3. For butterfish, estimation of the parameter p was not possible without applying a correction factor of 1.3, the ratio of the two codend mesh sizes, to the selectivity equation for the control codend. The application of this correction factor is based on the assumption of geometric similarity, whereby the selection curve of the control codend is assumed to be a scaled version of the selection curve of the experimental codend when selection depends only on the relative geometry of the mesh and the fish ( Millar and Fryer, 1999). For butterfish, the geometric similarity assumption is appropriate given the linear relationship between fish length (L) and maximum girth (G) defined as:G = −2.433 + 0.949L, r2 = 0.89 (Myer and Merriner, 1976). When the correction factor was applied, relative fishing efficiency was estimated as 0.30 for a logistic model fitted to catch data for the length class range of 37.5–187.5 mm. The fit resulted in a very large and positive trend in the deviance residuals for length classes greater than 137.5 mm and a smaller and negative trend in the 117.5–132.5 length classes (Fig. 3D). However, a curve fit through the catches of length classes greater than 137.5 mm resulted in unrealistically high estimates of p and L50 because catches composed of these larger length classes did not reach an asymptote. L50, SF and SR estimates were: 96 mm, 1.5 mm, and 26 mm, respectively.