The effectiveness of early generation testing for yield improvement in peanut (Arachis hypogaea L.) was evaluated in an attempt to increase the efficiency of an on-going recurrent selection program by reducing the time required for evaluation and selection. The objectives of the study were to (1) determine if the number of years required per cycle of selection could be reduced; and (2) estimate the amount of genetic variability remaining in the population for yield and four agronomic traits after four cycles of recurrent selection. Sixty-eight random-mated crosses, among the 40 highest yielding lines from the cycle 3 population, were evaluated for yield and agronomic traits in 1985 (S_0:1), 1986 (S_0:2), and 1987 (S_0:3). The results were compared to determine if early generation performance accurately predicts the performance of cross bulks in later generations. No significant correlations were found for yield of entries in the S_0:3 and S_0:2 or in the S_0:3 and S_0:1; however, there was a significant correlation between yield of entries in the S_0:2 and S_0:1 generations. No correlations were found between yield and any of the agronomic traits measured in the S_0:3; however, significant correlations were found in the S_0:2 between yield and most traits measured. Significant correlations were also found for entries in the S_0:3 and S_0:2 generations for each agronomic trait measured except number of seed per 20 pods. Early generation testings appears to be useful for selection of certain agronomic traits, but of limited value when selecting for yield. A decrease in genetic variation for yield in the population from the S_0:1 to the S_0:3 generation was observed; however, the number of crosses equal to or exceeding the yield of the check cultivar increased in later generations. All agronomic traits measured, except number of seed per 20 pods, showed significant genetic variation remaining in the population in both the S_0:2 and S_0:3 generations. The greater the genetic variation remaining in the population for a given trait, the more likely it is that further improvement of that trait will be realized with additional cycles of selection.