You read the name Sinningia cardinalis in a magazine like Gesneriads (or on this web site). That's the name of a species. It sounds very knowledgeable and authoritative. There are some plants which are clearly Sinningia cardinalis, and all other plants are not.
In particular, it makes it appear that the grouping of certain plants into a clump called Sinningia cardinalis and everything else in the universe into not-Sinningia-cardinalis is an operation which does not depend on human perception. Being Sinningia cardinalis or not-Sinningia-cardinalis is on a par with being nitrogen or not-nitrogen. Either something is, or it isn't. This is not a distinction which human beings created -- it is a fact of nature.
Now we do a thought experiment.
Thought experiments are so much easier than real ones, and get much less dirt under the fingernails. You can try this at home.
Start with Sinningia cardinalis and Sinningia eumorpha. We make the primary cross. It is half cardinalis and half eumorpha, so we call it c1e1. Whatever domain this plant is in, it's clearly not in CardinalisLand or EumorphaLand, but something else, which (for the moment) we call HybridLand. We don't care what HybridLand is -- all that matters is that it's not in CardinalisLand.
Now we cross the hybrid (c1e1) with S. cardinalis. The result is three-quarters cardinalis and one-quarter eumorpha, so we call it c3e1. Once again, it's not in CardinalisLand. Roughly one fourth of its genome is from Sinningia eumorpha.
Because S. cardinalis and S. eumorpha are in the same clade, these hybrids are very fertile, so we can do more crosses. We can cross c3e1 with S. cardinalis.
The result (c7e1) is 7/8 Sinningia cardinalis and 1/8 S. eumorpha. Is this in CardinalisLand yet?
No matter, we continue. We cross c7e1 with S. cardinalis to get c15e1, which is 15/16 S. cardinalis and 1/16 S. eumorpha, and then we cross that plant with S. cardinalis to get c31e1, which is 31/32 S. cardinalis and 1/32 S. eumorpha, and so on.
Eventually we reach a plant which is within the natural variation of what we have been calling Sinningia cardinalis.
The question is: where is the boundary? Is it something which nature drew, or human beings?
It is very hard to justify any answer other than "human beings". There is no objective way to tell when our sequence of c1e1, c3e1, c7e1, c15e1, c31e1, c63e1 and so on crossed the natural boundary from not-cardinalis into is-cardinalis. What this means is that Sinningia cardinalis is a human construct, an organization which people have imposed on the world.
It is not something which exists in nature apart from human perception.
As usual, the problem is sex.
If it weren't for sexual reproduction, the species might all stay confined within their own territories (CardinalisLand, EumorphaLand, and so on). No hybrids? Then no fuzzy boundary.
That tells us what to look for if we want a grouping that has an objective existence apart from human thought and ordering. What we want is a group with a genetic moat around it: no reproduction across the moat.
The family Gesneriaceae is such a group. Gesneriads do not cross with non-gesneriads. Therefore we do not have a sequence of individuals starting from a non-gesneriad and grading imperceptibly into a gesneriad. Because the Gesneriaceae are reproductively isolated, they are a grouping that actually exists, and existed before human beings came along.
Note that this does not guarantee there will be no disputes about what plants belong in the Gesneriaceae. That's a human-perception problem, not an intrinsic problem with the grouping itself. We could imagine a tree like this:
_______________ Gesneriads west _______ | | | | |_______________ Gesneriads east ____| | | |________________________ Rehmannia
If this tree represented the real relationships among these groups, it would not matter whether Rehmannia were placed in the gesneriad family. There would be five objective groups in this diagram:
The work of Perret et al. show that as presently constituted, the genera Sinningia, Paliavana, and Vanhouttea are not clades, and therefore cannot be one of those objective groupings we're looking for. However, the sinningia alliance as a whole appears to be a natural unit: there are no hybrids between plants in this alliance and plants outside it.
As for the subdivisions, the Dircaea clade would probably have an objective reality were it not for tetraploidy. If the only hybrids between Dircaea clade members and outsiders (such as S. pusilla) were sterile, then there would be nowhere to go from S. eumorpha x pusilla and therefore no fuzzy boundary. But the existence of fertile tetraploid hybrids muddies the water.
It's no good saying that tetraploid hybrids wouldn't exist without human intervention, and therefore aren't natural. First of all, we don't know that, and second, humans are part of nature.
Artificial modification of genomes to introduce genes from very different organisms is going to make things even more muddled. However, even that didn't start with human intervention: there are "dead" virus genes in the human genome, and some viruses steal pieces of their hosts' genomes, so there are probably bits of insect genome lurking in one of your chromosomes.
That darn sex.