Figure 913

According to Kosslyn's theory, images are constructed in parts, so one might first form (a) a skeletal image of a duck, and then (b) add a wing part to this initial skeletal image.

From there out it begins to get fuzzier. This is akin to the visual field which also has its highest resolution at the centre of the scene being viewed.

Third, the medium has a grain. The grain of a photograph or a VDU refers to the size of the basic dots of colour that make it up. If these dots are very large then the detail one can represent is limited, whereas if the dots are very small more detailed images can be represented. A good example of this is the comparison between a conventional Teletype computer screen and a typical PC monitor. The latter has the grain to depict different letter fonts and pictures in a manner that is impossible on the Teletype screen. Thus, the grain of the spatial medium determines what can and cannot be represented clearly. It also means that when an image is reduced in size then parts of it may disappear, because the grain may not be detailed enough to represent these parts. Specifically, a part of the larger image that was represented by a configuration of dots may, when the image is reduced, be represented by a single dot.

Finally, as soon as an image is generated in the medium it begins to fade and so, if the image is to be maintained in the medium, it needs to be regenerated or refreshed. A similar type of fading occurs with after-images in the visual system. When we look at bright lights and then close our eyes, we see after-images caused by the over-stimulation of our retinal cells. Although these after-images are not the same as visual images, they have this same quality of rapidly fading after they first appear.

Image andpropositional files

Returning to our duck, we have a fair idea of where she is represented but not how we come to represent her. In Kosslyn's computational model it is assumed that there are image files that represent the coordinates of dot-like points in the spatial medium. These image files can represent a whole object or various parts of an object. Specifically, some image files characterise a skeletal image that depicts the basic shape of the object, but lacks many of the object's details. These detailed parts of images may be represented in other image files, for reasons that will become apparent later. In terms of our example, the image in Figure 9.13a is a rough, skeletal image of the duck, while Figure 9.13b shows the addition of one of her parts (i.e., the wings).

The propositional files list the properties of ducks (e.g., HAS_WINGS, HAS_FEET) and the relationships between these properties and a "foundation part" of the duck (i.e., its body). The foundation part is that part that is central to the representation of the object and will be linked to the skeletal image file for the object. The propositional file for the duck might, thus, contain entries that relate the wing parts of the duck to the foundation part: for example, WINGS LOCATION ON_EITHER_SIDE BODY indicating that the wings are on either side of the body. Each of these parts would have a corresponding image file that contains the basic material for constructing the image of a given part in the spatial medium. Propositional files also contain more information about the rough size category of the object (e.g., very small, small, large, enormous) and information about superordinate categories of the objects (e.g., in the duck case, that BIRD would be the most likely superordinate; see Kosslyn, 1980, 1983, for details; and Chapter 10).

The information in the propositional files is connected to the image files. So, for example, the foundation part in the propositional file has a link or pointer to the image file that contains the skeletal image of the object. Similarly, the detailed parts of the object have links to image files containing images of these parts. For example, the wings-part is linked to an image file containing co-ordinate information for the construction of an image of a wing.

Imaging processes

Finally, when someone is asked to image a duck several processes use the propositional and image files to generate an image of the duck in the spatial medium. In the model, the main IMAGE process involves three sub-processes: PICTURE, FIND, and PUT. When asked to image, the IMAGE process first checks to see whether the object (i.e., the duck) mentioned in the instructions has, in its propositional-file definition, a reference to a skeletal-image file. If such a file is present then the PICTURE process takes the information about the co-ordinates of the image and represents it in the spatial medium (see Figure 9.13a). Unless the location or size of the image is specified (e.g., image a giant duck), the image is generated in the part of the spatial medium with the highest resolution and at a size that fills this region. The PUT process directs the PICTURE process to place the remaining image-parts at the appropriate locations on the skeletal image. For example, PUT might use the propositional information about the location of the wings to add them to the side of the skeletal image of the body. PUT, however, must use FIND to locate the objects or parts already in the image to which the new, to-be-imaged parts can be related. When the appropriate size and location of the wings are known they are added to the image (see Figure 9.13b).

In cases where more specific instructions are given, like "Does the duck have a rounded beak?" or "Image a fly on the tip of the duck's wing, up close" or "Rotate the duck 180 degrees", further processes called SCAN, LOOKFOR, PAN, ZOOM, and ROTATE operate on the image (see Kosslyn, 1983, Chapter 7; and Kosslyn, 1980, for more details). The names of these processes are self-explanatory and each one has been modelled as a set of specified procedures in the model that, for instance, SCAN and ROTATE images. These processes are used to explain the results of the mental scanning and mental rotation studies.

Empirical evidence for Kosslyn's theory

Kosslyn's work has several important and welcome features. First, by specifying computationally the processes and representations involved in imagery, he avoids the vagueness criticism. Second, the claims he makes for the properties of imagery are clear. Third, many of these detailed proposals are supported by empirical evidence. Consider some of the evidence for his proposals on limited extent and granularity, the fading of images and the area of high resolution in the spatial medium.

The image tracing task

Kosslyn (1975, 1976, 1980) has used an "image tracing task" to test his proposals on the limited extent of the spatial medium and on granularity. As in the duck example, in these experiments subjects were asked to image an object and then to try to "see" some property of the imaged object (e.g., "Can you 'see' the duck's beak?"). The critical manipulation in the experiment was the context in which the animal was imaged. The "target" animal (e.g., a rabbit) was imaged along with another animal that was either much larger or much

A schematic diagram of how the image of (a) an elephant and a rabbit, and (b) a fly and a rabbit might result in the rabbit being imaged at different levels of detail. Adapted from Ghosts in the mind's machine: Creating and using images in the brain by Stephen Kosslyn. Reproduced by permission of the author. Copyright © 1983 by Stephen M.Kosslyn.

Stop Anxiety Attacks

Stop Anxiety Attacks

Here's How You Could End Anxiety and Panic Attacks For Good Prevent Anxiety in Your Golden Years Without Harmful Prescription Drugs. If You Give Me 15 minutes, I Will Show You a Breakthrough That Will Change The Way You Think About Anxiety and Panic Attacks Forever! If you are still suffering because your doctor can't help you, here's some great news...!

Get My Free Ebook

Post a comment