Much research in recent decades has attempted to answer the question as to whether imagination of, for example, visual imagery, involves the same brain mechanisms as actual visual perception. How do images differ from actual perceptions, and how are they alike? This paper will summarize evidence which shows that visual imagery not only involves brain activity, but many of the same brain structures that are involved in visual perception are involved in visual imagery as well. A computer model of the cortical visual system and the effects of damage to that system are discussed in some detail. The interaction between visual perception and visual imagery is also briefly discussed, and its implications for perceptual disorders, such as hallucinations in schizophrenia.
Role of the Visual Cortex in Perception. The human striate cortex (called V1 or "primary visual cortex") in the occipital lobe of the brain has been long known to be an indispensable component of form perception in the visual system. In a classic study, Hubel and Wiesel (1968) showed that the V1 area in cats (primary visual cortex or striate cortex) is composed of cells that respond to definite qualities of stimuli, such as a point or line. The experimenters identified cells as "simple", "complex", or "hyper-complex", depending on the complexity of the stimulus to which they responded, with "hyper-complex" cells often responding only to lines in a certain orientation, for example. Few deny therefore that the V1 or striate cortex (Brodmann area 17 in humans) is involved in the 'nuts and bolts' of visual perception, with more subtle qualities of the visual stimulus, such as color, shading, etc. being covered by V2 and higher or "secondary" visual areas (Brodmann areas 18 and 19). When looking at an object, information is passed from the retina through the brainstem and the thalamus to the visual cortex. From the primary visual cortex (V1 or striate cortex) the information is passed to many distinct areas within the occipital cortex. All of the above areas are "retinotopically organized" (Ochsner and Kosslyn, 1999), meaning that the spatial structure of the object represented corresponds approximately to the image on the retina itself. A disproportion-ate area of the retina represents the center of the visual field, and this disproportion is represented in all of the other areas to which the information is passed. Pollen (1999) argues that the human striate cortex (V1) is not just a relayer of visual information from the brainstem and thalamus, but an "indispensable component" of a neural pathway concerned with achromatic form perception.
Kosslyn et. al. (1990) theorized that a series of subsystems must be involved in processing visual object recognition and identification. They used knowledge about what the visual system can do as a jumping off point. For example, we can recognize a human form in many different postures, such as standing, sitting, squatting, etc. these abilities had long go been postulated to rule out simple template theories of shape recognition (Neisser, 1967). We can also identify objects when they appear in different parts of the visual field, when they subtend different visual angles, when they vary in shape (eg. oak leaves), or when they have optional parts (eg., some chairs have arms, others do not). Their task was to produce a computer model that could perform all of these perceptual tasks. Research had shown that many cognitive processes could be modeled with neural networks. Kosslyn et. al. set out to model the visual recognition process using processing components each of which might correspond in the brain to a separate neural network. A second goal of the project was to consider numerous neurological syndromes which affect vision, and see if doing damage to the system at various points could shed some insight onto the problem of how lesions or dysfunctions of the brain could produce various neurological syndromes. Using computer simulations, they were able to show how dysfunctions can arise from the disruption of the various subsystems. To understand their model, it is first necessary to understand their theory of how the visual system of the brain functions, much of which is articulated in other studies.
Kosslyn and his colleagues have attempted to form a theory which could explain how the brain processes both perceived and imagined visual data. A set of retinotopically organized mapped areas are postulated to work together to separate figure from ground, and all of these areas grouped together have been designated as a single structure with a specific function, and have been given the name "visual buffer" (Kosslyn, 1994). This area consists of many more specialized components, one sensitive to wavelength (area V4), another to motion, and so on. Only a small amount of information in the visual buffer can be processed at one time. Thus an internal "attention window" selects patterns in the "visual buffer" for further processing. (Ochsner and Kosslyn, 1999; Kosslyn, 1994). Information selected by the "attention window" is sent over two parallel pathways in the cerebral cortex, the ventral, so-called "what" pathway in the inferior temporal lobe, which is concerned with properties of the object, thus determining what the object is, and the dorsal "where" pathway in the posterior parietal lobe, which is concerned with spatial properties of the object, thus determining where the object is with respect to other objects in the field (Ungerleider and Mishkin, 1982). In order to recognize an object (or to imagine it), we must have stored some sort of representation of that object in our brains. These representations are stored in perceptual encoding systems that store properties of an object from various sense modalities (Kosslyn and Koenig, 1992, Schacter, 1990). Visual perception and imagery are both postulated to use these mechanisms. The simplest form of perception is often referred to as a "bottom-up" process because it is essentially stimulus driven, whereas operations involving imagery are sometimes referred to as "top-down" mental processes, because they involve and are driven mainly by representations or images stored in higher visual or extravisual cortical areas, i.e., they are essentially image driven (Kosslyn, 1994). As will become apparent later, more complex forms of perception involve "top-down" as well as "bottom-up" mental processes. Moreover, both mental processes make use of the dorsal, parietal "where" pathway and the ventral, inferior temporal "what" pathway (Kosslyn et. al., 2001).
Having described the theory of Kosslyn and his colleagues of the functioning of the visual system, it will now be attempted to describe further, if only briefly, the computer simulated model of Kosslyn et. al. (1990), and what conclusions the experimenters were able to draw from it. A computer model was built to try to simulate using neural networks the known areas associated with visual perception in the macaque brain and their connections. A ventral and dorsal system ("what" and "where" pathways) were simulated just as are known to exist in the real primate brain. It is known that all connections in the visual system are two-way; that is to say that every structure in the visual system of the brain that sends information to another area also receives information from that area. A "visual buffer" is encoded to simulate, roughly, the two-dimensional geometry of the projection of the "object" (actually a 60 x 60 pixel array or crude "picture" of an object) on visual brain structures. An "attention window" represented the selective aspect of perception, since it is thought that the brain can only attend to one spatial region at a time. (Cave and Kosslyn, 1989: Posner et. al., 1980). "Objects" are encoded by the visual buffer as a 20 x 20 pixel array. Features of an encoded "object" are compared to other "objects" stored in memory. As one might expect, the program was very limited (in comparison to a primate brain) in what it could do. Obviously it could only identify two dimensional pictures, not three dimensional objects, and only a very limited number of pictures could be identified. The system was capable of performing only highly simplified versions of tasks. The "what" and "where" systems of the primate brain were divided into subsystems in the model, to enable a crude simulation of their functioning. Is the construction of two separate pathways the most efficient way to program these functions into a model? Interestingly, Rueckl, Cave and Kosslyn (1989) had previously produced computational models which showed that a single structure or neural network that identifies both the object's form and location is far less efficient than two separate neural networks, one for each computation. Thus the dual pathway system in the primate brain was simulated in the model.
The dorsal "where" system consisted of the following components: 1.) Spatiotopic mapping - "where" information in the visual buffer is retinotopic; i.e., determines "where" in relation to the center of the retina (fovea). 2.) Categorical relations encoding - codes spatial information into a long term associative memory, where it can be combined with information about object properties from the ventral system. 3.) Coordinate relations encoding - some objects can be identified only by noting subtle distance ratios, eg., distance between eyes, distance between nose and mouth, etc.
The ventral "what" system consisted of the following components: 1.) Preprocessing - To recognize objects from different visual angles, vantage points, etc., the model must find a way to process the data so that aspects of the object that remain the same under these conditions are recognized. 2.) Pattern activation - object identification is accomplished by comparing new objects against previously stored information. This subsystem contains visual representations that specify visual properties of previously seen shapes. 3.) Feature detection - makes judgments about characteristics of objects. 4.) Associative memory - stores information associated with previously seen objects: name, facts about object; the category to which it belongs, etc. A third major system of the model was programmed for "top-down" hypothesis testing. It consisted of the following subsystems: 1.) Coordinate property lookup - looks up in associative memory the properties the object should have. 2.) Categorical property lookup - looks up properties of categories. 3.) Categorical conversion - when a categorical spatial relation is looked up, the category must be converted to a specification of a location in space, which proved to be very complex. 4.) Attention shifting - adjusts the attention window as appropriate.
The "where" information from the dorsal pathway and the "what" information from the ventral system are transmitted into associative memory where they are stored in a short-term memory bank and compared to long-term memory banks containing information about the objects. The system, when functioning properly, could name and classify familiar objects shown in the pictures. There were only two categories, foxes and faces, and only eight pictures, including a partially occluded face and an unfamiliar twisted fox. The program looked for distinctive features, such as paws, or a human nose. If the picture was occluded or distorted, the program would not be able to match as many familiar features, but could still make the classification, although with somewhat less confidence.
Effects of Damage to the System To assess the effects of damage, processing could be disrupted within a system, or connections between systems could be broken. There were often competing subsystems or memory representations within the system, so that damage often resulted in a disruption of the normal balance between them, and "compensations" could often result by causing the subsystems to be used in different circumstances after damage. For example, a lookup function tries to find the best fit between the current picture and a stored memory of an object. If a connection is damaged, the system might find the second best fit instead. Considering complete damage to a subsystem, or the severing of data lines, there were 44 distinct categories of damage that could be done to the system, and the authors calculated that there were trillions of possible combinations of damage. Many combinations, however, produce the same result, because damage upstream often causes the system to fail before downstream connections even occur. There were two categories of damage according to the authors' classification: 1.) failure in the ability to represent and interpret perceptual units, and 2.) failure in the ability to represent and interpret spatial relations among units. In both categories there were deficits that the authors felt mimicked deficits seen in brain damaged or stroke patients.
In category 1 were the following deficits, which the authors contended were analogous to deficits seen in human patients: 1.) Visual object agnosia - the program lost the ability to name objects. 2.) Prosopagnosia (literally means "face blindness") - the program can classify the object but not name specific objects. A person with prosopagnosia can see individual features, but cannot put them together to identify a particular face (or fox). 3.) Metamorphopsia - Objects appear larger, or fragmented and compressed. This would occur in the program if the spatiotopic mapping system did not compute size correctly.
In category 2 were the following deficits: 1.) Simultagnosia - Inability to perceive more than one shape at a time. The model displayed this when the spatiotopic mapping system was partially damaged, so that all stimuli were assigned to the same location. 2.) Visuospatial disorientation - patients who fail to localize objects in space. Such a malfunction would occur in the program following damage to the visual buffer that results in degradation of the input so that perception is registered, but shape recognition is not possible. 3.) Disorders of visual search - Patients who have "paralysis of gaze" or visual scanning disorders - fixation on a stimulus without being able to release their gaze. This can be caused in the model by damage to the property lookup system, causing it to look up properties again and again, or by damage to the attention shifting subsystem. 4.) Hemineglect - Patients who ignore everything on one side of space, usually due to damage to the opposite parietal lobe. The model does not have bilateral symmetry, but a partial neglect of the visual input would be possible, which could be caused by a disorder of the attention shifting system.
The authors noted that many combinations of damage could produce similar deficits, and concluded that neurological testing may therefore have to be more subtle in the future to take into account and identify the many types of damage that could produce a given deficit.
How does the process of imagining an image compare with the process of perceiving the image? It has long been known that imagery involves the activity of brain structures, because lesions that affect perception also affect imagery. For example, the "neglect syndrome", which is defined as the failure to perceive or attend to stimuli on the side of space opposite a parietal lesion, was shown by Bisiach and Luzatti (1978) to apply to imagery as well as perception. Two Italian patients with left hemineglect were asked to imagine the famous Piazza del Duomo in Milan when facing the square while standing at opposite sides of the square; first at the north side of the square facing south, and then at the south of the square facing north. When asked to imagine themselves in the first position, they described only objects on one side of the square, whereas when asked to imagine themselves facing in the other direction, they described only objects on the opposite side of the square. In both cases they were describing only objects that would be on their right side. Moreover, Farah (1989b) showed that lesions to brain areas involved in specific aspects of visual functioning, such as color or localization, tend to affect imagery as well as perception. Mental visual imagery as measured by asking subjects to imagine taking a route involving several turns to a particular destination, was found to change the regional cerebral oxidative metabolism as well as regional cerebral blood flow in 25 cortical fields. (Roland et. al., 1987).
Experimenters have differed, however, on whether the primary visual cortex is equally involved when we simply imagine a stimulus. It may be that when we imagine a stimulus, the brain activates all of the brain structures that were involved in the original perception of the stimulus, or it may be that the primary visual cortex, (and perhaps secondary visual cortex as well), are activated only in visual perception but not during visual imagery. Roland and Gulyas (1994) have postulated that primary visual cortex may not be necessary at all, while Kosslyn and Ochsner (1994) took the opposite position. Roland and Gulyas argue that because some brain damaged patients can be found who have lost the capacity for visual imagery but not visual perception, that different brain structures must be involved. It is known that in imagined scenes, the parietal and temporal visual association areas are activated (Roland and Friberg, 1985). Unlike the primary and secondary visual cortical areas, these areas are not retinotopically organized. The argument put forth by Roland and Gulyas is that retinotopically organized areas are better suited for computation (as would presumably be involved in precise perception) than for representation (as would presumably be involved in imagining a stimulus). Thus the early, retinotopically organized visual areas might not be necessary for visual imagery. Kosslyn and Ochsler's (1994) position was based on a study by Kosslyn et. al., (1993) who found that V1 cortex was activated during PET scans of subjects who were asked to imagine letters on a grid, as opposed to subjects who were actually viewing the letters on the same grid. Indeed the precise coordinates of the activated regions were similar in the real and imagined stimulus conditions. Other studies seem to support Kosslyn and Ochsner's position. Using functional MRI (LeBihan, et. al., 1993), and positron emission tomography (PET) (Kosslyn et. al., 1995), activation of widespread regions of the occipital lobe including prestriate areas V2 and V1 have been demonstrated during the formation of imagery in experimental subjects. Thus an alternative explanation, which even Roland and Gulyas do not rule out, is that back projections from the areas involved in imagery may actually reconstruct the image on the "grid" presumed to be located in the retinotopically organized visual areas in the occipital cortex. Whether this happens or not could presumably depend on how precise an imagination of a stimulus is demanded. Klein et. al. (2000) showed that area V1, for example, was likely to be activated in visual imagery only when images with many details are formed and used in a task. However, the activity was much greater when subjects were asked to evaluate characteristics of an object, regardless of whether the object was real or imagined. Kosslyn et.al. (2001) showed that imagery activates retinotopically organized visual cortex (Brodmann areas 17, 18) in only some imaging tasks but not in others. The authors argue that these areas are more likely to be activated in tasks where the subject is asked to try to find high resolution detail in a mental image. Kosslyn (1980, 1994) therefore argued that visual mental images are "depictive", and indeed use retinotopically organized visual areas to represent an image. It appears that the same visual areas are used for either perception (stimulus driven or "bottom up" construction) or imaginary ("top down") construction of a form or shape, but different extravisual areas (Ochsner and Kosslyn, 1999). Visual areas activated will differ, however, on discrimination tasks, depending on what types of visual cues are used to make the discrimination. For example, Gulyas et. al. (1995), using rCBF and PET scans, showed that distinct areas in the visual cortex were activated depending on whether subjects were asked to perform a form discrimination or a color discrimination task. Even if retinotopically organized visual areas are involved in image generation, the question still remains whether a "depictive" image is identical in the brain to a perceived image. D'Esposito et. al. (1989), using functional MRI, found that when subjects were instructed to generate a mental image of a word, visual association cortex was activated, but not the primary visual cortex (V1). More recent studies have shown that a great deal of overlap occurs between brain structures used in visual imagery and those used in visual perception. Galis, et. al. (2004) had subjects either visualize or simply see faint drawings of simple objects, then were asked to make judgments about the images in the drawings. There was a great deal of overlap in brain structures activated in the perception vs. imagery conditions, but there was more overlap in the frontal and parietal areas activated than in the occipital lobe, where the retinotopically organized representation of visual stimuli are known to occur, perhaps giving more credence to Roland and Gulyas proposition that retinotopically organized visual cortex is at least not as necessary for visual imagery as for visual perception. However, the authors argued that the brain regions activated in common may be involved in performance of the tasks rather than in perception vs. imagining of the visual stimuli. The common areas activated in the frontal lobe, for example, could be explained by Kosslyn's (1994) argument that the frontal lobe is involved in shunting information from various brain areas representing the various sense modalities, and the same areas would be involved whether the information being shunted were perceived or imagined. The common parietal areas activated could be explained in a similar manner, since certain parietal areas are involved in the same cognitive control system as the aforementioned frontal areas. Mechelli et. al., (2004) found that, during visual perception, activation in non-striate visual cortex involved forward connections from early visual areas, whereas during visual imagery, activation involved backward connections from prefrontal cortex. However, there may be subtle hemispheric differences between operations involving visual perception and those involving imagery. The predominance of evidence favors left hemisphere dominance for mental imagery as opposed to perception (Farah, 1986, 1989b; Farah et. al., 1985), although some specific tasks such as mental rotation appear to be right hemisphere dominant (Farah, 1989b).
Does mental imagery interact with perceptual processing? Farah (1989a) showed that when subjects were asked to form an image of a letter "T" or "H" on a grid, stimuli falling within the grid points containing the letter were more likely to be perceived than those falling outside those grid points. Heil and Henninghausen (1993), however, showed that this was only true for "compact" images, as opposed to patterns that were disjointed, such as a pattern of squares as opposed to a letter. They argued that a "compact" figure, such as a letter, caused the subject to segregate figure from ground within a field, with more attention being directed to parts of the grid within the figure than to parts of the grid composing the ground. Less compact systems did not have this effect. Much research has been done in recent years to discover what implications the interaction between imagery and perception may have for hallucinations in schizophrenia. Hallucinations have been assumed to result from internally generated information being misinterpreted as being externally generated. Some have argued that increased vividness of imagery may make images less distinguishable from perceptions, thus making the distinction between them difficult (Johnson and Raye, 1981; Johnson et. al., 1993). This could account for hallucinations in schizophrenics. Aleman et. al., (2003) found a significant difference in performance on several tasks designed to measure imagery-perception interaction between hallucinating schizophrenics and normal controls, although the schizophrenics did not differ significantly from controls in tasks designed to measure their ability to form images and use them in a task. Aleman et. al., (2005) found a significant difference between patients and controls on an object imagery task, but not on a spatial imagery task.
In conclusion, it appears that visual perception and visual imagery may involve many of the same pathways in the brain. A computer simulated model of the cortical visual system demonstrates that damage to the system affects top-down (image driven) process as well as bottom-up (stimulus-driven) processes. Imagery is known to be associated with measurable brain activity. Cortical pathways have been identified which are involved in both visual perception and visual imagery. Much debate has occurred over whether retinotopically organized areas of the visual cortex are involved in imagery as well as perception. It appears that they are, the extent being determined by the precision and detail demanded by the particular imaging process or task. Furthermore, it has been shown that there are significant interactions between imagery and perception, both in normal and abnormal processes. Recent studies have attempted to assess the possible implications of these findings for hallucinations in schizophrenia.
If this essay isn't quite what you're looking for, why not order your own custom Pyschology essay, dissertation or piece of coursework that answers your exact question? There are UK writers just like me on hand, waiting to help you. Each of us is qualified to a high level in our area of expertise, and we can write you a fully researched, fully referenced complete original answer to your essay question. Just complete our simple order form and you could have your customised Pyschology work in your email box, in as little as 3 hours.
This Pyschology essay was submitted to us by a student in order to help you with your studies.
This page has approximately words.
If you use part of this page in your own work, you need to provide a citation, as follows:
Essay UK, Essay on Visual Imagery | Psychology. Available from: <http://www.essay.uk.com/free-essays/psychology/visual-imagery-neurobiological.php> [23-02-18].
If you are the original author of this content and no longer wish to have it published on our website then please click on the link below to request removal: