A mandatory programme monitoring the occurrence of Salmonella in pork at slaughterhouses and a serological monitoring of slaughter-pig herds has been implemented in Denmark since 1993 and 1995, respectively. All results are stored in a central database. From this, aggregated weekly results of serological and bacteriological samples collected in the period between January 1995 and July 2000 were extracted. In addition, the reported weekly incidence of human infections with S. Typhimurium covering the same time period was obtained. The times series were analysed for trends and cyclic variations by seasonal decomposition. The association between the incidence in humans and the prevalence of Salmonella in pigs and pork, and prevailing weather conditions, were analysed by using a general linear (glm) and a general additive model (gam). Explanatory variables were lagged to account for time elapsed between sampling, consumption, incubation period and case registration. The results of the seasonal decomposition showed an overall declining trend in all three time series; presumably an effect of the implemented Salmonella control measures. All time series exhibited a double peaked annual cycle. The seasonal variation of the prevalence in pork and the human incidence had a very similar course with a starting increase in the spring and a peak in August-September. The variables that were both biologically meaningful and statistically significant in both regression models were the prevalence in pork sampled 4 to 5 weeks before case registration, the seroprevalence, measured as the average prevalence of week 15 to 35 before case registration, and the air temperature lagged at 2 and 3 weeks. Limitations on inferences from overall surveillance data are discussed.