ON THE EXISTENCE OF INTERMEDIATE REPRESENTATIONS IN NUMERICAL PROCESSING

Do numbers presented in different formats activate exactly the same se mantic representations? Results of 6 experiments argue in favor of the hypothesis according to which different intermediate representations are activated, depending on the lexico-syntactic structure of the nume ral to be processed. First, equation verification is faster when the c alculation mimics the structure of the roman numeral proposed as a sol ution (e.g., VII = 5 + 2). Second, comparing the magnitude of 2 verbal numerals is more rapid when the 2 items share the same lexico-syntact ic structure (e.g., twelve hundred, fourteen hundred) than when they d o not (e.g., twelve hundred, one thousand four hundred). Third, when c alculating orally, participants tend to use the same verbal structure to express their response as the one used in the problem's addends. Th e implications of these results for the different views of numerical r epresentations are discussed.