Mental Imagery Debate

There is an ongoing debate between analog and propositional theories and this debate remains a major controversy in the field of psychology that deals with mental imagery. According to the analog theory, analog code contains mental images. Analog code or in other words pictorial or depictive representation is such a presentation that has resemblance to the physical objects. The images are contained in the code in such a manner that they are presented in the way they look in reality. The analog code has the following characteristics: it is of iconic nature, it looks like a map or a picture, and it is possible to manipulate mental images.

There is an experiment to support the theory. It deals with size and imagery. According to this experiment, it is necessary to imagine a rabbit and an elephant near it. Then the next objective is to say if this rabbit has eyelashes. The next task will be to imagine that there is a fly near the rabbit and after that say whether the rabbit has eyebrows besides eyelashes. It is easier to see eyebrows than the eyelashes. The experiment is meant to show that it is easier for a person to describe the details if the object is large.

Propositional coding (or descriptive representation) is such a language-like representation; besides, it is an abstract representation. The storage neither takes up some space nor is visual. There is no physical resemblance of the stimulus. Such code has the following characteristics: it is symbolic and descriptive. This theory can be supported by the so-called “part-whole relationship” experiment.

The participants of this experiment are asked to take a look at the picture. Later they are asked whether they remember shape depicted in the picture. If the participants did not notice or see the shape after the first time they had looked at the picture, they obviously will not remember it. When the experimenters ask them to draw this picture the participants are able to see the shape. This experiment proves that people are not able to mental reinterpret the images. It means that mental images should not remind of their physical representations, they should be coded in words.

Cognitive psychology is largely concerned with studying mental imagery. This field of psychology was accepted almost 15 years ago and since then the investigation of processes, related to reasoning and perception-like experiences, has been a central problem of the new mentalistic theory. The literature about the imagery contains many debates regarding this problem but not all debates are very significant or important if viewed from the theoretical point of view. It is possible to make arguments about discrete or continuous nature of images, or whether they are abstract or concrete, articulated or holistic, if they describe things or depict them. It still has not been clearly stated whether images have an entirely new cognition form or they are just a kind of a form that is used in the cognitive processes. The level that makes them different or the same is not clearly defined.

Arguments have been made regarding epiphenomenal nature of images or if they have a functional role in cognition. The latter problem can be addressed only after identification of a theoretical stand regarding properties of the images. First, it is necessary to understand what the meaning of epiphenomenal is and then something can or cannot be called epiphenomenal.

For instance, if an image is referred to what is experienced by a person when he or she imagines a scene, then it is obvious that it exists just like any sensation of consciousness (for example, tickle, pain, etc.). When an image is referenced to a specific theoretical concept that has certain properties and role during certain processes of cognition, it would be appropriate to ask a question regarding the fact if theoretical claims of some concepts are warranted and true instead of asking whether these concepts are epiphenomenal.

The main theoretical question that arises in this controversy would be whether it is necessary to make a postulate of some specific kinds of processes, just like those that are usually referred to, when the term of analog is mentioned. The focus of the discussion should be on the claim, referred to in the literature regarding imagery. This claim is the alleged spatiality of mental images. A set of experimental findings that are prototypical in their nature have to be discussed. It is considered that images, especially those that show “mental scanning” or “mental rotation” establish particular property of images.

It is doubtful that the only viable issue dividing the defenders of the debate, called “proposition versus images”, remains the question regarding specific cognition aspects that are associated with imagery and should be taken in consideration. In other words, the consideration should be made whether these aspects should be explained in terms of the processes operating via symbolic codes of rules and various representations (goals, beliefs, etc.) or if these aspects be viewed in form of properties of some representational media and mechanism that cannot be altered in arbitrary ways by tacit knowledge.

When discussing the question regarding epiphenomenal nature of images, Kosslyn et al. claim that “none of the models of imagery based on Artificial Intelligence research treat the images that people report experiencing as functional representations” (p. 536). The basis of this widely spread view lies partly in the misconception regarding what exactly is reported as imagery. According to Hebb, people report not properties of the image, they report objects that they imagine. The properties of imagined objects are shape, color, size, etc. It is of utter importance to make this distinction. Perhaps, the most general conceptual confusion that brings a lot of damage to the literature on imagery is the scope slip which makes an image of an object X, for example, possess property P and, in fact, it is object X with property P. Whereas it would be correct to say that it is the object X’s correct image with property P.

It is not just a figure of speech. It has a lot of weight in explanations of imagery phenomena, when it is necessary to consider usually accepted characters of images of “spatial” nature. There was an experiment, undertaken by Kosslyn, Ball, and Reiser, that involved “mental scanning” of images. It showed that the longer the distance from the image and the item on this image, which is the object of focus, the harder it is to see it and focus on it, not to mention how hard it is to describe that item. For instance, there is no doubt and dispute regarding Kosslyn’s et al. claim according to which “these results seem to indicate that images do represent metrical distance” (p. 537). The spatial extent of the images is asserted in the next sentence, though. In other words, the image has size or length rather than represents the images. Such transformation is necessary and important to the experiments conducted and promoted by Kosslyn and others.

Alternative 1 seems to appeal the facts regarding the world that are symbolically encoded as well as to the rules that are used to transform representations or to draw inferences. Such an approach is called the “cognitivist” one and it is advocated and promoted by Fodor, Chomsky, and others. The analog approach to mental processing and representation is presented by Alternative 2. The very term “analog” refers to a broad range of peculiarities of models as well as representations that cover almost everything, starting from representations’ mathematical continuity and finishing with simple requirements used for the representations that pass through intermediate stages that would be used by the actual system (e.g., Shepard). All these aspects partially represent the meaning of analog. Although, it would be possible to say that the only relevant analog’s aspect to the debate about imagery is the aspect of distinction between the two alternatives. An analogue process, presented by Alternative 2, is considered to be the one the behavior of which must be characterized in terms of internal lawful relations of properties in certain physical process instantiation, instead of the rules and algorithms of some processes. It does not matter if people use an “analogue representational medium” (e.g., Attneave) or a “surface display” (e.g., Kosslyn et al.), they still think it is a usual thing and take for granted the fact that this medium includes a system of rightfully connected properties or internal limitations. It is exactly the way of behavior for this set of relations and properties that will be the determining factor for the objects in this medium. These people particularly compare these accounts with those, presented by Alternative 1, which claims that the way a person behaves is more related to the person’s knowledge of the actual behavior and not the properties of the medium.

There are different concepts regarding analog processing but they have a lot in common. It is possible to make any process go through a certain sequence of states of the appropriate nature and these processes can be made purely verbal. This model, though, would not be counted as analog model, when such a mechanism is not naturally intended to go through such a sequence. The process would be considered as an analog, if the process of its passing through certain intermediary states is a needed consequence that internal properties of some medium or mechanisms have. If some mechanisms get artificial restrictions, according to which they would carry out the task in different ways, it is not possible to talk about analog process. Palmer, one of the researchers of these theories, made a similar distinction between non-analog and analog processes. This gives him an opportunity to conclude that these forms of processing will be distinguished by only biological evidence. If analog mechanisms are contrasted to those that work on the representations and tacit knowledge, it is possible to see that this distinction is functional and made by behavioral criteria. One of such criteria as an example will be discussed, whereas other criteria were discussed by Pylyshyn.

Picture theory, as the one of the imaginary theories, is recognized as one of the fundamental theories of imagery (and it is widely accepted). This theory is the part of human language, so people are not able to criticize it due to substantial bias. On the other hand, modern philosophers (the ones who represent 20^th and 21^st centuries) criticize it rather substantially but sometimes unjustifiably. They argue that people may not have experiences of quasi-visual nature.

It should be noted that picture theory in modern interpretation should not be recognized as a theory of folk origin. Stephen Kosslyn and other specialists in this area have created initial concepts of the imagery of mental state. Newer versions of the theory state that the mental image in its visual component can be presented as a pattern of two dimensions that affects neural sensation in the special (retinotopical) visual brain cortices. These patterns occur when a person sees something; they are isomorphic in terms of stimulation of optical eyes’ retina nerves. Kosslyn believes that images in our brain are mostly formed by internal sources than by sensory input.

The quasi-pictorial theory has allegedly empirical shortcomings but they are not to be discussed in this paper. Introspective as well as empirical evidence prove that there are ways that show the difference between the looking experience and the imagery experience. Some experiments provide the theory with strong evidence. During this experiment, the subjects are shown using pictures that allow multiple interpretations – picture of the duck-rabbit or the cube of Necker. If is very difficult to mentally interpret both images, whereas it is easier when one can physically see a picture in front of him/her. The person experiences the same difficulties when he/she draws the picture of him-/herself, based on the image he/she has in his/her mind.

Kosslyn claims that neural excitation’s pattern that he considers as the one, constituting the images in mental area, is just a picture (as in some metaphorical or extended meaning). It is a quasi-picture. This “quasi-picture” differs from usual pictures (pictures, photos, drawings, projected images, etc.) as it is not necessary to observe it physically to get data from it (and, probably, experience it). The pattern of neural excitation just reminds a picture because it is a representation of spatial relationships in the projection of two dimensions of the visual scene as the broaden pattern of sensation has topologically the same spatial relationships.

Kosslyn sticks to his opinion, as his theory does not make a full and literal identification of mental images and pictures. His theory is still full of controversies. There are a lot of empirical data gathered for the modern theory`s pictorial variation. The way humans utilize mental images in form of representations is the empirical data gathered for the theory. But, the researchers, who support this theory, pay little attention to the issue, regarding conscious experience of the imagery. Apparently, there are two major ways used to solve this problem. One way is the quasi-picture that appears in the cortical area and it is a conscious picture. Another way is that consciousness comes from an internal cognitive structure. In other words, this quasi-picture is sometimes quasi-seen by the mental eye, sort of speaking.

The problem of quasi-picture, being a conscious one, runs into consciousness problem: there is no knowledge how excitation patterns are formed in some parts of the human brain (this excitation is just a choreographed movement of ions and molecules through membranes and around them). It is unknown how these patterns evoke any conscious experience, not to mention the practice of spatial extension of isomorphic structure to the brain areas that form these patterns.

Anyway, it seems that Kosslyn prefers the option number two. He is one of the major modern advocates of the picture theory. According to him, a “mind’s eye” examines the quasi-pictorial image that has been evoked in the brain. It seems that the researcher builds explanations of his findings on this notion. Maybe then, just like usual visual consciousness occurs when people grasp visual information with their eyes, the imaginal consciousness occurs when the information from the internal picture is obtained through the mind’s eye.

The “mind’s eye” means that this theory is related to a homunculus argument. This mind’s eye belongs to a little human being inside a person’s head and this man experiences the internal world of these quasi-pictures in the same way that this person experiences the outside world with his usual eyes. Kosslyn’s theory of quasi-pictorial imagery reminds a more detailed variation of the imagery theory, presented by Descartes, although at some formal level. Descartes also saw mental images as some material quasi-pictures, formed in the human brain, and the conscious experience of them happened not because they were present in the brain, but because the soul somehow “saw” them. Kosslyn clarifies that it is not his intention to follow Descartes’ theory in which he attributes human consciousness of pictures, appearing in the brain, to some homunculus that lives there, because it is not a scientific approach. Although this quasi-pictorial theory does not necessarily mean Cartesian dualism, there is still some commitment to materialism of Cartesians. This notion means that person’s subjective experience is dependent on the brain structure, acting as an internal spectator (a conscious one) of the representations that happen in the brain.

The supporters of the quasi-pictorial theory appear to have a retort or two to the homunculus objection. One of them points out that a computer simulation was applied to the quasi-pictorial theory. Therefore, a computer cannot contain any homunculus inside. Although, it seems that this program of the quasi-pictorial theory simulation does not attempt to make a model of the imagery’s conscious nature. This program, just like the major part of Kosslyn’s work, is meant to explore and find out how some sorts of thinking in spatial way can be achieved via imagery and such operations like linear scanning across a quasi-picture, and its rotating is to be shown from a different angle. The program’s operations are performed because of the inner structures of data, which are meant to model some functional levels of those quasi-pictures of the brain that are supported by the theory. The results are displayed, as actual pictures appear on the computer screen in some relevant ways. The program does not attempt to make a simulation of the function of “mind’s eye”, so the consciousness, simulated or real one, does not play any part in it at all. There is no doubt regarding the fact that there is no homunculus in the computer, however, the system makes a model of consciousness; such modeling is made in cooperation with a human being, who plays the role of the homunculus.

There is no coincidence in the fact that the program does not model the imagery’s intentionality. The functioning of the program is in no way related to the patterns of two dimensions, produced and manipulated by it in the way that they seem like pictorial perceptions to humans who look at them (the programmers). People also have programmed them to look that way. They present nothing for the machine.

The second retort of the supporters of the quasi-pictorial theory is not to be easily dismissed. Maybe modern cognitive science is able to prove that there is no homunculus and the “mind’s eye” by proving that it is only a bundle of neural or computational processes. The original context of the development of the quasi-pictorial theory was in the paradigm of the “information processing” of the time when it was developed (the 1970’s).

According to the “information processing” theory, vision is such a way of perceiving visual data where it is first perceived using the eyes, then a series of processing stages occurs in the human brain and after that the data is transformed in a form, which is acceptable for behavior. Developed theories of information processing usually involve a lot of top-down modulations, which are characteristic for the data flow, provided by the organs of sense, and this flow is comes in the bottom-up direction. This flow direction suppresses and controls perception.

There is a usual and indirect assumption that the final representations’ set, ultimately produced by the processing system of visual information, causes conscious visual background. Although, some researchers of information processing theories tend to view visual consciousness as something that comes as a result of the general systems` work and not as an attachment to representations. The whole human organism together with sense organs and muscles responsible for behavioral response take part in information processing.

The theory of quasi-pictorial imagery claims that at the initial stages of processing visual information there are quasi-pictorial representations created in the human brain. They can derive from what a person actually sees or the actual visual input. They also can be obtained from information stored in memory, in other words what people imagine or remember. In both cases, they go through various processing stages because useful information will be extracted from them. These later stages are something that constitutes the function of the “mind’s eye”.

Therefore, an argument can be made that the role of the mind’s eye is independently motivated and principled in the broad sense of the vision theory. The progress does not stop and scientists work hard to investigate the subject in order to understand how visual information is processed through mechanisms that are considered computational as well as neural. The function of “mind’s eye” promises success in this never-ending program aimed at researching mechanisms of information processing.

Even if this theory of data processing finds the necessary framework that would help in understanding visual perception, although not all scientists agree with that, the processing results will be simple representations, instantiated like neural activity. Again, there is a problem whether such representations should be considered as mindful and complete or there is a need for another homunculus in order to read these representations and understand what they actually depict. Maybe this homunculus will be used in a series of data processing stages. This fact will only lead the researchers back to the initial results. The understanding of the mechanism of appearance of quasi-pictures and how they make imaginal experiences is still at its earlier stages.

Description theory, advocated by Zenon Pylyshyn, was an effort to realize how imagery phenomenon could be reviewed in terms of computational theory applied to the human mind. Pylyshyn states that computers (and their biological equivalent - brain) are designed to represent information more like the set of symbols or language than as referred to images or pictures. The language of thought, as an innate language known as mentalese, is hypothesized to be different from any actual language a person speaks. However, it is still like language. It has symbolic tokens representing something particular in the environment as words of any language do.

Thus, in case of real and “natural” languages (English and Chinese, for example) the correspondence between the words and their meaning (a “cow” means a cow, but not a dog, for example) is set up by the appropriate convention within society (Pylyshyn, Elcock, Marmor & Sander). However, a human-written computer program requires such a correspondence; it is regulated by a programmer. This fact leads to the certain difficulties in determining the process of words or phrases of mentalese’s referral to the real-world objects outside the mind. Such issues require extensive research and further exploration of the targeted area. According to the hypotheses the solution exists (Pylyshyn, Elcock, Marmor & Sander).

It is being theorized that in case computer-based systems have language interpretation schemes that describe images by words, human brain does the same – it represents visual images using descriptions. Considering such an approach, these mentalese descriptions should be recognized as the final product of mind information processing (Pylyshyn, Elcock, Marmor & Sander). Therefore, brain is responsible for the word-based representation of human experiences based on perception. It is clear that people do not percept descriptions of the brain as the actual descriptions. However, it is still assumed that these descriptions constitute the experience based on perceptions.

There is an opinion that people experience imagery of mental origin in the following cases: if mentalese-provided descriptions, which describe scenes of visual nature, are received from memory in greater extent than through eyes, for example, as input of sensory system, or in case these descriptions are created by the parts and bits of different descriptions retrieved from memory(Pylyshyn, Elcock, Marmor & Sander).

Pylyshyn states that due to the fact that descriptions can be not complete and can provide different information not fully (these may be not only the small details, but, in some cases, even information regarding the scene’s global structure), the description theory explains the human imagery much better than the above-discussed picture theory. On the other hand, among the major motivators of the theory is the idea that it can provide better explanation of computational representation of fundamental facts.

It should be noted, however, that the followers of the description theory did not pay too much attention to the issue regarding the process when images are experienced consciously. They are similar to the followers of the quasi-pictorial theory. It is clear that humans do not fully realize the human brain processes related to perceptions (or images), so it is impossible to state firmly that mentalese is something people know (Pylyshyn). However, human beings are aware of the fact that such a brain language of description images can be real and it is recognized consciously.

Considering the discussed theories, it is more plausible that human brain uses language descriptive information taken from the variety of sources – sensory systems, memory, etc. It means that descriptive theory is a contemporary theory, based on better understanding of brain processes; thus, it provides more solid and trustworthy grounding for mental imagery recognition in a human brain.