Toxoplasmosis is one of the most common food borne disease world-wide. Among food animals, sheep seems to having higher prevalence of Toxoplasma gondii infection. However, there is no consensus about the best cut-off for serodiagnosis in sheep. To estimate the more adequate cut-off value of Modified Agglutination Test (MAT) for serodiagnosis in sheep, a commercial ELISA kit was used as a golden standard. Evaluation of the optimal sensitivity and specificity was calculated using Youden's J-statistics. Values obtained were used to estimate the prevalence of sheep toxoplasmosis. One thousand four hundred and sixty seven blood samples were collected randomly from 160 farms from northern Portugal, representing approximately 10% of the ovine population from the region. All sera were tested for anti-T. gondii antibodies using the MAT. One hundred nine sheep (7.4%) presented a MAT titer > or = 1:80; 45 (3.0%) a MAT titer of 1:40; 97 (6.6%) a MAT titer of 1:20 and 1216 (83.0%) a MAT titer < or = 1:20. The best Youden's J-statistic was obtained at 1:20 titer (0.752), with 86.15% of sensitivity and 89.09% of specificity with negative and positive predictive values of 90.32% and 84.48% respectively, suggesting that the 1:20 was the most appropriate cut-off for serodiagnosis of toxoplasmosis in sheep. Assuming this cut-off, the prevalence of toxoplasmosis in the studied population was 17.1% and 92 (57.5%) of the 160 studied flocks having one or more positive sheep. Those results indicate that toxoplasmosis in Portugal should be considered in the differential diagnosis of abortions in sheep and neurological signs in lambs. Furthermore, while Portugal produces ovine meat for internal consumption and for exportation, isolation of T. gondii from ovine meat and further characterization of the isolates will be needed to understand the risk that ovine toxoplasmosis may represent for human health.