We analyze comparative morphemes (e.g., -er, more) as intervals that serve as default differentials in comparatives. We propose that a comparative sentence encodes an interval subtraction: the differential, which is an interval, is the result of subtracting the interval representing the position of the comparative standard on a scale from the interval representing the position of the comparative subject. We show that our analysis has at least two empirical advantages. First, it accounts for the semantics of comparatives with downward- entailing or non-monotone differentials in a very natural way, without making universal quantifiers inside of the than-clause take scope over the matrix clause. Second, it makes it possible to give a unified account for various uses of comparative morphemes (e.g., their use in comparative correlatives). We mainly focus on the first advantage in this presentation.