Two cave fish populations: one species, two, or three?

Verdict under the biological species concept (BSC): the BSC defines a species as a group of actually or potentially interbreeding populations that is reproductively isolated from other such groups. The resurgence pool is a direct test of that criterion, and it fails: 3 of 61 adults (about 4.9%) are hybrids and those hybrids are fertile.

• Fertile hybrids mean the two forms are not reproductively isolated in the sense the BSC requires, so strictly the Cave A fish and the adjacent surface fish are one species. The cave form is a morphologically extreme population - an ecomorph - not a separate species.
• What the data do not license: concluding that gene flow is free. A 4.9% hybrid frequency is low and could reflect rare interbreeding plus selection against hybrids (hybrid inviability in later generations, mate preference, habitat segregation). Hybrid presence is not the same as hybrid fitness: you would need F1 viability, backcross survival, or mating-choice data before claiming either free gene flow or effective isolation.
• Note also that the BSC is untestable between Cave A and Cave B, which are allopatric and never meet. Applying the BSC there is a category error; you must switch to a phylogenetic or diagnosability concept for that comparison.

Why the shared morphology is weak evidence

• Every shared cave character in the list is a loss - eyes gone, melanin gone, scales reduced. A single state such as 'eyes absent' can arise from different mutations in different lineages, so it carries almost no information about shared ancestry. It is the classic case where a character matrix built on regressions will mislead.
• Loss of eyes and pigment is under strong selection in permanent darkness (maintenance costs with no benefit, no need for photoprotection), so independent cave lineages worldwide converge on the same appearance. Shared absence is a homoplasy here, not a homology.

What the sequence data imply

• Each cave population is far closer to its own surface relative (0.4% and 6.8%) than the two cave populations are to each other (7.1%). If the cave form had evolved once and then split, Cave A and Cave B would be each other's closest relatives.
• The parsimonious reading is two independent colonisations of caves, each followed by parallel evolution of the same traits - convergence, not common descent.
• The 0.4% figure also tells you the Cave A split is recent, so the cave morphology evolved quickly relative to the mitochondrial substitution rate.

Several defensible options; each must be tied to a decision rule, not just listed.

Option A - nuclear markers (e.g. multiple microsatellite loci or ddRAD loci). Mitochondrial DNA is a single maternally inherited locus and can mislead through mitochondrial capture or incomplete lineage sorting. If nuclear markers show Cave A fish clustering with surface fish and low FST, the BSC verdict (one species) is supported; if they show strong clustering despite the hybrids, you would argue for a distinct species under a genotypic-cluster concept and publish two names.

Option B - reciprocal mate-choice and hybrid fitness experiments. If females assort strongly and F2/backcross survival is low, postzygotic and prezygotic isolation exist despite the 4.9% hybrids, which supports recognising the cave form as a separate species under a lineage concept. If mating is random and hybrids are as viable as parentals, the population stays one species.

Option C - wider geographic sampling of surface populations. The 0.4% divergence is measured against only the 'nearest' surface population. If sampling reveals a continuous cline rather than discrete clusters, the cave fish cannot be diagnosed as separate taxa.

The point is that the classification changes only if the new evidence bears on isolation or diagnosability; extra morphological characters of the same loss-type would not change the verdict.