Cognitive and learning styles research is limited by the lack of evidence supporting valid interpretations of style assessment scores. We sought evidence to support the validity of scores from 4 instruments: the Index of Learning Styles (ILS); the Learning Style Inventory (LSI); the Cognitive Styles Analysis (CSA), and the Learning Style Type Indicator (LSTI). The ILS assesses 4 domains: sensing-intuitive (SensInt), active-reflective (ActRefl), sequential-global (SeqGlob) and visual-verbal (VisVerb), each of which parallel a similar domain in at least 1 of the other instruments. We administered the ILS, LSI and CSA to family medicine and internal medicine residents and Year 1 and 3 medical students and applied the multitrait-multimethod matrix to evaluate convergence and discrimination. After 3 months participants repeated the ILS and completed the LSTI. A total of 89 residents and medical students participated. Multitrait-multimethod analysis showed evidence of both convergence and discrimination for ActRefl (ILS, LSI and LSTI) and SensInt (ILS and LSTI) scores. ILS SeqGlob and SensInt scores showed unanticipated correlation. No other domains met the criteria for convergence or discrimination. Test-retest reliabilities for ILS scores were 0.856 for SensInt, 0.809 for ActRefl, 0.703 for SeqGlob and 0.684 for VisVerb. Cronbach's alpha values were > or = 0.810 for LSI and 0.237-0.758 for LSTI. At least 9 participants misinterpreted the LSI instructions. These data support the validity of ILS active-reflective and sensing-intuitive scores, LSI active-reflective scores and LSTI sensing-intuitive scores for determining learning styles in this population. Cognitive style and learning style scores may not be interchangeable, even for constructs with similar definitions.