Identification of superior yielding populations in early generations of common bean ( Phaseolus mlgaris L.) would help eliminate poor populations, permit adequate sampling of promising populations, and save resources. Our objective was to determine the value of yield testing early generation bulks. Three checks and 33 populations were yield tested in a randomized complete‐block design with two replications at three locations in Colombia during three seasons. Each plot consisted of 4 rows; 3, 4, and 5‐m long in the F 2 , F 3 , and F 4 generations, respectively. Data were taken on yield, seed weight, maturity, anthracnose, angular leaf spot, and common bacterial blight. Populations and locations were considered random and generations fixed effects. Differences were observed among populations for all traits. Effects of locations were significant (P < 0.01) for yield, seed weight, and maturity. Positive correlations ( r ) were found for yield between the F 2 and F 3 , F 2 and F 4 , and F 3 and F 4 at Popayán; and all r values except between the F 2 and F 4 were positive and significant (P < 0.05) at Palmira. No associations were found at Quilichao. When data were pooled over any two or all three locations, all but one r value were positive and significant (P < 0.05). Heritability estimates obtained by the regression of the F 3 on F 2 were 0.29 for yield, 0.61 for seed weight, and 0.47 for maturity. Predicted gain from selection (20% selection intensity) for yield in the F 2 was 6.1% and actual gains in the F 3 and F 4 were 5.6 and 5.1%, respectively. Values for seed weight were >15% and for maturity <3%.