The research investigated whether a bilinguals’ second language (L2) is activated during a task involving only the first language (L1). We tested the hypothesis that the amount of L2 interference can vary across settings, with less interference occurring in testing locations where L2 is rarely used. In Experiment 1, we compared language processing for 50 Arabic–English bilinguals tested in Saudi Arabia and 49 Arabic–English tested in the United States. In the task, participants viewed a picture and judged whether a phoneme presented over headphones was part of the L1 picture name. The results showed no effect of testing location on processing. For both groups of bilinguals, we observed L2 interference in mean error rates, but not in mean response times. We also found evidence for L2 interference in correlational analyses between response times and (a) participants’ weekly L2 usage and (b) frequency of English picture names. A second experiment with 24 Arabic monolinguals supported the conclusion that the results with bilinguals were due to L2 interference. Implications for theories of bilingual memory are discussed.