental image, attend either to its visual features or to what it represents. To appropriate the terms "content" and "object," we can make the visual content the object of our attention. And, just as in perception, we have no consciousness proper of a mediating visual content qua representation unless we are aware either (1) that it at least potentially bears a relation to something else or (2) what, specifically, it does in fact refer to. This conforms to the analysis of visible images we gave at the beginning of this chapter, where "X represents Y to Z" was given as a formula indicating what is meant by apprehending an image.

What sort of circumstance can sustain these relations? It seems clear that this is quintessentially a circumstance brought about by the existence of conscious interpreters and that the apprehension of "objects" not physically present is a mental phenomenon without any obvious counterpart in the physical world. If, therefore, fundamental aspects of the mental realm are structured this way, we ought to admit the existence of objects that are part of this structure. Also, there seems to be a logical constraint on our description of the relations. If we assert the reality of the relations among X, Y, and Z, we ought to assert the reality of the relata -- including, therefore, mental objects.

However, such considerations about the relative being of mental images reach a stumbling block when we attempt to address simultaneously other desiderata, such as inexistence. There does not seem to be any clear way to explain this notion. Suppose we say, despite also wishing to claim that the subject and the object of consciousness are metaphysically distinct, that the object must (since it is "in" the subject) be a part or process within consciousness. Aristotle provides an interesting account of this possibility. He asks, "can a part of a substance, whether a contiguous part or a process within it, have relative being?" To this question Aristotle answered "no" (Categories, 8b15-25). A part of a substance does not conform to the requirement that when we cognize a relative thing we also cognize that to which it is related. A part of anything is not necessarily experienced as a part. For example, an old car part lying in a field is not necessarily apprehended as a part of any whole. Aristotle concluded that objects of knowledge and perception bore what was a singular type of fundamental relation to the substances in which they inhered: they were "in" the substance, but not as a part. Perhaps, then, this is the best that we can do to explain the notion of inexistence.

Let us now describe how these considerations come together in the form of a paradox. We have said that mental objects can be understood to be "in" the subject but not as a part. Yet, this is a strange, and perhaps unintelligible, way of putting it.⁽⁷⁾ Mental images themselves cannot literally be of a certain size, shape, or color. That is the sense datum fallacy. Yet we are conscious of presented shape and color properties that seem to inhere in an object, and the relations among subject, inner object, and what is represented seem to be real. We should not state, however, that simply because this is the form of our experience that the object itself unequivocally exists. That would be the existence error. Shall we then say that there exist no objects that have these properties and that there exist no objects that are literally "in" us, but that we are nevertheless aware of them as objects in consciousness? That would indeed be paradoxical.

To state the problem another way, we have in mental images consciousness of properties that we think of as belonging to an inner object, but we cannot assert that an inner object exists in any physical fashion. When it comes to defining how this is to be understood, it appears that we can state this only in a way that appeals to an intuitive understanding of the very notion we are trying to clarify. To say that the "object" is "in" the subject, but not as a physical part, is just to aver to the way in which mental objects are intuitively understood to be "in" consciousness.

One form of solution to this paradox can be derived from Santayana's philosophy of critical realism. Santayana distinguished between those things which have being or reality and those we should most properly say actually exist. The things that exist are beings in flux, determined by external relations and "jostled" by irrelevant events (Santayana, 1923, p. 42). The term "existence" was to be reserved for "facts and events believed to occur in nature" (Santayana, p. 47). There are, according to Santayana, also entities that have being but not existence. These are what he calls "essences," and they include the immediately given sensa of consciousness presented to us in perception and imagination. They include, in other words, what we have described as the appearance properties or pure visual content of mental images. These immediately given data of consciousness are not in nature; rather, they are the evidence by which I come to believe in the existence of exterior things that are in nature.

This account has the advantage that it coheres with a number of the desiderata we have listed. Mental images are real things that we encounter and we can also say that individuals have them. For what the facts and events of nature include, according to Santayana, are such facts as someone intuiting an image. What exists in the physical world is the person having the mental image, but not the mental image itself (Santayana, 1923, p.45). This allows us to say that it is true that someone has a mental image and that the mental image has objective being, but does not commit us to say that it has physical being.

This way of describing the facts also has the advantage that it defuses the question of how mental objects are "in" subjects. Objective essences are neither in nor out of the subject, and they are not parts of anything at all. They have their own objective being outside of time and space. The base fact of our world is that consciousness is able to know essences without altering the essences themselves.

On the other hand, this view is difficult to maintain in light of the desideratum of causal relevance. In Santayana's Platonic realm of essences, essences themselves are causally inert. How can they be effects or causes of human acts if they belong to an entirely different metaphysical realm?

I leave this final riddle unsolved. To solve it would be to give a more complete analysis of "nature," "physical," "cause," and other terms. In short, it would be to review the entire scope of the mind/body problem. Officially, I shall leave the metaphysical status of mental objects an unsolved problem. On the other hand, I believe that there are no metaphysical bars to exclude such objects.

C. The Search for an Alternate Resolution

If the above approach is rejected, there is another possibility -- and this seems to the most popular one to entertain at present. It might be possible to show that terms in Brentano's thesis, "reference," "content," "inexistence," and so on, could have meaning in a scientific theory of the mind that would reduce to one in which only physical terms would ultimately be implicated. This approach would seek to eliminate, not merely describe, the mystery of the mental realm by showing how the explanatory schemes of the physical and the mental are not really in conflict.

At the close of chapter 2, I indicated the rudiments of this approach and how it grew as a response to behaviorism. It is undoubtedly the single most influential approach in both psychology and philosophy in the latter half of this century. The fundamental philosophical impulse behind this approach was recognized by Wundt more than a century ago. He understood that philosophers would never accept the explanatory dualism he thought was inevitable.

The key to the contemporary effort to unify the sciences is the concept of information flow through the brain. There is an ordinary language description of how we acquire our mental abilities that roughly fits an elemental theory of information processing: we take in information, store it, process it, and utilize the results to perform actions. Note that if we substitute "images" for "information," we have just repeated the rudiments of the imagist theory of the British empiricists. There is a profound difference in the information processing analogy, however, because in this case we have no conscious access to the "information" that results in our various capabilities.

This ordinary language description of information flow is, as it stands, nothing but a suggestive metaphor. Can empirical science coupled with a new theory about "information" get past this metaphor? If the brain is regarded as an information processor, there should be no reason why we cannot learn the manner of its operation. The next two chapters discuss the theory and practice of this contemporary effort.

1. I owe this formula to Professor Charles Landesman.

2. The descriptivist proposal for descriptive replacement could be credible, however, if it could be shown that all mental images do and must have a definitive informational content, fully translatable into linguistic form. This possibility is explored in the next chapter.

3. Perkey's 1910 experiment is probably the most famous one demonstrating how subjects can be mistaken about the source of visual presentations. In the experiment, the subjects were fooled into thinking that an actual visual stimulus was the product of their own imagination. This is additional proof that the source of stimulation is not invariably the determining factor in our interpretation of visual presentations.

4. Again, pictorialist computationalists dispute all this. The original information in the image, they would say, is just stored in a different place or different form. Since it has not utterly disappeared, it can still be called either "an image" or "image information."

5. Though, as we stated earlier in this chapter, whether or not there are unconscious mental images in perception is a point that might be debated.

6. I owe this way of stating the "object problem" for mental images to Charles Landesman.

7. Aristotle equates, at one point, being in a subject but not as a part with "being incapable of existence apart from the said subject" (Categories, 1a22). This is a less unintelligible way of putting it, but it also locks in a certain metaphysical interpretation of the mode of existence of mental objects. This is point that might be pursued further.