OBJECTIVES: The Fatigue Impact Scale (FIS) is a 40-item multiple response questionnaire designed to assess fatigue, divided into cognitive, physical, and psychosocial functioning subscales. The purpose of the study was to improve the measure by removing differential item functioning (DIF) and reducing the number of items. METHODS: FIS data were available from 188 patients with multiple sclerosis (MS). These data were subjected to Rasch analysis (one-parameter logistic item response theory) using the RUMM programme. Fit to the Rasch model was examined via Chi2 statistics and assessments of DIF related to gender, age and MS type. RESULTS: FIS responses from the 188 MS patients were analysed (47/25% male; mean age 50.9, SD 10.5; 39.6% relapsing remitting, 36.9% primary progressive, 23.5% secondary progressive). Initial results showed that the 40-item FIS exhibited misfit and was not unidimensional. Several items exhibited DIF by age, gender or MS-type. For example, patients aged over 50 years scored significantly higher than patients aged 50 or younger (who had a similar level of fatigue) on the item ‘I have to rely more on others to help me'. DIF by MS-type indicated that answers to certain items are significantly influenced by the disease stage of the patient. After the removal of 9 items that either misfit or exhibited age or gender DIF, the reduced FIS fit the Rasch model (Chi2 p > 0.05), providing a unidimensional fatigue scale. The threshold map suggests that for some items the response categories did not discriminate in the way intended, suggesting that changing the response format may improve the scale. CONCLUSIONS: The analysis showed that it is feasible to derive a single unidimensional scale of fatigue from the FIS. The severity (logit) coverage of the scale is good but there remains item redundancy suggesting further scope for item reduction.