The tibial components in 143 patients with total knee replacements performed before 1988 were assessed for micromotion using roentgen stereophotogrammetric analysis (RSA) over a period of 13 years. The fixation of the prostheses remained clinically sound in all cases, although revision had been required for other reasons in seven. In a second group taken from all cases with RSA available on our full database to 1990, 15 tibial components had been followed by RSA from the insertion until, 1 to 11 years after the initial arthroplasty, they were revised for mechanical loosening of the tibial component; 12 of these comprised all the loosenings in the base group, thus making a total of 155 consecutive cases, while an additional three were inserted after the base material had been compiled. The mean migration in the first group was about 1 mm at one year, but subsequent migration was slower, reaching a mean of about 1.5 mm after ten years. About one-third migrated continuously throughout follow-up, while two-thirds ceased to migrate after one to two years. In the revision group, 14 components had migrated continuously and at one year significantly more than those in the first group. One revision case lacked the crucial one-year follow-up and could not be classified. These findings suggest that mechanical loosening begins early in the postoperative period. Clinical symptoms which necessitate revision, seen at this stage in 20% of abnormally migrating tibial components, may not appear until up to ten years after the operation.(ABSTRACT TRUNCATED AT 250 WORDS)