Bioassays are routinely employed for sediment quality assessment. In order to be able to effectively use Bioassays responses in regulatory and management frameworks, toxicity scores, which rank toxicity data in defined classes that are continuous and difficult to interpret, should be reliable and suitable tools to support decisions about the presence or absence of toxicity in tested samples and on how toxic a sample is. A statistical approach is needed to define thresholds for toxicity scores. The Minimum Significance Difference (MSD) criterion allowed the evaluation of toxicity thresholds for each test-matrix and organism pair, based on large sets of experimental data. The MSD values were normalized with respect to the control, ranked in ascending order, and the 90th percentile was identified; the Toxicity Threshold (TT) was calculated by subtracting the 90th percentile from 100 and the Toxicity Limit (TL) was estimated as the percentage of control response multiplied by TT. Taking into account sample responses normalized with respect to control (S), when S > TL, the sample is considered nontoxic; when S <or= TL, toxicity is statistically present. The MSD criterion was applied on a dataset of results from both published and ongoing projects using the validated embryotoxicity tests with the echinoid Paracentrotus lividus, the oyster Crassostrea gigas, and the mussel Mytilus galloprovincialis, on elutriates from sediments of the Venice Lagoon. Once the threshold defining presence or absence of toxicity had been calculated, four other toxicity classes were developed: low (for toxicity data expressed as percentage of effect), medium, high, and very high toxicity (for toxicity data expressed as TU50).