McGURK: PACIFIC HERRING OTOLITH RINGS 



not significantly different from (t = 2.2831, df = 4, 

 0. 10 > P > 0.20). This indicates that the starvation of 

 first-feeding larvae stopped ring production. The 

 1980B group had a rate which was not significantly 

 different from one of 1 ring/d (t = 2.3397, df = 2, 0.10 

 > P > 0.20) and not significantly different from a rate 

 of (t = 0.6989, df = 2, 0.50 > P > 0.90) or from the 

 rate of its parent feeding group, 1980A (F = 5.9185, 

 df = 1.3, 0.25 > P > 0.50). One reason for these 

 results is that the 1980B group had only three data 

 points for the regression, and the standard error of 

 the slope was therefore relatively high: 122% of the 

 value of the slope (Table 3). I conclude that starva- 

 tion for 5 d after a feeding period of 6-7 d has no effect 

 on the rate of ring deposition. The 1982B group had a 

 ring depoistion rate that was not significantly dif- 

 ferent from (t = 0.7843, df = 3, 0.40 > P> 0.50) and 

 which was not significantly different from the rate of 

 its parent feeding group, 1982A (F = 0.1352, df = 1, 

 3, P > 0.75). I conclude that starvation for 8 d after a 

 feeding period of about 25 d has no effect on the rate 

 of ring deposition, at least not in 25 1 enclosures. 



The average ring deposition rates were significantly 

 positively correlated with the average growth rates (n 

 = l,r = 0.83, 0.01 > P > 0.05) (Fig. 3). The regres- 

 sion of ring rate on growth rate was: 



Ring rate — 0.14 + 2.40 (growth rate). 



The residuals of this regression were not correlated 

 with container size, and there was no obvious 

 relationship with prey type. However, there was a 

 significiant positive correlation between the 

 residuals and the mean rearing temperature (n = 7,r 

 = 0.83,0.01 >P> 0.02). The midpoints of the tem- 

 perature range were used as an estimate of the mean 

 temperature (Table 1). A regression of ring deposi- 

 tion rate on growth rate and temperature increase the 

 multiple r to 0.99: 



Ring rate = -1.39 + 3.36 (growth rate) 

 + 0.14 (temperature). 



These results confirm the correlation between ring 

 deposition rate and growth rate found for Atlantic 

 herring larvae by Geffen (1982), who interpreted the 

 relationship as being curvilinear and linearized it by 

 transforming both variables with logarithms. In order 

 to compare the two sets of data the relationship be- 

 tween ring deposition rate and growth rate was 

 assumed to be linear. A covariance analysis of the 

 slopes of the two linear regressions indicated that 

 there was no significant difference between them at 

 the 0.05 probability level. Data from this study and 

 from Geffen's were pooled and a single linear regres- 

 sion was calculated (n = 12, r = 0.85, P < 0.001): 



Ring rate = 0.17 + 2.12 (growth rate). 



< 



Q 

 U 



z 

 cr 



LU 



i- 

 < 



DC 



O 



z 



DC 



1.2 



1.0 



0.8 



0.6 



0.4 



0.2 



0.0 



-0.2 



-0.1 0.0 0.1 0.2 0.3 0.4 0.5 



GROWTH RATE (MM/DAY) 



FIGURE 3. — Relationship between the average ring deposition rates 

 and the average growth rates of seven groups of Pacific herring lar- 

 vae. See text for regression equation. 



The influence of temperature on ring deposition rate 

 could not be compared between the two data sets 

 because the rearing temperature for Geffen's fish 

 was not constant over the rearing period. 



Plots of fish length on otolith diameter for the seven 

 populations were curvilinear, and the rate of growth 

 of fish length decreased with increasing otolith 

 diameter. Transforming otolith diameter with 

 logarithms best linearized the data, transforming 

 both variable with logarithms produced lower cor- 

 relation coefficients in all groups. Thus length was 

 regressed on log (otolith diameter) (Table 4, Fig. 4). 

 An analysis of covariance that included all seven 

 groups indicated that the slopes of the regres- 

 sions were significantly different from each 

 other at the 0.05 probability level. Inspection of the 

 slopes and their standard errors indicated that the 

 fed groups and 1980B had slopes of a similar value 

 and that 1980C and 1982B had slopes of a similar 

 value but that they were much lower than those of the 

 fed groups. The two groups were subjected to 

 separate covariance analyses, and in each group the 

 slopes were found to be not significantly different 

 from each other at the 0.05 probability level. The 



117 



