This study explored the effects of episodic and semantic retrieval, and feedback presentation, on learning of new complex material. Participants read a text divided into three parts, between which they engaged in (i) content-related (episodic retrieval) or (ii) general-knowledge testing (semantic retrieval), or (iii) reread the previous part. Participants in the two retrieval conditions were randomly assigned to either receive or not to receive feedback on their interpolated test achievement. Learning was measured through multiple choice questions whose distractors were designed specifically to enable capturing proactive interference.