In this study, we examined whether the lexical competition process embraced by most of models of spoken word recognition is sensitive to talker-specific information. We used a lexical decision task and a long lag priming experiment in which primes and targets sharing all except the last phoneme (e.g. /bagaR/ "fight" vs /bagaƷ/ "luggage") were presented in two separate blocks of stimuli. In Experiment 1 the competitor prime block was presented only once to listeners, and no modulation of the competitor priming effect as a function of a talker change between the primes and the targets was observed. However, attenuation in the competitor priming effect in case of a talker change between the primes and the targets was observed in Experiment 2 in which the competitor prime block was presented five times to listeners. We discussed our findings in reference to hybrid models of spoken word recognition in which the repetition of words with the same talker could be a key factor in the formation and access to talker-dependant representations.