In 1973, 502 employees in three metal industry plants, stratified by age, sex, and occupational class, were examined for physical workload by interview and observation at the workplace and for musculoskeletal disorders by questionnaire, interview, and clinical examination. Five years later, 88% of the sample were re-examined for their musculoskeletal status. The work of the blue collar groups was heavier as measured by indices of physical work load based on the observation and interview (physical strain, physical load, static phases, and stereotypy). Rheumatic symptoms, clinical findings in the musculoskeletal system, and chronic musculoskeletal diseases were more frequent in both female and male blue collar workers than in the respective white collar groups on the first occasion and the increase in morbidity during follow up was higher in the blue collar groups. At an individual level within the blue collar class, however, associations between indices of physical workload and musculoskeletal morbidity were weak or non-existent. The associations were weakened by selective movement of people with musculoskeletal disorders from heavy jobs to premature retirement or to lighter jobs. Physical strain and physical load were negatively associated with the incidence of long term musculoskeletal disorders in the female blue collar group.