The shape of the catch efficiency curves for 

 sockeye and chum salmon between 44 cm. and 

 62 cm. and for pink salmon between 38 cm. and 

 56 cm. is somewhat similar. The pink salmon 

 curve is displaced to shorter fish. Pink salmon 

 probably have greater girth per given length than 

 the other two species, although no girth measure- 

 ments of pink salmon were taken to verify this. 

 Differences shown in figure 10 probably 

 result from girth/length differences among the 

 three species. 



DISCUSSION 



Certain assumptions in applying Holt's method 

 were considered. One assumption was that stand- 

 ard deviations of selection curves should be similar. 

 A computation of the S values for each year from 

 1956 to 1960 checked this assumption. Table 16 

 lists jS values for each pair of mesh sizes. Sockeye 

 and chum salmon S" values are given for all years. 

 Pink salmon S values were computed only for odd- 

 numbered years; catches were small in even- 

 numbered years. 



Standard deviations of selection curves within 

 each species were reasonably similar in at least, the 

 larger mesh sizes, 3)1-, 4%-, and 5%-inch. Pink 

 salmon had slightly higher S values and sockeye 

 slightly lower S values in the paired 3%- and 4^- 

 inch mesh sizes than in the4&- and 5}4-inch (table 

 16). Chum salmon also varied only slightly be- 

 tween these sizes. In the 2%- and 3%-inch pair 

 of mesh sizes, however, the S values for sockeye 

 and chum salmon were low. The small 2%-inch 

 mesh was probably the main cause of these low 

 values. 



Table 16. — Standard deviation of mesh selection curves for 

 pink, sockeye, and chum salmon by year 



Another assumption was that the mean selection 

 length of salmon is proportional to mesh size. 

 Mesh size (perimeter) is directly related to the 

 fish's girth. Lander (1963) showed that the 

 girth and length of sockeye and chum salmon of 

 the high seas have a linear relation. Thus, the 

 relation of length of salmon and mesh size war- 

 rants using proportionality constants (K values). 



K values within species varied remarkably 

 little annually (table 17). Between species, K 

 values for pink salmon were lower than those for 

 sockeye and chum salmon, probably because 

 pink salmon have greater girth per given length. 

 As shown in table 17, mean selection lengths had 

 lower values in pink salmon than in the other two 

 species. 



All investigators did not use normal distribu- 

 tion for the mesh selection curve. Some used 

 a skewed mesh selection curve, tailing off to the 

 right, rather than a normal curve. Olsen (1959), 

 working with Newfoundland herring data, found 

 that logs of catch ratios followed a parabolic 

 line better than a straight line. His selection 

 curves, thus, are slightly skewed rather than 

 normal. Ishida (1962) used a mesh-size ratio 

 method in developing skewed selectivity curves 

 for salmon from the North Pacific. Gulland and 

 Harding (1961), using gill net catches of the 

 African catfish Clarias, obtained a skewed selection 

 curve with a long upper tail. The shape of 

 Clarias (long fish with a large bony head) and 

 the method of its capture (entanglement in 



Table 17. — Proportionality constants and mean selection 

 lengths for pink, sockeye, and chum salmon 



i if= Proportionality constant of fish length divided by mesh size. 



GILL NET MESH NET SELECTION CURVES FOR SALMON 



3S9 



