given by Lough (1975a). The calculated regres- 

 sion coefficients from a particular equation are 

 fitted by computer to a full quadratic equation in 

 temperature and salinity in order to print a con- 

 tour diagram of the response surface. Tempera- 

 ture and salinity scales on all plots were set to 

 range beyond the experimental conditions in 

 order to facilitate response comparison and to 

 allow the overall form of the surface to be vi- 

 sualized. Contours extrapolated beyond the ex- 

 perimental data lie outside the dotted lines. 



A summary of the multiple regression analyses 

 on survival after the various periods of rearing 

 and the response surfaces are given in Table 6 

 and Figure 4. The analyses indicated that after 

 20 days of rearing under the experimental condi- 

 tions S and S^ were the two most important vari- 

 ables in the model. T and S x T were of lesser 

 importance but still contributed significantly to 

 the model. Analyses of the later rearing periods of 

 C. magister emphasized the effect of temperature 

 and showed the decreasing importance of both S 

 and S^ and S x T. This trend is more evident 

 when one compares the response surface plots 

 from 20 through 50 days of rearing. After 20 days 

 of rearing, the response surface contours are 

 nearly circular, with a slight tilt to the main axis. 



FISHERY BULLETIN: VOL. 74, NO. 2 



indicating a small interaction effect. The axis of 

 the contours tilts progressively towards the 

 temperature axis until, at 50 days of rearing, the 

 contour axis is almost perpendicular to the tem- 

 perature axis. Also, the survival contours progres- 

 sively constrict about the temperature axis with 

 time showing the narrowing of the temperature 

 range tolerated by the larvae. Maximum survival 

 (80% contour) at 20 days is predicted to occur be- 

 tween 6.5° and 17.5°C and 21.5 and 35.01., while 

 at 50 days, maximum survival is predicted to 

 occur between 9.0° and 15.0°C and above 28.5'L. 

 The area of maximum survival (80% contour) 

 shifts somewhat during the 20- to 50-day period 

 from an initial low salinity-wide temperature 

 range to a high salinity-low temperature toler- 

 ance. However, when the 20- and 50-day survival 

 polynomials were tested by an analysis of 

 covariance (Ostle 1963:205), they were not found 

 significantly different in their response (Table 7). 

 In summary, salinity appears to exert an im- 

 mediate effect on C. magister larval survival, 

 while the effect of temperature becomes increas- 

 ingly important with time. 



Survival at a given temperature, salinity, and 

 time can now be estimated using the fitted equa- 

 tions. All of the fitted equations for the four time 



Table 6. — Multiple regression analyses of Cancer magister larval survival in 20 temperature and salinity 



combinations. 



362 



