The selectivity of square and diamond mesh codends with different mesh sizes (25mm, 30mm, 35mm, 40mm and 45mm) of two-stick swingnet was analyzed through covernet method selective experiments, in which Logistic curve was regarded as the selection curve and the maximum likelihood method as estimation method of the parameters of selectivity model. The result demonstrated that the 50% retention body length (or fork length) for Setipinna taty and Pampus argenteus of all diamond mesh codends were larger than those of all square mesh codends (3.79% and 32.15%, respectively), while that for Pseudosciana polyactis was smaller than that of square mesh codends (7.92%). Additionally, the selection ranges for the three species mentioned above of diamond mesh codends were larger than those of square mesh codends (20.76%, 5.36% and 45.99%, respectively). There exists some difference between the experimental result and the traditional opinion. In the discussion, we consider that the relatively flat body shape and the low swimming ability of Setipinna taty and Pampus argenteus mainly contribute to the difference. As a result, the traditional viewpoint that square mesh is superior to diamond mesh should not be adhered blindly to when we want to reform the construction of codend.