Direct, indirect and double antibody sandwich methods of ELISA were examined for their ability to detect and discriminate between granulosis viruses from Pieris brassicae, Agrotis segetum and Cydia pomonella and for their specificity in the presence of host material. The indirect method was the most sensitive, capable of detecting down to about 1 ng of dissolved capsules/ml compared with 10 ng/ml for the double antibody sandwich method and 25 ng/ml for the direct method. The double antibody sandwich method showed greatest discrimination between different granulosis viruses; the direct method was not quite so specific and the indirect method varied considerably depending upon the antibody concentration used although even with this latter method cross-reactions were restricted to baculo-viruses and did not occur with occluded insect viruses in other taxonomic groups. All three methods were highly specific in the presence of large amounts of host material and could be used to assay virus in diseased larvae, although at very high levels of contamination, equivalent to adding 100 healthy larvae to one diseased larva, only the double antibody sandwich method could still reliably detect virus though with reduced absorbance readings. Possible reasons why the ELISA shows much greater discrimination between different Baculovirus inclusion bodies than other serological techniques are discussed.