We examined two longitudinal hypotheses: the 1-year prevalence for night workers is consistently higher than for others; the time trend is steeper for night workers than for others. Subjects were 12,779 men (age 40-50) working at Electricité de France-Gaz de France, and recruited in the GAZEL cohort. Five questionnaires were sent to all subjects on 1989 through 1993. Self-reported occurrence of sleep disorder in the previous year was regressed on night work status in 1989 and covariates to estimate prevalence rate differences (PRDs). The estimated PRD for regular night workers versus others was 6.1% (95% CI, 1.1-11.1%). There was a common time trend of 1.4% per year (1.0-1.9%), and evidence of excess trend for regular night workers: 1.1% (-0.7-2.9%). Findings indicate that the effect of night work on sleep complaints manifests itself during the first year of night work and is permanent.