Four enrichment PCR protocols for detecting unlysed cells of pathogenic Yersinia enterocolitica were studied. First, the probability of detecting Y. enterocolitica cells of known concentrations by a multiplex PCR assay was determined, and it was found to follow a logistic regression model. From this model, the probability of detecting Y. enterocolitica at a specific concentration could be estimated; for example, the detection probability of 10 4 CFU/ml was estimated to be 85.4%. The protocols were evaluated on enrichment cultures inoculated with 10 2 CFU/ml Y. enterocolitica and 10 2-10 6 CFU/ml of a defined background flora. For each protocol, the time for sample withdrawal and the presence of background flora were studied with respect to PCR detection. The optimal point in time of sample withdrawal was found to be different for each protocol employed. Early detection was favoured by concentrating the target cells, and the most rapid PCR detection of Y. enterocolitica was achieved with enrichment in Yersinia-PCR-compatible-enrichment (YPCE) medium for 3 h at 25 °C, followed by a centrifugation prior to PCR analysis. For detection of Y. enterocolitica in the presence of high concentrations (10 6 CFU/ml) of background flora, a long incubation time followed by density centrifugation and a dilution step was most successful. The protocol that gave the most reliable PCR detection in the presence of 10 6 CFU/ml background flora included 24 h incubation in Yersinia-selective-enrichment (YSE) broth at 25 °C, followed by Percoll density centrifugation, and a 100 times dilution prior to PCR analysis. © 2002 Elsevier Science B.V. All rights reserved.