282 



FISHERY BULLETIN OF THE FISH AND WILDLIFE SERVICE 



a river. It would also be worthwhile to study 

 the relation between juveniles and the corre- 

 sponding year class when it enters the fishery 

 as adults. Probably many of the answers which 

 can be obtained from meristic counts will lead 

 to a better understanding of the biology of the 

 various shad populations. 



SUMMARY 



It is a commonly accepted theory that shad 

 from the different rivers on the Atlantic coast 

 return to the same river to spawn when they 

 reach sexual maturity. Tagging experiments have 

 offered considerable evidence to support this 

 theory. No shad tagged on the spawning ground 

 of one river system has ever been recaptured on 

 the spawning ground of another river system. 



If a group of fish return to the same spawning 

 ground year after year with little mixing from 

 other populations, it would be expected that the 

 fish within a river would be more like one another 

 than like the fish from other rivers. Thus, if 

 differences in some characteristics could be found 

 between rivers, and, if these differences were large 

 compared to the differences between years, the 

 conditions necessary to support a "racial" theory 

 would be present. 



Because of the selectivity of the fishing gear 

 used to obtain the samples of shad, it could not 

 be assumed that the samples were random. This 

 selectivity occurred in the size of the fish. The 

 various characters under consideration in this 

 paper were tested for a correlation with length; 

 when no consistent correlations could be found, 

 the samples were considered "representative," 

 even though they were not random. 



Analyses of variance of the various characters 

 provided evidence that there were differences 

 between fish from other rivers and, if differences 

 were present between years, they were of a much 

 smaller magnitude than the differences between 

 fish from other rivers. This contributed addi- 

 tional evidence to support the theory of a separate 

 population of shad in each major river. 



A large commercial shad fishery exists along the 

 coast of New Jersey, New York Bay, and Long 

 Island. The fishermen in these three areas catch 

 shad that are migrating to the Hudson or Con- 

 necticut Rivers. This ocean catch in some years 

 is approximately one-third the size of the river 

 catches and, therefore, should be included in any 

 management plan for the two rivers. To estab- 

 lish a management plan which would include the 

 ocean fisheries would require estimates of the 

 composition of the catches made at these various 

 locations. In the past, this would have had to 

 be done by tagging experiments. 



A discriminant function has been constructed 

 in this study which will classify correctly about 

 81 percent of a mixed population of Hudson and 

 Connecticut River shad. This function was con- 

 structed from data obtained from the Hudson 

 River in 1939 and the Connecticut River in 1945. 

 Data from a sample of Hudson River shad ob- 

 tained in 1940 were substituted into this dis- 

 criminant function. Out of the 105 fish, 16 were 

 classified incorrectly; this is in good agreement 

 with the theoretical 19 percent misclassifications. 

 Most of the individuals that are misclassified 

 fall close to the midpoint between the two popula- 

 tions. It is possible to reduce the number of these 

 mistakes by refusing to make a decision on the 

 individuals that lie close to the dividing line 

 between the two populations. This is equivalent 

 to classifying the individuals into three parts: 

 Hudson River, Connecticut River, and a third 

 part for which no decision can be reached. With- 

 out using this procedure the chance of misclassi- 

 fying an individual is 19 percent. By refusing to 

 classify 50 percent of the sample, it is possible to 

 reduce this error to 3.7 percent- 

 Several methods of estimating the relative 

 abundance or composition of a mixed population 

 are presented. These techniques could be used if 

 one is interested in the population composition of 

 a mixed sample rather than the identification of a 

 particular individual. 



