This study investigated the role of the syllable in visual recognition of French words. The syllable congruency procedure was combined with masked priming in the lexical-decision task (Experiments 1 and 3) and the naming task (Experiment 2). Target words were preceded by a nonword prime sharing the first three letters that either corresponded to the syllable (congruent condition), or not (incongruent condition). When primes were displayed for 67 ms, similar results were found in both the lexical decision and the naming tasks. Consonant-vowel targets such as BA.LANCE were recognised more rapidly in the congruent condition than in the incongruent and control conditions, while consonant-vowel-consonant targets such as BAL.CON were recognised more rapidly in the congruent and incongruent conditions than in the control condition. When a 43-ms SOA was used in the lexical-decision task, no significant priming effect was obtained. The results are discussed in an interactive-activation model incorporating syllable units.Keywords: syllable congruency, lexical decision, naming, masked priming, CV versus CVC syllables In recent decades, numerous studies have shown that phonological information is automatically activated during visual word recognition (see Frost, 1998, for a review). Masked priming (Forster & Davis, 1984) is a widely used paradigm to study phonological effects (e.g., Frost, Ahissar, Gotesman, & Tayeb, 2003;Grainger, Diependaele, Spinelli, Ferrand, & Farioli, 2003;Lukatela, Frost, & Turvey, 1998;Pollatsek, Perea, & Carreiras, 2005;Rastle & Brysbaert, 2006;Shen & Forster, 1999). Besides avoiding strategic processes from participants (Forster, 1998), this paradigm has made it possible to demonstrate that phonological effects are not confounded with orthographic activation (e.g., Frost et al., 2003;Lukatela et al., 1998). Moreover, phonological effects were obtained in tasks that did not involve postlexical phonological units, suggesting that these phonological effects arose from prelexical and lexical processes rather than articulatory processes (e.g., Lukatela et al., 1998).To take into account robust phonological effects, models of visual word recognition have to include a phonological coding of visual inputs. This feature requires determining which phonological units are activated during silent reading. In languages with clear syllable boundaries like Spanish, data have shown that syllables are involved in the processing of polysyllabic words (e.g., Alvarez, Carreiras, & Perea, 2004;Carreiras, Alvarez, & de Vega, 1993;Carreiras & Perea, 2002;Perea & Carreiras, 1998). In French also, there is evidence for the activation of syllable units during lexical access (e.g., Carreiras, Ferrand, Grainger, & Perea, 2005;Doignon & Zagar, 2005;Mathey & Zagar, 2002;Mathey, Zagar, Doignon, & Seigneuric, 2006). However, the results are less consistent than in Spanish since some studies failed to obtain syllabic effects during the processing of French words (e.g., Brand, Rey, & Peereman, 2003;Rouibah & Taft, 2001). To account...