A newly revised enrichment and agar-plating system was tested for selectivity and sensitivity in recovery of unstressed and cold-stressed Escherichia coli O157:H7 from foods. Various foods inoculated with known levels of enterohemorrhagic E. coli O157:H7 (EHEC) were tested by enrichment for 6 h at 37°C in modified tryptic soy broth (mTSB) base supplemented with vancomycin, cefsulodin and cefixime, referred to as EHEC enrichment broth (EEB). Subsequently, portions were spread-plated on sorbitol–MacConkey agar supplemented with tellurite and cefixime (TCSMAC). Further selective enrichment was also examined using immunomagnetic separation (IMS) from the EEB prior to spread-plating on TCSMAC agar. These methods were compared to a procedure of enrichment in mTSB (supplemented with novobiocin) at 37°C for 24 h followed by spread-plating of decimal dilutions on hemorrhagic colitis 4–methylumbelliferyl–B–D–glucuronide (HC–MUG) agar. The new enrichment isolation technique was found to be sensitive at a level of one EHEC organism per 10 g of food in four food types. This represents an approximate l00-fold to 1,000-fold enhancement in sensitivity over the comparative method for foods with high levels of competitive microflora. These enrichment-isolation protocols also were compared in analysis of naturally contaminated raw or undercooked ground beef samples implicated in foodborne illness. EEB-TCSMAC with and without IMS were combined with rapid biochemical tests, and with O157 latex agglutination and confirmation of toxin genes by polymerase chain reaction (PCR) to provide a completed test within 30 h of initiating testing. The new system was successful in 15 of 17 samples, where only 6 of 17 were found positive by the comparative technique.