Maillet and Checkley Effects of starvation on Brevoortia tyrannus 



157 



the mean calculated. If triplicate counts deviated by 

 ± 3 increments or more, the specimen was excluded 

 from fui'ther analysis. Five percent of the samples were 

 thereby excluded. Another 16% of the specimens were 

 excluded from the analysis because of our inability to 

 resolve increments in some specimens and/or improper 

 orientation for viewing specimens. Linear regressions 

 of mean increment count on days after first feeding 

 were computed for each experimental group. Stu- 

 dent's t was used to determine whether increment for- 

 mation is initiated at first feeding (i.e., regression in- 

 tercept = 0) and growth increments are formed daily 

 (i.e., slope = 1.0). Statistical power to detect a devia- 

 tion of 0.1 from a slope of 1.0 at /) = 0.05 level (two- 

 sided test) was estimated for each linear regression 

 (Rice 1987, Steel and Torrie 1980). Analysis of variance 

 was used to test the homogeneity among all slopes. 



The width of growth increments formed before, dur- 

 ing, and after starvation was measured for 5-10 lar- 

 vae from each experimental tank. Increment width was 

 measured from the jjerimetei- moving inward along the 

 maximal radius, the inner end of the radius being 

 defined as the center of the nucleus. This line was 

 consistently the best for enumerating increments. To 

 eliminate bias in measurement of increment width the 

 focal plane was adjusted, if necessary, for the measure- 

 ment of each increment. Increment width was aver- 

 aged for each larva over each of the three treatment 

 intervals (i.e., fed, starved, refed), except for the 1-day 

 starved treatment during starvation. Analysis of co- 

 variance (ANOCOVA) was used to compare the mean 

 width of growth increments between larvae from the 

 control and ti'eatment tanks during and after starva- 

 tion (Steel and Torrie 1980). Since the duration of the 

 starvation and recovery intervals varied with length 

 of starvation, the ANOCOVA tested for differences in 

 mean increment width between larvae from control and 

 treatment tanks within intervals. The interval duration 

 was identical between comparisons of mean increment 

 width of larvae from the control and treatment tanks. 

 Comparison of increment width between fed and 

 starved larvae within intervals was necessary because 

 of age-related trends in increment width. The covariate 

 included in the ANOCOVA was mean increment width 

 prior to the starvation and recovery intervals. The 

 ANOCOVA took into account any differences in the 

 width of growth increments of larvae allocated to the 

 experimental tanks and provided a more accurate com- 

 parison between individuals from control and treat- 

 ment groups. 



The relationships between sagittal radius, standard 

 length, estimated dry weight, and days after first 

 feeding were also investigated. We pooled all of the 

 larvae in the treatment groups (i.e., 1-, 2-, and 3-day 

 starved larvae) for analysis, except that starved larvae 



were excluded when estimating the dry weight— otolith 

 size relation since these measurements were not made 

 directly on individual larvae. Sagittal radius was mea- 

 sured from the center of the nucleus to the perimeter 

 along the maximal radius. Dry weight of individual lar- 

 vae was estimated from a linear regression of log- 

 transformed dry weight {DW, pig) on log-transformed 

 standard length (SL, mm) : In DW = -3.041 -i- 3.799 

 In SL , n = 195, r- = 0.95. This equation was derived 

 for laboratory-reared Atlantic menhaden larvae rang- 

 ing in size from 5 to 25 mm (Checkley et al., ms in 

 prep.). Linear and nonlinear regressions of standard 

 length and estimated dry weight on sagittal radius, and 

 sagittal radius on days after first feeding were com- 

 pared to determine the best predictive models. Average 

 growth rates {AVG, mm/day) of larvae for both age 

 groups were estimated from first feeding (ff) to experi- 

 ment termination hy.AVG = (SL - 4.7)/(days after ff), 

 where 4.7 is the average standard length (mm) at first 

 feeding (Powell and Phonlor 1986). 



Results 



Increment description and larva growth rate 



Sagittae were first obser-ved dui'ing embryonic develop- 

 ment and consisted of the dark and apparently pro- 

 teinaceous primordium (Fig. 1). Examination of sagit- 

 tae with the light microscope revealed that no growth 

 increments were formed during this period. At hatch- 

 ing, the sagitta resembled a flattened spheroid or hemi- 

 sphere with a mean radius of 4.8 ± 0.3 fxm, n = 15. The 

 first increment surrounding the primordium was 

 typically characterized by a wide discontinuous zone 

 and coincided with hatching. In some cases, 3-4 nar- 

 row, poorly defined increments were observed outside 

 the first increment. Limited resolution of the light 

 microscope did not always allow these increments to 

 be resolved, counted, and/or measured. The first prom- 

 inent increment in sagittae formed 3-4 days after 

 hatching and coincided with the initiation of exogenous 

 feeding. This prominent growth increment was used 

 as the starting point for subsequent counts along the 

 maximal radius. 



Mean size at hatching for this laboratory population 

 of Atlantic menhaden was 3.59 + 0.22 mm SL, n = 100 

 (range 3.08-4.00 mm). The average growth rate of 

 menhaden larvae from first feeding to experiment 

 termination was 0.368 ±0.006 mm/day (13-20 days) 

 and 0.360 + 0.004 mm/day (28-36 days). Size range for 

 the two age groups examined was 6.4-13.0 mm and 

 10.3-20.0 mm SL. This rate of growth slightly ex- 

 ceeded the estimate of 0.32 mm/day based on data of 

 Powell and Phonlor (1986) for Atlantic menhaden 

 larvae reared at 20°C. Larvae did not metamorphose 



