The purpose of the study was to describe the Functional Independence Measure (FIM) in Sweden and to analyse some aspects on its structure and the possibility to predict length of stay in a rehabilitation ward. It assesses the degree of dependence with 13 physical and 5 social and cognitive items, using a 7-level ordinal scale. Data are presented from a total of 312 patients and from 267 first admission patients with a mean age of 45 (SD 13) years, 66% being men, in rehabilitation medicine wards in three hospitals in Sweden. The patients were divided into six diagnostic groups. Ratings were made at admission and at discharge. The level of dependence in physical and social cognitive items was reduced during the stay at the ward. Using Rasch analysis, separate physical and social-cognitive items and personal measures were obtained on a linear scale. It was demonstrated that the relative order of the items was similar at admission and discharge. There were minor differences between diagnostic groups for the physical items, whereas more diagnostic-specific differences were seen for the social-cognitive items, for stroke patients with and without aphasia. Individual statistics were used for demonstrating FIM changes during the rehabilitation period. There was a high correlation between admission and discharge FIM values, and the admission FIM (physical items) accounted for up to nearly 50% of the variation in length of stay in a homogeneous sample such as stroke patients, but for less than 40% in the total sample. FIM can be used to follow changes during inpatient rehabilitation and for comparisons between different rehabilitation units.