Secondly, since estimates of additive genetic variance from single planting sites are 

 confounded by interactions of genotype and environment (Namkoong and others 1966) , present 

 estimates (table 6) may be subject to bias. But, genetic gains for ponderosa pine in southern 

 Idaho will accrue within seed zones, and, therefore, appropriate estimates of heritabilities 

 should be inflated by the effects of the genotype-environment interaction (Namkoong 1979) . 



Thirdly, estimates of the phenotypic variance apparently do not contain components of 

 variance due to genotype-environment interaction. It is likely, however, that some of the 

 effects of a genotype-environment interaction expected within seed zones are contained within 

 the interaction of blocks and families. Whereas the mean height of trees varied 63 cm among 

 planting sites, mean differences associated with blocks ranged from 67 cm to 105 cm at the 

 various planting sites. Significant mean differences among blocks (table 5) arise partially 

 because blocks were arranged according to aspect at each site. Thereby, blocks reflect a 

 variety of sites typical of the seed zone represented by each planting site. Therefore, the 

 interaction among blocks and families (table 5) likely reflects genotype - environment inter- 

 actions within seed zones as well as sampling errors. 



Finally, sampling errors are magnified by the experimental design. Only four trees were 

 planted in each block at each test site. As reflected by harmonic means {ki in table 1), an 

 average of two or fewer trees remain. Thus, family plots poorly reflect family means in each 

 block; experimental errors (block interactions) are magnified; and phenotypic variances are 

 greatly exaggerated in relation to the additive genetic variance. 



Thus, there is little doubt that the genetic components of variance (table 6) are only 

 approximate. But, particularly if each planting site is considered as a representative of a 

 single seed zone, errors of estimate can be absorbed without invalidation of statistical 

 analyses . 



Results of analyses of variance (table 5) and calculations of genetic components of 

 variance (table 6) are similar to those from age 11 (Rehfeldt 1980). Most of this corre- 

 spondence can be traced to the high correlation of individual tree heights (r = 0.86, 0.88, 

 0.91, and 0.83 at Jack's Creek, Boulder Creek, Idaho City, and Holcomb, respectively) and 

 family means (r = 0.81, 0.88, 0.89 and 0.81, respectively) between data at age 11 and age 16. 

 In fact, analyses of variance of the deviation from regression of 16-year on 11-year height 

 detected no differences among blocks, families, or their interaction; all of the variance was 

 attributable to experimental errors or to variation within family plots. Nevertheless, as 

 observed for ponderosa pine in California (Namkoong and Conkle 1976) , the proportion of the 

 total variance attributable to families within populations has decreased between ages 8, 11, 

 and 16. This reduction occurred because error variances (within plots and experimental 

 error) increased. In fact, while error components of variance doubled between ages 11 and 16, 

 family variance components increased in absolute size by only 50 percent. Consequently, 

 estimates of additive genetic variance are less; phenotypic variances are greater; herita- 

 bilities are less; and expected gains are smaller at age 16 than at age 11. 



Genetic gains in 16-year height that are expected after interpollination of trees chosen 

 under various intensities of family and individual tree selection are detailed in figure 1. 

 These gains were calculated for family selection by: 



AG = ih2 



and for selection of individuals within selected families by: 

 AG = 0.75ih? 



7 



