Even at high levels of proficiency, adult second language (L2) learners experience difficulty using grammatical gender online. Recent research has further suggested that these difficulties stem from processing problems rather than basic learnability problems: even high proficiency speakers routinely produce errors and show difficulty using gender-marked determiners to anticipate upcoming nouns, despite knowing nouns’ genders (e.g., Grüter et al., 2012). The cognitive locus of this difficulty, however, remains unknown. To address this, we used event-related potentials (ERPs) to investigate potential processing-based loci for these difficulties, focusing on the relative time course of retrieval of gender and phonology information during lexical retrieval. Based on L1 production research showing a timing advantage for retrieving gender versus phonological information (Van Turennout et al., 1998), we hypothesized that L2 processing difficulties may stem partly from delayed access of gender, relative to a word’s phonological code. Native German speakers (N=20) and native speakers of English who were proficient in L2 German (N=20) performed a covert picture naming task based on that used by Van Turennout et al. Participants viewed images of high frequency German nouns, and made decisions based on the grammatical gender (masculine or neuter) and initial phone (/b/ or /k/) of each noun. In a fully counterbalanced, within-subjects design, participants used one source of information (gender or phonology) to determine whether or not to press a button on a response pad, and the other source of information to determine which button to press in the case of a go-trial (left or right). Previous research using this paradigm has found that the relative timing of lexical information retrieval is sensitive to retrieval difficulty, showing delays as difficulty increases (e.g., Abdel Rahman & Sommer, 2003). We therefore adopted this paradigm to obtain a fine-grained estimate of the time course of gender information access in L2 populations. ERP analysis focused on the N200 component, which indexes response inhibition and has been used to provide an upper-bound estimate of the time at which lexical information is accessed in go/no-go tasks (cf. Schmitt et al., 2000). Results showed larger N200 effects for no-go versus go trials, but no significant differences in N200 peak latency for go/no-go decisions based on gender compared to phonology in native or non-native German speakers. Importantly, data also showed no delay in mean latency for the non-native German speakers (M=423.8 ms, sd=0.9) relative to native speakers (M=431.6, sd=10.7). First, our native speaker results do not replicate Van Turennout et al.’s lateralized readiness potential results: there was no evidence of earlier access to grammatical gender information compared to phonological information, suggesting that the gender-before-phonology information flow may not generalize to all situations. Second, the lack of any delay in access to grammatical gender information for the L2 population suggests that the difficulties experienced by L2 learners with grammatical gender cannot be directly tied to delays in the timing of gender access during online processing. Our results thus provide important constraints for theories of grammatical gender use and lexical access in L2 production.