This paper addresses a recent trend in the study of language variation and universals, namely to attribute cross-linguistic patterns to diachrony, rather than to other causal factors. This is an interesting suggestion, and I try to make the basic concepts clearer, by distinguishing clearly between language-particular regularities, universal tendencies, and mere recurrent patterns, as well as three kinds of causal factors (preferences, constraints, restrictions). I make four claims: (i) Explanations may involve diachrony in different ways; (ii) for causal explanations of universal tendencies, one needs to invoke mutational constraints (change constraints); (iii) in addition to mutational constraints, we need functional-adaptive constraints as well, as is clear from cases of multi-convergence; and (iv) successful functional-adaptive explanations do not depend on understanding the precise pathways of change.