Biomedical subjects
Publications and source records attributed to S M Kosslyn.
Cognitive neuroscience rests on findings, methods, and theory from three fields: experimental psychology, systems-level neuroscience, and computer science. The strong trend over the past few years has been for a greater integration across these fields. The influence of this interdisciplinary approach on current research on memory, perception, and language will be illustrated.
Results of 4 sets of neural network simulations support the distinction between categorical and coordinate spatial relations representations: (a) Networks that were split so that different hidden units contributed to each type of judgment performed better than unsplit networks; the reverse was observed when they made 2 coordinate judgments. (b) Both computations were more difficult when finer discriminations were required; this result mirrored findings with human Ss. (c) Networks with large, overlapping "receptive fields" performed the coordinate task better than did networks with small, less overlapping receptive fields, but vice versa for the categorical task; this suggests a possible basis for observed cerebral lateralization of the 2 kinds of processing. (d) The previously observed effect of stimulus contrast on this hemispheric asymmetry could reflect contributions of more neuronal input in high-contrast conditions.
Results of 4 experiments indicate that both within-modality and case-specific visual priming for words are greater when test stimuli are presented initially to the right cerebral hemisphere (RH). In contrast, neither within-modality nor case-specific explicit memory for words is greater when stimuli are presented initially to the RH. Priming is measured using word-stem completion, and explicit memory is measured using word-stem cued recall. In both cases, Ss first rate how much they like words, and then word stems are presented briefly to the RH (in the left visual field) or to the left hemisphere (in the right visual field). Results suggest that at least 2 separate systems encode the visual representations that produce priming. The system that is more effective in the RH is better at representing form-specific information, whereas another system that is not more effective in the RH does not distinguish among distinct instances of word forms.
Subjects were presented with two groups of characters and were to decide whether they were the same or different. The stimulus groups differed either by a single feature ("preattentive" trials) or by a conjunction of features ("attentive" trials). The two stimulus groups appeared at the corners of an imaginary square centered about the fixation point, falling either in the same or different hemifields. In two experiments, subjects evaluated both types of stimuli faster when they were presented in different hemifields than in the same hemifield. Subjects also compared pairs of single characters faster when they appeared in different hemifields. Finally, this different-hemifield advantage was eliminated when the characters appeared sequentially. These results indicated that two stimuli that appear simultaneously in the same hemifield engender either a competition for common processing structures or intrahemispheric inhibition.
Dyslexic and normal control subjects memorized simple line patterns inside a grid and subsequently judged whether an "X" would have fallen on the pattern had it been present in an empty grid. The patterns were letters and novel shapes. The grids were presented to the left visual field, to the right visual field, or in central vision. Dyslexic subjects had difficulty generating images of multipart patterns, but this deficit was limited to letters. The findings suggest that dyslexics may have selective difficulty integrating visual information stored in long-term memory.
Neuroanatomical, neurophysiological, and computational constraints are used to motivate a set of hypotheses about the functional organization of high-level vision. A set of processing subsystems is posited that underlies the later stages of visual object recognition and identification; these subsystems have been implemented in a running computer simulation model. The model is damaged in a variety of ways, and its performance on a set of tasks is observed. Dysfunctions arise from disruptions of the subsystems, disruptions of their interconnections, compensations by intact subsystems, and diminished activational capacity. The most common dysfunctions of high-level vision following brain damage are then reviewed, and accounts are offered by reference to the stimulation model. According to the theory and model, each type of dysfunction can arise from numerous underlying causes, all of which are potentially distinguishable by empirical methods.
Twenty-four-5-year-olds and 24 7-year-olds completed two divided-visual-field tasks; one task required subjects to categorize a dot as above or below a line, whereas the other required subjects to determine whether the dot was within 3 mm of the line. There was a relative left-hemisphere advantage for the above/below task and a relative right-hemisphere advantage for the distance task. The results indicate that distinct processing subsystems compute different kinds of visuo-spatial relations as early as 5 years of age.
Age differences were examined in 4 aspects of visual mental imagery, namely, image generation, maintenance, scanning, and rotation. The results suggested that one or more distinct processes are used to carry out each aspect of imagery, and that this is true for 5-year-olds, 8-year-olds, 14-year-olds, and adults. There was no evidence that younger children have fewer processing components, which become differentiated into more specialized subsystems over age. In addition, the results suggested that younger children are relatively poor at scanning, rotating, and generating objects in images, but are relatively good at maintaining images.
An embedded figures task was used to investigate hemispheric specialization for visual parsing. Error data from 40 normal right-handed males revealed a right-hemisphere advantage for parsing governed by the Gestalt laws of organization. In contrast, for parsing that violated those laws, subjects who could perform such parsing at better-than-chance levels showed a left-hemisphere advantage. A simple left hemisphere/analytic, right hemisphere/holistic dichotomy is too ambiguous to account for the results, because it fails to specify the level of visual structure (e.g. whole figures, intermediate components, or elementary segments) that is to be taken as a reference point for defining particular tasks as analytic or holistic; what is analytic from one point of view is holistic from another. We focus instead on principles of perceptual organization that characterize the results of perceptual processing, regardless of whether those results are seen as the outcome of analytic or holistic processing.
Analyses of human object recognition abilities led to the hypothesis that 2 kinds of spatial relation representations are used in human vision. Evidence for the distinction between abstract categorical spatial relation representations and specific coordinate spatial relation representations was provided in 4 experiments. These results indicate that Ss make categorical judgments--on/off, left/right, and above/below--faster when stimuli are initially presented to the left cerebral hemisphere, whereas they make evaluations of distance--in relation to 2 mm, 3 mm, or 1 in. (2.54 cm)--faster when stimuli are initially presented to the right cerebral hemisphere. In addition, there was evidence that categorical representations developed with practice.
Compared time to evaluate stimuli of varying sizes. When Ss expect an upcoming stimulus to be a certain size, response time increases with the disparity between expected and actual size. There are, however, 2 size adjustment processes, and they reflect 2 types of visual selection. In the first, a shape-specific image representation is used to separate a visual object from a superimposed distractor. These representations require the type of slow size scaling demonstrated in imagery experiments. The second size scaling process is faster and not shape-specific. At any given time the visual system is set to process information at a particular scale, and that scale can be adjusted to match an object's size. Because both selection mechanisms depend on size, they probably occur at a relatively low, spatially organized processing level. These findings lead to a new explanation for results that had been taken as evidence for attentional selection at the level of object representations.
Explore the source record for details and available documents.
In 49 sessions, pigeons failed to learn to sort a collection of 80 stimuli composed of a closed curve and a dot, divided into two categories, according to whether the dot was or was not inside the curve. Next, the pigeons were successfully trained, first with the insides of the curves shown in bright red, then with a darker red, and finally with a black matching the background outside the curve. After this stepwise procedure, the pigeons displayed a limited ability to sort novel curves and dot locations according to whether the dot was or was not inside the curve.
Although objects in visual mental images may seem to appear all of a piece, when the time to form images is measured this introspection is revealed to be incorrect; objects in images are constructed a part at a time. Studies with split-brain patients and normal subjects reveal that two classes of processes are used to form images--ones that activate stored memories of the appearances of parts and ones that arrange parts into the proper configuration. Some of the processes used to arrange parts are more effective in the left cerebral hemisphere and some are more effective in the right cerebral hemisphere; the notion that mental images are the product of right hemisphere activity is an oversimplification.
Explore the source record for details and available documents.
Explore the source record for details and available documents.
Imagery of congenitally blind and normally sighted subjects was compared in two experiments. In the first, subjects were asked to estimate how far away objects appeared in images. The results showed that blind subjects imaged objects "within arms' reach", and with only a slight tendency to image larger objects farther from them. In contrast, sighted subjects tended to image larger objects as if they were farther away. In addition, unlike the sighted, blind subjects' images also failed to overflow an "image space" of fixed size. Finally, blind subjects were, with one exception, unable to mimic successfully the responses of a sighted person, when explicitly asked to do so. In the second experiment, subjects pointed to the left and right sides of three objects imaged at three distances. Sighted and blind groups both showed a decrease in pointing span with the size of the object, but only the sighted subjects showed a decrease with increased image distance, in accordance with the laws of perspective.
Three general classes of algorithms have been proposed for figure/ground segregation. One class attempts to delineate figures by searching for edges, whereas another class attempts to grow homogeneous regions; the third class consists of hybrid algorithms, which combine both procedures in various ways. The experiment reported here demonstrated that humans use a hybrid algorithm that makes use of both kinds of processes simultaneously and interactively. This conclusion follows from the patterns of response times observed when humans tried to recognize degraded polygons. By blurring the edges, the edge-detection process was selectively impaired, and by imposing noise over the figure and background, the region-growing process was selectively impaired. By varying the amounts of both sorts of degradation independently, the interaction between the two processes was observed.