Most psycholinguistic models of lexical access, although making different proposals regarding nature of compet i-tors, postulate that word identification is the result of strong competitive mechanisms between simultan eously activated lexical candidates (see for example NAM, Luce and Pisoni, 1998; the revised Cohort model, Marslen -Wilson et al., 1996, TRACE, McClelland and Elman, 1986, or Shortlist, Norris, 1994). In that context, situation of speech-in-speech comprehension could be of great interest. In our studies, nature and language of back-ground noises were manipulated to identify information levels in which linguistic interferences can occur. Native speakers of French had to identify French target words inserted in babbles or in fluctuating noises generated in French, Breton, Irish, Italian, with signal-to-noise ratio of 0 or 5dB. Globally, performances are always better when background is noise rather than speech, revealing that linguistic information from babbles competes with target signal comprehension. The results also showed that at-5dB it is more difficult to understand French target words with babbles in French than in languages unknown to listeners, and that some languages interfere more with French than some others. These results will be discussed with a particular enhancement on the differences ob-served between intelligibility and lexical decision tasks.