Petrik et al.: Recruitment of Micropogontas undulatus 



959 



been a growing awareness that understanding how 

 demographic processes vary with habitat structure 

 will be critical for predicting population size in fishes 

 that occur in heterogeneous habitats (Levin, 1994). 

 In this study we examined patterns of abundance of 

 newly settled Atlantic croaker and demonstrated that 

 these fish use different estuarine habitats similarly. 

 The results of our mesocosm experiment suggest that 

 the pattern of similar recruitment in different habi- 

 tats results from a lack of preference for specific habi- 

 tats. In addition, when we investigated habitat dif- 

 ferences in postsettlement growth or survivorship, 

 we were unable to detect strong consequences of us- 

 ing one habitat over another. 



We found no evidence suggesting that habitat se- 

 lection by settling larvae and habitat-specific 

 postsettlement mortality are important in determin- 

 ing population size in croaker; however, this conclu- 

 sion is based on nonsignificant statistical tests rather 

 than explicit acceptance of the null hypothesis of no 

 difference between treatments. Recent reviews have 

 stressed the importance of power analysis in detect- 

 ing a type-II errors (Peterman, 1990; Reed and 

 Blaustein, 1995; Thomas and Juanes, 1996), and 

 because we wished to draw conclusions from "nega- 

 tive" results, power analysis was particularly impor- 

 tant. In this study, when a difference between treat- 

 ment means was not detected, we examined power 

 in an attempt to determine our ability to accept the 

 null hypothesis. If the power of the test was too low 

 to accept the null hypothesis ( /5<0.05 ), the number of 

 replicates required to achieve this power level was 

 calculated. For example, no difference was detected 

 in mean croaker number between plots from which 

 predators were excluded and control plots, but the 

 power of this experiment was low (1-/3=0.06). Suffi- 

 cient power to be able to accept the null hypothesis, 

 would have required 550 replicates. Our level of rep- 

 lication was inadequate because of the extreme vari- 

 ability in croaker densities among experimental 

 plots — variation likely produced by a combination of 

 stochastic settlement and habitat-specific mortality. 

 The effect of this variability was to weaken the power 

 of our experiments to detect small, but real differ- 

 ences among treatments. Although our experimen- 

 tal design precluded the detection of small treatment 

 effects, the high number of replicates required to 

 detect these small effects suggests that other pro- 

 cesses are likely to be more important in determin- 

 ing variability in abundance. 



Although many estuarine species select vegetated 

 over unvegetated habitat at settlement (Orth et al., 

 1984), in some cases, initial patterns of settlement 

 have little to do with habitat selection by individual 

 organisms. Settlement may occur at the first suit- 



able habitat encountered regardless of specific at- 

 tributes of that habitat (Bell and Westoby, 1986). 

 Additionally, current patterns may exclude delivery 

 of competent larvae to some habitats (Morgan et al., 

 1996); therefore, even ideal habitats may seldom re- 

 ceive recruits. In such cases, larvae do not select 

 against a habitat, instead that habitat is never an 

 available choice. By experimentally providing habi- 

 tats, and by using a blocked sampling design such 

 that all habitats were available in a particular loca- 

 tion, we eliminated the possibility that settling 

 croaker would not have the opportunity to choose a 

 habitat. In our field sampling and experiments, 

 croaker had the opportunity to choose between veg- 

 etated and unvegetated habitats, but they did not 

 consistently choose one habitat over another. By con- 

 trast, in an identical experiment performed at the 

 same time and in the same study site, pinfish 

 (Lagodon rhomboides) showed strong responses to 

 habitat, food supply, and predators (Levin et al., 

 1997). Pinfish occurred in much higher densities in 

 vegetated than in unvegetated habitats and also gi'ew 

 faster in grass habitats supplemented with food than 

 in unsupplemented or unvegetated habitats. In ad- 

 dition, the presence of predators reduced pinfish 

 numbers by 50%. The pinfish and croaker occupying 

 experimental plots were similar in size ( 15-25 mm 

 SL), and at this size the diets of the two species are 

 similar (Darcy, 1985; Soto et al., 1998). Thus, it is 

 likely that the lack of response by croaker to the habi- 

 tat attributes we investigated is the result of char- 

 acteristics of the species rather than an artifact of 

 sampling or experimental design. 



Selection for specific habitats at settlement may 

 overwhelm variation in larval supply, thus produc- 

 ing variability in recruitment that is associated with 

 the preferred habitat. This appears to be the case 

 for pinfish (Levin et al., 1997). Although croaker of- 

 ten form part of fish assemblages within seagi'ass 

 (Rooker et al., 1998), they appear to have broad mi- 

 crohabitat preferences, and our results suggest that 

 there is no strong fitness consequences for croaker 

 using vegetated versus unvegetated habitats. As a 

 result, resources associated with the benthic habi- 

 tat seem unlikely to determine population size of 

 newly recruiting croaker. Rather, where and when 

 larvae that are competent to settle are delivered 

 should determine population size in croaker. The 

 contrasting results for croaker and pinfish may re- 

 flect a more general difference in the processes de- 

 termining population sizes of fish. For fishes, such 

 as pinfish, where settling larvae select specific habi- 

 tats and postsettlement processes reinforce initial 

 settlement patterns, spatial and temporal variabil- 

 ity in habitat should be a strong predictor of future 



