In this study on regional accent perception we conducted two experiments to examine a hypothesis based on previous work on talker variability (Pisoni et al., 1998), and on compressed speech (Dupoux et al., 1997), which predicts that accent-related variations would elicit a perceptual cost in the course of word recognition. In both experiments participants were asked for a lexical decision on the last item of sentences uttered in a familiar or an unfamiliar regional accent. The first experiment examined whether the effect of regional accent was sensitive to sentence length. Two accents were used, Besan?on (familiar) and Toulouse (unfamiliar), with ten disyllabic test items and six legal nonwords. These words were appended onto a set of 192 sentences which were split equally between the two accents and produced by two female speakers per accent. Three blocks of different sentence length were used. After testing thirty-five monolingual French-speaking participants from the Franche- Comté region we found that familiar accents were processed faster than unfamiliar accents (559 ms versus 574 ms). Moreover a significant interaction between accent and sentence length revealed that the effect of the accent increased in relation with sentence length. These findings suggest that unfamiliar regional accents elicit a cost in word recognition, possibly reflecting a normalisation process, emerging mainly after long utterances. A second experiment was conducted to determine whether this processing delay was due to the normalisation process per se, or to the disruption of this mechanism caused by the random presentation of accents. The procedure of the second experiment was identical to the first, departing only in the presentation order of stimuli. Instead of randomly alternating the accents from one sentence to the other participants were presented with blocks of sentences uttered in each of the accents. Twentyfour monolingual French-speaking subjects from the Franche-Comté region were tested in this experiment which showed no significant effect of accent familiarity (Franche-Comté: 593 ms; Toulouse: 595 ms). These results indicate that when participants are presented with the same unfamiliar regional accent for a certain period of time the processing cost observed in the previous experiment habituates. The implications of these findings and their impact on the understanding of accent variability processing are discussed, particularly in light of previous research on compressed speech.