264 



Fishery Bulletin 93(2). 1995 



Numbers of individuals of selected demersal nek- 

 ton species were also recorded for each 1 -minute in- 

 terval. This provided a consistent estimate of rela- 

 tive abundance. Nekton included fishes, cephalopods, 

 and macrocrustacea. As no voucher specimens were 

 collected, identifications were made visually with the 

 assistance of specialists familiar with the groups 

 (noted below in the Results section). Most of these 

 forms could be confidently identified only to genus 

 from the videotapes. 



Data analysis 



Data from dives at Cape Hatteras and at Cape Look- 

 out were analyzed separately. Cape Hatteras dives 

 spanned a depth range that included two faunal zones 

 (upper and middle slope) identified by Haedrich et al. 

 ( 1980) and Wenner and Boesch ( 1979). In general, these 

 authors found differences between continental slope 

 communities above 700 meters and those below. Thus, 

 Cape Hatteras dives 2629 and 2630 (Table 1) were con- 

 ducted on the upper slope, whereas dive 2623 was con- 

 ducted on the middle slope. Dive 2627 crossed the 

 boundary identified by Haedrich et al. ( 1980). All Cape 

 Lookout dives were conducted on the middle slope. 

 Depth was not included as a variable in the statistical 

 analyses detailed below, because it was not recorded 

 for each 1-minute videotape segment. Potential effects 

 of biotic zonation were investigated separately by com- 

 parison of species distribution with dive depth. 



Statistical analysis of habitat choice by the identi- 

 fied nekton followed Felley and Felley (1986, 1987) 

 and Felley et al. ( 1989). All statistical analyses were 

 conducted with the SAS program (SAS Institute, 

 1988). The steps in the analysis were as follows: 1) 

 calculation of species' mean abundances for environ- 

 mental variables; 2) calculation of a correlation ma- 

 trix among species' mean abundances; 3 ) factor analy- 

 sis of the correlation matrix; 4) comparison of vari- 

 ances of sampling units and of numbers of individu- 

 als of a species on the artificial variables (factors) 

 generated by the factor analysis. These steps were 

 accomplished as follows. 



First, we calculated means of environmental vari- 

 ables for each species as each variable's mean over 

 1-minute intervals, weighted by the number of indi- 

 viduals of that species seen in each interval. Thus, a 

 species' mean abundance for a variable represented 

 the value of that environmental variable in inter- 

 vals where the species was most likely to be found. 

 The species' mean abundance was considered the 

 species "preference" for the variable, assuming that 

 these nektonic species select their habitat. 



Second, species' mean abundances for the environ- 

 mental variables were used to construct a correla- 



tion matrix among the variables. Note that this cor- 

 relation matrix implies standardizing each variable 

 using a "mean of means" and a "standard deviation 

 of means." A high correlation between two variables 

 is seen when species tend to occur in habitats with 

 contrasting values for both of these variables. For 

 example, an analysis might include some species 

 found typically in shallow gravel areas and some 

 preferring deep sandy areas. This analysis would 

 generate a high correlation between such environ- 

 mental variables as depth and substrate particle size. 

 Thus, patterns of habitat use by species are reflected 

 in patterns of interrelations among the variables. 

 This is an analysis of species associations with par- 

 ticular environments, and the data contain no infor- 

 mation about why a particular species is occurring 

 more often in one habitat type than another. 



Third, factor analysis (principal component analy- 

 sis with Varimax rotation, Mulaik, 1972) was per- 

 formed on the correlation matrix. Factor analysis 

 resolves patterns of interrelationships among vari- 

 ables into a smaller set of composite variables (fac- 

 tors) to which observed variables (species mean abun- 

 dances) are correlated. Sets of interrelated variables 

 correlating highly with a factor are variables reflect- 

 ing similar patterns of habitat use among the spe- 

 cies. Each factor represents a particular trend in 

 habitat use, an axis differentiating among sets of 

 species that are likely to be found in habitats with 

 contrasting conditions for the variables that define 

 the factor. The example above might produce a fac- 

 tor defined by depth and substrate particle size. 



Species' values, or scores, for a factor can be calcu- 

 lated by using a factor scoring function. Species with 

 contrasting scores are those found most often in con- 

 trasting environments relative to that factor. To con- 

 tinue the example, species more likely to be found in 

 deep water over sandy substrate would have factor 

 scores that contrasted with those of species more of- 

 ten found in shallow water over coarse substrates 

 (i.e. positive vs. negative scores on the factor). Species 

 with intermediate scores may be characteristic of in- 

 termediate environments on that factor, or they may 

 be found over the entire range of environments reflected 

 by a factor (because the species' score is the weighted 

 mean of scores of intervals where it was found). 



Note that only a subset of the species analyzed in 

 the example may in fact select habitat based on en- 

 vironmental variables related to depth and substrate. 

 Though a trend in habitat use may be identified for 

 a species assemblage, not all species in the assem- 

 blage may select habitat according to that trend. 

 Further analysis is required to determine which spe- 

 cies show evidence of active selection according to a 

 particular habitat trend. 



