How does lexical selection function in highly-proficient bilingual speakers? What is the role of the non-response language during the course of lexicalization? Evidence of cross-language interference was obtained by Hermans, Bongaerts, De Bot and Schreuder (1998) using the picture-word interference paradigm: participants took longer to name the picture when it appeared with a word phonologically related to the target's translation than when it was accompanied by a control (phono-translation interference). Here we report two picture-word interference experiments that explored whether the same pattern of interference is present for highly-proficient bilinguals producing speech in L2. The results of the two experiments partially replicate those of Hermans et al. (1998), revealing the reliability of the phono-translation effect. We discuss some possible loci of this effect during the lexicalization process, and its implications for models of bilinguals speech production. (Contains 1 footnote.) [This research was supported by two grants from the Spanish Government and also by the McDonnell grant "Bridging Mind, Brain & Behavior".]