uncompensated hyperbaric dissolved gas pressure 

 with a Weiss saturometer. 



Results 

 Estimation of Heritability 



Initial mortalities occurred 8 h after the test fish 

 were placed in 130''? air-supersaturated water. All 

 fish were dead after 132 h. The grand means of 

 time to death (hours) in the three tanks were 22.62, 

 24.66, and 25.04. Two fish died in the control tank. 



Because of counting errors at time of marking, 

 individuals per tank- family ranged from 5 to 54. 

 Few tank-families varied greatly from the ex- 

 pected number per family (28.4) as was indicated 

 by the harmonic mean (26.8). Because of the 

 une(iual numbers of individuals per tank-family 

 and the large number of observations, an un- 

 weighted means analysis of variance was used 

 (Kempthorne 1957). First, the unweighted means 

 of each family in each tank were subjected to an 

 analysis of variance (Table 1). Second, all obser- 

 vations were used in a one-way analysis of 

 variance to compute the within tank-families 

 (error) sum of squares. Because the distribution of 

 time to death for fish in each tank followed a pois- 

 son distribution, a square root transformation of 

 time to death was applied before the analysis of 

 variance was carried out. The square root trans- 

 formation is the most appropriate for poisson data 

 (Bartlett 1936). 



Variance components were estimated as: 



'^.= {MS^^-MS,jpr)/f = 0.041 



o/,,^ = {MSj^f^r-i^/ri,)MS^) = 0.024 



<r,2 = MS 



w 



= 2.800. 



r2- 

 M 



= {MSM-MSt)/tf^ 0.027 



The additive genetic variance, \\ , was es- 

 timated as 4f''^f, and the total phenotypic variance, 

 Vp, as a'~j+ aj, + v^^^j, -i- «r|,. Heritability, //", or 

 V^/Vp, was 4(0.027)72.892 = 0.037. The standard 

 error was 0.022 (Dickerson 1959). 



Survival for each tank-family after 36 h of ex- 

 posure to the test conditions ranged from 53.8 to 

 (V^. The angular transformation was applied to 

 these data before the analysis of variance 

 described above was performed. The theoretical 

 binomial variance, 821. was then used as an es- 

 timate of error variance. Heritability of resistance 

 to gas bubl)le disease estimated in this analysis 

 was: V^/Vp = 30.68/840.87 = 0.036. 



St(Kk Comparisons 



Resistance togas l)ubble disease by offsi)ring of 

 fall chinook salmon from Little Goose Dam and 

 Trask River Salmon Hatchery exposed to \2T/c 

 supersaturation in Experiment I differed 

 markedly. Time to 509f mortality, a\eraged from 

 two replicates, was 73.5 h for Trask River fish and 

 154 h for fish from Little Goose Dam; the 

 difference of 80.5 (SE = 3.39) was highly sig- 

 nificant. 



The difference in time to 507f mortality in 

 Experiment II (Tal)le 2) between Columbia River 

 and Trask River stocks (22.25 h; SE = 6.37) was 



Table 1. -Analysis of variance of the square root of time to death for juvenile fail chinook salmon 



exposed to air-supersaturated water. 



w 



where: m = number of males 



1 = number of females per male 

 t = number of tanks (replicates) 

 N . . . = total number of individuals 



n„ = harmonic mean number of Individuals per tank family = 26.82 



'Error mean square obtained in separate analysis (see Kempthorne 1957:459). 



936 



