The scorpion family Typhlochactidae Mitchell, 1971 is endemic to eastern Mexico and exclusively troglomorphic. Six of the nine species in the family are hypogean (troglobitic), morphologically specialized for life in the cave environment, whereas three are endogean (humicolous) and comparably less specialized. The family therefore provides a model for testing the hypotheses that ecological specialists (stenotopes) evolve from generalist ancestors (eurytopes) and that specialization (in this case to the cavernicolous habitat) is an irreversible, evolutionary dead-end that ultimately leads to extinction. Due to their cryptic ecology, inaccessible habitat, and apparently low population density, Typhlochactidae are very poorly known. The monophyly of these troglomorphic scorpions has never been rigorously tested, nor has their phylogeny been investigated in a quantitative analysis. We test and confirm their monophyly with a cladistic analysis of 195 morphological characters (142 phylogenetically informative), the first for a group of scorpions in which primary homology of pedipalp trichobothria was determined strictly according to topographical identity (the "placeholder approach"). The phylogeny of Typhlochactidae challenges the conventional wisdom that ecological specialization (stenotopy) is unidirectional and irreversible, falsifying Cope's Law of the unspecialized and Dollo's Law of evolutionary irreversibility. Troglobitism is not an evolutionary dead-end: endogean scorpions evolved from hypogean ancestors on more than one occasion.
More at ScienceDaily. Those with broader knowledge of different species can weigh in, but for some reason I always felt skeptical of the idea that specialization was a dead end...though I suppose it could be analogized to an asexual clonal lineage which is at some adaptive peak. A naive idea I had in mind is that specializations are at least as likely to be gain of function mutations as loss of function mutations, and it is easier to go from functional variants to non-functional ones. But in any case, biological generalizations almost always have a lot of exceptions, the question is the extent of the exceptions, and whether that makes the generalization useful or not.
Citation: Cladistics, 26:2, doi:10.1111/j.1096-0031.2009.00277.x