Reliability, Classification Stability, and Response-Set Bias of Alternate Forms of the Learning-Style Inventory (LSI-1985)

Abstract

This study is an extension of previous research comparing the standard LSI-1985 with a scrambled version of the instrument. The two forms of the LSI were examined for internal consistency, test-retest reliability, and classification stability. The internal consistency of the standard LSI was greater than that of the scrambled version while the test-retest reliability of the scrambled version was higher. After a five-week interval, classification stability for the two instruments was 56% for the standard LSI and 53% for the scrambled version. This level of classification stability represented a modest improvement over chance. Compared to the scrambled version of the LSI, the standard LSI yielded more extreme classification changes. Many of the differences between the standard and scrambled versions of the LSI may be due to a responseset bias of the standard LSI. Overall, the results suggest that further revision of the LSI would be desirable to reduce the response-set bias and improve the psychometric properties of the instrument.