Undoubtedly, tight links between space and number processing exist. Usually, findings of Spatial-Numerical Associations (SNA) are interpreted causally, i.e., that spatial capabilities aid or are even a fundamental cornerstone of mathematical skill. In this book chapter, we question this seemingly ubiquitous assumption.To start with, there is no robust and prevalent correlation between SNA in general and math abilities. After presenting an extended taxonomy for different SNA subtypes, we show that only some SNA subtypes correlate with math abilities, whereas others do not. We argue that these correlations are not conclusive for several reasons. (i) Their correlations vary (i.e., stronger SNA sometimes is related to better math ability, and sometimes to poorer math ability). (ii) The correlations might not show a genuine relation between space and number; rather mediator variables might explain the correlations. For instance, SNA tasks often involve an interference component tapping cognitive control functions (as in multi-digit number processing) or some relatively advanced reasoning skills or strategies. (iii) Finally, the direction of causality (if it exists) is far from resolved. While conventional theories suggest that spatial-numerical abilities underlie arithmetic skill, we argue that vice versa arithmetic abilities instead underlie performance in some spatial-numerical tasks used to assess spatial-numerical representations.On the other hand, benefits conferred by SNA trainings on math abilities seem to reinforce the claim that SNA underlies math abilities. We contend that tasks used in such trainings may tap several cognitive operations required in arithmetic, but not built-up fixed SNAs themselves. Therefore, we argue that using space is a powerful tool, especially for instructing and learning multi-digit numbers; however, this does not necessarily imply an internalized fixed mental number line.