Should we judge empirical generalisations as mere regularities, separately from theory? The paper argues that this may bring out the links with theory better and gives illustrative examples. It also notes how in pursuing empirical generalisability more explicitly, we will have to shift how we design and analyse our studies.