A simple principle predicts a tendency, or vector, toward increasing organismal complexity in the history of life: As the parts of an organism accumulate variations in evolution, they should tend to become more different from each other. In other words, the variance among the parts, or what I call the “internal variance” of the organism, will tend to increase spontaneously. Internal variance is complexity, I argue, albeit complexity in a purely structural sense, divorced from any notion of function. If the principle is correct, this tendency should exist in all lineages, and the resulting trend (if there is one) will be driven, or more precisely, driven by constraint (as opposed to selection). The existence of a trend is uncertain, because the internal-variance principle predicts only that the range of options offered up to selection will be increasingly complex, on average. And it is unclear whether selection will enhance this vector, act neutrally, or oppose it, perhaps negating it. The vector might also be negated if variations producing certain kinds of developmental truncations are especially common in evolution.

Constraint-driven trends—or what I call large-scale trends of the fourth kind—have been in bad odor in evolutionary studies since the Modern Synthesis. Indeed, one such trend, orthogenesis, is famous for having been discredited. In Stephen Jay Gould's last book, The Structure of Evolutionary Thought, he tried to rehabilitate this category (although not orthogenesis), showing how constraint-driven trends could be produced by processes well within the mainstream of contemporary evolutionary theory. The internal-variance principle contributes to Gould's project by adding another candidate trend to this category.