514 



Fishery Bulletin 100(3) 



Table 3 



Results of one-way ANOVA comparing growth increments among size-intervals at the harbor wall from the two sample periods. 

 Subsequent post hoc Student-Newman-Keuls multiple-range test results are summarized in the matrix below; an asterisk indi- 

 cates that molt increments differ significantly (P<0.05) between the two samples being compared. 



Source of variation 



df 



88 



MS 



Between groups 

 Within groups 

 Total 



13 

 486 

 499 



332.57 

 258.48 

 591.06 



25.58 

 0.53 



48.1 



<0.001 



Size class (mm CLi 



1971-72 



1996-97 



15-20 20-25 25-30 30-35 35-40 40-45 45-50 15-20 20-25 25-30 30-35 35-40 40-45 4.5-50 



1971-72 



1996-97 



15-20 

 20-25 

 25-30 

 30-35 



35-40 

 40-45 

 45^50 



15-20 

 20-25 

 25-30 

 30-35 

 35-40 

 40-45 

 45-50 



P<0.0001 in all cases). Three females at the harbor wall 

 >50 mm CL were excluded from all analyses (unshaded 

 circles, Fig. 2A) because they were the only specimens of 

 this size. Further, females >50 mm CL might have had 

 smaller than expected molt increments due to approach- 

 ing maturity (Pollock, 1973). 



Two-way, fixed effects ANCOVA comparing these linear 

 regressions indicated no interaction between the factors 

 site and sex; nor was there a significant sex-effect (Table 

 4). However, there was a significant site-effect: the rela- 

 tionship for samples from Mouille Point had a significantly 

 higher regression constant than that for samples from the 

 harbor wall (Table 4). These results imply first that there 

 were no differences in length-specific growth increments 

 between the sexes within sites. The second implication is 

 that molt increment increased with carapace length at 

 the same rate, irrespective of sex or site. Finally, juvenile 

 lobsters of any given size or sex at Mouille Point grew sig- 

 nificantly more during each molt stage than coiTesponding 

 specimens at the harbor wall (see also Table 2 and Fig. 2). 



Log-linear analysis indicated significant first-, second-, 

 and third-order interactions among the factors site, size- 

 category, sample month, and molt state (Table 5). How- 

 ever, the only significant third-order interaction was that 

 among the independent factors site, size-category, and 

 sample month. This interaction simply confirmed that 

 size-frequency distributions differed between sites over 



the sample period, as would be expected if growth rates 

 differed. Of the three remaining nonsignificant third- 

 order interactions, the most important for our purposes 

 were those including the site factor. These imply that the 

 temporal pattern of molting is similar between sites, as is 

 the size-specific pattern of molting. These results are sup- 

 ported by graphical representations of the data (Fig. 3), 

 which suggest that for both sites 1) the proportion of molt- 

 ers per size-category decreases as size increases, and 2) 



