Magnitudes along different dimensions (e.g., space and time) tend to interact with each other in perception, with some magnitude dimensions more susceptible to crossdimensional interference than others. What causes such asymmetries in cross-dimensional interaction is being debated. The current study investigated whether the representational noise of magnitudes modulates the (a)symmetry in space-time interaction. In experiments using different formats of length, we showed that dynamic unfilled lengths result in a higher representational noise than either static unfilled length or static filled length. Correspondingly, we observed that the time-on-space effect was larger for dynamic unfilled lengths than for static unfilled length or static filled length (and it did not differ between the latter two). Further correlational analyses showed that the susceptibility of a target dimension to the influence of a concurrent dimension increased as a function of participants' representational noise in the target dimension (e.g., noisier length representations, a larger effect of stimulus duration on length reproduction). In all, our study showed that the representational noise of space and time modulate the way the two dimensions interact. These findings suggest that cross-dimensional magnitude interactions arise as a result of memory interference, with noisier magnitudes being more prone to being nudged by concurrent magnitudes in other dimensions. Such memory interference can be seen as a result of Bayesian inference with correlated priors between magnitude dimensions.