Abstract: Models taking genetic gain, relatedness, delay at generation turnover and breeding population size into account have been developed to optimize selection age and generation interval. Relatedness (expressed as group coancestry) and average breeding value for the breeding population are merged into a joint index ("group merit"). The negative impact of group coancestry (like potential inbreeding depression) is expressed in a scale compatible with breeding value. Group merit measures the desirable characteristic of a breeding population. Annual increase of group merit is the criterion for comparing alternatives. Optimum generation interval is when annual group merit increase is highest. Generally the optimum selection age becomes higher when increase in relatedness is considered. We quantify the influence of relatedness penalty, early-mature genetic correlation, breeding population size and delay at generation turnover on optimum selection age. A reasonable large population counteracts the increase of relatedness, and thus favors early selection. Early selection can have a negative impact if a small early selection gain does not compensate for the build-up of relatedness at generation turnover. Conditions for this to occur are quantified. Early selection requires sufficient high juvenile-mature correlation to have a positive effect; this requirement can be reduced by using a large breeding population. The methods developed were applied to a number of situations relevant to forest tree improvement.