Abstract : Imprecisions related to numerical expressions are pervasive in human communication. The way they are propagated in calculations is still an issue. Fuzzy logic is an attempt to account for human imprecise reasoning. In this paper, a comparison between human imprecise calculation and fuzzy arithmetic is experimentally performed. An empirical study has been conducted to collect real intervals resulting from imprecise products and additions from participants. Fuzzy intervals are elicited from these data and fuzzy arithmetic is applied to the collected imprecise operands. Comparisons of the fuzzy intervals show that the fuzzy product and addition do not fit well they way human beings perform these operations on imprecise operands. Moreover, they show that the participants, rather than taking into account the imprecisions in the calculation, realise exact calculation and then approximate the exact result.