Diversification of egg-deposition behaviours and the evolution of male parental care in darters (Teleostei: Percidae: Etheostomatinae)


Natasha B. Kelly, Osborn Memorial Laboratories, Department of Ecology and Evolutionary Biology, Yale University, 165 Prospect Street, New Haven, CT 06520.
Tel.: +1 203 432 4077; fax: +1 203 432 2374; e-mail: natasha.kelly@yale.edu


Male-only care is the most frequent parental care behaviour in teleost fishes, but little is known about its evolutionary origins and patterns of diversity in species-rich lineages. Darters are a clade of North American freshwater fishes that contain both nonparental care species and species with male-only care. In darters, paternal care takes the form of egg-guarding and other egg-tending behaviours that are dependent on the female mode of egg deposition. Male care has been hypothesized to evolve independently in darters at least three times, and it has been thought to be irreversible. We investigated the diversification of egg-deposition behaviours and the evolution of complex male care using published descriptions of darter reproductive behaviours and a multilocus molecular phylogeny that included all 146 species for which reproductive behaviours are known. We find support for two origins of male-only care behaviour. One origin of paternal care occurred relatively early in the radiation of Etheostoma and is characteristic of a recently discovered clade, Goneaperca. The other origin of male-only care occurred much more recently in a derived clade of Nothonotus. Our analyses of character diversification demonstrate reversals from care to noncare and multiple transitions between egg-deposition behaviours that are not associated with parental care.

Parental care in vertebrates has originated multiple times over the course of their evolutionary diversification. Among vertebrate lineages, mammals exhibit predominantly female-only care (approximately 90%) and birds predominantly biparental care (approximately 90%), whereas in teleost fishes, when care does occur, the majority of species provide male-only care (approximately 50%), followed by female-only care (approximately 30%) and biparental care (approximately 20%) (Gross & Sargent, 1985). Teleosts therefore offer representations of nearly all forms of care and provide excellent models for the study of parental care evolution.

The origin of male-only care has attracted much interest (Trivers, 1972; Williams, 1975; Dawkins & Carlisle, 1976; Maynard Smith, 1977; Ridley, 1978; Gross & Sargent, 1985; Tallamy, 2000), due primarily to the fact that the evolution of female-only care and, to a lesser extent, biparental care may be explained by anisogamy and parental investment theory (Bateman, 1948; Trivers, 1972), which takes into account the importance of the probability of parentage (Queller, 1997), more intense sexual selection on one sex and the effect of mortality on the operational sex ratio (Kokko & Jennions, 2008). However, this theory does not consider that males may both compete for mates and provide care for offspring, as is seen in many species of fish with male-only care (Breder & Rosen, 1966). The ability to both provide parental care and simultaneously compete for mates is proposed as an important factor maintaining male-only care in teleosts once it has evolved (Blumer, 1979).

Despite the attention the study of parental care in teleost fishes has received, there is no unified explanation for the evolution of parental care or the phylogenetic distribution of this trait among the major lineages of teleosts (Lindström & St Mary, 2008). Several factors suggested to either promote or inhibit the evolution of male-only care may counteract each other. For example, competition for resources has been proposed to promote the evolution of male-only care (Clutton-Brock, 1991), whereas reduced foraging opportunities have been proposed to inhibit it (Trivers, 1972; Williams, 1975). High predation pressure (Clutton-Brock, 1991) and sexual selection (Tallamy, 2000) have also been proposed to promote the evolution of male-only care, but the influence of these factors may be counteracted by the proposed inhibitory effects of increased mortality (Magnhagen & Vestergaard, 1991; Kokko & Jennions, 2008) and lost mating opportunities (Trivers, 1972; Williams, 1975; Queller, 1997; Balshine-Earn & Earn, 1998). The relative strength of each of these proposed factors may determine whether male-only parental care can evolve in a lineage. However, prior to investigating the relative influence of these factors on the evolution of male-only care, the evolutionary patterns of reproductive strategy diversification need to be assessed from comparative studies of teleost clades that exhibit male-only care as a reproductive strategy.

Comparative studies investigating the evolution of parental care in ray-finned fishes have done so at the phylogenetic scale of the major clades (Teleostei: Ah-King et al., 2005; Actinopterigii: Mank et al., 2005; Mank & Avise, 2006), limiting evolutionary inferences to identified transitions in parental care among major inclusive lineages. Analyses using comparative phylogenetic methods suggest that external fertilization (Mank et al., 2005) and male territoriality (Ah-King et al., 2005) are associated with the evolution of male-only care, whereas internal fertilization appears to be correlated with female-only care (Mank et al., 2005). Biparental care has also been suggested to have originated from an ancestral state of no care (Mank et al., 2005), rather than an intermediate state between male- and female-only care as previously hypothesized (Gross & Sargent, 1985). The reason that these analyses were conducted at the scale of major inclusive clades was largely due to (i) the lack of behavioural data scored for a large fraction of individual species for a given clade (although there are some studies that concentrate on that small fraction for which data exist: e.g. Hanel et al., 2002) and (ii) the lack of resolved phylogenies for a given species-rich clade.

One of the few lineages of teleosts for which both behavioural data and species-level phylogenies are available, Cichlidae, has been the subject of multiple comparative phylogenetic studies examining the evolution of and transition between different forms of parental care (Goodwin et al., 1998), and the effect of sexual selection on the evolution of parental care (Gonzalez-Voyer et al., 2008, 2009). In contrast to conclusions resulting from comparative analyses of major ray-finned fish clades (Mank et al., 2005), female-only care appears to have originated multiple times from an ancestral state of biparental care in cichlids (Goodwin et al., 1998). Thus, studies performed at a broad phylogenetic scale among the major ray-finned fish clades appear to lack the resolution to assess interspecific variation in parental care traits and patterns of their evolution at finer phylogenetic scales. Comparative studies of cichlids also concluded that transitions between parental care type are dependent on the intensity of sexual selection, more intense sexual selection being associated with the evolution of female-only care and moderate sexual selection with biparental care (Gonzalez-Voyer et al., 2008). Although cichlids provide an excellent framework for the study of the evolution of biparental and female-only care, exclusive male care occurs in only one species, Sarotherodon melanotheron (Aronson, 1948), making it impossible to explore the evolution of this reproductive strategy within this group. In this study, we present analyses of a teleost lineage with a well-resolved species-level phylogeny and behavioural data described for a large fraction of species in the clade that we believe is a potential model system for the study of the evolution of a range of behaviours, including, as we demonstrate, male parental care: the darters.

Darters as a model system for the study of the evolution of male parental care

Darters are a species-rich clade of freshwater teleosts classified in Percidae, which include approximately 40 species that exhibit male-only parental care. There are approximately 250 species of darters, and the clade is endemic to eastern North America with the greatest species diversity concentrated in the streams and rivers of the Interior Highlands and Eastern Highlands (Page, 1983; Etnier & Starnes, 1993; Near et al., 2011). Many of these streams harbour multiple sympatric darter species that partition habitat use by substrate type, flow regime or position within the water column (Page & Swofford, 1984; Greenberg, 1991; Welsh & Perry, 1998). Most species are sexually dimorphic with males exhibiting either bright coloration or melanic pigmentation during the breeding season. Female darters exhibit a range of egg-deposition behaviours that include egg-burying, egg-attaching, and egg-clustering and clumping that are associated with male parental care (Page, 1983, 1985). Parental care is absent in species that bury eggs under the substrate or attach eggs, either singly or in small numbers, to rocks or vegetation (Page, 1985). The attachment of eggs to the underside of the spawning substrate, in a monolayer or in clumps, is associated with male egg-guarding behaviour (Page, 1985).

Page (1985) hypothesized that the phylogenetic distribution of egg-deposition and egg-guarding behaviours within darters indicated that male-only care had arisen multiple times and that male egg-guarding originated from two different ancestral states. In Page’s (1985) evolutionary hypothesis, egg-burying is the ancestral state for darters that transitioned to egg-attaching with no egg-guarding behaviour, as observed in multiple subclades of Etheostoma, or to the attachment of eggs in clumps to the substrate in combination with male-only egg-guarding as observed in a subclade of Nothonotus. Within Etheostoma, the attachment of eggs in the absence of egg-guarding was hypothesized as the ancestral state that diversified to a derived state of attachment of eggs to the substrate in a monolayer in combination with male-only egg-guarding as seen in the subclades Boleosoma and Catonotus (Page, 1985). Thus, there are two character diversification pathways proposed for the evolution of male-only care in darters. (i) a two-state transition: burying with no care diversified to clumping eggs with care and (ii) a three-state transition: burying with no care diversified to egg-attaching with no care that subsequently diversified to egg-clustering in a monolayer with care. Page’s (1985) model provides testable hypotheses for exploring patterns for the evolution of male-only parental care in darters, namely did care originate once or on multiple occasions in darters and did this behaviour evolve along multiple evolutionary pathways?

Page (1985) developed the hypotheses of egg-deposition diversification in darters using a phylogenetic perspective based on external morphological characters. Male-only care was hypothesized to have originated three times in darters: once in a subclade of Nothonotus and once in each of the Etheostoma subclades Boleosoma and Catonotus. Page (1985) further hypothesized that transitions between deposition and care modes were unidirectional, so that attaching behaviour without egg-guarding must have evolved on multiple occasions within Etheostoma and that once egg-guarding behaviour originated in a lineage, it would not transition back to a no care egg-deposition mode. A recent molecular phylogeny of darters includes a newly discovered clade named Goneaperca that contains both Boleosoma and lineages that comprised previous concepts of Catonotus (Near et al., 2011), raising the possibility that male-only egg-guarding behaviour has a single evolutionary origin in Etheostoma.

The number of darter species for which behavioural data are available has increased from 68 to 146 in the last 25 years. In addition, a DNA-inferred time-calibrated phylogeny that includes 245 of the 248 darter species based on both mitochondrial and nuclear genes is available to serve as the basis for phylogenetic comparative analyses. This phylogeny indicates that the clade Catonotus is paraphyletic and that the lineage Stigmacerca, in which all species exhibit male parental care, was removed from Catonotus. Regardless, all lineages of Etheostoma that exhibit male parental care, Boleosoma, Catonotus, and Stigmacerca, were resolved along with Psychromaster in a newly discovered clade called Goneaperca. The increase in our knowledge of darter reproductive behaviour and an objective data-driven phylogenetic hypothesis provides a unique opportunity to evaluate Page’s (1985) seminal hypotheses regarding the diversification of egg-deposition behaviours and the evolutionary origin of male-only care in darters. As such, we investigate (i) whether male parental care has evolved multiple times in darters, (ii) if male care did evolve multiple times, did it do so along distinct pathways? (iii) whether transitions from one egg-deposition/parental care state are uni- or bidirectional.


Scoring egg-deposition behaviours in darters

We used information on egg-deposition mode, spawning site and male spawning territoriality to explore the evolution of male-only care. The main sources providing descriptions of male and female spawning behaviour, including female egg-deposition behaviour, male courtship, male mate-guarding, male site-guarding and egg-deposition site, were the studies of Winn (1958), Page (1983), Jenkins & Burkhead (1994), and Simon & Wallus (2005). Patrick L. Rakes and John R. Shute of Conservation Fisheries Inc. (CFI) based in Knoxville, Tennessee, USA (conservationfisheries.org), provided information on the observations of the egg-deposition behaviours for 15 darter species (see Supporting Information). CFI carries out captive propagation of imperilled North American freshwater species, as well as species closely related to these threatened and endangered species (Rakes et al., 1999). We also ran species-specific searches under both scientific and common names in ISI Web of Science and Google Scholar.

Darter egg-deposition behaviours and their implied male care behaviours were classified into three character states: egg-buriers (egg-burying with no care, eggs buried under the substrate), egg-attachers (egg-attaching with no care, eggs attached to rocks or plants above the substrate) and egg-guarders (egg-attaching with care, eggs attached to the underside of a rock or piece of wood above the substrate and guarded by the male). To simplify the analysis, we grouped together descriptions of egg-clumping and egg-clustering behaviours that are associated with male-only care into a single category: egg-guarding. We scored egg-deposition behaviour data for 146 darter species (Supporting Information).

Phylogenetic methods

A set of posterior Bayesian time-calibrated molecular phylogenies using external molecular evolutionary rates was generated using the computer program beast version 1.5.3. The phylogeny was inferred from DNA sequences of the mtDNA cytochrome b and two nuclear genes, S7 ribosomal protein intron 1 and recombination activating gene (RAG) 1 exon 3. Mitochondrial DNA sequences were treated as missing data in combined data set analyses for species with mtDNA genomes of heterospecific origin (see Near et al., 2011 for further description and rationale of methods used to construct phylogenetic trees). The phylogenetic data set included 245 of 248 recognized darter species and was similar to the phylogeny presented in Near et al. (2011). We used a subset of the posterior BEAST-inferred time trees (10 000 trees) for our comparative analyses. Species lacking egg-deposition data were removed from the phylogeny using the Ape package in the computer program R (http://www.R-project.org: script available on request). The resulting phylogeny included species from every named subclade (Near et al., 2011), and almost every deep node in the phylogeny is subtended by species sampled in our analyses (Fig. 1). To provide a heuristic assessment of likely character state transitions, we performed preliminary maximum likelihood ancestral state reconstructions of egg-deposition behaviours on selected individual phylogenetic trees using Mesquite version 2.74 (Maddison & Maddison, 2009).

Figure 1.

 Time-calibrated maximum-clade credibility phylogeny containing 245 of 248 recognized darter species inferred from a three-gene DNA sequence data set (see Near et al. 2011). Species that were included in our analysis and thus for which we have egg-deposition mode data are marked with a black circle.

Bayesian ancestral state reconstruction and estimation of character state transition frequencies

The evolution of egg-deposition behaviour and parental care in darters was reconstructed using a set of BEAST-inferred posterior time trees and the reversible-jump Markov chain Monte Carlo (MCMC) method in the BayesMultistate module of BayesTraits (http://evolution.rdg.ac.uk: Pagel et al., 2004). BayesTraits accounts for phylogenetic uncertainty in the ancestral state reconstructions by sampling the posterior distribution of equally credible phylogenetic trees and estimates the ancestral state of the most recent common ancestor of selected taxa regardless of whether they form a monophyletic group. This allows the incorporation of information on trait evolution from all trees in the posterior distribution, which allowed us to estimate the prior probability for each state at ‘deep nodes’ and the rates of character change across the tree. The analysis was run for 5 000 000 iterations, sampled at every 100 iterations, and a burn-in equal to 50 000. Rate deviation priors were selected to achieve an approximately 0.20 acceptance rate. Our analysis used a rate deviation of 0.02 and a gamma reverse-jump hyperprior (mean: 0.00, 10.00; variance: 0.00, 10.00).

In some cases, the two most probable states at a node were resolved with mean probability of approximately 0.5, or all states were assigned an equivocal mean probability of approximately 0.33. In these cases, nodes were fixed in BayesTraits using the ‘fossil’ command to determine which state has greater support when the ancestral state reconstruction is equivocal. Bayes factors (BF) were calculated and used to compare MCMC runs in which the node in question was constrained for each possible state. Each BF was defined as the difference between the highest harmonic mean log likelihood from five independent MCMC runs for each state. For BFs, values > 2 were taken as positive evidence, > 6 as strong evidence and > 10 as very strong evidence for the better fitting model (Kass & Raftery, 1995).

BayesTraits does not provide posterior distributions of the number or location of character state transformations across the phylogeny. Thus, we explored the historical pattern of egg-deposition character state diversification in darters using the Bayesian stochastic-character mapping procedure outlined by Huelsenbeck et al. (2003) as implemented using the computer program simmap version 1.5 (Bollback, 2006). We generated 500 maps on each of our 10 000 trees, resulting in 5 million runs. Because we were exploring the evolution of a character with more than two states, we used an equal prior for the bias parameter and used a branch length prior in which the branch lengths reflect the rate for our rate parameter. The set of posterior time trees were scaled to a tree depth of 1.0 before applying the priors.


Bayesian ancestral state reconstruction and estimation of character state transition frequencies

The mean posterior probabilities for the ancestral state reconstructions at key nodes in the darter phylogeny are illustrated in Fig. 2a,b. The Bayesian character state reconstruction analysis resulted in egg-burying behaviour having the highest probability as the ancestral state for the most recent common ancestor (MRCA) of darters (mean probability = 0.7488 ± 0.0147), transitioning to both egg-attaching in Etheostoma and clumping with egg-guarding in Nothonotus (Fig. 2a). A single origin of male-only care in a derived lineage of Nothonotus has the highest probability (mean probability = 0.9957 ± 0.0058; Fig. 2a), and within Etheostoma, there is an inference for a single origin of male-only care from the MRCA of Simoperca and Goneaperca (mean probability = 0.7911 ± 0.0046; Fig. 2b) to the MRCA of Goneaperca, which contains Boleosoma, Catonotus, Stigmacerca and Psychromaster (mean probability = 0.9348 ± 0.0043; Fig. 2b). Although support for the MRCA of Etheostoma and Nothonotus is equivocal as to an attacher or burier in the MCMC analysis (mean probability approximately 0.50), egg-burying was very strongly supported when the node was fixed using the fossil command in BayesTraits (BF = 51.79; see Table 1 for BF values for each equivocal node). No species of Nothonotus is known as an egg-attacher, and our analysis optimizes the MRCA of this clade as an egg-burier (mean probability = 0.9635 ± 0.0184; Fig. 2a). Egg-attaching without male care has the highest probability as the ancestral state for the MRCA of Etheostoma (mean probability = 0.8211 ± 0.0162; Fig. 2a), suggesting that egg-attaching has a single origin relatively early in the diversification of this lineage.

Figure 2.

 Time-calibrated maximum-clade credibility phylogeny containing 146 darter species for which egg-deposition behaviours are documented. Coloured circles denote the egg-deposition character state for each species. Mean posterior probability values for ancestral state reconstructions are illustrated at key nodes in the darter phylogeny. The phylogeny is presented in two parts, labelled (a) and (b).

Table 1.   Egg-deposition mode and Bayes factor support for equivocal nodes. All nodes show very strong support (BF > 10) for the egg-deposition mode listed.
NodeEgg-deposition modeBayes factor
Etheostoma–NothonotusEgg-burying/no care51.79
E. artesiae–E. nuchaleEgg-burying/no care13.49
E. exile–E. nuchaleEgg-attaching/no care13.72
PsychromasterEgg-burying/no care40.08

The analysis of character state changes reveals that there have been multiple reversals from attaching without care to burying without care in Etheostoma (Fig. 2a,b). There is support for a complete reversal to burying in the Etheostoma subclades Iriperca (mean probability = 0.9915 ± 0.0173; Fig. 2b) and Doration (mean probability = 0.9841 ± 0.0170; Fig. 2b). Although there is support for burying in the MRCA of Oligocephalus (mean probability = 0.6989 ± 0.0140; Fig. 2a), there are a reversal from burying to attaching in the MRCA of the subclade Astatichthys (mean probability = 0.8560 ± 0.0127; Fig. 2a) and a transition from attaching to burying in the MRCA of E. swaini and E. collettei (mean probability = 0.9834 ± 0.0233; Fig. 2a).

Within Goneaperca, there are two reversals from the ancestral state of egg-attaching with care to attaching or burying without care in Boleosoma and Psychromaster (Fig. 2b). The MRCA of Etheostoma vitreum and a clade containing Epodostemone and Elongimanum has a high probability as an egg-guarder (mean probability = 0.8416 ± 0.0038; Fig. 2b), but there is a transition to egg-attaching in E. vitreum. No species of Psychromaster exhibits egg-guarding, and there are resolved transitions between egg-attaching and egg-burying within this lineage (Fig. 2b).

The rate of change between egg-burying to egg-attaching with no care was similar to the reversal rate (mean of Bayesian posterior probability: q01 = 0.0471, q10 = 0.0521). However, the change from egg-attaching with no care to egg-attaching with care had a higher transition rate than transitions from egg-burying to egg-attaching with care (q12 = 0.0105, q02 = 0.0053). There was a higher rate of the transition from egg-attaching with care to egg-burying than that observed in the transition of egg-attaching with care to egg-attaching with no care (q21 = 0.0167, q20 = 0.0027).

Historical transitions between character states

The simmap analysis predicts a mean of 14.46 transitions between egg-deposition character states on the darter phylogeny. The transition from egg-burying with no care to egg-attaching with no care is predicted to occur an average 4.49 times, and the reversal to burying 5.53 times (Fig. 3). Transitions between attaching with egg-guarding and both attaching with no care and egg-burying are predicted to occur much more infrequently: attaching with no care to attaching with guarding 0.98 times; burying to attaching with guarding 1.21 times; egg-attaching with guarding to egg-attaching with no guarding 1.37 times; and egg-attaching with guarding to egg-burying 0.89 times. The proportion of time spent in a given state along the trees analysed was similar for egg-burying (proportion = 0.46) and egg-attaching with no care (proportion = 0.38). The proportion of time along the trees analysed spent in egg-attaching with guarding (proportion = 0.17) was smaller than egg-deposition states without care, implying that this parental care has either evolved less frequently or more recently than the other two states. As simmap produces multiple maps containing the possible timing of transitions between states but does not produce a composite map, we present a representative map that matches the ancestral states supported by our Bayesian analysis to provide an example of where these transitions may occur (Fig. 4).

Figure 3.

 Histograms showing the estimated number of transitions between egg-deposition states from 5 million character maps generated in simmap (Bollback, 2006). This shows the variability in the number of transitions due to uncertainty in the tree topology, branch lengths and character mapping by using 10000 trees sampled from the posterior distribution generated by beast and 500 samples from the posterior for each map.

Figure 4.

 Darter phylogeny with one possible evolutionary history of egg-deposition and parental care states using stochastic character mapping. This mapping represents one of five million simulations using simmap reconstruction.


Results from the Bayesian analyses of evolutionary transitions of egg-deposition character states in darters are consistent with the predictions outlined by Page (1985). In particular, there is support that the ancestral condition in darters is egg-burying, and egg-burying is the ancestral state to both egg-attaching with no care, observed in Etheostoma, and egg-clumping with male-only care, observed in Nothonotus (Fig. 2a,b). Egg-attaching with no care is optimized as the ancestral state to egg-attaching with male-only care in Goneaperca, specifically in Boleosoma, Catonotus and Stigmacerca. However, although we find support for multiple origins of male-only care, our analyses suggest that it has only originated on two occasions rather than the three proposed by Page (1985). We also find support for multiple reversals between egg-deposition states, including the reversal from egg-guarding to nonguarding in Goneaperca, suggesting that the evolution of male-only care in darters is not unidirectional as suggested by Page (1985). The phylogenetic and ancestral state analyses indicate that once the egg-guarding behaviour was lost, it was not regained (Fig. 2b).

While our analysis supports two separate origins of male-only care in darters, it also suggests that there are differences in the likelihood that male-only care will evolve along each of the pathways proposed for its evolution. The transition rates for egg-burying with no care diversifying to egg-attaching with no care and its subsequent diversification to egg-attaching with care (three-state pathway) are higher than the transition rate for egg-burying with no care diversifying to egg-clumping with care (two-state pathway). This suggests that egg-attaching with guarding evolves more rapidly along the three-state character diversification pathway proposed by Page (1985) than along the two-state pathway observed in Nothonotus.

There are two conclusions about the evolution of egg-deposition and complex care behaviours in darters revealed by our phylogenetic-based analyses. First, Page’s (1985) hypothesis that the evolution of male parental care has originated multiple times in the course of the diversification of darters is confirmed. Egg deposition with egg-guarding takes different forms in Nothonotus and Etheostoma (deposited in clumps versus a monolayer). We find support for these two forms of male parental care arising from different ancestral states, egg-attaching with no care versus egg-burying with no care. Second, our analyses reveal that the transitions between egg-deposition behaviours have occurred far more frequently than previously hypothesized (Page, 1985), particularly in the case of reversals from character states traditionally considered derived to those considered ancestral. Our analyses discovered that changes in character state that involve reversals from egg-attaching with no care to egg-burying have occurred multiple times in the evolutionary history of darters (Fig. 2a,b). Within the mostly egg-guarding clade Goneaperca of Etheostoma, reversals from egg-guarding (or egg-attaching with care) to egg-attaching with no care and egg-burying with no care have also occurred, specifically in Psychromaster, E. vitreum and E. tuscumbria (Fig. 2b).

Consideration of the habitat utilized by darter species suggests that further investigation into the correlates of ecology and life history of darter species with egg-deposition behaviours may illuminate factors that influence the transition between care and no care. For example, Etheostoma vitreum occupies sand-run habitats in rivers and streams and is often buried in the sand–gravel substrate, but spawns communally on the surface of an object and exhibits no male territoriality or egg-guarding behaviour (Winn & Picciolo, 1960), despite being phylogenetically nested in a clade of darters that largely exhibit egg-attaching with male-only care (Fig. 2b). The reduced availability of suitable nesting sites in sand-run riverine habitats may have facilitated the origin of communal spawning and the loss of egg-guarding behaviour in this lineage.

We suggest that further research into darter reproductive behaviour and ecology may also increase our understanding of the factors influencing the convergent evolution of care in darters. Egg-burying behaviour observed in Nothonotus differs from burying in Etheostoma, Percina and Carnipellucida. Within Etheostoma, eggs are scattered randomly and rapidly in the substrate (Reeves, 1907), whereas in Nothonotus, females remain buried for several minutes and eggs are deposited as a single mass in large clumps (Ross & Wilkins, 1993), attached to stones in small clusters (Warren et al., 1986) or buried as large clumps (Mount, 1959). This may indicate that while we classify Nothonotus as an egg-burier, it may be more accurate to categorize the behaviour as more similar to that of an egg-attacher. Certainly, the attaching or clumping of eggs within the substrate is a logical precursor to the above-substrate egg-clumping behaviour with male care observed in a derived clade of five species of Nothonotus (Fig. 2a). The territorial behaviour and site-guarding observed in many Nothonotus egg-burying species may also have been important in the evolution of male egg-guarding in this clade (Warren et al., 1986).

Despite the incomplete sampling of darter species in our analyses, it is unlikely that the conclusions regarding the origin of male-only parental care will be greatly modified by the discovery and description of reproductive behaviours for more darter species. For example, Goneaperca is the sister lineage of Simoperca, and all species of Simoperca, for which reproductive behaviour is known, are egg-attachers (Fig. 2b). It is unlikely that the handful of Simoperca species for which egg-deposition behaviour is not known (e.g. Etheostoma planasaxatile) are not egg-attachers. If any of these unsampled species of Simoperca turn out to be egg-buriers, they would be nested well towards the tips of the phylogeny so that the ancestral state estimate of Simoperca would likely remain egg-attaching, as it is in our analysis (Fig. 2b). Therefore, the unlikely scenario that some unsampled species of Simoperca is not an egg-attacher is unlikely to alter our conclusion that egg-attaching is the intermediate condition to the male-only care observed in Goneaperca. This is not to say that descriptions of egg-deposition behaviour in more darter species are not needed. The discovery of egg-guarders in any of the undescribed species on the tree or of egg-attachers in those lineages thought to contain only egg-buriers (e.g. Percina; Carnipellucida) would raise the possibility of multiple origins for these behaviours and could only aid in our understanding of how these behaviours evolve.

Our analyses of the evolution of male parental care in darters using densely sampled species-level phylogenetic trees reveal two evolutionary pathways from no care to male-only care. These comparative analyses support a historical reconstruction of character diversification that is sequenced as burying without care diversifying to egg-attaching without care and in turn diversifying to egg-attaching with male-only care in Etheostoma, and support for egg-burying without care diversifying to egg-clumping with care in Nothonotus. To investigate the evolutionary scenarios and behavioural ecological factors associated with these inferred pathways of character state diversification, additional research and documentation of male territorial behaviour is needed. Warren et al. (1986) posited that male territorial behaviour and site-guarding may have been important in combination with the mode of egg deposition in the evolution of male guarding behaviour. Male darters, when territorial, exhibit one of the two forms of territoriality: site-guarding, in which males guard a physical site and females enter that site to lay eggs; and mate-guarding, in which males guard a distinct area around the female which moves as she moves. Either of these forms or some combination of both behaviours can be hypothesized to have facilitated the transition, or acted as an intermediate step, between modes of egg deposition, and ultimately the evolution of male parental care. Unfortunately, we were unable to analyse this hypothesis due to the lack of available data on male territoriality. Another valuable direction for investigating the evolution of reproductive strategies in darters is determination of correlations between habitat use and egg-deposition behaviours, including male-only care.

Future directions

At the start of this paper, we detailed the still unresolved questions regarding the role of ecological and behavioural factors in the inhibition or promotion of the evolution of male-only care. We argued that a better understanding of the evolutionary patterns of reproductive strategy diversification was required before these questions could be resolved. We believe that this study provides that understanding and demonstrates the suitability of darters as a system for the investigation into the relative influence of these ecological and behavioural factors. Additional studies of darter ecology, life history and reproductive and territorial behaviour are required before these factors can truly be explored within darters. However, coupled with high species diversity, variation in egg-deposition behaviour and male-only care, and the availability of resolved and time-calibrated phylogenetic hypotheses for the clade, our analyses illustrate the utility of darters as a rewarding system to investigate the evolution of parental care in teleost fishes. Beyond parental care, the powerful framework of this resolved species-level phylogeny, and the decades of natural history research available for darters, has incredible potential to expand our knowledge of the evolution and co-evolution of behaviours, life histories and ecological traits.