FISHERY BULLETIN: VOL. 85, NO. 2 



DISCUSSION 



The one-way ANOVA's which we performed to 

 test the significance of differences in parameter 

 means among matrices at any given samphng date 

 can demonstrate heterogeneity among matrices. If 

 there is no significant heterogeneity, we probably 

 can conclude safely that there was no effect of treat- 

 ment on that parameter at that sampling date, 

 assuming that equivalent levels of the parameter 

 prevailed before application of the treatment (which 

 was not always true). If, on the other hand, the one- 

 way ANOVA demonstrates significant differences 

 among matrices, this result does not necessarily im- 

 ply that the treatment was the cause. Replication 

 in these ANOVA's is generated from subsamples 

 within each individual matrix. These subsamples 

 taken from within a given matrix are not indepen- 

 dent because of their spatial proximity. Consequent- 

 ly, matrices can diverge in various ways from one 

 another over the course of an experiment, caused 

 by extraneous events that act on the scale of the plot 

 (matrix) to destroy independence among subsam- 

 ples. This experimental design would be termed 

 pseudoreplication (Hurlbert 1984), and permits a 

 test of whether plots differ significantly and does 

 not allow an unambiguous assignment of observed 

 differences to the treatment applied (but see 

 Stewart-Oaten et al. 1986). For that reason, we 

 replicated both our control matrices and our intense- 

 kicking matrices in each habitat. These permit us 

 to use a priori contrasts, with replication of 2 sep- 

 arate, independent plots, to test unambiguously 

 whether the most important treatment (intense clam 

 kicking) was responsible for observed changes. Ap- 

 preciation of the differences between these two sorts 

 of analyses is necessary to interpret properly the 

 results of this study. 



Although we designate our heavier clam-kicking 

 treatment "intense", it probably falls well short of 

 the effort that commercial clammers would apply 

 to a productive seagrass bottom; we took only an 

 estimated 50% of the clams legally available for 

 harvest (Table 2). Consequently, the intensity of 

 harvest that we applied in the seagrass is not un- 

 reasonably high. In the sand-flat system, we took 

 approximately 100% of the estimated numbers of 

 legally available clams in our intense treatments. 

 Although higher than the percent taken in the sea- 

 grass, this probably better approximates the fish- 

 ing intensity that is applied to productive unvege- 

 tated areas by commercial clammers. Efficiency of 

 returns remained high even in the high-intensity 

 kicking matrices, as compared with hand raking. In 



the sand flat, light kicking produced an average of 

 8.1 clams per minute and intense kicking 6.2 clams 

 per minute, compared with a return of only 0.9 

 clams per minute from hand raking (Table 2). In the 

 seagrass bed, light kicking yielded an average of 8.1 

 clams per minute and intense kicking 16.1 clams per 

 minute, in contrast to a return of only 0.4 clams per 

 minute from hand raking (Table 2). Thus, efficiency 

 of harvest, defined as clams caught per unit of time, 

 was clearly greater by over an order of magnitude 

 with the mechanical technique than with the tradi- 

 tional hand method. The improved efficiency dur- 

 ing clam kicking in the seagrass as harvest inten- 

 sity increased from taking about 15% to about 50% 

 of available clams is probably caused by the gradual 

 removal of seagrasses which, when present, reduce 

 the efficiency of clamming. 



To test whether hard clam harvest affects its own 

 recruitment in the area of harvest, we counted new 

 recruits (<2.5 cm in length, Peterson et al. 1983b). 

 Recruitment, when estimated in this fashion, con- 

 founds both larval (and postlarval) settlement with 

 subsequent early mortality from time of settlement 

 until October. Consequently, we do not directly test 

 the hypothesis that natural densities of adult hard 

 clams inhibit larval settlement in their vicinity. Fur- 

 thermore, our clam harvest treatment not only 

 removes many larger hard clams, but it also disturbs 

 the bottom sediments. Consequently, there are 

 several plausible mechanisms by which our clam 

 harvest treatments may affect October recruitment 

 of hard clams: 1) reduction of adult hard clam den- 

 sity may affect hard clam settlement (positively, if 

 negative adult-larval interactions predominate, as 

 suggested by most past studies: Woodin 1976; 

 Williams 1981; Peterson 1982b) or survivorship from 

 settlement until October (no a priori prediction from 

 the literature on what direction this effect may 

 take), or 2) disturbance of the bottom may alter hard 

 clam settlement (positively, if hard clam larvae 

 select disturbed sediments, which seems unlikely, 

 or negatively if hard clam larvae avoid disturbed 

 sediments) or early survivorship (negatively, if the 

 clam harvest buries small clams too deeply to 

 reemerge or if disturbance has removed protective 

 seagrass or shell materials and thereby made 

 juvenile hard clams more vulnerable to predators 

 (Peterson 1982a; Summerson and Peterson 1984)). 



Our data on hard clam recruitment are sufficiently 

 ambiguous to preclude any definitive answers to the 

 question of how clam harvest affects subsequent 

 recruitment. In the sand flat, there was no signifi- 

 cant effect of harvest treatment, but the 2 intense- 

 ly kicked matrices yielded only 50% of the recruits 



294 



