160 



Fishery Bulletin 88(1), 1990 



Figure 3 



Mean increment width ol' Atlanlic ineiihailen larvae from 

 control (fed, ■) and treatment (1-3 day starved, O) tanks 

 before, during, and after the respective starvation inter- 

 val, (a) 1-day starved, (h) 2-day starved, and (c) 3-day 

 starved larvae. Vertical bars represent standard errors. 



increased with longer starvation, and both age classes 

 of larvae responded similiarly to these stressful events. 

 Starved larvae in both age classes displayed narrow 

 [1.4±0.1 fim, n =60, (0.7-2.2 ^^m, range), 13-20 days; 

 1.8 ±0.3 ^m, n =63, (1.0-3.2 j^m), 28-36 days] poorly 

 defined increments while control larvae exhibited wider 

 [2.1±0.2Mm, n = 30, (1.2-2.8 fim), 13-20 days; 2.7 ±0.6 

 /im, n =57, (1.3-4.6 j:.im), 28-36 days] well-defined in- 

 crements during the starvation interval. After starva- 

 tion, mean increment width of starved larvae in both 

 age classes increased during the 3-6 day recovery in- 

 terval (Fig. 3). 



The results of the ANOCOVA, which adjusted for 

 mean increment width prior to the initiation of star- 

 vation, indicated that mean increment width differed 

 significantly (])<0.05) between individuals from control 

 and treatment tanks during starvation in both age 

 classes (Table 2). This result suggests that differences 

 in mean increment width arose during the starvation 

 interval and were not due to differences prior to this 

 interval. However, during the recovery interval, no 

 significant differences (;)>0.05) were observed between 

 larvae from control and treatment tanks when adjusted 

 by the covariate. This result suggests that differences 

 observed between individuals from control and treat- 

 ments during the recovery interval were the result of 



differences formed during the starvation interval, and 

 indicates that increment width of larval Atlantic men- 

 haden responds rapidly to short-term variations in 

 feeding. 



Relationship between sagittal size and body 

 size and sagittal size and age 



Visual inspection of residuals for regressions of stan- 

 dard length and estimated dry weight on sagittal 

 radius, and sagittal radius on days after first feeding, 

 indicated that nonlinear models were superior to linear 

 models in all cases. Asymptotic regressions of standard 

 length on sagittal radius were fit separately for larvae 

 from control and pooled treatments (Fig. 4a). The 

 regressions were highly significant (p<0.0001) and 

 residuals were distributed at random over the entire 

 size range examined, indicating that the models fit the 

 data well. The regression coefficients for fed larvae 

 were slightly larger than for starved larvae. A logistic 

 function was used to regress estimated dry weight on 

 sagittal radius (Fig. 4b). Data were log-transformed to 

 stabilize the variance and reduce the inlluence of larger 

 values on the regression. Starved larvae were excluded 

 from this relationship since dry weights were estimated 

 from standard length. The regression was highly 



