In this paper, we present a machine learning based approach for estimating antecedents of anaphorically used personal pronouns in Turkish sentences using a decision tree classification technique coupled with the ensemble learning method. The technique learns from an annotated corpus, which has been compiled mostly from various popular child stories.