Mating preferences, sexual selection and patterns of cladogenesis in ray-finned fishes


Judith E. Mank, Evolutionary Biology Centre, Department of Evolutionary Biology, Uppsala University, Norbyvägen 18D, SE-752 36 Uppsala, Sweden.
Tel.: +46 (18) 471-6461; fax: +46 (18) 471-6310; e-mail:


Evolutionary theory predicts that sexual selection may increase taxonomic diversity when emergent mating preferences result in reproductive isolation and therefore speciation. This theory has been invoked to explain patterns of diversity in ray-finned fishes (most notably in the cichlids), but the theory has not been tested comparatively in fish. Additionally, several other unrelated factors have been identified as promoters of cladogenesis, so it is unclear how important sexual selection might be in diversification. Using sister-clade analysis, I tested the relationship between the presence of sexually selected traits and taxonomic diversification in actinopterygiian fishes, a large clade that shows substantial diversity in mating preferences and related sexually selected traits. In all identified sister-families that differed with regard to the proportion of species manifesting sexually selected traits, sexual selection was correlated with increased diversification, and this association was significant across all sister clades (P = 0.02). This suggests that sexual selection, when present, is a substantial driver of diversification in the ray-finned fishes, and lends further empirical support to the theoretical link between mating preferences and accelerated cladogenesis.


Changes in mating preferences can theoretically lead to speciation events if the emergent mating preferences result in assortative mating and reproductive (prezygotic) isolation. Increased prezygotic isolation would be expected to accelerate taxonomic diversification compared with post-zygotic barriers to gene flow, which can require long periods of time to accumulate. This suggests that sexually selected traits, common manifestations of the pressures of mating preferences, may be associated with taxonomic diversification (Darwin, 1871; Lande, 1981, 1982; West-Eberhard, 1983). This correlation has been supported by some large-scale comparative analyses, primarily in birds and invertebrates (Barraclough et al., 1995; Gleason & Ritchie, 1998; Masta & Maddison, 2002; Polak et al., 2004).

The theory linking sexual selection (as a proxy for mating preferences) and cladogenesis has been anecdotally invoked to explain observed patterns of diversity in some clades of ray-finned (actinopterygiian) fishes (McMillan et al., 1999; Mendelson, 2003). The explosive radiation of the cichlid fishes has been repeatedly linked to mating preferences for sexually selected behavioural and morphological ornaments (Dominey, 1984; Knight et al., 1998; Maan et al., 2004). Sexual selection via mating preference has been documented in shaping a variety of traits in the ray-finned fishes, including dichromatism (Endler, 1980; Houde & Endler, 1990), breeding tubercles (Kortet et al., 2003, 2004) and elongated fins (Basolo, 1990; Meyer, 1997). Despite these clear documentations linking mating preferences and sexually selected traits (usually, though not always displayed in males) for species or small clades of Actinopterygii, some comparative studies in the ray-finned fishes have failed to uncover the expected manifestations of female preference (Ritchie et al., 2005; Mank et al., 2006), casting doubt as to pervasiveness and magnitude of female preference through the clade.

Additionally, the relative importance of sexual selection in explaining patterns of actinopterygiian diversity is unclear as several other unrelated factors have been recently identified as promoting diversification in this clade. Changes in genomic architecture (Hoegg et al., 2004; Mank & Avise, 2006a), key innovations such as internal gestation (Lydeard, 1993; Mank & Avise, 2006c) and antifreeze genes (Bargelloni et al., 1994; Eastman & McCune, 2000; Near et al., 2004), as well as vicariance (Murphy & Collier, 1996, 1997; Hurwood & Hughes, 1998; Planes & Fauvelot, 2002; Near et al., 2003) have all been demonstrably linked to increased cladogenesis in the ray-finned fishes. It is therefore likely that mating preferences, if a major factor at all, act in combination with other unrelated causes to create the current patterns in actinopterygiian diversity.

Despite the evidence for and against sexual selection as a driver of diversification in ray-finned fishes, the theory has not been tested across the clade, probably because of a combination of problems with the underlying phylogeny and the difficulty in gathering sufficient data on sexually selected traits. The recent construction of a well-resolved provisional supertree (Mank et al., 2005) for the Actinopterygii has partially resolved this problem and provides the necessary phylogenetic framework for a broad-scale comparative analysis. This supertree framework, in conjunction with numerous species accounts and field guides, offers the first opportunity to test the role of mating preference in promoting taxonomic diversification across the Actinopterygii. Using sister clade comparisons identified from the supertree, I tested the relationship between the presence of sexually selected traits, a proxy for mating preference, and increased diversification in the ray-finned fishes.

Materials and methods

From the actinopterygiian supertree (Mank et al., 2005), a 90% consensus tree from 25 000 most parsimonious trees, I identified all potential monophyletic sister families. Sister families are pairs of families that are taxonomically more related to one another than they are to any other family, and are equally old by definition (Cracraft, 1981). This type of sister clade comparison therefore automatically corrects for shared ancestry and divergence time in the assessment of any correlative relationship.

For each of these sister taxa, I first assessed whether there were any manifest sexually selected traits, using a family level compendium (Breder & Rosen, 1966). Sister families that completely lacked evidence of sexual selection were removed from any further analysis. These sister clades are uninformative regarding the relationship between manifest sexual selection and diversification because any quantitative cladogenetic differences would have to be due to other factors.

For the sister families with some degree of manifest sexual selection, I searched numerous field guides, species accounts, and aquarium references for descriptions of sexually selected traits (available in the Supplemental materials) in all currently taxonomically recognized species (Eschmeyer, 1990, 1998; Froese & Pauly, 2004). The sexually selected traits included here have been used in previous comparative studies of sexual selection in the Actinopterygii (Mank et al., 2005, 2006; Mank & Avise, 2006b; Mank, in press), and have been shown in the clade to be the result of mating preferences. These traits include extended or elongate fins or rays (Harrington 1997; Marcus & McCune 1999; Kuwamura et al. 2000), breeding tubercles (Kortet et al., 2003, 2004), and sexual dichromatism (Reimchen, 1989; Houde & Endler, 1990; Stott & Poulin, 1996; Amundsen & Forgren, 2001). I omitted from this analysis sexually dimorphic traits that may be the result (at least partially) of natural selection, such as body-size dimorphisms (Hamon & Foote, 2005; Schutz & Taborsky, 2005) and gonopodia (Langerhans et al., 2005). The nature of ichthyological data also prevented incorporation of some putatively sexually selected traits that are simply too poorly documented in a wide variety of actinopterygiian species, such as olfactory cues (Shohet & Watt, 2004; Milinski et al., 2005).

Numerous fish species are described on the basis of a single preserved type specimen, often collected long before the taxonomy is evaluated and described. As colour patterns often rapidly fade in preservation jars, it is not possible to ascertain from preserved museum type specimens whether sexual dichromatism, the most common manifestation of sexual selection in fish, exists. In order to avoid underestimating the incidence and importance of sexual dichromatism in the dataset, I did not use species accounts based solely on preserved specimens. This strategy presented an alternative problem, as some families are described almost entirely based on pickled individuals. I therefore removed all sister families that were insufficiently characterized (< 10% of recognized species described in detail) from further analysis.

There is not yet a complete and comprehensive molecular phylogeny for the Actinopterygii, or a large sub-clade, with which to estimate branch-lengths and divergence times. Additionally, it is not possible to estimate branch lengths and divergence times on the current actinoptergyiian supertree (Mank et al., 2005) because of the amalgamated nature of the underlying dataset. I therefore analysed the identified sister families according to the recommendations of Barraclough et al. (1995), Nee et al. (1996) and Vamosi & Vamosi (2005). For each sister family, I calculated the proportion of species that exhibited sexually selected traits, as well as determined the current number of recognized species (Eschmeyer, 1990, 1998). Under the null expectation that female preference does not influence patterns of diversity, we would expect families with a higher proportion of species with manifestations of sexual selection to be no more or less taxonomically diverse than their sister families. I evaluated the data against this null expectation with a randomization test for matched pairs according to Nee et al. (1996) and Barraclough et al. (1995), which I solved probabilistically rather than with repetitions. The randomization test is similar to the Wilcoxon sign test, which is not applicable to small numbers of comparisons. The randomization test computes the probability that the observed patterns of diversity, in this case the correlation between greater taxonomic diversity and a higher proportion of manifest sexual selection, is due to chance alone across all the analysed clades. Although it is arguably possible to use a one-tailed test of significance, I used the more conservative two-tailed test.

Two sister families showed very similar proportions of species with sexually selected traits (6% in both comparisons). I therefore performed two different analyses. I first performed the randomization test including these clades. However, previous analyses of this type in avifauna considered differences this small to be due to sampling error (Barraclough et al., 1995). I therefore also performed the randomization test treating these comparisons in the same manner as the comparisons that lacked manifest sexual selection entirely. In this second analysis, I did not consider these two comparisons in the randomization test, as comparisons with the same degree of sexual selection in both sister families are uninformative, because any differences in diversification must be due to factors other than mating preferences.


Of the 66 potential sister families, 42 (64%) lacked sexually selected traits entirely. I was unable to find sufficient data for another 15 identified sister clades.

The nine informative sister clades are shown in Table 1, and are distributed across seven taxonomic orders. Of the 806 species characterized in these 18 taxonomic families, 36% exhibited sexually selected traits. This is most definitely an overestimate of the incidence of sexually selected traits for the entire Actinopterygii, as all sister clades that lacked manifestations of mating preferences are not included. The percentage of species in a given family exhibiting manifestations of sexual selection ranged from 0 to 89, as shown in Table 1. Sexual dichromatism was the most common trait, present in 75% of species that manifested sexual selection, followed by elongate rays or fins (29%). Breeding tubercles were not documented in any of the species surveyed for this analysis.

Table 1.   Sister-families analyzed in this study.
 Sister families (Order) Number of species*Number of characterized species (%)Proportion manifesting sexual selection Direction of correlation†
  1. *Number of recognized species according to Eschmeyer (1990).

  2. †Direction of correlation between sexual selection and species diversity.

  3. ‡These comparisons can be treated as negative associations (and included in the randomization test), or omitted from the randomization test on the grounds that the degree of sexual selection in the sister families differ only by very small amounts.

I.Poeciliidae30955 (18)0.42+
Anablepidae (Cyprinodontiformes)1511 (73)0.00
II.Goodeidae479 (19)0.89+
Profundulidae (Cyprinodontiformes)55 (100)0.20
III.Melanotaenidae6716 (24)0.69+
Bedotiidae (Atheriniformes)1111 (100)0.27
IV.Belonidae348 (24)0.13+
Scomberesocidae (Beloniformes)44 (100)0.00
V.Monacanthidae10728 (26)0.39+
Balistidae (Tetraodontiformes)4218 (43)0.00
VI.Bothidae15732 (20)0.41+
Cynoglossidae (Pleuronectiformes)13636 (26)0.00
VII.Labridae481133 (28)0.52–‡
Scaridae (Perciformes)9536 (78)0.58
VIII.Gobiidae1426227 (16)0.22–‡
Eleotridae (Perciformes)16145 (28)0.28
IX.Characidae1113116 (10)0.58+
Alestiidae (Characiformes)11126 (23)0.31

Seven of the nine comparisons showed a positive correlation between taxonomic diversity and increasing proportions of sexually selected traits. The perciform comparisons (Gobiidae–Eleotridae and Labridae–Scaridae) showed only small differences in the proportion of species with manifest sexual selection (6% in both comparisons). These were the only comparisons that showed a negative association between sexual selection and taxonomic diversification. It is difficult to know whether these small differences are real, or are due to sampling error. Previous analyses of this type in birds considered differences this small to be nonsignificant (Barraclough et al., 1995), although for the sake of completeness I analysed the data both including and excluding these clades.

If these clades are excluded from the randomization test (accepting the assumption that the small differences in proportion of species exhibiting sexually selected traits are because of sampling error and therefore sexual selection is acting roughly equally on each sister family), the association between sexual selection and taxonomic diversity is highly significant (P = 0.008). If however, the clades are included in the randomization test, the association remains significant (P = 0.02).


This analysis supports the theoretical link between sexual selection and taxonomic diversification, through the presumed intermediate mechanism of shifting mating preferences. Additionally, these findings are concordant with previous comparative work in birds (Barraclough et al., 1995), suggesting that sexual selection acts in a similar manner throughout the vertebrates to accelerate cladogenesis. As the observed pattern was significant across several taxonomically diverse actinopterygiian orders, this work implies that mating preference is another mechanism to explain some of the rapid diversification exhibited by many clades of ray-finned fishes (Johns & Avise, 1998; Clements et al., 2003; Ruber et al., 2003; Ruber & Zardoya, 2005).

Despite the positive association between sexual selection and increasing taxonomic diversification, it should be noted that emerging mating preferences are not likely to be a major factor driving cladogenesis across the entire actinopterygiian clade. Of the families analysed, only 13% showed any manifestations of sexual selection via mating preferences. As the vast majority (87%) of families lacked any recognized sexually selected traits, other mechanisms must be acting to drive their taxonomic diversification. There are several factors other than mating preference that have been demonstrated to drive cladogenesis in ray-finned fishes, including genome duplication (Hoegg et al., 2004; Mank & Avise, 2006a), and key innovations like internal gestation and anti-freeze proteins (Lydeard, 1993; Bargelloni et al., 1994; Mank & Avise, 2006c). These other mechanisms of diversification suggest that mating preference is but one of a complex suite of factors that explain differential rates of speciation. It is however surprising, given these numerous other factors that have been demonstrably linked to taxonomic diversification, that sexually selected traits are so significantly correlated with cladogenesis. This implies that when present, mating preferences are a powerful evolutionary force acting to increase prezygotic barriers to reproduction.

Additionally, these other cladogenetic factors may account for why the association between sexual selection and taxonomic diversification was not concordant across all nine comparisons (although it is difficult to know whether the comparisons that showed a negative association are because of sampling error or reflect significant findings).

There are several caveats to this analysis that are worth careful consideration. First, the evidence described here is indirect. Not only are comparative analyses strictly correlative in nature, my analysis relies on a proxy for mating preferences, i.e. manifestations of sexual selection. Sexual dichromatism and elongate fins, the sexually selected traits in this study, have been shown to be the result of mating preference in the ray-finned fishes (Basolo, 1990; Houde & Endler, 1990; Maan et al., 2004). However, the sexually selected traits used here are only indirect indicators of mating preferences, and as mating preferences are not the only way in which sexual selection can influence cladogenesis (Arnqvist et al., 2000), it is difficult to parse out the specific effects of mating preference from other sexual selection factors in this analysis.

Other caveats are more related to the current state of available data. The actinopterygiian supertree used to identify possible sister clades is, by its very nature provisional, as it is an amalgam of all the applicable and robust phylogenetic information available in the current literature. Although the supertree is the best estimate of actinopterygiian phylogenetic relationships given the current systematic data, as more relevant phylogenetic information is published, it may be prudent to revisit this topic with a new and improved supertree. Also, my analysis relies upon extant taxa, and does not account for differential extinction rates, which could be a potential source of noise in the data. However, as sexually selected lineages may experience an elevated extinction risk compared with sexually monomorphic lineages (McLain et al., 1995, 1999; Kokko & Brooks, 2003; Morrow & Pitcher, 2003; J.E. Mank, in review), it is logical to conclude that extinction rates would obfuscate the relationship between sexual selection and cladogenesis rather than spuriously suggest it. Finally, roughly a quarter of identified sister clades were insufficiently characterized and were excluded from this study, and these omissions only hint at the lack of information available for the Actinopterygii. It is conceivable that emerging systematic databases, such as FishBASE (Froese & Pauly, 2004) will eventually solve this problem, and make a more complete analysis possible in the future.

Despite these caveats, this analysis is useful as it suggests a relationship between manifestations of sexual selection and taxonomic diversification, and lends supports to theories linking mating preferences to cladogenesis (Darwin, 1871; Lande, 1981, 1982; West-Eberhard, 1983) in another large vertebrate clade. Shifting mating preferences have been suggested to explain several actinopterygiian radiations (Dominey, 1984; McMillan et al., 1999; Danley & Kocher, 2001; Jones et al., 2003; Mendelson, 2003), although this analysis is the first to test comparatively the role of sexual selection in taxonomic diversification across the ray-finned fishes.


This work was supported by an American dissertation fellowship from the American Association of University Women. I would like to thank Mark Kirkpatrick, John Wares, and two anonymous reviewers for helpful comments on the manuscript.