In this research we investigate the relevance of a distinction between
an exact memory and an inferential memory. In the two studies, on 140
fourth- and fifth-graders and on 456 adult students respectively, the
subjects are first informed of a numerical material: an arithmetic ve
rbal problem or a mental calculation (e.g. 86 + 47 + 93 + 52). Then th
ey are tested both on exact data and on data that can be inferred from
the information previously given. The research does not clearly suppo
rt an independence between the two forms of memory. However, it shows
evidence for a lower sensitivity to priming of the exact memory. Furth
ermore, the second study demonstrates that the mental arithmetic split
effect (Ashcraft, 1992) can be differentiated following the arithmeti
cal operations: additive operations are more sensitive to it than mult
iplicative operations.