A neuroblastoma assay for determination of brevetoxins in shellfish was developed together with a method for sample cleanup that allows separation of brevetoxins from most of the components that cause matrix interference in the assay. This improved assay method was applied to a range of shellfish samples with different characteristics. Extracts of naturally contaminated and nontoxic shellfish together with extracts spiked with known amounts of toxin were tested. The results demonstrated that brevetoxins could be reliably detected in shellfish extracts at concentrations below the regulatory limit. Brevetoxin activity was detected in 15 of 23 samples from 5 separate toxicity incidents in which shellfish tested positive in the neurotoxic shellfish poisoning (NSP) mouse bioassay. Twelve of these positive NSP results came from 2 toxicity incidents. Yessotoxin was the major contributor to toxicity in 2 other incidents, although some samples contained both yessotoxin and brevetoxin. The sample from the remaining incident contained an unidentified toxin bioactivity, together with gymnodimine. In contrast to earlier versions of the neuroblastoma assay, gymnodimine was not detected by this modified method.