590 



Fishery Bulletin 9! [4), 1993 



FA tests for traits not normally distributed, 

 further analyses were canned out twice, once 

 with and once without such traits. A second 

 order ANOVA was applied to test for influ- 

 ences of sex and year class on growth rate. 

 Finally, fish were catergorized as fast or slow 

 growers, according to whether growth rate 

 fell above or below the median for the year 

 in question, and a three-way MANOVA was 

 performed to test for effects of growth rate, 

 sex, and year on FA and right-left otolith 

 shape differences. 



Results 



For both the 1980 and 1984 year classes, 

 we found that the distributions of weight, 

 width, and length did not differ significantly 

 from normality (Table 1). These variables 



Table 1 



Tests for normality of (L-R)/(L+R) for various 

 traits used in developmental instability. (L in- 

 dicates measure on the left, R the correspond- 

 ing measure on the right otolith ). 



Width of otolith in = 147) 

 Mean = 0.002 

 Skew = 0.184 

 Kurtosis = 1.333 

 Shapiro-Wilk W = 0.990 

 P = 0.9456 



Weight of otolith in = 135) 

 Mean = 0.003 

 Skew = 0.242 

 Kurtosis = 0.683 

 Shapiro-Wilk W = 0.994 

 P = 0.9965 



Length of otolith (n = 135) 

 Mean = 0.001 

 Skew = 0.049 

 Kurtosis = 2.030 

 Shapiro-Wilk W = 0.975 

 P = 0.1981 



Growth rate of otolith ( 1980 year class; n = 79) 

 Mean = -0.001 

 Skew = -1.567 

 Kurtosis = 6.124 

 Shapiro-Wilk W = 0.981 

 P = 0.0001 



Growth rate of otolith ( 1984 year class; n = 69) 

 Mean = 0.002 

 Skew = -0.281 

 Kurtosis = 0.248 

 Shapiro-Wilk W = 0.981 

 P = 0.7044 



can, therefore, be used in analysis of FA. Growth rate was not 

 normally distributed. The two-way ANOVA indicated no influ- 

 ence of sex (7^=1.42; df=l, 141; P=0.234) and no significant inter- 

 action effect of sex X year (F=0.36; df=l,141; P=0.550) on growth. 

 The main effect of year, however, was strong (F=14.20; df=l,141; 

 P=0.0002). This last result, along with the possibility that growth 

 rate might influence asymmetry of other characters, provided the 

 rationale for incorporating growth rate into the three-way 

 MANOVA, as noted above. Results were qualitatively the same — 

 no differences with respect to which comparison deviated signifi- 

 cantly — whether growth rate was or was not included in the 

 MANOVA, and so only one result set, that including this vari- 

 able, is presented in this paper (Table 2). There was a highly 

 significant difference in developmental instability, but only be- 

 tween the years. Univariate analysis showed that FA was greater 

 for the El Nino fish with respect to otolith weight and shape, but 

 not width or length (Tables 3 and 4). 



To examine the possibility that the between-year differences 

 were not due to El Nino effects, we examined otoliths of 4-year- 

 old fish from another year class (1977) unaffected by El Nino 

 events. As with the above two groups, fish sampling took place 

 during the first two weeks of September. These otoliths were 

 analyzed for right-left asymmetry in weight and shape. Results 

 of a Multiple Analysis of Variance for all three year classes, with 

 fixed effects of year and sex, are given in Tables 5 and 6. The 

 results, in conjunction with Tukey tests for differences among 

 years (Table 7) show clear increases in right-left differences in 

 both shape and weight for the 1980 El Nino year over 1984 and, 

 in shape, over 1977 as well. Differences between the two years 

 not affected by El Nino (1977 and 1984) were not statistically 

 significant. 



Discussion 



The observation that only two of the five right-left asymmetry 

 measures responded to stress at a statistically detectable level 

 raises the question of consistency. If developmental instability is 

 to be a useful indicator of stress, we need to know what measures 

 will be relatively sensitive to stress and which will not. Highly 



