The current study investigated the robustness of priming from a masked speech priming method introduced by Kouider and Dupoux [(2005). Psychol. Sci. 16, 617-625]. In this procedure, a compressed spoken prime is embedded in auditory masking stimuli and presented immediately prior to an uncompressed auditory target. The degree to which spoken stimuli could be compressed without significant data loss was first determined. Using this compression level, repetition and form priming were measured for the target words with a high versus low number of phonological neighbors. The results indicated that robust masked speech priming occurred only for word targets that had few neighbors.