We measure the mass functions for generically red and blue galaxies, using a z < 0.12 sample of log M* > 8.7 field galaxies from the Galaxy And Mass Assembly (GAMA) survey. Our motivation is that, as we show, the dominant uncertainty in existing measurements stems from how ‘red’ and ‘blue’ galaxies have been selected/defined. Accordingly, we model our data as two naturally overlapping populations, each with their own mass function and colour–mass relation, which enables us characterize the two populations without having to specify a priori which galaxies are ‘red’ and ‘blue’. Our results then provide the means to derive objective operational definitions for the terms ‘red’ and ‘blue’, which are based on the phenomenology of the colour–mass diagrams. Informed by this descriptive modelling, we show that (1) after accounting for dust, the stellar colours of ‘blue’ galaxies do not depend strongly on mass; (2) the tight, flat ‘dead sequence’ does not extend much below log M* ∼ 10.5; instead, (3) the stellar colours of ‘red’ galaxies vary rather strongly with mass, such that lower mass ‘red’ galaxies have bluer stellar populations; (4) below log M* ∼ 9.3, the ‘red’ population dissolves into obscurity, and it becomes problematic to talk about two distinct populations; as a consequence, (5) it is hard to meaningfully constrain the shape, including the existence of an upturn, of the ‘red’ galaxy mass function below log M* ∼ 9.3. Points 1–4 provide meaningful targets for models of galaxy formation and evolution to aim for.