PubMed HealthSearch

Biomedical subjects

W Garrison

Publications and source records attributed to W Garrison.

2 recordsLinked to original sources

Reading comprehension test item difficulty as a function of cognitive processing variables.

This study reports the results of a componential analysis of items comprising Sections A and C of Form Z of the reading comprehension portions of the California Achievement Tests (CAT) (Tiegs & Clark, 1963). A set of problem components or attributes characterizing the test items in terms of manifest content, psychologically salient features, and processing demands was developed, including methods for their quantification. The contributions of these components to task difficulty were then evaluated using linear regression methodology. Item difficulty indices were transformations of the familiar proportion-correct item score, obtained from data gathered during the spring of 1989 from 158 deaf examinees. Variation in the item difficulty values was substantially accounted for in terms of a small number of predictor variables (R2 greater than or equal to .90). Implications of the results for construct validity and interpretation of test scores are discussed.

Cognition

Simple discrimination isn't really simple. A confirmatory analysis of the Speech Sound Pattern Discrimination Test.

This paper reports the results of a componential analysis of items comprising the Speech Sound Pattern Discrimination Test (SSPDT). The SSPDT, developed by Bochner et al. (1986), uses a closed-set sentence discrimination task to assess the auditory speech processing skill of severely and profoundly hearing-impaired individuals. A set of components reflecting differences in the phonetic and task-related characteristics of the test stimuli was developed, and the contributions of the components to discrimination task difficulty were evaluated using linear regression methodology. Discrimination task difficulty indices were transformations of percent-correct scores, resulting from fit of the SSPDT data to the Rasch measurement model. Three of the hypothesized components (one spectral, one temporal, and one task-related) entered a stepwise regression solution. These components have an intrinsic role in the construct validity of the instrument. The structure of the discrimination task, however, is more complex than might be suspected, because 'same' or matching test stimuli showed advantages in ease of discriminability compared with their 'different' or non-matching counterparts. The study findings will facilitate development of an enlarged item bank, and aid in the interpretation of test scores.

Adult