The architecture of the language processing system for speakers of more than one language remains an intriguing topic of research. A common finding is that speakers of multiple languages are slower at responding to language stimuli in their non-native language (L2) than monolingual speakers. This may simply reflect participants' unfamiliarity with words in the L2, however it may also be the reflection of interference from competing lexical alternatives both across and within the participants' multiple languages. In the current studies (one behavioral, one electrophysiological) we investigate how interference from phonologically similar words within the L2 alone may account for problems in auditory language comprehension in non-native speakers. To this end a cross modal lexical priming (CMLP) paradigm was implemented, which allowed us to look for effects of spoken word primes embedded in sentences on the recognition of target stimuli. Specifically, we investigated whether a word such as "mouse", which has a close phonological neighbor, "house", would show a modulating effect on recognition of a word semantically related to "house" but not to "mouse" (i.e., "roof"). We hypothesized that L2 speakers, less efficient at categorizing phonemes in their L2 would show a difference in the processing of "roof" preceded by "mouse" as compared to "roof" preceded by another unrelated word, such as "lamp", due to a residual co-activation of the phonological neighbor "mouse". Furthermore, L1 speakers, highly proficient at recognizing phonemes in their native tongue, should show no such effect. The results of both studies clearly support our hypothesis, indicating that phonological neighbors in the L2 may greatly interfere with L2 word recognition.