A species is a group of organisms that interbreed with one another and not with anything else. That is the only definition that designates a boundary the organisms themselves maintain instead of one delineated by an observer. Every other criterion -- similarity of form, of ecology, of genome -- requires somebody to decide how much difference is enough, and different somebodies decide differently. Consequently, counts of species generated by those criteria vary with the counter. Reproductive isolation is not like that. Either gene flow occurs or it does not, and in the absence of such flow the two populations will diverge regardless of what anyone thinks. The definition also explains rather than merely describes: it identifies the reason for the resemblance among members of a species -- they are drawing from a common pool -- and the reason that resemblance terminates at the point it does. A criterion that answers the question of why the pattern exists is worth more than one that documents the pattern accurately, and no other candidate does the first at all. The objection that the criterion cannot be applied everywhere is irrelevant: a definition is not a procedure. It specifies what a species is; whether one can determine that, in a particular case, is a question concerning evidence, not the concept. Nobody relinquishes a definition of a chemical element because some specimens are too small to assay.
[[PASSAGE B]]
The criterion is excellent and its domain is small. It has nothing to say about organisms that do not interbreed. That category encompasses the bacteria and archaea and a good many eukaryotes besides -- by most reckonings the majority of life. It has nothing to say about fossils, since no one can test gene flow between populations that have been rock for forty million years, and most of the history we want to name actually happened in the fossil record. And it becomes imprecise exactly where the interesting cases are: populations that hybridise occasionally, or across part of a range, or in one direction only. Confronted with those, the criterion does not deliver a verdict; the observer decides how much gene flow is too much. That is the discretion the criterion was supposed to eliminate. What we have is a definition that works beautifully for contemporaneous, sexual, well-separated populations and is silent or unhelpful elsewhere, presented as though it were the general account. Nor will it do to answer that the difficulty is evidential, not conceptual. In the hybridising cases the evidence is abundant and the criterion still returns no verdict, because it has no way of specifying how much flow is compatible with two species. That is a gap in the concept, and calling it a gap in the data does not close it. A special case is not made general by being clear.
