Automated tailored testing using Raven's Matrices and the Mill Hill Vocabulary tests: a comparison with manual administration

Abstract Three studies were carried out using automated and tailored versions of the Raven's Progressive Matrices and Mill Hill Vocabulary tests. In the first study, four groups of 20 normal subjects were tested on two occasions being given either the automated version on both occasions, the Standard on both occasions, or automated on one and Standard on the other. Results suggested a high test-retest reliability for the automated tests, together with high levels of validity as measured by correlations between the automated and conventional versions of the tests. Time to complete the automated version was less than half that for the conventional test, but there proved to be a small but consistent discrepancy in absolute test score between the Standard and automated version of Progressive Matrices. This suggests that automated tests should not use norms based upon Standard versions without further validation. Experiments 2 and 3 showed that the automated versions of the tests were highly suitable for using with both elderly subjects and brain damaged patients. Further evidence of the validity of the automated tests was provided by the tendency for left brain-damaged patients to show impaired Mill Hill Vocabulary performance, while right hemisphere damage was reflected in impaired performance on the Matrices test.