and photographed in the field). Toller collected and preserved vouchers from fishery landings when possible and photographed specimens representing new records. Toller's vouchers were processed into the fish collection at the NMNH and are also available in the online database.
Some species were only taken during fishery bottom trawl surveys by the M/V Oregon (1958, 1959) and the R/V Pillsbury (1969) on Saba Bank. Species represented by voucher specimens in museum collections are included in our comprehensive species list.
A non-metric multidimensional-scaling (MDS) ordination of the 12 rotenone and roving stations based on a Bray-Curtis similarity matrix of incidence data was used to illustrate similarities and differences of fish assemblages found on the four atoll habitat types: fore reef, patch reef, lagoonal, Small Bank (Figure 2). There were significant differences among habitat types (ANOSIM Global R = 0.96, P = 0.001). Differences between fore-reef stations and patch-reef stations were most pronounced (ANOSIM, Pairwise R = 0.92, P = 0.002). Fore-reef assemblages ranged from 39 to 60 species per station while the patch-reef stations had 26 and 32 species. Fore-reef stations were 50% similar to each other, and patch-reef stations were up to 60% similar. Fore-reef assemblages were not significantly different than Small Bank, or the lagoonal habitat (ANOSIM, Pairwise R = 0.973, P = 0.11). Low sample size (one station with 39 species) in the lagoonal habitat limits this comparison.
There were significant differences in the fish assemblages when classified by depth - shallow, middle, and deep (ANOSIM Global test, R = 0.618, P = 0.006). Differences in species composition were most evident between mid-depth and deep sites (ANOSIM Pairwise test, R = 0.829, P = 0.008). Fore-reef sites were typically at mid to shallow depths (20 to 34 m) whereas the patch-reef sites were typically deep (35 to 38 m). There is some evidence of habitat heterogeneity and vertical zonation for fish assemblages on Saba Bank, but more sampling is necessary to discern whether habitat or depth best explains the differences among groups.
The horizontal axis in the MDS plot represents a gradient in species richness with patch-reef and lagoonal sites having lowest richness ( Figure 2). The vertical axis in the MDS plot illustrates differences related to depth with deepest station (Small Bank) higher in the vertical axis, and shallower sites lower in the vertical axis. The two-dimensional stress value was low in the MDS (stress = 0.1) indicating a slight chance of misrepresentation.
A principal components analysis (PCA) was used to understand which species assemblages were responsible for differences among stations. The first three components explained 43.9% of the variation. The first component (18.6% of variation explained) was dominated by ubiquitous species and most common species. Strongest loadings (negative) on the first principal component (PC1) included the four species found at all stations (Halichoeres garnoti, Serranus tigrinus, Stegastes partitus, and Thalassoma bifasciatum) and 13 species found at all but one to four of the stations. These 13 common species are rarely found at the small bank, patch-reef, and lagoonal stations and, therefore, PC1 also serves to define the fore-reef sites. For example, Acanthurus bahianus, Coryphopterus glaucofraenum, Coryphopterus dicrus, and Scarus taeniopterus were found at all fore-reef sites but rarely at the Small Bank, lagoonal and at least one of the patch-reef sites. The strongest positive loadings on PC1 are from 11 species only found at the lagoonal site. Principal components 2 and 3 correspond strongly to the horizontal or species richness component on the MDS with strongest positive