In a previous study, we reported the performance of a PCR assay amplifying 285-bp of the invA gene of Salmonella spp. through an international ring-trial involving four participating laboratories [Int. J. Food Microbiol. 89 (2003) 241]. Based on the validated set of primers and recent advancements in PCR technology, we have designed two specific PCR assays for detecting Salmonella spp. We have compared PCR-enzyme-linked immunosorbent assay (PCR-ELISA) and LightCycler real-time PCR assay (LC-PCR) with the standard ISO 6579 bacteriological reference method. The two PCR tests incorporated an internal amplification control (IAC) co-amplified with the invA gene of Salmonella to monitor potential PCR inhibitors and ensure successful amplification. The selectivity study involved 84 Salmonella and 44 non-Salmonella strains and the samples tested were represented by 60 artificially-contaminated samples of fish, minced beef and raw milk, and 92 naturally-contaminated milk and meat samples. When using either PCR-ELISA or LC-PCR assays, only Salmonella strains were detected. PCR-ELISA and LC-PCR assays gave with pure Salmonella cultures the same detection limit level of 10(3)CFU/ml, which corresponds respectively to 50 and 10 cells per PCR tube. Data on artificially contaminated samples indicated that both PCR methods were able to detect after enrichment less than five Salmonella cells in 25 g of food, giving 100% concordance with the ISO 6579 reference method. The results on naturally contaminated samples demonstrated that despite certain inhibition problems, LC-PCR and PCR-ELISA assays were highly specific and sensitive, and provide a powerful tool for detection of Salmonella in food samples.