A quantitative genetic modelf or arbitrary numbers of alleles and loci with additive, dominance, and additive by additive epistatle effects was employed. Utilizing descent measures and the genetic components, the selection response from selecting among self‐fertilized progenies in generation g, and measuredin selffertilized relatives in generation g', where the last common ancestor was in generation t, was formulated. Permanent response, g' = ∞, led to considerable simplifications. The effect of the number of generations, t, of selfing before testing on the permanent response relative to the response from selecting among homozygotes,t = ∞, was evaluated numerically for a model with two loci, two alleles each. The result varied within a narrow range depending on gene frequencies and effects with a mean relative response of 0.68, 0.83, 0.92, 0.96, 0.98 for t = 1,2,3,4,5, respectively. Selection response based on self‐progeny evaluation was also formulated for the outbred performance of the offspring of randomly mated g's. On a per generation basis, the numerical evaluations gave about the same overall picture for t = 1 and t = 2 and less response for t > 2. If time is of no importance, the response generally increased with t, but there are parameter combinations of gene frequencies and effects for which the response is slightly negative. These responses were compared to those from evaluating full‐sib progenies from crossing pairs of self progenies, which give exactly the same result for t = 1 and t = 2 on a per generation basis. Self‐progeny evaluations generally lead to more selection response than fullsib progenies except for those parameter combinations for which the response is slightly negative, and then it is very small for the full‐sib progeny evaluation. Permanent response is slightly less than immediate response for inbred performance, and varies more with gene frequencies and effects when t is small. Permanent response for outbred performancec an be muchl ess than immediate response depending on the additive by additive variance.