Numerosity processing in dot comparison tasks is influenced not only by the number of the dots but also by continuous magnitudes, such as the size of the dots and their convex hull (the largest contour around all dots). Our group has demonstrated that certain continuous magnitudes affect numerical comparisons in opposite ways. While we tend to judge dot arrays with a larger convex hull as more numerous, we tend to judge dot arrays with smaller rather than larger dots as more numerous. It has been shown that during development, children become increasingly precise in comparing such non-symbolic numerosities. According to the classical view, this improvement is the result of the sharpening of internal number representations. However, this view has recently been challenged by Piazza et al. (2018) who proposed that this developmental trend is driven by a so-called filtering process, i.e. children get increasingly better at filtering out irrelevant continuous magnitudes. In the current study we investigated whether convex hull and dot size are affected differently by developmental changes. To do so, we implemented a dot comparison task in which dot size and convex hull were separately manipulated and compared performance and congruency effects of 10-year-old children and adults. Our results suggest that, depending on the type of continuous magnitude, the increase in performance can be explained by both sharpening and filtering.