A major issue in speech production research is the question of how speakers retrieve words from the so-called mental lexicon. Current models of lexical retrieval converge on the assumption that category associates of a target word are semantically activated during speech planning. However, the question of whether these competitors are also phonologically activated is less agreed on. Past research has addressed this issue by testing whether lexical retrieval of a picture name (e.g. sheep) affects, or is affected by, the processing of a word that is phonologically related to a semantic category associate to the picture name (e.g. goal, phonologically related to goat). Behavioral studies have failed to obtain such so-called mediated priming effects, but have been questioned on the grounds of possibly insufficient task sensitivity. As such priming effects have reliably been obtained with event-related brain potentials in word recognition, we used this approach for testing these effects in lexical retrieval during speech planning. Our results consistently demonstrate the absence of mediated priming effects, putting strong constraints on the activation flow in the mental lexicon during speech planning.