*5.4. Speciation*

Speciation reflects the interactions among several biotic, abiotic, and historical events that have operated over various temporal and spatial scales. Most models that attempt to explain population divergence assume a role of geography (e.g., allopatry, parapatry) combined with niche conservatism or divergence to generate species [344–346]. For example, the linearity of the Andes results in elongate geographical ranges and reduces potential contact and gene flow among parapatric populations [347,348] (Figure 233). Pleistocene glaciations have produced recurrent fragmentation, isolation, and reconnection of montane forests and their faunas [349–353], but see Bush and Oliveira [354]. Species that today occupy the lowlands of the Chocó and Amazon ecoregions may had been separated by the uplift of the Andes, creating the opportunity for large-scale, co-occurring allopatric speciation.

**Figure 233.** Speciation modes. Effect of the linearity of the Andes in diversification processes. Modified from Guayasamin et al. [194].

Speciation may or may not be coupled with an ecological shift (Figure 234). The tendency of populations to retain their ancestral ecological niche and failing to adapt to the new environmental conditions facilitates lineage divergence when ecological barriers are present [344–356]. Recent studies demonstrate that closely related centrolenid species have, in most cases, similar abiotic ecological niches and that elevational shifts are rare in the family [3,28] (Figure 233). Then, lineage divergence in glassfrogs seems to be driven mostly by allopatric speciation coupled with niche conservatism [28]. There are, however, clear exceptions where clades are adapted to very different climates, meaning that niche divergence has occurred at least a few times (e.g., *Hyalinobatrachium* is mostly a lowland clade, whereas *Centrolene* and *Nymphargus* are montane clades). The relevance of niche conservatism in glassfrog diversification is at odds with studies that predicted and showed that, in amphibians, the primary causes of speciation were adaptation to climates (elevated regions vs. lowland regions) coupled with fragmentation of the once contiguous lowlands; in other words, allopatric speciation with ecological evolution [357,358] (Figure 234).

**Figure 234.** Speciation modes. (**i**) Ecological speciation in elevational gradients. (**ii**) Non-ecological speciation in the same elevational band.

Giving that our phylogeny includes the most complete taxon sampling so far, we are presented with the opportunity to search for specific lineages that match the different scenarios that explain tropical diversity. We acknowledge that many more species of glassfrogs remain to be discovered, changing the topology that is the backbone of our interpretations; however, we expect that these variations will be relatively minor and that the general patterns of speciation will hold. Also, we work under the assumption that geography is a primary player in the process of speciation and that closely related species are likely to be geographically nearby; this assumption is supported by the relatively low vagility that amphibians have in relation to other vertebrates [259,359–362], and that speciation is likely to produce sister species that are found nearby (allopatry), adjacent (parapatry), or in the same place (sympatry) [345,363]. Thus, below, we summarize the speciation hypotheses that are relevant for the Ecuadorian Andes and lowlands and include the examples that fit these scenarios.


**Figure 235.** Niche conservatism in glassfrogs at the generic level (modified from Guayasamin et al. [1]). Note that distribution of genera is restricted to particular biogeographic regions, suggesting that closely related species have similar climatic requirements. Number of species per genera are as follows: *Centrolene* = 24, *Nymphargus* = 41, *Hyalinobatrachium* = 32.

**Figure 236.** Potential river valleys that promote speciation in Ecuador.


## *5.5. Pending Taxonomic Problems and Candidate Species*

We have identified a total of 24 candidate species (Table S6), as well as numerous pending taxonomic problems, which we describe below:

• *Centrolene buckleyi:* Different sources of evidence (i.e., genetic, acoustic) sugges<sup>t</sup> that *C*. *buckleyi*, as currently defined, is a species complex [2,20,98]. We find three lineages (*Centrolene* sp. *Ca02*, *C*. sp. *Ca04*, and *C*. sp. *Ca05*) that are morphologically similar to *C*. *buckleyi*. None of these candidate species are sister to populations from the neotype locality of *C*. *buckleyi*. The extensive distribution of *C*. *buckleyi* in the high Andes of Colombia and Ecuador provide multiple opportunities for isolation and speciation. A taxonomic evaluation of this species complex is greatly needed, especially in the face of population declines that seem to have affected this species [91,103].


#### *5.6. Evolution of Translucency, Parental Care, and Humeral Spines*

*Translucency:* The widespread occurrence of translucency in glassfrogs is puzzling. A recent study (Barnett et al. 2020) [373] found that perceived luminance of glassfrogs changes depending on the immediate background, a change that is more pronounced on the legs, suggesting that camouflage is through edge diffusion. The strategy of disrupting the typical frog body outline might be even more efficient in species that exhibit complete ventral translucency and where internal organs are covered by reflective iridophores (Figures 13A and 102). It has also been shown that the dorsal green coloration in glassfrogs has similar reflective properties as photosynthetic leaves [374], also supporting the relevance of camouflage as an antipredatory mechanism. The venters of all glassfrogs are partially or completely transparent; therefore, it is reasonable to assume that this feature appeared in the ancestor of the clade. Also, complete ventral transparency has evolved multiple independent times within Centrolenidae (e.g., *Hyalinobatrachium*, *Chimerella*, *Vitreorana*).

*Parental care:* Until very recently, parental care in glassfrogs was considered to be rare and, when present, provided exclusively by males [4,17]. However, Delia et al. [25], based on detailed observations of 40 species, demonstrated that parental care is widespread in Centrolenidae (Table 8). In species thought to lack parental care, Delia et al. [25] observed that, just after oviposition and fertilization, females exhibit a short brooding behavior; this behavior significantly reduces embryonic mortality (experimentally tested in *Cochranella granulosa* and *Teratohyla pulverata*). Even though we still lack information on parental care for most glassfrogs species, the results by Delia et al. [25] have produced a major shift in what we thought we knew about parental behavior in this frog family. Ancestral reconstructions sugges<sup>t</sup> that the most recent common ancestor of glassfrogs exhibited a short, female-only parental care, from which some species (mostly *Hyalinobatrachium* and some *Centrolene*) have evolved prolonged, male-only parental care [25]. Additionally, the repeated evolution of complex male care is always associated with reductions in egg jelly and changes in oviposition sites [375]. Female mate choice and the evolution of parental care is an area of glassfrog biology that still needs further research.

**Table 8.** Parental care in glassfrogs. **Short-term maternal care** = immediately after oviposition, female provides brooding to egg clutch for several hours; after this initial brooding, clutches remain unattended. **Prolonged male care** = male provides parental care to egg clutch for several days. **Prolonged maternal care** = female provides parental care to egg clutch for several days. Each type of parental care is coded as Absent (0), Present (1), or Unknown (?). Terminology and data are summarized from Delia et al. [25].



**Table 8.** *Cont.*


**Table 8.** *Cont.*

*Humeral spines:* Male glassfrogs are highly territorial and they have evolved very unique behaviors and structures that are specifically associated to male-to-male combats (Table 9). The most conspicuous morphological trait that males use during fights are humeral spines [52,103], which are bony processes that project from each humeral bone (Figure 14), and that are present only in males of several glassfrog species (see Guayasamin et al. [1]). Although humeral spines are a rarity among amphibians, they have evolved multiple independent times in Centrolenidae [2], suggesting that they provide a selective advantage. Armaments (i.e., humeral spines) in glassfrogs probably only allow males to obtain or defend a territory [52,103], and most likely have no direct role in attracting females, which presumably choose a mate based on the quality of his territory, and acoustic and behavioral displays.

**Table 9.** Centrolenid species in which combat behavior has been documented, coded by the type of combat behaviors: **Primitive**: Combat in axillary amplexus-like position or wrestling on the leaves; **derived**: Combat dangling by hind limbs; both: Primitive and derived. Modified from Rojas-Runjaic and Cabello [384].


a the only complete fight observed lasted 3 min, but the longest observed fight lasted 26 min, beginning mid-conflict. \* Complete observation of combat behavior from the beginning to the end.
