days and making the readings without 

 knowledge of the sample sources. Readings 

 were also made on the same material by 

 independent observers. Such controls sub- 

 stantiated the comparisons of different 

 tests, provided a basis for evaluating the 

 limits of experimental error in these tests, 

 and suggested refinements in technique for 

 control of such error. 



EXPERIMENTAL RESULTS AND DISCUSSION 



Several serums from humans^/, cattle , 

 horses, sheep, pigs, goats and fish, all 

 preserved by freezing, were titrated 

 against sockeye cells. Included were a 

 number of cross matches (approximately 700) 

 of cells and serums (1 in k) between sock- 

 eye samples from various localities. All 

 of the latter gave negative results, an 

 observation that is in contrast to those 

 that have been made on the brown bullhead, 

 the white croaker ( Genyonemus lineatus, 

 Ayres) and skipjack (Cushing and Durall 

 1957)- Some brown bullhead serums (o out 

 of 16) were found to react strongly with 

 sockeye cells at 1 in k dilution. However, 

 these serums did not differentiate among 

 sockeye from different areas. There also 

 appeared to be no relation between the 

 positive reactions and the blood group of 

 the brown bullhead involved. 



It is of interest that bovine serum 

 which proved so useful in detecting indi- 

 vidual differences in chickens and in skip- 

 jack reacted very poorly with sockeye cells. 



Pig and horse serums were found to 

 agglutinate sockeye cells more strongly 

 than the other serums examined. As ti- 

 trations of several single pig serums with 

 the cells of different fish showed differ- 

 ences in the strength of reactions with 

 these cells, attention was concentrated 

 on this observation. While the optimum 

 temperature of agglutination for this 

 serum was approximately four degrees, it 

 was practical to work only at room temper- 

 atures which consistently remained at 22° 

 to 2U°C. 



2/ Human typing serums kindly supplied by 

 Hyland Laboratories, Los Angeles, 

 California. 



A standardized procedure was adopted 

 for the comparison of readings of different 

 sets of samples. This procedure consisted 

 of testing the cells of each fish against 

 four dilutions (1:16, 1:32, l-.bk and 1:128) 

 of heat- inactivated (56*C for 3° minutes) 

 serum from a single pig designated W.P. 4. 

 The scores of the readings at the latter 

 three dilutions were added to give a single 

 total score. Table 1 presents an example 

 of the protocol just described. Figure 1 

 presents a summary of scores from many such 

 protocols, which combines the samples into 

 three groups. The first of these consists 

 of samples taken from the rivers of Bristol 

 Bay, the second of samples taken from rivers 

 of the eastern Pacific coast from Kodiak 

 Island to the Fraser River, the third of 

 samples taken in the Columbia River System. 

 An analysis of variance was performed 

 (Table 2), which shews that the three 

 groups differ significantly. This analysis 

 was performed by the method given in 

 Snedecor, 1957, page 283, which takes into 

 account the differences in variances. This 

 is the first time that it has been possible 

 to demonstrate a statistically significant 

 serological difference between populations 

 of individuals within a single species of 

 fish. 



It should be borne in mind that the 

 salmon in each of these areas are homing 

 toward a variety of separate and distinct 

 spawning grounds. In spite of this it 

 seems reasonable that genetic exchanges 

 may be more frequent within the Columbia 

 and Bristol Bay groups than between them, 

 and that this would plausibly account for 

 the differences observed. 



Consideration of all the individual 

 scores of the above groups suggests that 

 these form a trimodal distribution (Figure 

 1). Some statistical justification for 

 the validity of this suggestion was ob- 

 tained by applying a test for bitan- 

 gentiality (Haldane 1952). 



Additional evidence of antigenic 

 heterogeneity was obtained from absorption 

 experiments. Samples of the serums of 

 several pigs were absorbed with strong and 

 weakly reactive cells. In general, two 

 capacities for antibody removal were ob- 

 served. Several absorptions indicated 

 that weaker reacting calls (Bristol Bay) 

 were less effective in removing antibodies 

 than stronger reacting cells (Columbia River). 



