Recent reviews of accounting education research have called for continued assessment of learning style inventories in accounting contexts (Geiger and Boyle 1992; Rebele et al. 1998). This study presents a critical evaluation of the psychometric properties of the Canfield Learning Styles Inventory (LSI), proposed by Francis et al. (1995) as an instrument for possible use by accounting educators. The study administered two versions (standard and scrambled) to 531 accounting majors from eight universities, applying a test‐retest strategy after a 4–5 week interval. The versions were examined for internal consistency reliability, test‐retest reliability, classification stability, and construct validity. We found moderate internal consistency (item analyses scores weaker than those reported by Canfield [1988]) and a substantial amount of learner‐type classification instability across the two administrations. In terms of construct validity, two‐factor solutions were not consistent with those reported by Canfield (1988). Further, three‐factor solutions also were not consistent with Canfield's three learning domains. Therefore, along with limited theoretical support and the lack of empirical justification, we find little support for the use of the Canfield LSI in accounting education research.
