Paralytic shellfish toxins were quantified in whole tissues of the mussel Mytilus galloprovincialis exposed to blooms of the dinoflagellate Gymnodinium catenatum in Portuguese coastal waters. A validated liquid chromatography method with fluorescence detection, involving pre-chromatographic oxidation was used to quantify carbamoyl, N-sulfocarbamoyl and decarbamoyl toxins. In order to test for any matrix effect in the quantification of those toxins, concentrations obtained from solvent and matrix matched calibration curves were compared. A suppression of the fluorescence signal was observed in mussel extract or fraction in comparison to solvent for the compounds dcGTX2 + 3, GTX2 + 3 and GTX1 + 4, while an enhancement was found for C1 + 2, dcSTX, STX, B1, dcNEO and NEO. These results showed that a matrix effect varies among compounds. The difference of concentrations between solvent and matrix matched calibration curves for C1 + 2 (median = 421 ng g⁻¹) exceeded largely the values for the other quantified compounds (0.09-58 ng g⁻¹). Those differences were converted into toxicity differences, using Oshima toxicity equivalence factors. The compounds C1 + 2 and dcNEO were the major contributors to the differences of total toxicity in the mussel samples. The differences of total toxicity were calculated in ten mussel samples collected during a 10-week blooming period in Portuguese coastal lagoon. Values varied between 53 and 218 µg STX equivalents kg⁻¹. The positive differences mean that the estimated toxicity using solvent calibration curves exceed the values taking into account the matrix. For the toxicity interval 200-800 µg STX equivalents kg⁻¹ an increase was found between 44 and 28%.