FISHERY BULLETIN: VOL. 83, NO. 4 



individuals collected during 1979 and of specimens 

 used for gut analysis also were recorded. A complete 

 description of collecting methods and the study site 

 are presented in Grossman (1982, in press b). 



Collection dates were assigned either to the Up- 

 welling or the combined Oceanic-Davidson Current 

 (ODC) oceanographic seasons, based on water tem- 

 peratui'e and a calculated upwelling index (see Gross- 

 man in press b). The Upwelling period is character- 

 ized by frequent upwelling of nutrient-rich waters 

 and high productivity (Bolin and Abbott 1963; Gar- 

 rison 1979). In contrast, the ODC period exhibits 

 minimal upwelling, frequent downwelling, and low 

 productivity (Bolin and Abbott 1963; Garrison 1979). 

 Data from individuals collected during the latter 

 seasons were pooled because these periods were not 

 readily distinguishable (Grossman in press b). Con- 

 sequently, comparisons of 0. snyderi food habits and 

 length-weight relationships represent contrasts 

 between a high productivity period (Upwelling, all 

 months from April to August, plus March 1980) and 

 the low productivity ODC period (October-February, 

 plus March 1979). Monthly mean water temperatures 

 ranged from 9.7° to 11.9°C during the Upwelling 

 period from 11.2° to 14.2° C during the ODC period 

 (Grossman in press b). 



Length-Weight and Growth Calculations 



Regressions of individual weight versus standard 

 length were calculated separately for males and 

 females collected in the Upwelling and ODC periods. 

 Regression equations were of the form W = aL'', 

 fitted as In M^ = In a + 6 In L. Slopes (b) of regres- 

 sion equations for males and females were compared 

 for each period by analysis of covariance (ANCOVA). 

 Seasonal differences between slopes also were ex- 

 amined for each sex. 



Regression slopes for the two sexes and periods 

 were divided by their correlation coefficients to 

 estimate slopes for geometric mean functional 

 regressions of weight versus length (Ricker 1973). 

 Functional regressions are recommended for 

 describing relations between weight and length 

 because both variables are subject to natural 

 variability (Ricker 1973). Functional regression 

 slopes were used to derive ^/-intercepts from the 

 equation y-int. = y - {b/r)x (Ricker 1973). These 

 slope estimates were employed to calculate instan- 

 taneous rates of growth in weight (G) as 



G = (blr) (In L2 - In Lj), (Ricker 1975), 



where b/r = slope of the functional length-weight 

 646 



regression; 

 Lj = mean length at the time t; and 

 L2 = mean length at time ^ -1- 1. 



Growth rates were calculated for individual year 

 classes which were separated using length-frequency 

 distributions. lb compare growth rates for males 

 and females, ^tests for paired samples were used. 

 Sex ratios for the 1979 samples were tested for 

 deviations from unity using chi-square tests cor- 

 rected for continuity. 



Dietary Analyses 



Stomach contents were removed from a subsam- 

 ple (597 total) of the 0. snyderi collected between 

 January 1979 and July 1981. Ninety-seven percent 

 of stomachs examined contained food. Prey were 

 identified to the lowest possible taxon and weighed 

 (wet) to the nearest milligram. The weights of each 

 prey type were pooled separately for males and 

 females from each collection. Diets were then calcu- 

 lated as the percent of the total stomach content 

 weight attributable to each prey typa The number 

 of fish containing each prey type (i.a, prey frequen- 

 cy of occurrence, henceforth FO) also was recorded. 



Dietary data were pooled across collections wdthin 

 the Upwelling and ODC periods for each sex, to com- 

 pare feeding habits between sexes and seasons. 

 Because the study period spanned 3 yr, it also was 

 possible to quantify among year variability in 

 seasonal and sex-specific feeding habits. Sex-linked 

 ontogenetic dietary changes were examined by 

 separating males and females into 20 mm size classes 

 and comparing prey consumed by each size class in 

 different seasons and years. Niche breadth was 

 calculated as I l/p,;^ (Levins 1968), where p, is the 

 proportion of the total prey weight comprised of the 

 ith prey type. Only prey types constituting at least 

 1% of the total prey weight were included in niche 

 breadth calculations. Niche breadth was compared 

 among size classes by using the Wilcoxon signed- 

 rank test. 



RESULTS 



Length-Weight Relationships 



The slope for the male length-weight regression 

 for the Upwelling period was significantly greater 

 than that for females (ANCOVA, F = 12.875, P < 

 0.001; T^ble 1). Slopes of length-weight regressions 

 for males and females collected during the ODC 

 period were not significantly different (ANCOVA, 



