Quantitative estimation of physical work load requires breakdown of jobs into smaller entities. The objective was to study the inter-rater reproducibility of the contents, frequency and duration of work tasks assessed by telephone interview. Two researchers interviewed 18 industrial workers with an interval of from 2 to 3 weeks in a balanced and blinded design. Altogether 114 tasks were identified, 68 of which were recorded by both interviewers. The tasks were classified into regularly occurring (n = 34) and occasional (n = 80). The outcome was the total duration of the tasks per day computed from the data on frequency and duration. Validity of the interview was studied against prestructured diaries filled in by nine workers. The interviewers' assessments of the overall contents of the tasks were rated as 'similar' or 'very similar' for 17 of the 18 workers. Both interviewers detected all 34 regularly occurring tasks. The intraclass correlation coefficient (ICC) of the total duration between the interviewers was 0.81 for regularly occurring tasks. ICCs of the total duration between the diary and the two interviewers were 0.90 and 0.91. However, in many cases the workers could not give a numerical value for duration or frequency. A telephone interview can be used as a first step in exposure assessment in epidemiological studies on risk factors of musculoskeletal disorders. These results show that a single telephone interview can give reproducible and valid information of the frequency and duration of tasks occurring daily. For occasional tasks interview methods should be developed further.