This is a retrospective study on absenteeism in a large teaching hospital covering the calender years 1982 and 1983. The number of personnel under study is 2.700 divided in six professional categories: administrative employees, qualified nurses, non-qualified nurses, laboratory technicians, caterers, cleaners. A 30% sample was chosen by the method of stratified sampling. Data were collected for each person on personal and professional factors. Absence data were collected for each person and were analysed by computer techniques. Females form the majority of the sample (82%). The overall absence rate was 9%; on excluding maternity leave it was around 6%. Female workers have double the pregnancy rate of employees and double their absence rate (maternity leave excluded). Both younger age groups have a high frequency of absence. A consistantly higher absenteeism was observed in females than in males even after the exclusion of pregnancy. Nationality had only a slight influence. The effects of seasonal variations was observed: increase in the winter and decrease in the summer. The factor education showed a direct relationship with absenteeism, the higher the education, the lower the absenteeism and vice versa. In order to detect which factors significantly influence absenteeism, each isolated factor was first considered. 'Marital status' appeared to be the most significant. On the multifactorial analysis, only education was found to be highly significant. Profession appeared also significant but is correlated to education. From these data, the probability of absenteeism can be determined. Propositions are made for an effective registration of absences and some approaches to reduce absenteeism are discussed.