FISHERY BULLETIN: VOL. 83. NO. 3 



did not result in homogeneity of variances). Except 

 for the residuals for three small larvae, analysis for 

 residuals showed no trends. 



For larvae with 60 increments and fewer (Fig. 7), 

 Bartlett's test showed homogeneity of variance. 

 These data were better fitted by two lines (P < 0.01); 

 the slopes were significantly different (P < 0.01). 

 The change in length was 0.34 mm/increment and 

 0.24 mm/increment for early- and late-hatched lar- 

 vae, respectively. 



The von Bertalanffy growth equation fit the early- 

 hatched larval data poorly (Table 4). L^ was esti- 

 mated at 113.22 mm, with a 95% confidence interval 

 of 16.37 to 210.06 mm. The von Bertalanffy growth 

 function could not be fitted (solution would not con- 

 verge) to the late-hatched larval data. 



The 1978-79 data could not be tested under 

 assumptions that increment deposition could vary 

 from 0.5 to 1.0 increment/d. Almost all of the calcu- 

 lated hatch dates for late-hatched larvae, estimated 

 on deposition rates of 0.5 increment/d, overlapped 

 the classification division date. Too few points were 

 left for analysis. 



DISCUSSION 



Evidence from the Gulf of Maine supports the 



hypothesis that increase in length for herring larvae 

 hatched early in the spawning season is greater than 

 for larvae hatched late in the season. These differ- 

 ences were evident both under assumptions of daily 

 otolith increment deposition and for deposition of 

 one increment every other day. Before these dif- 

 ferences are assumed to be due to differences in 

 growth, however, there are other hypotheses which 

 should be considered that could explain these obser- 

 vations. Differences could be the result of within- 

 season changes in otolith increment deposition rates, 

 or of differential mortality due to selective predation. 

 If there are within-season changes in otolith incre- 

 ment deposition rates, growth (change in length at 

 age) could actually be similar, but the calculated 

 growth rates would appear to be different because 

 they are expressed as change in length per incre- 

 ment count. In order for this hypothesis to explain 

 the above results, larvae born early in the season 

 would be required to put down fewer increments per 

 time period than would larvae born late in the 

 season. The data allow a test of the hypothesis that 

 larvae lay down fewer than 1 increment/d during the 

 early part of the year. When estimated hatching 

 dates are calculated for larvae caught early in the 

 season, under the assumption that one increment 

 was deposited every other day, some of these larvae 



35 -1 



LARVAL HERRING GROWTH 

 1978-1979 



TIME OF HATCH 



EARLY • 

 LATE A 



10 



20 



I r \ 



30 40 



INCREMENT COUNTS 



50 



60 



Figure 7. - Regression plot of length-at-otolith increment count for Atlantic herring. Only lengths for larvae with 60 or fewer otolith incre- 

 ments have been included for analysis. Data from Townsend and Graham (1981). 



296 



