FIGURE 1. — Northwest Atlantic area sampled by NMFS Groundfish Survey. In the present study, data collected from the Middle 

 Atlantic Bight area (strata 1-12, 61-76) between fall 1967 and spring 1976 were examined. 



(2-16 stations per stratum). At each station a tow 

 of l /2-h duration at a speed of 3.5 kn was made 

 along the bottom. A standard #36 Yankee trawl 

 was utilized except during the spring cruises from 

 1973 to 1976, when a modified high-opening #41 

 Yankee trawl was used. The fishes captured were 

 identified, counted, and weighed by species. A 

 bathythermograph cast was made at each station. 

 Further details of sampling design and sample 

 processing may be found in Clark and Brown 

 (1977) and Grosslein (1969). 



Analyses 



Clustering 



Catch data were initially analyzed separately 

 for each of 18 cruises, using numerical classifica- 

 tion (clustering). Assemblages of fishes were de- 

 fined by computing a similarity coefficient, Srv k )> 

 among species from the species-station matrix and 

 subsequently classifying species into clusters or 

 groups (Sneath and Sokal 1973). Stations were 

 clustered in the same manner from the inverted 

 matrix, and species and station (site) groups were 

 compared by nodal analysis (Lambert and Wil- 

 liams 1962). Matrix values entered were counts of 



individuals, as biomass measurements are overly 

 influenced by the presence of relatively rare but 

 large, motile fishes (which are poorly sampled by 

 trawls) in the collections. 



The similarity coefficient used was the Can- 

 berra metric (Lance and Williams 1967), which is 

 particularly effective when the organisms under 

 study are contagiously distributed (Clifford and 

 Stephenson 1975) as most fishes are. Also, to 

 further reduce the effects of contagion, the numer- 

 ical abundance data were transformed [log 10 (x + 

 D] before analysis (Taylor 1953). Species were 

 eliminated from cluster analysis if they occurred 

 at <5 < 7c of the stations occupied during a sampling 

 period. Although this is a more severe data reduc- 

 tion than is commonly employed, examination of 

 the raw matrix and trial runs at various cutoff 

 levels indicated that species occurring below this 

 level showed highly inconsistent distributions. 



The clustering strategy used was flexible fusion 

 with beta set at the conventional value of -0.25 

 (Boesch 1977). Calculations were performed on an 

 IBM 370-115 3 at the Virginia Institute of Marine 

 Science using the Fortran IV program COMPAH 



3 Reference to trade names does not imply endorsement by the 

 National Marine Fisheries Service, NOAA. 



296 



