This study utilizes a word association (WA) paradigm to infer similarities and differences between processes used to access the mental lexicons of native speakers (NS) and Japanese nonnative speakers of English (NNS). Three hypotheses were examined: a) grammatical word stimuli will elicit proportionately fewer paradigmatic responses than will content words; b) The proportion of phonologically-related responses will increase when stimuli are presented aurally rather than in written format;and c) NNS responses to infrequent words will not differ from responses to common words if a loan word equivalent exists in their first language (L1). Generally speaking, results concurred with established findings. Where results failed to validate the hypotheses, cognitive models are outlined to account for the data. In particular, a process model involving access to explicit knowledge of grammar rules is presented to account for the fact that NNS were less likely to respond to grammatical word stimuli with syntagmatic responses than were NS (χ2 = 15.22, p < .001, df = 1). Also, during aural presentation, only NNS responses, not NS responses, displayed more phonological similarities to their stimuli, suggesting the NNS rely on phonological cues in the absence of semantic knowledge. Similarly, NNS produced fewer semantic associates to low-frequency nouns with loan word equivalents than they did to commonly occurring nouns (χ2 = 3.89, p < .05, df = 1). In fact, NNS produced marginally more semantic responses to low-frequency nouns without loan word equivalents at all. A model postulating competition between cognitive processes that precipitate semantic responses and those instigated by the salience of phonological similarities between the stimuli and their loan word equivalents is proposed.