PETERSON ET AL.: IMPACT OF MECHANICAL CLAM HARVESTING 



significantly higher initial values (finer sediments), 

 although the differences among matrices were 

 small. Percent organic content did not differ sig- 

 nificantly {P > 0.05) among seagrass matrices in a 

 one-way ANOVA on angular-transformed propor- 

 tions. 



The results of this initial sampling in spring 1980 

 prior to any application of clam harvest treatments 

 imply that the sand-flat matrices were initially quite 

 homogeneous. Consequently, any treatment effects 

 can be expected to appear as significant differences 

 that emerge among matrices in some or all sam- 

 plings after application of the treatments. However, 

 the initial differences among seagrass matrices im- 

 ply that treatment effects may not be so readily 

 identified. For those variables that exhibited initial 

 differences among matrices, we performed two dif- 

 ferent tests of the effects of treatment. We per- 

 formed simple ANOVA' s to test for differences 

 following treatment and we also, by subtraction of 

 matrix means for spring 1980, adjusted the data 

 from each matrix for initial differences and tested 

 by ANOVA for significant changes in the differences 

 among matrices. The first approach is appropriate 

 if one believes that initial differences among 

 matrices do not reflect intrinsic between-matrix dif- 

 ferences that require adjustment, whereas the sec- 

 ond approach assumes that initial differences among 



matrices would be expected to persist or recur in 

 the absence of any treatment. An examination of 

 how replicate matrices vary over time helps resolve 

 which test procedure is more appropriate, but we 

 performed both tests to provide a more robust set 

 of conclusions. 



Although all matrices in each habitat were chosen 

 to be homogeneous in surface appearance, our Octo- 

 ber 1985 estimates of coarse shelliness of the sur- 

 face (0-12 cm) sediments demonstrated that sea- 

 grass control I had almost 10 times the amount of 

 coarse shell than any of the other seagrass matrices. 

 The average ( ± SE) mass of shell fragments >3 mm 

 in the top 12 cm of the 0.25 m^ area in seagrass 

 control I was 5,257 g ( ± 701) compared vdth a range 

 of 375 (±70) to 777 (±135) g across the other 5 

 seagrass matrices. This substantially larger amount 

 of shell (P < 0.001 in a one-way ANOVA) seemed 

 to be present during the entire experiment. Because 

 surface shell fragments could greatly influence sea- 

 grass growth and especially M. mercenaria recruit- 

 ment and survival (see Castagna and Kraeuter 

 1977), this physical anomaly of seagrass control I 

 renders it a questionable control for the various 

 treatment matrices. Similar data on surface shelli- 

 ness taken from the sand matrices in October 1985 

 revealed no significant differences (P > 0.05) among 

 matrices in a one-way ANOVA, with mean (±SE) 



Table 3. — Contrasts among replicate matrices within each habitat before application of harvest treatments. Data are sample means ( ± SE) 

 from spring 1980 (22 Feb. -6 May). Sample sizes appear in Table 1 . Superscripts A and B indicate significant differences among matrices 

 in Duncan's test at a = 0.05, with those means sharing capital letter superscripts not differing significantly. Where ANOVA was non- 

 significant, no means differ significantly. 



'Recruits defined as <2.5 cm in length (see Peterson et a! 1983b for size data on year class as support). 



2* - P < 0.05, " - P < 0.01, *** - P < 0.001. NS - P > 0.05 in one-way ANOVA comparing matrices before experimental initiation. 



287 



