The phenomenological experience of lexical retrieval often involves repeated, active attempts to retrieve phonologically and/or semantically related information. However, the influence of these multiple retrieval attempts on subsequent lexical retrieval is presently unknown. We investigated the influence of passively viewing or actively retrieving different types of information at the critical moment preceding lexical retrieval through a novel priming paradigm. Participants attempted to retrieve target words (e.g., abdicate) from low-frequency descriptions (e.g., to formally renounce a throne). Target retrieval was preceded by passive viewing (Experiment 1), or active retrieval of the prime word (Experiments 2-6). Primes were either "both" semantically and phonologically related (e.g., abandon), only phonologically related (e.g., abdomen), only semantically related (e.g., resign), or unrelated (e.g., obvious) to the target word. When primes were passively viewed, phonological facilitation in target retrieval accuracy was observed. In contrast, when participants actively attempted to retrieve primes from their definitions, no phonological facilitation was observed. Successful retrieval of semantic and both primes facilitated subsequent target retrieval, whereas, failure to retrieve semantic and both primes inhibited subsequent target retrieval. These facilitatory and inhibitory influences of prime retrieval for semantic and both primes were independent of feedback on retrieval performance (Experiment 4) and participants' overall knowledge of the primes and targets (Experiment 5), and also did not extend to retrieval from episodic memory (Experiment 6). The findings are consistent with ongoing retrospective processes during target retrieval, which reengage prime retrieval success or failure and consequently produce benefits and costs during repeated retrieval from semantic memory.