Use relative frequency histograms instead of ordinary frequency histograms when the two distributions come from samples of different sizes (different total numbers of observations). Because a relative frequency histogram puts the y‑axis on a common 0–1 (or 0–100%) scale, it “normalizes” each bar to show proportions rather than raw counts, which makes the shapes directly comparable even if one data set has, say, 30 observations and the other has 300.