Dehaene et al. (1993, Experiment 6) presented evidence that the mental number line is left-to-right oriented with respect to representational associations and not with respect to left and right hands. Here we tried to replicate the study of Dehaene et al. (1993) in a larger sample (n = 32) using four different stimulus notations (Arabic numbers, number words, auditory number words, and dice patterns). As in the study by Dehaene et al. (1993), the spatial numerical association of response codes (SNARC) effect was examined with an incongruent hand assignment to left/right response keys (crossed hands). In contrast to Dehaene et al. (1993), we did not observe a SNARC effect in any condition. Power analyses revealed that n = 32 should have been large enough to detect SNARC effects of usual size. Furthermore, time-course analyses revealed no SNARC slope in faster and slower responses, so that the null effect could not be due to relatively slow responses with crossed hands. Joint analyses with previous data (Nuerk et al., 2005b) revealed significantly steeper SNARC slopes with congruent hand assignment, and no interaction between hand assignment and notation. Altogether, these findings suggest that the results of Dehaene et al. (1993) only hold under specific conditions. Differences between studies are discussed. We suggest that spatial context has an influence on the SNARC effect and that hand-based associations (and not only representational associations) are relevant for the SNARC effect.