Donald Homa

Limitations of exemplar-based generalization and the abstraction of categorical information.

An evaluation of exemplar-based models of generalization was provided for illdefined categories in a category abstraction paradigm. Subjects initially classified 35 high-level distortions into three categories, defined by 5, 10, and 20 different patterns, followed by a transfer test administered immediately and after 1 wk. The transfer patterns included old, new, prototype, and unrelated exemplars, of which the new patterns were at one of five levels of similarity to a particular training (old) stimulus. In both experiments, increases in category size and oldnew similarity facilitated transfer performance. However, the effectiveness of old-new similarity was strongly attenuated by increases in category size and delay of the transfer test. It was concluded that examplar-based generalization may be effective only under conditions of minimal category experience and immediacy of test; with continued category experience, performance on the prototype determines classification accuracy. Categories are said to be ill defined (Neisser, 1967) when it is not obvious what dimensions characterize a category, and the variety among the potential members of a category is essentially infinite. Examples of ill-defined categories are quite diverse and would include the natural categories, musical style, hand-written letter As, and the class of sound patterns associated with a specific spoken word. How the human organism learns ill-defined categories, and how this knowledge is transferred to novel situations, has been a topic of considerable attention over the past 10 years. Posner and Keele (1968, 1970) argued that a prototype or central tendency is abstracted during the classification of distorted but related patterns. In their experiments, the subject initially sorted dot-pattern stimuli into a number of categories, with each category represented by a different reference pattern (objective prototype). ClasThis research was supported b y National Institute of

Perceptibility of schematic face stimuli: Evidence for a perceptual Gestalt

The perceptibility of face, scrambled face, and single-feature stimuli was investigated in three experiments. Stimuli were presented tachistoscopically, followed by a visual noise mask and a forced-choice test of one of three features (eyes, nose, and mouth). In Experiment I, two processing strategies which have been proposed for word perception (involving expectancy and redundancy) were investigated for the stimuli employed here. In Experiments II and III, experimentally induced familiarity was studied for its effect on recognition and perception, and an immediate and delayed perceptual test was employed. Across all three experiments, perception of single-feature and face stimuli were consistently superior to scrambled faces; in Experiment III, differences between single features and faces were eliminated. The effects of perceptual expectancy, internal feature redundancy, familiarity, guessing biases, etc., were shown to be insufficient to account for the superiority of face to scrambled face stimuli. It was argued that the perceptibility of nonredundant features are enhanced when those features are aligned in a well-defined form class. The view that familiarity operates directly on recognitive processes but indirectly on perceptual ones was discussed.

On the Nature of Categories

Publisher Summary This chapter focuses on the general taxonomy of category types, which is specifically directed toward the dichotomy of categories into ill-defined and well-defined types. The proposed taxonomy of category types allows results to be summarized and interpretative discrepancies to be isolated. The proposed taxonomy may also help systematize important findings from diverse areas that are essentially concerned with categorical problems, for example, medical diagnostics and the acquisition of finite-state grammars. The acquisition and utilization of conceptual information may be broadly conceived as the problems in sampling theory in which an organism attempts to comprehend the meaning of a category by sampling its members from the category domain. The chapter emphasizes on learning variables and the critical role they play in the shaping and modification of concepts. A number of recent findings in diverse categorical areas are presented in the chapter in which it is shown that variable manipulations critically determine the interpretation.

Reallocation of Visual Attention

Attentional movement time was investigated within and across hemifields in a discrimination task in which retinal acuity was controlled. Ss discriminated targets in a two-alternative, forced-choice latency paradigm. In Experiments 1-3, costs were mediated by distance, even though it was varied independently of acuity. In Experiments 4 and 7, with distance held constant, costs were equivalent for crossing the vertical and horizontal meridians and for crossing 1 and 2 meridians. However, crossing 1 meridian produced costs that were less than the costs for shifts to unexpected locations in the same quadrant, partially replicating the inhibitory aftereffect of Tassinari, Aglioti,Chelazzi,Marzi,and Berlucchi (1987). An explanation based on a combination of analog movement (Tsal, 1983) and attentional distribution (Downing & Pinker, 1985; Hughes & Zimba, 1987; Tassinari et al., 1987) models was discussed.

Duration judgment and the experience of change

Predictions based on storage size, processing effort, and change models of time estimation were tested in five experiments. The first of these presented subjects with stimulus patterns that varied on dimensions of sensory-event number and uncertainty. Subjects estimated the duration of time periods using the reproduction method. Duration estimates were most accurately predicted by the number of sensory events in each pattern. This relationship was generally positive, although the specific function relating these variables was dependent upon clock duration. The change model seemed to fit these data best. Experiment 2 demonstrated that the relationship between sensory change and judged duration was not due to the total time of sensory input. Experiments 3, 4, and 5 tested the effects of several different types of change. In Experiments 3 and 4, the number of sensory events was held constant, but the regularity of their spatial presentation was varied. In both experiments, duration judgments were positively related to the number of changes that occurred. Because the manipulations used in these experiments produced differences in the visual complexity of stimulus patterns, the argument could be made that storage size or processing effort accounted for the size of duration judgments. Experiment 5 tested the effects of change while holding the visual complexity of stimulus patterns constant. A positive relationship between duration judgments and number of changes was again found.

Expanding the search for a linear separability constraint on category learning

Formal models of categorization make different predictions about the theoretical importance of linear separability. Prior research, most of which has failed to find support for a linear separability constraint on category learning, has been conducted using tasks that involve learning two categories with a small number of members. The present experiment used four categories with three or nine patterns per category that were either linearly separable or not linearly separable. With overall category structure equivalent across category types, the linearly separable categories were found to be easier to learn than the not linearly separable categories. An analysis of individual participants’ data showed that there were more participants operating under a linear separability constraint when learning large categories than when learning small ones. Formal modeling showed that an exemplar model could not account for many of these data. These results are taken to support the existence of multiple processes in categorization.

As easy to memorize as they are to classify: The 5- 4 categories and the category advantage

Recently, it has been suggested that some categories commonly used in category learning research are eliciting primarily item-level memorization strategies. A new measure of generalization, the category advantage, was introduced and used to test performance on the popular “5–4” categories. To estimate a category advantage, performance on a standard category learning task is compared with performance in an identification task, where participants learn a unique response to each stimulus. Once corrected for differences in chance expectancy, the advantage shown for the category learning task represents the degree to which participants capitalize on the natural similarity structure of the categories. In Experiment 1, the category advantage measure was validated on structured and unstructured categories. In Experiments 2 and 3, the 5–4 categories failed to produce a category advantage when tested with either of two stimulus types, suggesting that these categories elicit predominantly memorization.

Long-term memory for pictures under conditions of thematically related foils

Memory for pictures was investigated under conditions of difficult foil discriminability and lengthy retention intervals. The foils preserved the theme of the studied stimulus, but differed in the number and quality of nonessential physical details. In each experiment, subjects viewed colored photographs, black-and-white photographs, elaborated line drawings, and unelaborated line drawings, followed by an old/new (Experiment 1) or a four-alternative forced-choice (Experiment 2) test given either immediately, 1 day, 1 week, or 4 weeks following study; Experiment 3 replicated Experiment 1 but with a 12-week delay. For the old/new procedure, performance was best on colored photographs, with performance differences among the four stimulus types still significant after 4 weeks. For the forced-choice test, performance on colored photographs and unelaborated line drawings was best, with performance differences among stimulus types also still significant after 4 weeks. A confusion analysis indicated that errors were based on physical similarity, even after 12 weeks. These results refute the hypothesis that the memorial representations for pictorial variations converge to a common, thematic code after lengthy delays; instead, non-thematic, analogue information is encoded and preserved for lengthy time periods.

Organization and long-term memory search

A recognition RT paradigm was used to assess a number of plausible search strategies in LTM for categorized lists. List length was determined by factorially combining two, three, or five categories with two, three, or five words per category, and test items could be one of four types: (1) positive, (2) a repeated positive, (3) a related negative, or (4) an unrelated negative. It was found that RT increased linearly with category size for both positive and related negatives (about 30-40 msec/item); whereas the increase was much smaller for the unrelated negatives, especially for three and five categories (about 9 msec/item). With an increase in the number of categories, RT increased at the rate of about 40 msec/category for all three test items. A theory of .high-speed scanning for categorized material was proposed in which a serial and exhaustive search of the categories, is first undertaken. If a positive category match is found, a serial and exhaustive search within the contents of the positive category is initiated; if no category match is found, the search is simply terminated. Some evidence was presented that categories recently probed may provide for a short-circuiting of the category search.

Influence of manipulated category knowledge on prototype classification and recognition

The recognition and classification of category members was explored, following a variable number of learning trials. In Experiment 1, subjects received 1 or 9 learning trials, followed by a recognition-then-classification test containing old, new, prototype, and foil patterns. In Experiment 2, subjects received 1, 6, or 12 trials, and made either classification or recognition judgments. In each experiment, classification accuracy for all item types was at near-chance -performance after a single trial but steadily increased with increased learning trials. On the transfer test, oldness judgments were highest for the category prototype after a single trial. However, with increased learning trials, oldness judgments increased for old instances and decreased for the category prototype and new instances. We suggest that false recognition of the category prototype, especially after a single learning trial, need not reflect an abstraction process. We discuss the possibility that an abstracted prototype may emerge with additional learning as an unfamiliar, ideal point.