Four F3 populations of chickpea (Cicer arietinum L.) were simultaneously evaluated for yield in an F3 yield trial and in single plant progeny rows. Ten high yielding, 10 low yielding and 10 randomly sampled lines, along with 10 lines visually selected for yield from the progeny rows, were retained for further evaluation. The lines from each of the four selection groups in each population were bulked and evaluated in a replicated yield trial at three locations and four environments. The bulk of visually selected lines was not superior in yield to the bulk of randomly sampled lines at all locations. The present results indicate that an early generation yield testing selection procedure is more efficient than visual selection for yield improvements in chickpea.