KENNEDY ET AL.: TEMPERATURE-TIME RELATIONSHIPS 



Mercenana mercenana 

 cleavage stages 



Mercenana mercenana 

 straight  hinge larvae 



Figure 1. — Mercenaria mercenaria cleavage stages. Response 

 surface generated from multiple regression analysis of per- 

 centage mortality on temperature and time. Refer to Table 1 

 for appropriate temperatures. 



Mercenana mercenaria 

 trochophore larvae 



Figure 3. — Mercenaria mercenaria straight-hinge larvae. Re- 

 sponse surface as in Figure 1. Refer to Table 3 for appropriate 

 temperatures. 



mortality associated with exposure in the block 

 at normal temperatures (Kennedy et al., 1974). 

 Percentage mortality data for each developmental 

 stage were determined as for Mulina lateralis 

 (Kennedy et al., 1974) and are presented in Tables 

 1 to 3. The stepwise multiple regression program 

 transformed these data to arcsine square root of 

 the percentage mortality to allow the distribution 

 to approximate the normal. First, second, and 

 third order terms for main effects (temperature, 

 time) and all possible interactions were 



'Figure 2. — Mercenaria mercenaria trochophore larvae. Re- 

 '\ sponse surface as in Figure 1. Refer to Table 2 for appropriate 

 .temperatures. 



scrutinized. Only those terms (variables) withF ^ 

 3.96 (P = 0.05, d.f. 1, 80) were entered in the final 

 equation. The program selected variables making 

 greatest reduction in residual sum of squares until 

 no further variables satisfied the acceptance 

 criteria. The final empirical models appeared to be 

 good predictive equations for all three stages 

 (Table 4). The derived constant and variables 

 selected for each stage are presented in Table 5, 

 along with other statistics. The equations incor- 

 porating these constants and variables allow cal- 

 culation of predicted percentage mortality for dif- 

 ferent combinations of temperature and time. The 

 resulting estimates are in transformed form and 

 must be converted to untransformed values ( Sokal 

 and Rohlf, 1969). Figures 1 to 3 were constructed 

 using these equations and seven temperature 

 levels to outline the basic pattern of the estimated 

 response surface. 



For each stage, the coefficient of determination 

 ranged between lY7c to SCK^r when all the variables 

 were selected (Table 5), indicating that most of 

 the variation in mortality can be explained by 

 these variables (Steel and Torrie, 1960). For 

 cleavage stages and trochophore larvae, T^, by 

 itself, was the best single predictor of percentage 

 mortality. This was also true for straight-hinge 

 larvae although T^ was eventually eliminated 

 by the program as new variables entered. In 

 combination with the other variables in the final 

 predictive equation, T^ continued to be the most 

 useful variable in estimating or predicting per- 

 centage mortality for trochophore larvae, as 



1163 



