594 HENRY D. GERHOLD 



traits would in effect be indirect selection for yield. Falconer (I960) 

 and Searle (1965) have outlined conditions under which indirect selection 

 may be more advantageous than direct selection, and Scheinberg (1967) 

 has developed a method for computing the sampling variance of relative 

 efficiency estimates using variance-covariance components from sibship 

 data. 



There are some obvious opportunities for achieving more rapid progress 

 in white pine breeding by using indirect selection. For example, if height 

 growth to age 20 gives a good estimate of yield at maturity, age 80, and 

 if hg were not greatly inferior to hy/rgy 2 , the genetic rate of gain per 

 unit of time would be multiplied several fold. Furthermore, if blister 

 rust is the major determinant of yield in areas where the risk of infec- 

 tion is high, and if h^ measured under uniform rust nursery conditions 

 is much larger than h y measured in a highly variable forest environment, 

 the efficiency of improving yield may be increased by selecting for 

 blister rust resistance instead of, or in addition to, measuring height 

 growth or yield itself. A similar proposition could be stated for 

 weevil resistance. In such cases the assumption that resistance genes 

 found in a nursery will be useful in forest plantings needs to be con- 

 firmed. 



If faster growth rate, blister rust resistance, and weevil resistance 

 are all valid selection objectives, how may selection best be applied to 

 give maximum improvement of economic yield? Lerner (1958, p. 176) and 

 Falconer (1960, p. 324) have given theoretical and experimental informa- 

 tion about selection for more than one trait. Tandem selection, in which 

 one trait is selected for after another, and independent culling levels 3 

 in which individuals are rejected if they fail to meet any of the minimum 

 standards set for each trait, are considered to be less effective than 

 index selection, which is based on a total score combining genetic and 

 economic values of all traits. For three traits, the relative selection 

 efficiencies are 1.0 for tandem, 1.7 for independent culling levels, and 

 2.0 for index selection, according to Lerner (1958), who made the com- 

 parison under several simplifying assumptions. Falconer (1960) points 

 out that the advantages of an index are less if family selection is 

 practiced, and suggests that little efficiency will be lost if each 

 phenotypic value in this situation is weighted only by its relative 

 economic importance. 



Multiple trait selection also has several potential disadvantages or 

 pitfalls of which breeders should be aware. Even though total economic 

 gain is expected to be superior, the gain for each component trait will 

 not be as great because its selection intensity will be lower. To 

 illustrate, if the best 1% of the individuals in a larger population are 

 selected for each of two uncorrelated traits, the effective selection 

 differential for each will be 1.75 a instead of 2.67 a. The effective- 

 ness also can be lowered by negative genetic correlations, and if these 

 are pleiotropically determined it may not even be feasible to reach an 

 improvement goal involving two traits. There is a tendency for white 

 pine weevils to attack the taller trees, and this might (or might not) 

 cause r W g to be negative. Genetic correlations may change in different 

 environments or in successive generations as a result of selection 

 pressure. This complicates the design of progeny tests. The assigning 

 of economic weights can also be troublesome, especially in view of the 

 long-time span over which the values of timber crops must be predicted 

 and because changing disease and insect control measures must be 

 anticipated. These few examples illustrate the complexity of constructing 



