While self-organization has been studied for some time and is now a familiar concept in science, there is as yet no general theory nor even a generally accepted definition. Much of what has been done has been within the context of dissipative structures, so much so that for many people the two ideas have become totally conflated. But the concept of self-organization is much broader; it encompasses systems described by catastrophe theory, for example. Here we draw on the experience of catastrophe theory to suggest features that a theory of self-organization should have, and we illustrate this by the problem of segmentation. We point out that self-organization provides an alternative to natural selection as an explanation of order and organization in biological systems.