The assessment and specificity of environmental and genotype-environmental components of variability

Abstract

The statistical and biometrical genetical advantages and disadvantages of using dependent and independent assessments of the environmental values in the joint regression analysis of genotype-environmental interactions have been investigated. The material consisted of 82 inbred lines produced by nine successive generations of selfing from a random sample of F2 individuals from the cross between varieties 1 and 5 of Wicotiana ruslica, these two varieties and their F2. All were grown in the eight environments produced by two planting densities in each of four sowing dates. Ten of these inbred lines, chosen to be a stratified sample of the 82 lines for the character final height, were grown in the 16 environments produced by all combinations of presence and absence of N, P, K and Ca fertilisers. Eight individually randomised plants of each family were grown in each environmental treatment and flowering time, linear growth rate, leaf length and final height recorded. In addition to the usual dependent assessment of the environments, the main experiment provided three sources of independent assessment, namely, replicate samples of individuals of each inbred line, replicate samples of inbred lines and the parental varieties 1 and 5. As far as the significance of the heterogeneity of regression and remainder items in the joint regression analysis and the ranking of the inbred lines on the basis of their linear regression coefficients are concerned, it made little difference whether the dependent or any one of the three independent measures of the environmental values was used. The independent measures, however, raised statistical problems where these were based on few observations and problems of interpretation at the biometrical genetical level. Epistatic components of mean performance over all environments for flowering time, of linear sensitivity to environmental differences for growth rate and of non-linear sensitivity for final height have been detected. The correlations over lines between mean performance and linear sensitivity were generally low and non-significant and the number of effective factors of the largely independent genetical systems controlling these two aspects of the phenotype have been estimated. The stratified sample of 10 inbred lines, on the other hand, showed a positive correlation between mean performance and linear sensitivity in the very poor environments produced by the N, P and K treatments in the absence of calcium. This is expected since the performances of all the lines must ultimately converge as they approach the worse environments and during convergence these two aspects of the phenotype must be correlated. The relative linear sensitivities of these lines to the different kinds of environmental variables revealed a high degree of specificity. © 1973, The Genetical Society of Great Britain. All rights reserved.