Estimates of broad-sense heritability for roasted flavor attribute of peanut (Arachis hypogaea L.) range from 9 to 24% on a single-plot basis. Response to selection is determined by the narrow-sense heritability, calculated from estimates of additive genetic variance which are not available for this trait. One way to assess the additive component of genetic variation is to determine how much of the total phenotypic variation can be predicted from genetic contributions of ancestors of the individuals measured. From 1986 to 1991, samples of 128 peanut cultivars and breeding lines were obtained from peanut research programs representing the three major production areas in the U.S. Samples were roasted to a nearly common color, ground into paste, and assessed for roasted flavor and fruity attribute by a trained sensory panel. CIELAB L* color was also measured for use as a covariate in statistical analysis to adjust for small differences in color. The sum of squares associated with the 128 genotypes accounted for 11% of the total phenotypic variation. Ancestry of the lines was traced back to 47 progenitors for which no further pedigree information was available. Eight progenitors made ancestral contributions that were linearly dependent on the other 39. Ancestral effects accounted for 53% of the genotypic variation, i.e., 6% of the phenotypic variation. Despite shortcomings of this 6% figure as an estimate of narrow-sense heritability for roasted flavor, no other estimates are extant. The residual (nonadditive) variation among genotypes after accounting for ancestral (additive) effects was highly significant. Multiple regression model-building techniques were used to identify 13 ancestors exerting significant effects on roasted flavor. Jenkins Jumbo, F231 (a cross of Dixie Giant with Small White Spanish 3x-2), and Improved Spanish 2B were the only ancestors among the 13 that were common to 40 or more of the 128 lines tested. Jenkins Jumbo was the single most important ancestor, exerting a negative effect on flavor (b = −1.25±0.19). Its progeny would be expected to have roast flavor scores reduced by |b|/2 = 0.62 units and grandprogeny by |b|/4 = 0.31 units. All but four of the 13 ancestors deemed important had deleterious effects on flavor.