it was desirable to examine other lakes, especially for pygmy smelt, to see whether they 

 fell reasonably well into one of the categories. We try to accomplish this and other goals 

 in the present paper. 



Luey, Krueger and Schreiner (1982) examined proteins of Osmerus samples using 

 electrophoretic methods. Samples from Lake Huron (originating from rainbow smelt in Green 

 Lake, Maine) were closer to a sample of the western European smelt from Lake Ijssel, 

 Netherlands, than to Osmerus mordax dentex from Alaska. The genetic distances suggested 

 that the Alaskan samples were specifically distinct from the two Atlantic basin collections 

 which were closer than would be expected for congeneric species. This contrasts with 

 Klyukanov's (1975) results; he found 7 osteological and meristic characters which linked 

 0. m. mordax to 0. m. dentex, but only 2 which linked 0. m. mordax to 0. eperlanus. 



MATERIALS 



One hundred and twenty-six specimens of both types taken from three geographical areas 

 were examined (Table 1). The samples from Lake Utopia were taken during the populations' 

 spawning runs and were composed of sexually mature individuals. All samples were fixed in 

 10% formalin and preserved in 50% isopropyl alcohol. An additional sample of 14 pygmy 

 smelts from Meach Lake, Gatineau Park, Quebec, which originated from a transfer from Utopia 

 Lake, New Brunswick, were examined in the final stages of this study, but not included in 

 all the analyses. 



METHODS 



Counting and Measuring 



Body measurements were made with Helios dial calipers to the nearest 0.1 mm, according to 

 the methods of Legendre (1952). Vertebral tounts, including the urostyle, were made from 

 X-rays taken on a Softex EMB machine using Kodak Type M2 film. Gill raker counts, including 

 the rudiments, were made under a binocular microscope. Scale counts were made along the 

 midlateral line to the caudal flexure. 



Statistical Methods 



The principal components analysis (PCA) program of Davies (1971) was translated from 

 FORTRAN to BASIC and run using a Hewlett-Packard 9845 minicomputer with 12 digit precision. 

 Measurements were converted to log to the base 10. The meristic values and log measurements 

 were standardized by subtracting the mean value for that character and dividing by the 

 standard deviation of that character. The PCAs were run using these values in correlation 

 matrices . 



12 



