In a recent study, Rastle and Coltheart (1997) reported the intriguing finding that pseudowords comprising 5 letters and 5 graphemes (TRULPS) were read more rapidly than matched pseudowords made of 5 letters and 3 graphemes (GAUCH). They proposed to account for this result in the framework of the Dual-Route-Cascaded model, in which the assembly mechanism proceeds sequentially letter-by-letter. The inverse effect of the number of graphemes is explained as follows: A multi-letter grapheme (CH) disrupts performance because the phonemic unit for the first letter, which is activated earlier, inhibits the activation of the phonemic unit for the whole grapheme, which appears later. In a replication of this study with French pseudowords, we manipulated both the number of letters and the number of graphemes. We observed the expected positive effect of number of letters, along with a positive but not significant effect of number of graphemes. A post-hoc analysis showed that the English and French data could be reconciled by considering the influence of grapheme frequency, which was not controlled in RC97. The implications of these data for current formulations ofthe dual-route theory will be discussed.