Perception of grammatical structure by good and poor readers: an individual differences multidimensional scaling approach

Abstract

The primary purpose of the present research was to delieate differences between "good" and "poor" readers in terms of their perception of and ability to use the grammatical structure in English sentences. The paradigm chosen for this investigation involved the collection of proximity measures from good and poor readers for sentences constructed to differ only in grammatical structure. These proximity measures were analyzed using an individual differences multidimensional scaling model and various other statistical techniques in order to discover the latent patterns of perceived sentence relationships which are characteristic of good and poor readers. The analyses also afforded an opportunity to compare the relative appropriateness of a phrase structure model and a transformational model of the perception of grammatical structure and to evaluate the utility of individual differences multidimensional scaling as a tool to be used in the study of psychological models of grammatical structure. Two experiments were completed. In the first, subjects were asked to make estimates of the relative confusability of English sentences differing only by negative, passive, and interrogative transformations; by paraphrasing; or by combinations of these variations. In the second experiment subjects were asked to transform auditorily presented sentences into some specified transformation of those sentences. The latency in performing each transformation was recorded and treated as a proximity measure between the form of the sentence presented and the form of the transformation specified. The data from the two experiments were analyzed using individual differences multidimensional scaling, multiple discriminant analysis, classification analysis, and other statistical techniques. On the basis of the first experiment it was determined that (1) the spatial configurations derived from the multidimensional scaling of the confusability estimates collected from good and poor readers are significantly and meaningfully different, (2) the pattern of perceived sentence relationships derived from the confusability data is more consistent with the phrase structure model than the transformational model, and finally, (3) individual differences multidimensional scaling is a powerful tool for investigating good poor reader differences in the perception of grammatical structure. The latency data from the second experiment contained such a large proportion of error variance that it was difficult to derive any meaningful results. Proposed changes in the paradigm for that experiment which should reduce the level of error variance were discussed.