FISHERY BULLETIN: VOL. 85, NO. 2 



mass of shell fragments >3 mm ranging from 28 

 (±7) to 157 (+121) across the 6 sand-flat matrices. 

 Our field plots were closed to all commercial and 

 recreational shellfishing during the 4 years of the 

 experiment by proclamation of the North Carolina 

 Division of Marine Fisheries to avoid disruption of 

 the experiments. However, on 7 occasions out of 50 

 days of observation, we observed clammers within 

 the boundaries of our plots: 5 times in seagrass con- 

 trol matrix I and once in both the seagrass raking 

 matrix and the intense-kicking II matrix. This 

 represents significantly more illegal clamming in 

 control I than would be expected by chance alone 

 (P < 0.01 in a binomial test). Thus, the seagrass con- 

 trol I matrix may not represent a true control for 

 our experiment. 



Posttreatment Sampling 



Mercenaria mercenaria Recruitment 



In the sand-flat habitat there were only two Octo- 

 bers during which M. mercenaria recruits were 

 sampled: October 1980 after the initial application 

 of the clam harvest treatments and October 1981 

 after both treatment applications. In neither sam- 

 pling did a one-way ANOVA on log {x + l)-trans- 

 formed counts (which removed heteroscedacity in 



Cochran's tests) reveal significant (a = 0.05) vari- 

 ation in average density of recruits among sand-flat 

 matrices (Table 4). Furthermore, a two-way ANOVA 

 on log (x + l)-transformed counts from both time 

 periods, done to increase the power of the test of 

 matrix differences, also failed to reveal any signifi- 

 cant variation in average recruitment among sand- 

 flat matrices. Despite the failure to demonstrate 

 statistical significance in M. mercenaria recruitment 

 among sand-flat matrices, the average density of 

 recruits in the control matrices during these two 

 Octobers was more than double (on untransformed 

 scale) the average density in the 2 high-intensity 

 clam kicking matrices (Fig. 3). Some of this differ- 

 ence may have been present even before treatments 

 were applied (Fig. 3), but it is also possible that the 

 high local variability in recruitment lowers the 

 power of this test of harvest treatment to a degree 

 that even a twofold difference is undetectable. 



During 4 Octobers, M. mercenaria recruitment 

 was estimated in the seagrass habitat (Table 4). One 

 of these, October 1980, fell after the first harvest 

 treatment (which Table 2 shows to have been very 

 light in the seagrass plots) but before the second, 

 more intense treatment. The other 3 samplings 

 came in successive years, increasingly far from the 

 actual time of application of the harvest treatments. 

 Because of the preexisting significant differences 



Table 4. — The impact of clam harvesting on recruitment of Mercenaria mercenaria. Entries are mean 

 numbers (± SE) of recruits per % m^. Recruits are defined as all individuals <2.5 cm in length in October 

 of each year. For 1980 and 1981 , n = 36 samples from each treatment matrix in each habitat, whereas 

 for 1982 and 1983, n = 9 for seagrass and for sand flat. 



' * - P < 0.05, " - P < 0.01 in one-way ANOVA's on each date and two-way ANOVA's over all dates, reported in the 

 unweighted average column. These analyses were performed on log-transformed data, which eliminated or reduced 

 heteroscedacity in Cochran's tests Superscripts A and B indicate significant differences in Duncan's test at o = 0.05. 

 No Duncan's test results are given for the unweighted averages in the seagrass bed because the two-way ANOVA ex- 

 hibited highly significant (P < 0.001) interaction between date and treatment. 



288 



