Does language really matter when doing arithmetic? Reply to Campbell (1998)

Campbell (1998) has questioned the conclusion of Noël et al. (1997) and has argued that alternative analyses of their data provide strong evidence that arithmetic performance is subject to reading-based interference and provide some support for the language-specificity of number-fact memory. We consider that Campbell reached conclusions different from those we had obtained because (1) he performed his analyses on a different data set (i.e. including also the table-unrelated errors), (2) he has given a double weight to the naming errors and (3) he has multiplied the analyses without correcting the corresponding P values. We thus consider that there exist interactions between language and performance in simple multiplication tasks, but that the current data can easily be explained without postulating that such interactions operate at the level of the retrieval stage. In other words, we consider that there are not definitive arguments, as yet, in favour of the hypothesis of modality-specific arithmetical-fact networks.