METHODS 



The specimens that were examined 

 ranged from 26 to 489 mm . in standard length; 

 the majority were either (young) or I (yearling) 

 in age. In order to count the number of gill 

 rakers present on tiie first left branchial arch, 

 it was necessary to remove the arch from the 

 fish. Gill rakers were counted using a binocu- 

 lar microscope; they were tabulated as follows: 

 the number present on the upper arm plus one 

 at the angle of the arch plus the number on the 

 lower arm . All developed gill rakers were 

 counted. At the extreme ends of the arch, 

 tubercles sometimes occurred. If higher than 

 the diameter of tlieir base they were counted; 

 if not, they were considered as rudiments and 

 not counted. 



Some of the specimens had been put in 

 a deep freeze and later were transferred to 

 formalin. In these cases mucus had built up on 

 the gill arches and it was necessary to scrape it 

 off before counts could be made. TTie same 

 situation existed when specimens had been poor- 

 ly preserved. Gill raker counts were not made 

 on fish below 26 mm . due to the difficulty of re- 

 moving the entire gill arch. 



Comparisons between samples were 

 made by the following statistical procedures: 

 analysis of variance, analysis of covariance, 

 t - test and chi- square test and were in accord- 

 ance with the methods followed by Snedecor 

 (1946) and Steel and Torrie (ms.). Chi -square 

 tests were run on samples that had a binomial 

 distribution and on distributions that approached 

 normality but the resulting level of significance 

 was the same as when an analysis of variance 

 was used. 



Statistical significance is designated in 

 the tables as follows: N.S. - not significant; 

 * - significant at the five percent level; ** - 

 significant at the one percent level . If the age 

 of specimens in the sample is not stated in the 

 table title, it is designated after the individual 

 sample as: yg. - young or yr. - yearling. 



Owing to environmental variation Rounse- 

 fell and Everhart (1953) considered it advisable 

 to use the same year class in comparing samples 

 of fish. In this study of gill-raKer counts of 



striped bass, the same year classes are not 

 always available from all the areas studied. 

 Therefore, when several year classes from one 

 river or area were compared and the differences 

 found to be non- significant, these samples were 

 combined and used to make comparisons with 

 other areas. 



When the term "by observation" is used 

 it indicates that a previous t - test with less 

 divergence between the means had been made 

 and the results were significant. 



In this paper a population is considered 

 in the same sense as Mayr, Lindsley and Usinger 

 (1953:308) use the term local population, that is, 

 "the individuals of a given locality which poten- 

 tially form a single interbreeding community." 

 The term subpopulation is used to denote the 

 specimens from the James, York -Rappahannock 

 and Upper Bay Rivers in the Chesapeake Bay fol- 

 lowing the findings of Raney (ms.). 



RESULTS 



Variation in the number of gill rakers 

 with age.- -In order to determine if the number of 

 gill rakers changes with age, a study was made 

 of counts from localities where young of the year 

 and older fish were available. The results are 

 based on fish classified as O, I and a few II. 



The best way to determine if there is a 

 change with size or age is by use of statistical 

 regression. A plot of the data was made on a 

 graph and from this it was ascertained that the 

 regression was linear. Therefore using the 

 formula Y = y"+ b (X-Jc), a linear regression 

 line was calculated (Snedecor, 1946). 



To get a clear picture of any possible 

 change in gill raKer number, fish from one year 

 class and one locality were selected. One of the 

 underlying assumptions of a regression, which 

 is a moving average, is that the variances of the 

 individual samples be homogeneous. This was 

 verified by Bartlett's test of homogeneity of 

 variance . 



The results of several samples and com- 

 binations of samples are shown in table 1 . First 

 a regression on the individual samples was run. 

 In all the rivers or river systems (Rappahannock, 



