Category B

Black Belt Memory

10x Your Memory Power

Get Instant Access

Figure 1 Two categories of stick figures. The figures in category A have at least two of the following three attributes: square body, shaded body, and long legs. The figures in category B have at least two of the following attributes: round body, white body, and short legs. Subjects may be inducing these rules during training, or they may be basing categorization judgments on similarity to previously experienced items in the two categories.

according to rules. Also, the application of rules requires working memory, an ability that has been strongly associated with prefrontal (especially dorso-lateral) function. In order to determine whether an item satisfies a list of conditions, one must maintain these conditions in a short-term buffer and actively test the item against each condition. Putting demands on working memory by requiring subjects to perform a concurrent task has been shown to induce subjects to shift to an exemplar-based strategy from a rule-based strategy. The fact that young children do not use a rule application strategy as readily as adults do is also supportive of the idea that prefrontal cortex is important for rule-based classification. The frontal lobes mature relatively late in development, continuing into adolescence.

Classification rules tend to focus on individual attributes of stimuli. In addition to working memory, the application of rules requires the ability to selectively attend to attributes of the stimuli and to shift attention between different attributes. Neuroimaging findings are consistent with this idea in that posterior parietal brain regions are active during rule application. These regions are typically active in selective attention tasks and are damaged in patients exhibiting the neglect syndrome, especially when the lesion is on the right side.

III. EXEMPLAR-BASED CATEGORIZATION: GENERAL OR SPECIFIC?

Although people often make classification judgments on their previous experience with examples of the category, the type of exemplar-based information could take different forms. After experiencing many examples of a category, such as chairs, we may have formed an idea of what a typical chair looks like, and we may compare new examples to this typical chair and classify these new items as chairs depending on their similarity to this typical chair. This typical example, or category prototype, represents the average of experienced category exemplars. When additional exemplars are encountered, the representation of the prototype is modified to reflect the new experience. By this view, category knowledge takes the form of abstracted information that is distinct from memories of the individual exemplars. In support of this prototype view is the fact that category prototypes or average members of categories are much easier to classify than members that are very dissimilar from the prototype. People are much faster to acknowledge that a robin is a bird rather than a penguin or chicken. In fact, in artificial categories generated in the laboratory, speed and accuracy of classification can be predicted quite well by the similarity of an item to the prototype (Fig. 2).

However, such effects do not necessarily mean that people are using prototypes to make classification judgments. According to another view, people base their classification judgments on the average similarity to previously encountered items stored in memory. This view is appealing because of its parsimony in that one does not need to hypothesize a separate base of category-level knowledge. This view can also account for the superior classification of prototypes because the prototype is generally the item that is similar to the largest number of category members since it is the average of all experienced examples.

Based on experimental data, it is difficult to distinguish between the possibility that people use information general to the category to classify items or that they make comparisons to specific examples stored in memory. These two views often make similar predictions about which items are easiest and most difficult to classify. One approach to addressing this question has been to test patients with amnesia on category learning tasks. Because these patients have sustained damage to medial temporal lobe or midline diencephalic regions, they exhibit a profound deficit in declarative memory. Declarative memory refers to the

Figure 2 Illustration of a category based on a prototype. The top dot pattern serves as the prototype, and the patterns below are all distortions created by moving each dot in the prototype in a random direction. After studying a series of such distortions, subjects are very likely to classify the prototype as being part of the category.

Figure 2 Illustration of a category based on a prototype. The top dot pattern serves as the prototype, and the patterns below are all distortions created by moving each dot in the prototype in a random direction. After studying a series of such distortions, subjects are very likely to classify the prototype as being part of the category.

conscious memory of facts and events. Declarative memory would be accessed on a recognition test of previously seen examples of a category. Because of their declarative memory deficits, amnesic patients should not be able to learn categories normally if this ability depends on declarative memory for examples. If, however, amnesic patients are able to learn to classify items based on category membership despite poor memory for the individual training examples, it would suggest that category-level information could be learned independently of exemplar memory.

In fact, there is evidence that amnesic patients can learn at least some types of categories normally. For example, amnesic patients have been shown to be able to perform normally on an artificial grammar learning task. In this task, subjects view a series of letter strings that were formed according to a finite-state rule system (Fig. 3). This rule system allows only certain letters to follow other letters. After viewing the strings, subjects are told for the first time that the items they had just seen all followed a set of rules and that their task is to classify a new set of items as following these rules or not. Although subjects typically believe that they did not learn anything about the grammatical rules, they nevertheless can reliably classify new items at a level that is significantly higher than chance. Amnesic patients perform as well as normal subjects on this task, although they are severely impaired in recognizing the letter strings that were used during training. Thus, it appears that subjects acquire information

Figure 3 A rule system used to generate an artificial grammar. Letter strings are generated by traversing the diagram along the arrows from the ''in'' arrows to the ''out'' arrow. An additional letter is added as each arrow is traversed. For example, DGJ, BDG, JDGBDG, and GBDG are examples of letter strings formed according to the artificial grammar. In contrast, DDJ, BGDJ, and JDGB cannot be formed according to the rule system.

Figure 3 A rule system used to generate an artificial grammar. Letter strings are generated by traversing the diagram along the arrows from the ''in'' arrows to the ''out'' arrow. An additional letter is added as each arrow is traversed. For example, DGJ, BDG, JDGBDG, and GBDG are examples of letter strings formed according to the artificial grammar. In contrast, DDJ, BGDJ, and JDGB cannot be formed according to the rule system.

about the grammatical category that is independent of memory for individual exemplars.

The findings from amnesic patients suggest that information about categories can be acquired that is distinct from explicit memory for exemplars. However, these results do not preclude specific exemplar-based classification in other paradigms. It is likely that when subjects are given a small number of highly distinctive and memorizable exemplars, they will consider these in making classification judgments.

Was this article helpful?

0 0
Unraveling Alzheimers Disease

Unraveling Alzheimers Disease

I leave absolutely nothing out! Everything that I learned about Alzheimer’s I share with you. This is the most comprehensive report on Alzheimer’s you will ever read. No stone is left unturned in this comprehensive report.

Get My Free Ebook


Post a comment