Recently, a growing consensus has emerged regarding the instructional flexibility of the mental format of non-symbolic magnitudes (e.g., quantities of objects’ collections, just-learned height, animals’ size). Indeed, while it is well-established that symbolic magnitudes are stably organized according to a left-to-right orientation, this organization is less clear regarding non-symbolic magnitudes. Flexibility has been invoked for accounting for patterns of response times generally observed in tasks involving comparative judgements and deviating from the standard SNARC (i.e., faster left-sided responses to small magnitudes vs. faster right-sided responses to large magnitudes). We reinterpret flexibility in terms of the Semantic Congruity Effect (SCE): a general tendency for extreme, rather than intermediate, magnitudes to be detected more readily amongst a pair of elements belonging to the same semantic category (i.e., small-small/large-large) when the comparison task requires judging greater/lesser. A major prediction of SCE is a reversed SNARC for pairs of magnitudes displayed in incongruent spatial orientation with the left-to-right mental format. Right-sided responses should be faster for small magnitudes, and vice-versa for large magnitudes, when paired with an intermediate magnitude. We re-analysed the data of some previous studies and show that, in line with our results (Fantoni et al., ECVP2017), a reversed SNARC did reliably occur – consistent with SCE. This undermines preceding interpretation of results in the context of comparative judgements based on SNARC flexibility. In this context SCE provides a more general model than SNARC, being SNARC incidental to the spatial properties of a pair.