A test protocol for the detection and enumeration of Clostridium botulinum in fish and sediment samples with specific identification of neurotoxin types A, B, E and F was developed. Specific amplification products generated by polymerase chain reaction (PCR) formed the basis of identification of the toxin-producing organism, whereas quantification of the results was achieved with the most probable number (MPN) method. Twenty-six C. botulinum strains studied with PCR assays after enrichment in trypticase-peptone-glucose-yeast extract (TPGY) broth gave identical results as with the mouse bioassay. The suitability of the detection method for food and environmental surveys was assessed by running it on 32 samples of rainbow trout inoculated with spore loads ranging from 10(2) to 10(6) C. botulinum type E spores per kg. The organism was detected in all samples, and MPN estimates corresponded well to inoculum levels. In order to assess possible natural contamination, 16 fish and 16 visceral samples of rainbow trout, as well as ten aquatic sediment samples were tested. Of these, eight (80%) of the sediment samples were positive, with estimated spore counts of C. botulinum type E ranging from 95-2710 per kg sample.