The protein phosphatase inhibition assay for okadaic acid, the major DSP toxin, modified to use the fluorescence substrates methylumbelliferyl phosphate (MUP) and fluorescein diphosphate (FDP), was compared to the assay using p-nitrophenylphosphate (p-NPP) and the bioluminescence assay using luciferin phosphate (L–P). Under the standard assay conditions used okadaic acid inhibited the enzyme activity dose-dependently with IC50 values of 1.5 nM (MUP) and 1.2 nM (FDP). This compares to IC50 values of 0.9 and 6 nM using L–P and p-NPP respectively. CDP-star, a chemiluminescence substrate, was not hydrolysed by the enzyme. Decreasing the enzyme concentration lowered the IC50 for the colorimetric method (IC50=2 nM [p-NPP], 0.75 nM enzyme) but no shift was observed with fluorimetry. However at enzyme concentrations <1.5 nM (standard assay) the error margin was too great for routine analysis. The method using fluorimetry allowed detection of okadaic acid concentrations to levels ≤1 μg/100 g of mussel tissue which is well below the limit of 20 μg/100 g (mouse bioassay) set by some regulatory agencies. Determination of the toxin content in naturally contaminated mussels in three separate experiments gave coefficients of variance ranging from 16 to 29% (MUP) and from 8 to 78% (p-NPP). Multicomparison studies showed that concentrations of okadaic acid in naturally contaminated mussel samples determined by fluorescence generally agreed with those obtained using ELISA and LC–MS procedures, and with the mouse bioassay. However using the mouse bioassay as the standard, values determined by the ELISA, PP-2A and LC–MS all scored false negative results compared to those for the mouse bioassay in the range 20–40 μg/100 g mussel, and at the limit of the mouse bioassay the values by the other three methods were substantially less. With few exceptions the methods scored okadaic acid with highest to lowest values in the following order: mouse bioassay>ELISA>PP-2A>LC–MS. The fluorimetric assay was both more sensitive and accurate than the colorimetric assay (the latter showed a propensity towards false positives in the region 20 μg/100 g), and the moderate increase in equipment cost appears to be outweighed by the performance of the method.