greater in all seasons, number of specimens was larger in 

 all seasons except spring, and mass was greater in all 

 seasons except fall. The seasonal patterns for the in- 

 dividual years were quite similar to the preceding. 



The surf zone and tidal pool are related, but quite 

 different habitats. Their ecological relationship can be 

 described as that of a severe, but seasonally relatively 

 stable, environment (the surf zone) with a depauperate 

 fauna which "regularly" contributes segments of its pop- 

 ulations in a partly random fashion to a transient, un- 

 stable environment (the tidal pool) which reappears at 

 predictable intervals unless certain conditions, whose 

 appearances are largely unpredictable, intervene. 



Analysis of Abundance of Fishes 



Abundance in terms of numbers of species and 

 specimens and total weight per collection was analyzed 

 by season, year, and sampling location. A two-factor ran- 

 domized complete-block design was used as the ex- 

 perimental model for each of the dependent variables of 

 interest: number of species, number of specimens, and 

 total weight of fishes per collection. Analysis of variance 

 was employed for each of the three dependent variables 

 in testing the null hypothesis that the effects of sampling 

 location (factor 1, 2 levels; surf, tidal pool), season (fac- 

 tor 2, 4 levels; fall, winter, spring, summer), interaction 

 between sampling location and season, and year (block, 2 

 levels; October 1969 through September 1970, October 

 1970 through September 1971) were nonsignificant. 

 Throughout "nonsignificant" means P>0.05; 

 "significant," 0.01 < P< 0.05; and "highly significant," 

 P<0.01. The number of collections (observations) made 

 for the various treatment combinations was: 



Fall 



Winter Spring Summer 



Year 1: 











Surf 



6 



7 



6 



6 



Tidal pool 



5 



4 



4 



5 



Year 2: 











Surf 



6 



6 



7 



6 



Tidal pool 



5 



3 



6 



3 



It should be noted that the surf and tidal pool collections 

 made on 10 October 1971 (the start of a third year) were 

 not included in these analyses. Because the cell frequen- 

 cies were unequal, an unweighted-means analysis was 

 used as the computational procedure with the harmonic 

 mean (n h -= 4.97) of the number of observations being 

 used as the effective number of observations per cell for 

 computation of main effects and interactions (Winer 

 1962). 



One of the underlying assumptions of the analysis of 

 variance is that of homogeneity of variance. Each set of 

 data was examined for correlation between treatment 

 means and their within-treatment variances. This ex- 

 amination suggested transformations producing 

 homogeneity. An analysis of variance was then con- 

 ducted. 



The within-cell means and variances of the number of 

 species per collection for the various treatment com- 



binations were examined for correlations, but none were 

 found. The within-cell variances for the number of 

 species per collection ranged from 0.25 to 6.97. Bartlett's 

 test for homogeneity of variance gave aX 2 value of 11.04 

 for 15 df which was not significant. We then proceeded 

 with the analysis of variance on the original data, X = 

 number of species of fishes per collection. 



Similar examinations were conducted for the number 

 of specimens and total weight of fishes per collection. In 

 both cases it was apparent that the variances were 

 proportional to the means and were also heterogeneous. 

 Bartlett's test was employed to test the hypothesis of 

 equal within-cell variances for both the number of 

 specimens and total weight of fishes per collection. The 



results of Bartlett's test were X 2 



15df 



= 93.19 and 



* 2 i5df = 99.78, respectively. Both X 2 values were highly 

 significant and we therefore rejected the hypothesis of 

 homogeneity. It appeared that the heterogeneity was of 

 the regular type for both the number of specimens and 

 total weight of fishes per collection, i.e., the variability 

 within the several treatments was proportional to the 

 treatment means. For this reason the logarithmic 

 transformation, log (X + 1), was used to make the means 

 and variances independent and to stabilize the within- 

 cell variances. The logarithmic transformation was made 

 on the number of specimens of fishes per collection, and 

 Bartlett's test was conducted to determine if this 

 transformation homogenized the variances. The result of 

 Bartlett's test on the within-cell variances for X = log 

 (number of specimens of fishes per collection + 1) was 

 * 2 iMf = 13.78. Since this was nonsignificant and it 

 appeared that no relationship remained between the 

 means and variances of the transformed data, we then 

 conducted the analysis of variance on the transformed 

 data, X = log (number of specimens of fishes per collec- 

 tion + l).The logarithmic transformation was also used 

 for the total weight of fishes per collection. Bartlett's test 

 resulted in X 2 15df = 12.85 for the within-cell variances of 

 the transformed data. This was nonsignificant and the 

 analysis of variance was then conducted on the 

 transformed data, X = log (total weight of fishes per 

 collection + 1). 



The results of the analysis of variance for the number 

 of species of fishes per collection (Table 7) indicate that 

 the block or year effect is not significant. The nonsignifi- 

 cant location-by-season interaction shown in Table 7 in- 

 dicates that the surf and tidal pool differences in number 

 of species per collection are constant across season. The 

 location-by-season means of number of species of fishes 



Table 7. — Analysis of variance of X = number of species of fishes per 

 collection. (• Significant, P< 0.05; •* highly significant, P<0.01; ns 

 not significant, P> 0.05.) 



Source 



SS 



df 



MS 



Blocks (years) 



4.92 



1 



4.92 



1.248 ns 



Location 



124.85 



1 



124.85 



31.669 *• 



Season 



39.60 



3 



13.20 



3.348* 



Location by season 



17.04 



3 



5.68 



1.441 ns 



Error 



299.61 



76 



3.94 





Total 



486.02 



84 







10 



