244 



Fishery Bulletin 88(2), 1990 



pored lateral line scale and continuing posteriorly to 

 the plica associated with the last pored scale (Fig. 3). 

 Because plicae are much easier to count than scales, 

 the folds themselves were counted between the two end 

 points. If, however, the regular serial arrangement of 

 plicae was interrupted (i.e., size and direction of slope 

 of the plicae become irregular), the pored lateral line 

 scales were counted in order to maintain consistency 

 in counts between specimens. Plicae counts made in 

 this manner are much more consistent. 



Meristic data were divided into four geographic 

 regions in order to describe and analyze geographic 

 variation: Labrador, Quebec-Nova Scotia, Maine- 

 Massachusetts, and New York-North Carolina. Since 

 three species of Ammodytes may be present in Green- 

 land waters, Greenland specimens were analyzed sep- 

 arately. Conclusions made with regard to the other 

 geographic regions were applied to Greenland collec- 

 tions in an effort to determine which species are pre- 

 sent in this region. 



To determine if two species could be distinguished 

 objectively, principle component analysis (PCA) was 

 conducted on a covariance matrix for meristic char- 

 acters (plicae, vertebrae, dorsal, anal, and pectoral fin 

 rays, and gill rakers) for 332 individuals. These speci- 

 mens, all with a complete complement of meristic 

 values, were pooled from locations throughout the 

 geographic range, exclusive of Greenland. An attempt 

 was made to give equal representation to each area; 

 however, there were not many offshore collections 

 from Labrador available. Since the data set contained 

 both individuals with low meristics and those with high 

 counts, the assumption was made that both forms were 

 present within the sample. An individual's score on the 

 first component was determined by entering its values 

 for meristic characters into an equation. Component 

 scores were plotted and two groups are clearly present 

 (Fig. 4). The point of least overlap was chosen as the 

 boundary between the two species. Data points were 

 then coded using earlier tentative identifications (based 

 on observed meristic trends) to determine the effec- 

 tiveness of the methodology. The equation and the 

 value of the score at the boundary between the two 

 species were cross-validated by testing an additional 

 54 specimens. 



Results 



Number of lateral plicae was the most useful single 

 character in separating the two species, even though 

 plicae numbers varied considerably between individuals 

 and slightly from side to side in individual specimens 

 (Table 1). Of 723 specimens included in the analysis, 

 98.8% could be separated into species at a line of 



Figure 4 



Histogram of principal component analysis (PCA) scores for si.x 

 meristic characters oi Ammodytes. Ranges: A. am.ericanus 1199- 

 1378, A. duhius 1378-1572. Component scores were calculated using 

 the following equation: (8.09 x number of plicae) + (2.44 x number 

 of dorsal rays) + (2.31 x total number of vertebrae) + (1.23 x total 

 number of gill rakers) + (1.03 x number of anal rays) + (0.45 x 

 number of pectoral rays). 



separation between 124 and 125 plicae (Fig. 5). Ammo- 

 dytes americanus had fewer plicae, ranging from 106 

 to 126 (x 117.4) with virtually no geographic varia- 

 tion; A. dubius had more plicae (124-147, x 132.1) 

 and exhibited geographic variation. Plicae means for 

 A. dubius ranged from 131.2 in the New York-North 

 Carolina region to approximately 132-136 from Massa- 

 chusetts northward. 



Number of vertebrae, a standard character examined 

 by previous researchers interested in Amnwdytes, also 

 proved to be a relatively good character in separating 

 species. Although some overlap in vertebral numbers 

 exists, particularly south of Nova Scotia, 92.3% separa- 

 tion was achieved at a line of separation between 68 

 and 69 vertebrae (Table 2). Ammodytes americanus had 

 fewer vertebrae, with mean values ranging from 65.2 

 in specimens from New York to North Carolina and 

 increasing to a mean value of 67.2 in specimens from 

 Quebec to Labrador. Ammodytes dubius characteris- 

 tically had more vertebrae than A. americanus, with 

 means ranging between 70-71 in the southern portions 

 of its range (Maine to North Carolina) to 73.8 in the 

 Quebec-Nova Scotia region and 71.9 off Labrador. 

 There is a distinct increase in vertebral number {x 

 72.9) in specimens of A. dubius taken off the Scotian 

 Shelf and northward. 



Dorsal ray counts were also effective in separating 

 specimens. Using this character alone, 92.3% of the 

 specimens could be separated at a line of separation 

 between 59 and 60 dorsal rays. Am.modytes americanus 

 had relatively consistent dorsal ray counts throughout 

 the entire geographic range (Table 3; 52-61, x 57.4); 

 however, in comparison, number of dorsal rays for A. 

 dubius was higher for specimens collected north of 



