When is a species a species?
The answer is most unsatisfactory: "when a competent taxonomist says so".
Two problems arose. Some biologists looked for a deeper meaning to relationships than "they have always been like that." The other was that sometimes one could not create separate piles of specimens - the genus Rubus (Brambles) proved to be impossible.
Darwin solved both these problems with his theory of evolution. Species are units of evolution and the intermediates indicate either incipient species (subspecies) or coalescing species (hybrids). However, solving the problems of "why a species is a species" did not help much with "when is a species a species". If two taxa (taxon: any defined group) do not co-occur and therefore cannot interbreed in nature, should they be one or two species? And how different must they be before they can be considered as two separate species?
These are intractable problems, not helped by plants which are obviously different being able to hybridize (e.g. Di thymelaeoides and Ls oleifolium) or when certain species readily form hybrids (e.g. Pr longifolia). Since Linnaeus we have gone through several different philosophies of species concept: nominalist species (man-made, arbitrary constructions), biological species (interbreeding units), evolutionary species (units carrying genes through time), etc. In effect then, the ultimate decision is still essentialist, but with the history of the species included. Ultimately, a species is still what a competent taxonomist says it is. The only difference is that now taxonomists have more tools (methodological and philosophical) for defining their species. And that is why we will always have the John Rourkes and the John Beards able to convincingly argue a completely different set of species from the same situation.
Some argue that the population is the only true unit of evolution. All categories above it are "Nominalist." A population is a group of potentially interbreeding individuals. Demes are groups within the population which regularly breed with one another. A subspecies comprises populations which are morphologically distinct, but which readily interbreed with other subspecies where they meet. A species is a group of populations which can interbreed with one another, but not with populations of other species. Precise definitions depend on the philosophy adopted and the exact unit considered (co-adapted genes, clades, populations, etc.). But no matter how precise the definition, in reality there are many exceptions and no set of rules are adequate to cover all cases. This is essentially true of plants where development is continuous (as opposed to animals where a malformed kidney cannot simply be jettisoned and regrown from somewhere else).
In practice the following conventions are used below the species level:
subspecies: morphologically-distinct populations occurring in discrete areas. Where subspecies are widely separated or have been separated for a long time they may validly be considered species;
varieties: morphologically distinct populations which co-occur geographically with other varieties;
forms: morphologically distinct individuals which commonly occur in nature (albinos, etc). This is the lowest unit described by taxonomists.
race: a subspecies, variety or form for which the rank is uncertain. (Note the careful use of a neutral term for the variation within our own species, and the subsequent connotations the term has acquired].
ecotype: a population within a species with genetic adaptations to a specific habitat (e.g. able to tolerate heavy metals on a specific soil).
cultivars: a cultivated variety - a variety produced artificially and not normally found in nature.
In the Proteaceae Rourke and Levyns rarely describe subspecies and almost never consider varieties and forms. Beard regularly uses all the categories. Williams conservatively uses subspecies and varieties, additionally using the latter both as defined above and for situations where a single population is geographically isolated (which Rourke would probably ignore and Beard would have as a separate species). These differences probably stem from different backgrounds. In the Cape with hundreds of species the naming of each variety (Vogts documented 81 for Protea cynaroides!) would be a nightmare, which a Cape taxonomist would never attempt. With only 33 species in the tropics, the situation is far more tractable and recognizing varieties and forms becomes an option. However, the current philosophy seems to be to ignore taxa below species: this possibly has its origins with the stigma attached to human races. Numerous cultivars have been formally described for the cut flower trade.