Variation within categories
MetadataShow full item record
This thesis concerns the memory storage of category information. It presents an analysis of several aspects of variation among category exemplars. For each of these aspects, the effects on the way that a category is remembered following experience with exemplars and non¬-exemplars were assessed. In addition, the study considers the connec¬tion between cognitive processes and physiological activity. In the experiments described in Chapter 5, typicality of value variants was varied in traditional category learning problems with defining values. Typicality of value variants appeared to influence learning performance. The effect, however, differed for relevant and irrelevant values. Typicality of relevant values facilitated category learning, whereas typicality of irrelevant values did not affect categorization performance. The latter discrepancy can be explained by interaction between selective attention and the developing category representation (see 8.4). Different aspects of within-category variation in categories without defining values are the focus of Chapter 6 and 7. Two types of categorization models that are often contrasted are relevant in this respect. These are frequency models and prototype-distance models (see Chapter 2). In accordance with these models, two main aspects of within-category variation were investigated, i.e. frequency of dimensional values and the typicality of the variants with which they occur. A variety of methods was employed to examine their influ¬ence on category learning and its results. Frequency of values and variant typicality appeared to independently influence categorization performance. It was concluded that frequency and prototype-distance models are complementary rather than contradictory. They explain dif¬ferent aspects of variation within the same categories. Regarding specific frequency effects, the experiments showed how frequency of values not only in the focal, but also in the contrast category influences category representation. The latter aspect is an important one that should be, but is not always accounted for in categorization models. A task effect appeared with respect to the influence of frequency in the contrast category. Implications are discussed in section 8.1.2. The question whether or not conjoint occurrence of values affects categorization, represents another theoretically controversial issue (see 3.1.2.). From the present results it appeared that the typical¬ity of an irrelevant value is enhanced when it correlates with a value that is important to categorization. The latter aspect may offer a solution to the above-mentioned controversy (see 8.1.3.). In all, subjective importance of values to categorization was clearly demon¬strated to be determined by these frequency variables. Another vari¬able, delay of testing, showed that distinctive values are more resis¬tant to decay. Finally, an experimental manipulation related to frequency was whether or not subjects received additional feedback information on frequency. Additional information had no beneficial effect on category learning. In section 8.1.1. this finding is discussed in connection with automatic encoding of frequency information. With variant typicality the problem of category boundaries was raised. This is an often neglected issue in models of categorization. In the present study, the influence of range of variant typicality experienced during learning was investigated. As expected, broad rather than narrow range experience resulted in a larger extension of the focal category, with better categorization of atypical focal category boundary exemplars. However, atypical boundary items of the contrast category were not categorized better. The range of variants also influenced item typicality judgments. Within the focal category these judgments diverged relatively more following small range experi¬ence. Narrow as compared with broad range experience furthermore had the effect of polarizing the categories, i.e. the same physical dis¬tances were judged differently depending range experience. Finally, the relationship was investigated between information processing in categorization tasks, and physiological activity as measured by the amplitude of the Galvanic Skin Response (GSR). In a variety of ways it was shown in Chapter 5 and 6 that the higher the subjective information value of the feedback, the higher the GSR. It is argued that GSR measurement as an indication of uncertainty reduc¬tion provides a useful supplementary instrument in cognitive research.