Dilutions of faecal samples spiked with Yersinia enterocolitica O:3 were analysed using immunomagnetic separation (IMS) followed by PCR. In 10% faecal dilutions with added Y. enterocolitica cells, the limit of detection was 200 cells g-1 faeces. Faecal samples from 38 pigs were analysed by IMS-PCR in parallel with detection and quantification of Y. enterocolitica O:3 using cold pre-enrichment culturing. Of the 15 culture-positive samples, only two were detected with IMS-PCR. These two samples contained 40-400 Y. enterocolitica O:3 cells g-1 faeces; the highest level found in the investigation. This indicated that the low sensitivity of IMS-PCR was due to low amounts of cells in the faecal samples. Swab samples from 195 pig tonsils, taken on a slaughterline were examined using IMS-PCR and culture detection. Of 164 culture-positive samples, 60 were positive with IMS-PCR. In addition, IMS-PCR was positive for three culture-negative samples. Forty-five of the samples were further examined by IMS-PCR after 7-10 d of cold pre-enrichment. All 31 culture-positive samples as well as five culture-negative samples were detected by IMS-PCR. From these data it can be concluded that IMS-PCR can be used to detect Y. enterocolitica O:3 cells after pre-enrichment, but direct detection needs further optimization of the sample preparation procedures.