Spatial-numerical associations (SNAs) are a hot topic in the field of cognitive psychology. An important index to explore SNAs is the spatial-numerical association of response codes (SNARC) effect (i.e., faster responses to small numbers using left effectors, and the inverse for large numbers). Previous studies have verified the universality of the SNARC effect and its flexibility in direction, and a variety of theoretical explanations have been proposed. In addition, the SNARC effect is also flexible in the processing stage at which it occurs, which might be caused by: (1) disparities in the comprehension of additive-factor logic, (2) observation from a single point, (3) different types of Simon effects were adopted as the measure index, and (4) different tasks were adopted. Combining the above reasons, a new two-stage processing (spatial representation of magnitude, spatial representation to response selection) model was proposed. It was proposed that different interference factors acting on the two stages might be the core reason for the flexibility of the SNARC effect. Future research could focus on comparisons of different tasks and the adoption of various interference factors to verify the two-stage processing model and combine cognitive neuroscience technologies to further elucidate the neural mechanism underlying the flexibility of SNAs.