The cross modal repetition priming paradigm was used to investigate how potential lexically ambiguous no-release variants are processed. In particular we focus on segmental regularities that affect the variant's frequency of occurrence (voicing of the critical segment) and phonological context in which the variant occurs (status of the following word-initial segment). Primes consisted of carrier words ending in a segment likely (voiced; e.g., BAND) or unlikely (voiceless; PLANT) to be produced in no-release form followed by a consonant or vowel onset context word. Each carrier word had an embedded lexical competitor (embedded word) formed by the segments prior to its final consonant (e.g., plan in plant). Productions of these embedded words (true Priming;Phonemes;Word Recognition;Speech Communication;Language Processing;Investigations;Undergraduate Students;Auditory Stimuli; embedded word) were also used as primes. Both true embedded words (Experiments 1a-1c) and carrier words (Experiments 2a-2c) were used as visual targets. The results are discussed in terms of the contribution of probabilistic speech events to theories of spoken word recognition and lexical competition. (Contains 6 tables and 3 notes.)