Objective . To evaluate the reliability and validity of a Dutch version of the Arthritis Impact Measurement Scales 2 short form (AIMS2‐SF) and examine the agreement between the AIMS2 and AIMS2‐SF in rheumatoid arthritis (RA) patients. Methods . Data were collected from 587 RA patients from three studies. Patients completed the Dutch‐AIMS2, Modified Health Assessment Questionnaire (M‐HAQ), and Visual Analogue Scale for pain (VAS‐pain), and clinical data were collected to calculate the Disease Activity Score 28 (DAS28). Short‐form component scores were calculated from the AIMS2 long-form data. In addition, a Modified Symptom component score was calculated by replacing item 42 with item 38 as was suggested by Haavardsholm et al . [<cross-ref type="bib" refid="r7">7</cross-ref>] for the Norwegian version. Results . The internal consistency of the Physical, Symptom and Affect components was good (Cronbach's α= 0.75–0.87), moderate for the Role component (α=0.62) but rather low for the Social Interaction (0.51) component. Replacing item 33 with item 31 of the long-form AIMS2 increased internal consistency for the Social Interaction component to 0.63. Test–retest reliability of the AIMS2‐SF components was high (intraclass correlation coefficients >0.70). Mean scores of the AIMS2‐SF were generally close to those from the AIMS2, but the limits of agreement were rather wide. Both the Modified Symptom and Modified Social Interaction components showed better agreement than the original short‐form components. Plots of differences between AIMS2 and AIMS2‐SF against the mean of the two scores for the five components showed that the differences varied over the range of the measurements. Factor analysis confirmed the three‐factor structure, with a physical, psychological and social dimension that has been found for the Dutch‐AIMS2 long form. Correlations of the AIMS2‐SF components with M‐HAQ total score, functional class, VAS‐pain and DAS28 were very similar to the correlations for the original AIMS2. Conclusion . ...