The recently introduced implicit priming task (Meyer, 1990, 1991) for the study of word production processes has already provided an impressive number of findings which are taken as the main support for the principle of serial encoding in production. However, prior results can as well be explained by an episodic memory retrieval account which does not resource to production processes. In experiment 1, this episodic memory account is tested in a picture naming version of the task, which minimizes episodic memory contributions. The implicit priming effect is replicated and therefore the standard production account is supported. Experiment 2 found a same-sized implicit priming effect in a standard implicit priming task using the same materials as in experiment 1, which was nevertheless statistically non-significant. Between-experiment comparisons showed that the memory components of the standard version of the task introduces noise in the data, and makes the picture naming version more suitable to study implicit priming effects in experimentally naive participants. Key words: language production, phonological encoding, implicit priming, picture naming.