The Phylogenetic Distribution of a Female Preference

Home , Southern platyfish, Xiphophorus continens, Xiphophorus montezumae

University of Nebraska - Lincoln DigitalCommons@University of Nebraska - Lincoln

Faculty Publications in the Biological Sciences Papers in the Biological Sciences

1996

Alexandra Basolo University of Nebraska - Lincoln, [email protected]

Follow this and additional works at: https://digitalcommons.unl.edu/bioscifacpub

Part of the Life Sciences Commons

Basolo, Alexandra, "The Phylogenetic Distribution of a Female Preference" (1996). Faculty Publications in the Biological Sciences. 45. https://digitalcommons.unl.edu/bioscifacpub/45

This Article is brought to you for free and open access by the Papers in the Biological Sciences at DigitalCommons@University of Nebraska - Lincoln. It has been accepted for inclusion in Faculty Publications in the Biological Sciences by an authorized administrator of DigitalCommons@University of Nebraska - Lincoln. Syst. Biol. 45(3):290-307, 1996

THE PHYLOGENETIC DISTRIBUTION OF A FEMALE PREFERENCE

ALEXANDRA L. BASOLO Nebraska Behavioral Biobgy Group, University of Nebraska, Lincoln, Nebraska 68510, USA; E-mail: [email protected]

Abstract.—Robust phylogenetic information can be instrumental to the study of the evolution of female mating preferences and preferred male traits. In this paper, the evolution of a preexisting female bias favoring a sword in male swordtail fish and the evolution of the sword, a complex character, are used to demonstrate how the evolution of mating preferences and preferred traits can be examined in a phylogenetic context. Phylogenetic information suggests that a preference for a sword arose prior to the evolution of the sword in the genus Xiphophorus and that the sword was adaptive at its origin. A phylogenetic approach to the study of female preferences and male traits can also be informative when used in conjunction with mate choice theory in making predictions about evolutionary changes in an initial bias, both prior to the appearance of the male trait it favors and subsequent to the appearance of the trait. [Mate choice; female preference; preexisting biases; Priapella; swordtail; sword; Xiphophorus.]

The study of the evolution of mating who have genotypes of higher viability. preferences is an area of much contention. Fisherian processes models stress the im- Until relatively recently, there was dis- portance of linkage disequilibrium be- agreement over the importance of female tween the female preference and the sex- mate choice. In the past 15 years, however, ually selected trait; the preference evolves many studies have shown that females ex- as a correlated response to male trait evo- hibit mating preferences and that these lution. Correlated effects models suggest mating preferences can have an important that female preferences evolve as a corre- effect on male mating success. Many mod- lated result of male preference evolution. els have been proposed to explain the evo- That is, males have a preference for a trait lution of female preferences; these models in females, the genetic basis of the prefer- can be assigned to one of five primary cat- ence in males is shared with females or is egories: direct benefits (Trivers, 1972; Hal- correlated for some other reason, there is liday, 1978; Thornhill and Alcock, 1983; no significant cost to the expression of this Kirkpatrick and Ryan, 1991), good genes preference in females, and thus a female (Williams, 1966; Zahavi, 1977; Hamilton preference for the trait in males evolves as and Zuk, 1982; Andersson, 1986; Pomian- a correlated effect of evolution of the male kowski, 1988), Fisherian processes (Fisher, preference for the trait in females. 1930; Lande, 1981; Kirkpatrick, 1982), cor- Models stressing the importance of good related effects (Darwin, 1871; Lande and genes, correlated effects, Fisherian process- Arnold, 1985; Halliday and Arnold, 1987; es, and some cases of direct benefits re- Arnold and Halliday, 1992), and preexist- quire genetic variation in the preference ing biases (Barlow, 1977; West-Eberhard, and some component of male fitness, and 1984; Kirkpatrick, 1987a; West-Eberhard et they predict coevolution of the preference al., 1987; Endler and McLellan, 1988; Ba- and trait. Little experimental data have solo, 1990b, 1995b; Kirkpatrick and Ryan, been produced that support one of these 1991). Direct benefits models suggest that revolutionary models to the exclusion of females receive a benefit from mating with the others. None of these models predict a certain male. For example, they may re- the evolution of a preference prior to a trait ceive a resource directly or they may gain (although any of them could result in pref- access to a resource for either themselves erences for traits not yet present). The fifth or their offspring. Good genes models sug- model, the preexisting bias model, suggest that female preferences evolve under gests that there are female biases resulting selection for females to mate with males from evolutionary processes independent 290 1996 BASOLO—PHYLOGENY OF A PREFERENCE 291 of sexual selection on the specific trait un- to describe how their interaction might af- der study and present prior to the appear- fect the direction of selection. Ryan (1990) ance of that male trait. (By convention, suggested that male traits may arise that mating preferences are presented in terms capitalize on biases in the female sensory of females and preferred traits are pre- system, using the term "sensory exploita- sented in terms of males, but the ideas are tion." Basolo (1990b) proposed a critical interchangeable regardless of sex.) test using phylogenetic information and a A unique prediction of a preexisting bias stepwise approach for testing preexisting model is that the preference appeared pri- biases models and suggested a broadening or to the origin of the trait. This prediction of the category of biases to include higher can only be tested using the comparative levels of integration and information pro- method, and to implement the compara- cessing. tive method good phylogenetic informa- Preexisting biases can evolve for a num- tion is essential. In this paper, the impor- ber of reasons (Basolo, 1990b, 1995a, tance of robust phylogenetic information 1995b), but regardless of the origin of the to the study of the evolution of female bias, preexisting bias models postulate that preferences and preferred male traits is a male trait can arise and be favored by a demonstrated. First, I discuss the preexist- previously established bias. However, fur- ing bias model of sexual selection in terms ther evolution of the bias may occur fol- of both female preference and male trait lowing its origin, and after the male trait evolution. Second, I discuss the use of phy- arises and is selected by the preexisting logenetic information to test a hypothesis bias, other sources of sexual selection, e.g., of sexual selection, using my own work on good genes, direct benefits, correlated ef- swordtail fish as an example. Third, I dis- fects, Fisherian processes, and other pre- cuss how phylogenetic information can be existing biases, may act in conjunction utilized to generate new hypotheses once with the existing bias to further elaborate we know the phylogenetic distribution of or modify the preference. In addition, the a preference and trait; specifically, how initial bias can be changed as a result of new hypotheses can be formulated about natural selection or other sources of evo- how a preexisting bias should differ be- lutionary change. tween lineages in which the male trait has A clear demonstration that a preexisting and has not arisen. bias has played a role in the evolution of a male trait requires that three criteria be PREEXISTING BIASES met: (1) females prefer a conspecific male trait, (2) the absence of the male trait is the Sexual Selection ancestral state within a group, and (3) in a The idea of preexisting biases affecting species possessing the ancestral state for the direction of selection is not new, but it the male trait, females prefer the male trait has recently been formalized. Barlow even though it was not generally present (1977) suggested that signals should in the evolutionary history of the species. evolve to forms that are best perceived by To adequately test a preexisting bias mod- the receiver. West-Eberhard (1984) sug- el, a robust phylogenetic hypothesis is es- gested that sensory traps can be used by a sential. signaler to capitalize on receiver response that has been selected in another context. Adaptive Male Trait Evolution Kirkpatrick (1987a, 1987b) suggested that Mate choice models address the evolu- intrinsic biases affect female perception tion of mating preferences, but they also and can affect the direction of male trait make predictions concerning the evolution evolution. Endler and McLellan (1988) con- of male traits. In the case of preexisting sidered how the environment and sensory biases, male traits that arise and increase properties of the receiver affect signal re- in frequency because of a preexisting bias ception and used the term "sensory drive" are always by definition adaptive in the 292 SYSTEMATIC BIOLOGY VOL. 45

context of mating. Using phylogenetic in- erence has been investigated theoretically formation, it is possible to determine when with a model of visual recognition (Arak a female bias arose. At any time after the and Enquist, 1993; Enquist and Arak, appearance of a bias, an "adaptive field" 1993); simulations of a neural network for a male trait exists; that is, mate choice model demonstrate that in some cases a re- will favor the male trait if it arises. Muta- sponse to certain patterns not previously tions that occur at low frequencies and re- encountered can be stronger than a result in new male traits may be lost before sponse to patterns that were initially se- selection can ever act on them or before lected to elicit a response. These unexpect- selection can have a maintaining effect. ed biases may evolve concealed and However, when an adaptive field in the nonfunctioning until a trait arises that is form of a preexisting bias exists for this recognized and selected by them. The de- new variant, there is a higher likelihood gree of similarity or dissimilarity of a male that the trait will increase in frequency. If trait among lineages resulting from the there is no counterbalancing selection, the same initial preference, or from an altered preexisting bias will result in an increase form of the initial bias, would depend on in the frequency of the male trait (predic- available variation, the influence of other tions concerning the fate of the male trait sources of selection, and genetic drift. if it arose prior to the origin of the bias Preexisting biases can be defined by bi- favoring it are not considered here). ases in the sensory system, which have Armed with phylogenetic information, one been shaped by natural selection, sexual can make predictions concerning the di- selection in another context, or other rection of selection, should the appropriate sources of evolutionary change or by a male trait arise. Therefore, although female combination of these. To establish that a preferences based on preexisting biases bias is specifically a sensory bias, a fourth may or may not have an adaptive origin, criterion must also be met: it must be es- in a sexual selection context the male traits tablished that there is a bias in a specific selected by such biases are undeniably sensory modality that is predictive of the adaptive at their origin. direction of sexual selection (Basolo, 1995a). To my knowledge, these four cri- Bases of Biases teria have been met in only one case. Ryan Preexisting biases can exist at any level et al. (1990) found that Physalaemus pustu- along the signal reception and processing losus and P. coloradorum share basilar pa- continuum from stimulation of a periph- pillae with peak sensitivity at 2,200 Hz, eral sensory system to actual synthesis of which phylogenetic evidence suggests is input from sensory systems, including the ancestral condition. Physalaemus colora- higher level integration and cognitive pro- dorum males do not produce calls at this cessing. The possible bases for an initial frequency, but P. pustulosus males produce bias are numerous; a subset of the possi- chucks that appear to have evolved to "ex- bilities are presented here. First, the basis ploit" this bias in the peripheral auditory of a preexisting bias may be as simple as system. Based on the phylogenetic infor- preference for rare, novel traits (but see Ba- mation available for female peak sensitiv- solo, 1990b). Second, preferences for one ity, one would predict that females of both male trait may have a pleiotropic effect, re- species may prefer male calls near 2,200 sulting in selection on a different male trait Hz over calls at other frequencies. Ryan (Ryan, 1990). Third, females may show and Rand (data published and discussed stronger mating responses to males with by Kirkpatrick and Ryan, 1991) found that more conspicuous traits simply because P. coloradorum females prefer synthetic males with these traits are easier to detect. calls with an added chuck. With these Fourth, evolutionary change in an estab- combined results, the four criteria sup- lished preference could result in a new porting a sensory basis to a bias are satis- bias or preference. The plasticity of a pref- fied. However, one must be careful when 1996 BASOLO—PHYLOGENY OF A PREFERENCE 293 specifically invoking a sensory bias as the The character states for nine caudal fin basis of a preference. Sensory biases have and sword characters for the genus Xz- been targeted as the basis of a bias in two phophorus are presented in Table 1 (based other situations (Clark and Uetz, 1992; on Rosen, 1979; Kallman and Bao, 1987; Proctor, 1992). Both appear to involve male Rauchenberger et al., 1990; Basolo, 1991, exploitation of female prey detection mech- pers. obs.). One component of the sword is anisms. Whereas a specific sensory bias at elongation of a specific set of rays (Rauch- the peripheral sensory level may explain enberger et al., 1990; Basolo, 1995a) at the the results in these additional studies, such base of the caudal fin (E), ranging in length as was found for the Physalaemus complex, from 0.6 to 1.0 times the body length (stan- it may be that a bias involving a search dard length). A second component, lower image would involve higher order process- black stripe (L), consisting of melanophore ing. Higher level biases may involve recep- coloration extending from the proximal tion and integration of neural signals from base of the sword to the distal tip along several to many peripheral sensory the ventral margin of the rays, occurs in 12 sources. Because the level of a bias may be species of swordtails. A character found in difficult to establish directly, caution some populations of the southern platy- should be exercised when assigning a basis fish, X. maculatus, resembles the proximal to a bias; the bias may be more complex third of this lower black stripe. A third than stimulation of several neurons at the component of the sword, the proximal peripheral sensory system level. portion of the upper melanophore stripe, is derived from the character grave DEMONSTRATING A PREEXISTING BIAS IN spot (G: a black caudal fin pigment pattern POECILIID FISHES that is sex limited in some species but not Background on Xiphophorus and Priapella in others), which is shared by 13 of the 14 The genus Xiphophorus is a group of swordtail species, although its expression freshwater, livebearing fish (Poeciliidae), is reduced in 2 species (X. birchmanni and which has historically been divided into X. continens; Rauchenberger et al., 1990). In two groups, the unsworded platyfish and the 11 swordtails that have an upper black the swordtails, most species of which pos- stripe, the grave spot gives rise to the prox- sess a sword. The sword does not develop imal portion of the stripe (PU: proximal until sexual maturity, at which time rays at upper stripe) in 7 species, and in the re- the lower margin of the caudal fin in males maining 4 species the grave spot gives rise begin to elongate and acquire pigmenta- to the proximal portion of the stripe, which tion. The proposed sister genus of Xi- continues to develop distally to the tip of phophorus is Priapella (Rosen, 1979), whichthe sword (PU plus MDU: midway to dis- consists of five described species, all of tal upper stripe). Of the seven species in which lack a sword. which the grave spot gives rise only to the proximal portion of the upper stripe, one Nature of the Sword is in the southern swordtail group (X. cle- The term "sword" has been used by dif- menciae) and the other six are in the north- ferent investigators to describe different ern swordtail group. Of these seven species structures but was originally used by of swordtails, five always have a distal up- Heckel (1848) to describe the lower caudal per stripe, which develops from the distal appendage of the green swordtail, Xz- tip of the sword proximally (DMU: distal phophorus helleri. This species is a memberto midway upper) and meets the proximal of the southern swordtail group, all mem- portion of the stripe midway, and two spe- bers of which have a complex sword con- cies occasionally have DMU (X. nigrensis, sisting of a number of component charac- X. multilineatus). A final sword component ters. Most species in the second group of consists of orange, yellow, and/or green swordtails, traditionally called the north- coloration (C) situated between the two ern swordtails, also have a complex sword. melanophore stripes. This coloration con- 294 SYSTEMATIC BIOLOGY VOL. 45

TABLE 1. Twenty-three described species of Xiphophorus with character states for nine caudal fin and sword characters. + = presence of the character; ca = carotenoid coloration; pt = pterin coloration. - = absence of the character but does not indicate whether or not character has been secondarily lost; ? = no available information because species has not been tested. An asterisk indicates that character state is obscured by presence of E. States in parentheses indicate that a species is polymorphic for that character.

Characters* Species PU MDU DMU Platyfish meyeri couchianus — — — gordoni - - maculatus -\ — — milleri ^ - - variatus - - - evelynae - xiphidium + + - - andersi + + - - Swordtails Northern nigrensis *V * + (+ca) multilineatus 't * + (+ca) pygmaeus 'h (+) (+ca) cortezi 't * + (+ca) malinche "1- * + (+ca) birchmanni "h (+) (+ca) continens * - - - nezahualcoyotl "i * + (+ca) montezumae t * + (+ca/pt) Southern clemenciae \ * + +ca alvarezi V * + (+ca/pt) PMHC t * + +ca/pt signum t * + +pt helleri t * + (+ca/pt) a p = genetic ability to express protrusion, unexpressed unless hormonally induced; P = expressed protrusion; E = elongation; C = yellow, green, and/or orange coloration; G = grave spot; L = lower melanophore stripe; PU = proximal upper melanophore stripe; MDU = distal upper melanophore stripe, developing from midway to distal tip of sword; DMU = distal upper melanophore stripe, developing from distal tip of sword proximally. b A color pattern resembling the proximal % to % of L is expressed in some males and females. Whether this partial stripe is homologous to L is unknown. c Designation given by Rosen (1979:361) to a swordtail species in the southern clade, "clearly allied to helleri and signum ... possibly a separate entity from helleri." sists of a central field of pterin and/or ca- branching patterns, but it could also indi- rotenoid pigments (Kallman and Bao, cate that one branching pattern is ancestral 1987). The wealth of information available and the other evolved from it.) for the genus Xiphophorus indicates that the The term "sword" has been used to de- sword is a composite trait, consisting of a scribe structures in other genera (e.g., in set of distinct component characters (some fish hobbyist publications to describe cau- of which may not yet be designated). (The dal fin extensions in guppies, Poecilia reti- branching patterns of the rays constituting culata) that do not appear to be homolo- the swords differ between the northern gous to the sword found in swordtails, and southern swordtails, which may indi- based on caudal rays counts, branching cate that the character E is a composite patterns, ontogeny, and the position of col- character itself. This difference would sug- or pigments (Rauchenberger et al, 1990; gest that there are two independent Basolo, unpubl. data). 1996 BASOLO—PHYLOGENY OF A PREFERENCE 295

Within Xiphophorus, the term "sword" has also been used to describe a short, unpigmented protrusion (P) found naturally in two species of platyfish and to describe a short, unpigmented protrusion that can be hormonally induced (p) in two other species (Gordon et al., 1943; Zander and Dzwillow, 1969). In his revision of the genus Xiphophorus, Rosen (1979) referred to the structure P as a caudal appendage (described as the prolongation of the ventral- most principal caudal fin rays in adult males) and used the term sword within quotation marks for describing this short prolongation, indicating it was not a true sword. Such a structure is also seen occasionally in two of the three swordtails lacking a sword (X. birchmanni, X. pygmaeus). This structure is never seen in the other swordtail lacking a sword, X. continens. In this paper, P is considered a possible ancestral state for the sword component character E but is not equivalent to the complex sword. For this paper, a structure consisting of characters E, L, C, PU, and either MDU or DMU is defined as a sword. (b) Preference for the Sword in Swordtails To address the first criterion of the preexisting bias model, I examined female preference for sword length in the green swordtail, X. helleri, a species in which males possess swords that vary in length. Females preferred conspecific males with longer swords to those with shorter swords (Basolo, 1990a). The sword is thus a target of female choice in this sworded species. Evolutionary History of the Sword Sexual selection arising from a preexisting bias requires that the origin of the pref- (c) erence precede the origin of the male trait. To address the second criterion of the pre- FIGURE 1. Three phylogenies showing proposed existing biases model, phylogenetic infor- relationships of Xiphophorus and other poeciliids. (a) mation must be assessed to establish the Composite phylogeny based on biochemical, genetic, evolutionary history of the sword. and morphological characters, with biogeographic Four phylogenies have now been pro- considerations (Rosen, 1979; Rauchenberger et al., 1990). (b) Phylogeny based on morphological and be- posed for Xiphophorus, three of which are havioral data (Haas, 1993). (c) Phylogeny based on depicted in Figure 1. One is a composite molecular evidence (Meyer et al., 1994). * = all south- phylogeny (Fig. la), constructed by impos- ern swordtails except X. clemenciae. 296 SYSTEMATIC BIOLOGY VOL. 45 ing Rauchenberger et al.'s (1990) phyloge- 1987; Rauchenberger et al., 1990) suggests ny at the appropriate node of Rosen's that this character is shared between the (1979) phylogeny (consolidated by Basolo, two swordtail groups. Regarding the 1991). This phylogeny, the culmination of sword, Rauchenberger et al. (1990:21) con- nearly 50 years of work by a number of cluded that the distribution of the charac- researchers (Gordon, 1947; Rosen and Gor- ter long sword (involving character E but don, 1953; Atz, 1962; Miller and Minckley, not character P) "is most parsimoniously 1963; Rosen and Bailey, 1963; Kallman and interpreted as a synapomorphy uniting the Atz, 1966; Kallman, 1971, 1975, 1989; Ro- northern and southern swordtail groups." sen, 1979; Borowsky, 1984; Morizot and Si- However, they conceded that the form of ciliano, 1984; Rauchenberger et al., 1990), the long sword in the northern swordtails was derived using biochemical and genetic may not be homologous to the form of the hybridization data and a suite of morpho- long sword in the southern swordtails and logical characters, with biogeographic con- that these two "evolved in parallel from a siderations. This phylogeny suggests that common ancestor with a gene pool that swordtails are a monophyletic group and permitted the evolution of a long sword" that the components constituting a sword (Rauchenberger et al., 1990:21). Therefore, appeared progressively, with a completely although parsimony suggests that the assembled sword appearing relatively late northern and southern swordtails are unit- in the evolutionary history of the group ed by the general development of long pig- (Basolo, 1991, 1995a). The Rosen/Rauch- mented swords in males, Rauchenberger et enberger et al. composite phylogeny in its al. allowed that there is evidence concerning entirety, representing the relationships of the shape of the sword and differences in 22 of the 23 described species of Xiphopho- the development of the upper melano- rus (the platyfish X. andersi is not repre- phore stripe between the two groups of sented) is shown in Figure 2a, with the swordtails that indicates that the sword character states of sword components de- may not be homologous for all compo- picted. Rauchenberger et al. (1990) listed nents. Some evidence suggests that the the lower melanophore stripe component sword may have evolved further after the of the sword (L) as a character that unites divergence of the northern and southern the two swordtail clades: the northern and swordtails clades (primarily the differ- the southern swordtails. However, the on- ences in the development of the upper togeny of sword development suggests black distal portion of sword coloration that the evolution of the upper melano- and the branching pattern of the most phore stripe is not straightforward. Of the elongated rays of the sword). Nonetheless, 14 species of described swordtails, 3 do the southern and northern swordtails are not have an upper stripe (X. pygmaeus, X. united as a monophyletic group, and they birchmanni, X. continens). However, two of share the presence of a sword composed these three species have a grave spot (G). of elongation (E), the lower black melano- The proximal portion of the upper stripe phore stripe (L), an interior field of color- (PU) appears to be a synapomorphy of the ation (C), and the proximal portion of the northern and southern swordtails, but the upper black stripe (PU). Using parsimony, condition for the distal portion of the up- the composite phylogeny suggests that per black coloration is not clear. DMU ap- swordlessness is the ancestral state, be- pears to be the primitive condition because cause the platyfish and the sister genus it is shared by one species each from the Priapella lack a long pigmented sword. In southern and northern swordtail clades, addition, Priapella species lack character P, whereas the upper stripe portion MDU which is found in the platyfish X. andersi may be the derived condition for the re- and X. xiphidium. Therefore, this character maining four species in the southern likely arose after the divergence of Xi- swordtail clade. Available information for phophorus and Priapella. From the distribu- sword coloration (C) (Kallman and Bao, tion, character P and the genetic ability to 1996 BASOLO—PHYLOGENY OF A PREFERENCE 297

£ 2

=3 'I sfiliiiliilsiiiflio i I c £0 u io -c

(a)

(b)

FIGURE 2. Two proposed Xiphophorus phylogenies with caudal fincharacter s and sword component characters mapped onto two proposed phylogenies. 1 = characters PU, G, E, DMU, L; 2 = characters PU, E, DMU; 3 = characters PU, C, E, DMU, P, L; 4 = characters E, P, PU, DMU. • = gain of character; • = loss of character, (a) Rosen (1979)/Rauchenberger et al. (1990) composite phylogeny. (b) Meyer et al. (1994) phylogeny, with G, C, and L not depicted because interpretation of character appearance is equivocal. express this character (character p = unex- treatment, whereas X. variatus, X. couchi- pressed genetic ability to produce a pro- anus, and X. gordoni do not) appear to have trusion; some female and male X. macula- arisen after the divergence of the desert tus and X. milleri that do not normally have platyfish from the rest of Xiphophorus. a protrusion develop one with hormone Xiphophorus continens has secondarily lost 298 SYSTEMATIC BIOLOGY VOL. 45 character P along with all component char- present state for X. continens, which is the acters constituting a sword. For the sword- lack of P and of any other sword charac- tails X. birchmanni and X. pygmaeus, char- ters; their conclusion should have been acter P (^5 mm prolongation) and the three gains and two losses.) In essence, color components of the sword (L, PU, C) however, their analysis suggests only that are occasionally expressed and thus ap- character P likely is ancestral for the genus pear either to be in the state of evolution- Xiphophorus (see also Wiens and Morris, ary loss or to have reached some equilib- 1996) and has been lost once and regained rium state at which they are maintained at at least twice within Xiphophorus. To inves- a low frequency tigate the evolution of the sword using this A second phylogeny based on courtship phylogeny, characters E, C, L, PU, MDU, behavioral traits and morphological traits and DMU must be considered; based on differs in its placement of several species the distribution of E, PU, MDU, and DMU, (Haas 1993; Fig. lb); the southern platyfish the most-parsimonious conclusion is that X. maculatus is placed basal to the rest of these characters are not ancestral to but the genus and the platyfish X. xiphidium have multiple origins within the genus and X. andersi are grouped together. De- Xiphophorus. The conclusions for characters spite these differences, the sword still ap- C, L, and G based on this phylogeny are pears to be a derived composite character equivocal. There are three equally parsi- within Xiphophorus, arising after the diver- monious interpretations for G, each involv- gence of the platyfish and the swordtails. ing four evolutionary events; two suggest This phylogeny suggests that P arose after that G is a derived condition within Xi- the divergence of the swordtails and the phophorus and one suggests that G is an- two platyfish species X. andersi and X. cestral for Xiphophorus. There are three xiphidium from the rest of the platyfish. equally parsimonious interpretations for L, Based on this phylogeny, there are two each involving five evolutionary events; equally parsimonious explanations for two suggest that L is a derived character character p: (1) p is a primitive character for Xiphophorus and one suggests that L is for the genus Xiphophorus and has been ancestral for Xiphophorus. There are two secondarily lost by both X. variatus and the equally parsimonious interpretations for C, desert platyfish or (2) p has arisen three each involving four evolutionary events; separate times. one suggests that C is derived within Xi- In a third phylogeny (Meyer et al., 1994; phophorus and the other suggests that C is Fig. lc) based on partial sequences from ancestral for Xiphophorus. Therefore, based three genes, the swordtails are not a mono- on this phylogeny, the distribution of four phyletic group and the platyfish are placed sword components suggests that a com- within the swordtails (data for the sword- posite sword is not ancestral to the genus. tail PMH [see Table 1] were not included In addition, the most-parsimonious analy- in this analysis). Meyer et al. (1994) used sis of the distribution of E, PU, DMU, and their phylogeny to investigate male caudal MDU suggests that none of these sword fin elaboration in Xiphophorus. In their components have occurred in any platyfish analysis, they called different structures a lineage. sword: not only those structures consisting In another phylogenetic treatment of of E, C, L, PU, and either MDU or DMU Xiphophorus, Borowsky et al. (1995) used but also those defined here as P. With these arbitrarily primed polymerase chain reac- two different criteria for a sword, they con- tion (AP-PCR) data to investigate the rela- cluded that a sword is ancestral for the ge- tionships of 14 of the 23 known species of nus Xiphophorus, and the distribution of the Xiphophorus to produce an unrooted tree. character states based on this phylogeny They converted both the Rosen/Rauch- (Fig. 2b) suggests that a sword defined as enberger et al. composite phylogeny (Fig. such has been gained three times and lost 2a) and the molecular phylogeny based on once. (However, Meyer et al. omitted the three genes (Fig. 2b; Meyer et al., 1994) to 1996 BASOLO—PHYLOGENY OF A PREFERENCE 299 unrooted trees and compared each to the 800 i (a) unrooted tree they generated; they con- 600 - cluded that their tree was "very consistent" with the composite phylogeny and 400 "less so" with the three-gene molecular phylogeny. Because this AP-PCR molecu- 200 - lar phylogeny is unrooted, its utility in determining the evolution of the sword components as such is poor and will not be o discussed further in this context. 0 800- (b) The evidence that a preexisting bias has jD played a role in the evolution of the sword CO 600- in the genus Xiphophorus, as well as any investigation on the origin and subsequent 400- change in the bias, critically depends on CD 200- the phylogeny. Problems with conflicting Q. phylogenies at one taxonomic level, how- CD ever, may be circumvented by examining the character states in outgroups to obtain 8OO-1 a clearer picture of character evolution. (C) Members of Priapella, the proposed sister 600 genus of Xiphophorus, lack any sword component and lack the caudal fin characters 400 G and P. In addition, composite swords are not known to occur in any other species of 200 poeciliid fishes. Some traits similar to sword components are found in other species, but their structure, rarity, and scat- Sword No sword tered phylogenetic distribution indicate Sword Status that they are homoplastic rather than ho- FIGURE 3. Preference for conspecific males with ar- mologous characters. Thus, the sword ap- tificial swords for three species, (a) Xiphophorus ma- pears to be a derived structure that arose culatus (six of six females preferred conspecific males after the divergence of Priapella and Xz- with artificial swords; two-tailed binomial test, P = 0.03). (b) Xiphophorus variatus (six of six females pre- phophorus. Consequently, by referring to ferred conspecific males with artificial swords; two- phylogenetic relationships outside of the tailed binomial test, P = 0.03). (c) Priapella olmecae (13 genus Xiphophorus, the second criterion for of 14 females preferred conspecific males with artifi- the preexisting biases model is satisfied cial swords; two-tailed binomial test, P < 0.01). [This comparison is for illustrative purposes. Because ex- and in full agreement with all proposed perimental methodology varied, the data may not be phylogenies; parsimony suggests that the strictly comparable.] absence of the sword is the ancestral condition for the Priapella/Xiphophorus group (Rosen, 1979; Meyer et al, 1994). Priapella to assess the evolutionary history of the preference favoring a sword. Preference for conspecific males with Evolutionary History of the Preference swords was found in two species of platy- Favoring a Sword flsh, the southern platyfish, X. maculatus Sexual selection arising from a preexist- (Basolo, 1990a; Fig. 3a), and the variable ing bias requires that the origin of the pref- platyfish, X. variatus (Basolo, 1990c, 1995b; erence precede the origin of the male trait. Fig. 3b), despite the absence of a sword in To address the third criterion of the pre- both species (this result has been verified existing bias model, I tested sword pref- for X. variatus by Haines and Gould erences of female platyfish and female [1994]). In these choice tests, the character 300 SYSTEMATIC BIOLOGY VOL. 45

state of males was manipulated by surgi- The sword appears to have evolved ei- cally attaching artificial swords to males. ther subsequent to the divergence of platy- To control for surgery and swimming ef- fish and swordtails (Figs, la, lb) or in the fects, paired males each received a lower common ancestor of platyfish and sword- caudal extension, but one test male re- tails when P is scored as presence of a ceived a complete sword consisting of sword (Fig. lc). When P is assumed to be elongation, a lower stripe, a complete up- a sword (Meyer et al., 1994), the critical test per stripe, and interior yellow coloration, of the preexisting bias model to explain and the other received a clear attachment preference favoring a sword therefore in- of equal length. This clear attachment sim- volves testing preferences of females in ulates absence of a sword but controls for outgroups to Xiphophorus. Considering surgery effects and the influence of the at- Priapella as the closest relative to Xiphopho- tachments on swimming and courtship be- rus, three hypotheses can be generated havior. The results for these two platyfish concerning the distribution of a preference species satisfy the third criterion of the favoring a sword: (1) a homologous pref- preexisting bias model for the composite erence is shared by Priapella and Xiphopho- phylogeny and the phylogeny based on rus, (2) Priapella lacks a preference favoring morphological and behavioral traits (Figs, a sword, and (3) Priapella has a preference la, 2a, lb) but not for the three-gene mo- favoring a sword, but this preference is not lecular phylogeny (Meyer et al., 1994) homologous to that found in Xiphophorus. when the presence of P is considered as The conflict generated by the conclu- the presence of a sword (Figs, lc, 2b). Con- sions drawn from the three-gene molecu- sidering the first two phylogenies, it ap- lar phylogeny and the other proposed phy- pears that a preference for a sword arose logenies can be resolved by determining in the ancestor of platyfish and swordtails, the character state for the character sword whereas the completed composite sword preference for the proposed sister genus arose after the divergence of platyfish and Priapella. Preference tests were conducted swordtails. to determine whether P. olmecae females Although the three-gene molecular phy- preferred conspecific males with artificial logeny suggests an alternative scenario for swords to conspecific males without male sword evolution when the presence swords. Like female platyfish, female P. ol- of character P is considered as presence of mecae preferred conspecific males with ar- a sword, it sheds no light on the evolution tificial swords to males without artificial of the preference for a sword. This phylog- swords (Fig. 3c; see also Basolo, 1995b). eny suggests that a "sword" evolved in the These results suggest that a preference fa- ancestor of platyfish and swordtails, after voring a sword is an ancestral condition the divergence of Priapella and Xiphophorus. shared by the sister taxa Priapella and Xi- However, it does not distinguish between phophorus, and this conclusion is consistent the two possible hypotheses for the rela- with all three of the proposed rooted phy- tive time of origin of the preference: (1) the logenies (Fig. 1). Regardless of relation- preference evolved simultaneously with ships within Xiphophorus, these results the sword or (2) the preference evolved suggest that the preference favoring a prior to the origin of the sword. Based on sword evolved prior to the origin of the this molecular phylogeny, it is impossible sword itself. A less parsimonious expla- to accept one of these hypotheses to the nation for the presence of a shared pref- exclusion of the other. Thus, although the erence in the four species tested is that the molecular phylogeny provides support for preference favoring a sword has evolved an alternate hypothesis for the evolution- several times. However, this interpretation ary history of a "sword," it provides no requires at least two independent evolutionary events: the evolution of a prefer- support for any of the five models for the ence in the Xiphophorus clade and the evo- evolution of the preference favoring a lution of a similar preference in the sword. 1996 BASOLO—PHYLOGENY OF A PREFERENCE 301

Priapella clade. A preexisting bias hypothesis, in which a single evolutionary event accounts for a shared preference, is a more parsimonious explanation. Whether the ancestral bias resulting in the demonstrated preference favoring a sword in the two species of platyfish and Priapella is specific for a sword or for something else has yet Trait to be addressed. By referring to phylogenetic relationships outside of the genus Xiphophorus, res- olution of the problem presented by the differing phytogenies is attained. Priapella and Xiphophorus share the ancestral con- Bias dition of sword preference, and the sword arose later, after the appearance of the FIGURE 4. Hypothetical distribution for the strength of a female preference (bias) prior to the ap- preference. With the combined phyloge- pearance of a male trait and after the male trait has netic evidence, the three criteria to support arisen if there has been coevolution of the preference a preexisting bias model for the evolution and the trait. of a preference for a sword are satisfied. In addition, phylogenetic information indicates that there was an "adaptive field" for existing bias has been documented, it then the sword when it appeared. becomes possible to use a phylogenetic approach to test whether other processes TESTING OTHER HYPOTHESES IN A have modified the bias after the origin of PHYLOGENETIC CONTEXT the male trait favored by the bias. Given a The data suggest that a preexisting bias phylogeny, robust inferences about when has played a role in the establishment of a the bias arose in relation to when the male sword preference and in the subsequent trait arose can be made, and we can make evolution of the sword: (1) female green predictions about how the preference may swordtails prefer conspecific males with vary among taxa. If other processes have longer swords, (2) the sword is either a modified a preference after the origin of shared derived character of swordtails or the trait, the strength of the preference a shared derived character of Xiphophorus, could either be stronger or weaker in taxa and (3) a preference favoring a sword is a in which males possess the trait than in shared primitive state for Xiphophorus and taxa exhibiting the primitive conditions of Priapella. This knowledge of the phyloge- the presence of the female bias and the ab- netic distribution of the preference and the sence of the male trait. sword allows us to propose and test ad- For poeciliid fishes, Fisherian processes, ditional hypotheses concerning the evolu- good genes models, and some direct ben- tionary history of the preference favoring efits models predict that the strength of the a sword. preference favoring a sword should be The existence of a preexisting bias does greater in species with swords than in spe- not necessarily preclude a role for other cies with the preference but without processes in the evolution of a mating pref- swords in their evolutionary history (Fig. erence. For example, a preexisting bias may 4). These revolutionary hypotheses would select for a male trait once it arises, but as thus predict that the strength of the pref- the male traits evolves, other processes erence favoring the sword will be greater may subsequently modify the preference, in swordtails than in Priapella. Alternative- depending on the nature of genetic varia- ly, if there is a cost associated with ex- tion in the female bias and the fitness con- pressing the preference or if the trait and sequences of such variation. Once a pre- preference become genetically linked and 302 SYSTEMATIC BIOLOGY VOL. 45

800 n trate that in P. olmecae, an unsworded genus, the preference favoring a sword is an

600- increasing function of sword length. PREDICTIONS 400- Once the strength of a preference is determined for a group lacking the favored 200- trait, this preference function can be used to investigate how the preference has changed since the trait appeared. In the

24 26 28 30 32 34 36 38 case of Xiphophorus, female preference functions in swordtails such as X. helleri Sword length (mm) can be determined. If the bias has not been FIGURE 5. Strength of female preference for sword modified subsequent to the divergence of length in Priapella olmecae (from Basolo, 1995b). The Priapella and Xiphophorus, the preference amount of time females spent in courtship was sig- function in Priapella and in swordtails nificantly correlated with male sword length (n = 12, should not differ. However, if evolutionary r = 0.9, P < 0.001). processes have influenced preference evolution after the divergence of Priapella and Xiphophorus, we would predict that the there is a cost to expressing the trait, the preference function in Xiphophorus will dif- bias in Priapella could be stronger than that fer from that in Priapella. in Xiphophorus. One way to quantify the In addition, once a preference has been strength of a female preference is to deter- found in a group lacking the male trait, the mine the preference function (Wagner et strength of the preference can be com- al., 1995), which describes variation in the pared with that in more distantly related strength of a female's response in relation groups. From such analyses, information to expression of the male trait. The steeper can be obtained concerning bias origin and the slope of the preference function, the changes in the strength of the preference more strongly females prefer greater val- prior to trait evolution. With phylogenetic ues of the trait. inference concerning the distribution of the Based on one or a combination of the preference and the strength of the prefer- coevolutionary models, we might predict ence, we can determine whether the initial that the preference functions based on bias has changed since its origin. Thus, male sword length would be stronger in phylogenetic information concerning the swordtails than in Priapella. I examined the strength of a preference can be important relationship between male sword length in examining the origin of a bias and in and the amount of time females spent in examining changes in an initial bias. If courtship with sworded males in P. olmecae such changes are found, phylogenetic in- by attaching artificial swords to conspecific formation may also be useful in generating males. Females were tested using a single hypotheses concerning why such changes stimulus design; in these choice tests, a may have occurred. single male and a companion female were presented to each test female. Individual Possible Costs of Preference Expression females were randomly assigned males of Numerous costs could be associated variable sword lengths, and each test fe- with expression of a preference for a male male and male were only used once. In trait. If the degree of preference expression these tests, the sword preference in P. ol- is positively correlated with search cost mecae was positively correlated with sword (time and/or energy involved in finding a length; as sword length increased, the preferred mate), then searching for a male strength of the preference increased (Fig. 5 with a preferred trait would be costly to [from Basolo, 1995b]). These results illus- the female. In addition, searching could in- 1996 BASOLO—PHYLOGENY OF A PREFERENCE 303 crease predation risk, if females with the troducing coloration to the fin elongation, preference take longer to find mates and a preference for coloration may have thus expose themselves to a greater pre- evolved that strengthened the overall pref- dation risk than do females who do not erence for the sword (this is one of nu- have the preference. merous possible scenarios). By comparing If females with a bias favoring a male female preference functions across taxa for trait are more likely to mate with males various sword components, it may be pos- with the trait once it has arisen, this will sible to determine if some components of lead to a genetic correlation between the the preference coevolved with the sword trait and the preference. If the preference components they favor. In this way, the is open ended, i.e., if females with the bias comparative method presents the potential prefer the most extreme variation of the to track the historical pathway of change traits (in the case of swordtails, longer and in components of female preferences and more colorful swords), arid if the trait is male traits. costly, then the bias may decrease in Whether the preference for the compos- strength as an indirect result. ite sword is general for a structure lacking The examples given here are just several one or several of the components of a com- of the possible ways in which expression plete sword is not known. Likewise, of a preference could carry a cost. By using whether or not this preference will favor phylogenetic information about the character P (called a sword by other re- strength of the preference in sworded and searchers; Meyer et al., 1994) is also un- unsworded species, we can determine known. whether changes in a bias occur prior to the appearance of the male trait it favors, PREEXISTING BIASES AND THE ADAPTIVE as neural network simulations suggest EVOLUTION OF THE SWORD (Enquist and Arak, 1993), and whether Robust phylogenetic information can changes occur after the male trait arises, as provide information about the adaptive suggested by both revolutionary models value of a male trait when it arises. Al- and models in which there is a cost asso- though this point seems obvious, in light ciated with preference expression. of current criticism concerning the useful- ness of using phylogenies to test hypoth- Examining the Bases of Bias eses of adaptation (Reeve and Sherman, The Xiphophorus group appears to be an 1993; Frumhoff and Reeve, 1994), a brief excellent example of the accumulation over discussion relative to sword evolution time of a suite of characters resulting in an seems appropriate. elaborate male trait: the genetic ability to For the genus Xiphophorus, phylogenetic produce a short protrusion arose but was information suggests that the evolution of not expressed, ability was followed by ac- the sword was a result of adaptive selec- tual protrusion expression, and expression tion in the form of female mate choice. was then followed by elongation of varying Phylogenetic hypotheses suggest that the degree with several types of pigmentary preference favoring a sword arose before coloration. By examining the appearance the sword itself appeared. Thus, although of the different components of the sword it is not known whether the ancestral pref- along with variation in the strength of the erence for a sword arose via adaptive pro- preference among taxa, one can examine cesses, nonadaptive processes, or a com- whether the preference was strengthened bination of the two, it does appear that with each additional component and if sword evolution was adaptive and thus and to what degree each of the sword com- that the sword is an adaptation in a sexual ponents contributes to the preference fa- selection context. voring a sword in swordtail species. For Frumhoff and Reeve (1994) identified a example, the initial bias may have favored number of difficulties in using phylogenies fin elongation. Once a mutation arose in- to test hypotheses of adaptation. The fol- 304 SYSTEMATIC BIOLOGY VOL. 45 lowing are addressed here, using the evo- sult of the same preexisting bias and in lution of the sword as an example: (1) the fact may be an identifying characteristic of danger of using parsimony, (2) the prob- male trait evolution in closely related spe- lem of homoplasies and parallel evolution cies when a preexisting bias is involved. of similar traits in related taxa, (3) physi- Third, although in many cases male ological constraints on character evolution, traits may be physiologically constrained (4) the original versus current function of to evolve down one pathway, this does not the character under consideration, and (5) seem to be the case for at least some com- the appropriateness of using phylogenies ponents of the sword. In the case of the to test hypotheses of adaptation when the upper melanophore stripe, there appears selective environment is not expected to to be at least two separate pathways to the have remained invariant. expression of the distal portion of this stripe. First, a danger in using parsimony is In the case of yellow, orange, and/or green that a character that is gained and then re- coloration layered between the two mela- peatedly lost and regained may be incor- nophore stripes, there is variation both rectly determined to be a single gain when within and across swordtail species in the it is present in an extant group. In the case pigment-cell composition of this layer of the sword, this danger is decreased be- (Kallman and Bao, 1987; Basolo, pers. cause we are considering a composite obs.). character. That is, the sword is not a single Fourth, the statement that "reconstruct- character; it consists of component char- ed phylogenies cannot reliably be used ... acters that appear to have accumulated to determine whether a character original- over time. Although other poeciliids have ly spread in ancestral populations through traits that resemble components of the natural selection for its current function" sword, it is unlikely that these are homol- (Frumhoff and Reeve, 1994:172) does not ogous to sword components, and no other appear to be true when considering the poeciliid has a complete sword. However, evolution of male traits via preexisting bi- in one species of swordtail all sword com- ases (using the same assumption as Frum- ponents appear to have been lost, and in hoff and Reeve used, the reconstructed two additional species several sword com- phylogeny closely corresponds to the true ponents appear to have been secondarily phylogeny). If the condition exists in which lost. In the case of the preference, it is not a species has both the trait and preference known whether the preference for the for the trait and a closely related species sword found in four poeciliid species thus lacks the male trait but shares the ancestral far is homologous or not. condition of the preference, then there is Second, Frumhoff and Reeve (1994) ex- good inferential support that the trait pressed concern about the possibility of spread at least in part through sexual se- parallel evolution producing homoplastic lection for its current function (mating suc- but similar traits. However, preexisting bi- cess). ases are expected to favor the independent Fifth, in the case of preexisting biases, evolution of similar traits in different we are considering the association of a groups. Although a number of lineages male trait and a selective component of the may share a preexisting bias for a male biotic environment, i.e., females of the trait, these different lineages are not ex- same species. Although the original envi- pected to experience identical mutations. ronment, i.e., the ancestral females, may Thus, although the genetic blueprint for have had the preference at a frequency of the male trait is not expected to be the <1 and there may have been variation in same in most cases, the selective pressure the strength of the preference, demonstra- via the preexisting bias is expected to be tion of the preference in current females of similar. In this way, male traits that resem- both a species with the male trait and a ble one another but are not homologous sister species lacking the male trait (ances- may be very common when they are a re- tral state) is strong evidence that the male 1996 BASOLO—PHYLOGENY OF A PREFERENCE 305

trait has evolved as an adaptation. In the as preexisting biases. These apparent pre- case of the sword, even if sword preference existing biases may actually be residual or has changed in frequency or strength over retained primitive preferences (Basolo, evolutionary time, the direction of the pref- 1995a), as appears to be the case in at least erence, i.e., for males with swords, has not one study (Ryan and Wagner, 1987) in changed. which preferences for male traits were Consequently, although female prefer- shown in species that had secondarily lost ences based on a preexisting bias favoring the trait. However, knowledge of both the a sword may or may not have an adaptive phylogeny and the distribution of a pref- origin, the condition at the origin of the erence and trait can allow discrimination sword was undeniably adaptive for this of true and apparent preexisting biases. male trait. Thus, in the case of preexisting Once a residual preference has been de- biases, we have a general model to estab- tected, it is possible to test whether the loss lish that a male trait spread at least in part of the trait has been accompanied by a re- for an adaptive function. In the Xiphopho- duction in the strength of the preference, rus group, a source of selection favoring a as would be predicted by a number of trait was present prior to the trait itself and models of sexual selection. For example, a male trait subsequently arose and capi- coevolutionary models suggest that with talized on this preexisting bias. Such adap- the loss of a trait, there would be a corre- tive selection driven by preexisting biases sponding decrease in the strength of the in females may also explain the evolution preference. Alternatively, selection may of male traits in frogs (Ryan et al., 1990), have caused reduction in the preference, jumping spiders (Clark and Uetz, 1992), which in turn could result in the loss of water mites (Proctor, 1992), and Uganda the trait. For example, for Xiphophorus, the kob (Deutsch, 1994). phylogenies suggest that some components of the sword have been secondarily CONCLUSIONS lost. In such cases, we can ask whether the The results presented here confirm the preference has weakened with the loss of idea that a preexisting bias resulting in the sword components. preference favoring a sword is a primitive There are numerous ways in which character for Xiphophorus or some larger knowledge of trait distribution obtained unsworded group of fishes and thus likely from robust phylogenetic inferences can be played an important role in the evolution useful in generating new hypotheses. In of the sword in this genus. As these results the case of preexisting biases, knowledge illustrate, knowledge about phylogenetic concerning the evolution of the sword in relationships can allow us to test hypoth- swordtails allows for hypotheses to be eses that would otherwise be difficult, if tested concerning the evolution of the pref- not impossible, to test. In addition, knowl- erence favoring the sword. By referring to edge of phylogenetic distribution of traits the phylogeny, hypotheses can be gener- may suggest new testable hypotheses. ated and tested concerning the origin of Without knowing that the preference is a the preference, changes in the preference shared ancestral trait for the Xiphopho- over evolutionary time, what specific male rus/Priapella complex, whereas the sword traits the bias has influenced in different is a derived trait for Xiphophorus, we would lineages, and whether the bias has not think to investigate how the preference changed since the appearance of the might have changed since the origin of the sword. male trait. Caution is needed when using the com- ACKNOWLEDGMENTS parative method to attribute a female pref- I thank J. Endler, K. Kallman, D. Leger, and W. E. erence to a preexisting bias model. Unless Wagner, Jr., for helpful conversations on the ideas pre- a robust phylogeny is available, female sented here, W. E. Wagner, Jr., for input on experimen- preferences may be incorrectly categorized tal design and for constructive criticism of this manu- 306 SYSTEMATIC BIOLOGY VOL. 45 script, and D. Cannatella, J. Losos, D. McLennan, and ENQUIST, M., AND A. ARAK. 1993. Selection of exag- an anonymous reviewer for their critical reviews of gerated male traits by female aesthetic senses. Na- this manuscript. This work was supported by NIMH ture 361:446-448. National Research Individual Service Award F32- FISHER, R. A. 1930. The genetical theory of natural MH10089, an Animal Behavior Society Research selection, 2nd edition. Dover, New York. grant, NIMH B-START award MH55288-01, and NSF FRUMHOFF, P. C, AND H. K. REEVE. 1994. Using phy- RPG 9509211. logenies to test hypotheses of adaptation: A critique of some current proposals. Evolution 48:172-180. REFERENCES GORDON, M. 1947. Speciation in fishes; distribution in time and space of seven dominant multiple al- ANDERSSON, M. 1986. Evolution of condition-depen- leles in Platypoecilus maculatus. Pages 95-132 in Ad- dent sex ornaments and mating preferences: Sexual vances in genetics (M. Demerec, ed.). Academic selection based on viability differences. Evolution 40:804-816. Press, New York. GORDON, M., H. COHEN, AND R. F. NIGRELLI. 1943. A ARAK, A., AND M. ENQUIST. 1993. Hidden preferences and the evolution of signals. R. Soc. Philos. Trans. hormone-produced taxonomic character in Platypoe- 340:207-213. cilus maculatus diagnostic of wild Platypoecilus xiphidium. Am. Nat. 77:569-572. ARNOLD, S. J., AND T. HALLIDAY. 1992. Multiple mating by females: The design and interpretation of se- HAAS, V. 1993. Xiphophorus phylogeny, reviewed on lection experiments. Anim. Behav. 43:178-179. the basis of the courtship behavior. Pages 279-288 ATZ, J. W. 1962. Effects of hybridization on pigmen- in Trends in ichthyology (J. H. Schroder, J. Bauer, tation in fishes of the genus Xiphophorus. Zoologica and M. Schartl, eds.). Blackwell, London. (N.Y.) 47:153-181. HAINES, S., AND J. GOULD. 1994. Female platys prefer BARLOW, G. W. 1977. Modal action patterns. Pages long tails. Nature 370:512. 98-134 in How animals communicate (T. A. Sebeok, HALLIDAY, T. 1978. Sexual selection and mate choice. ed.). Indiana Univ. Press, Bloomington. Pages 180-213 in Behavioural ecology: An evolu- BASOLO, A. L. 1990a. Female preference for male tionary approach, 1st edition (J. R. Krebs and N. B. sword length in the green swordtail, Xiphophorus Davies, eds.). Sinauer, Sunderland, Massachusetts. helleri (Pisces, Poeciliidae). Anim. Behav. 40:332-338. HALLIDAY, T., AND S. J. ARNOLD. 1987. Multiple mat- BASOLO, A. L. 1990b. Female preference predates the ing by females: A perspective from quantitative ge- evolution of the sword in swordtail fish. Science 250: netics. Anim. Behav. 35:939-941. 808-810. HAMILTON, W. D., AND M. ZUK. 1982. Heritable true BASOLO, A. L. 1990c. Preexisting biases and the evo- fitness and bright birds: A role for parasites? Science lution of the sword in the genus Xiphophorus. Am. 218:384-387. Zool. 30:80A. HECKEL, J. 1848. Eine neue Gattung von Poecilien mit BASOLO, A. L. 1991. Male swords and female prefer- rochenartigem. Anklammerungs-Organe Sitzber. K. ences. Science 253:1426-1427. Acad. Wiss. Wien. Math. Nat. Cl. 1:289-303. BASOLO, A. L. 1995a. A further examination of a pre- KALLMAN, K. D. 1971. Inheritance of melanophore existing bias favouring a sword in the genus Xi- patterns and sex determination in the Montezumae phophorus. Anim. Behav. 50:365-375. swordtail, Xiphophorus montezumae cortezi Rosen. BASOLO, A. L. 1995b. Phylogenetic evidence for the Zoologica (N.Y.) 56:77-94. role of a pre-existing bias in sexual selection. Proc. KALLMAN, K. D. 1975. The platyfish, Xiphophorus ma- R. Soc. Lond. B 259:307-311. culatus. Pages 81-132 in Handbook of genetics, Vol- BOROWSKY, R. 1984. The evolutionary genetics of ume 4 (R. C. King, ed.). Plenum, New York. Xiphophorus. Pages 235-310 in Evolutionary genetics KALLMAN, K. D. 1989. Genetic control of size at ma- of fishes (B. J. Turner, ed.). Plenum, New York. turity in Xiphophorus. Pages 163-184 in Ecology and BOROWSKY, R. L,. M. MCCLELLAND, R. CHENG, AND J. evolution of livebearing fishes (Poeciliidae) (G. K. WELSH. 1995. Arbitrarily primed DNA fingerprint- Meffe and F. S. Snelson, Jr., eds.). Prentice-Hall, En- ing for phylogenetic reconstruction in vertebrates— glewood Cliffs, New Jersey. The Xiphophorus model. Mol. Biol. Evol. 12:1022- KALLMAN, K. D, AND J. W. ATZ. 1966. Gene and chro- 1032. mosome homology in fishes of the genus Xiphopho- CLARK, D. L., AND G. W. UETZ. 1992. Morph-indepen- rus. Zoologica (N.Y.) 51:107-135. dent mate selection in a dimorphic jumping spider: KALLMAN, K. D, AND I. Y. BAO. 1987. Female het- Demonstration of movement bias in female choice erogamety in the swordtail, Xiphophorus alvarezi Ro- using video-controlled courtship behavior. Anim. sen (Pisces, Poeciliidae), with comments on a nat- Behav. 43:247-254. ural polymorphism affecting sword coloration. J. DARWIN, C. R. 1871. The descent of man, and selec- Exp. Zool. 243:93-102. tion in relation to sex. J. Murray, London. KIRKPATRICK, M. 1982. Sexual selection and the evo- DEUTSCH, J. C. 1994. Lekking by default—Female lution of female choice. Evolution 36:1-12. habitat preferences and male strategies in Uganda KIRKPATRICK, M. 1987a. The evolutionary forces act- kob. J. Anim. Ecol. 63:101-115. ing on mating preferences in polygynous animals. ENDLER, J. A., AND T. MCLELLAN. 1988. The processes Pages 67-82 in Sexual selection: Testing the alter- of evolution: Towards a newer synthesis. Annu. Rev. natives (J. W. Bradbury and M. B. Andersson, eds.). Ecol. Syst. 19:395-421. Wiley, New York. 1996 BASOLO—PHYLOGENY OF A PREFERENCE 307

KIRKPATRICK, M. 1987b. Sexual selection by female omy and evolution of male genitalia in poeciliid choice in polygynous animals. Annu. Rev. Ecol. fishes. Zoologica (NY.) 38:1^7. Syst. 18:43-70. RYAN, M. J. 1990. Sexual selection, sensory systems, KIRKPATRICK, M., AND M. J. RYAN. 1991. The evolu- and sensory exploitation. Oxf. Surv. Evol. Biol. 7: tion of mating preferences and the paradox of the 156-195. lek. Nature 350:33-38. RYAN, M. J., J. H. Fox, W. WILCZYNSKI, AND S. A. LANDE, R. M. 1981. Models of speciation by sexual RAND. 1990. Sexual selection for sensory exploita- selection on polygenic traits. Proc. Natl. Acad. Sci. tion in the frog Physalaemus pustubsus. Nature 34: USA 78:3721-3725. 66-67. LANDE, R. M., AND S. J. ARNOLD. 1985. Evolution of RYAN, M. J., AND W. E. WAGNER, JR. 1987. Asym- mating preferences and sexual dimorphism. J. metries in mating preferences between species: Fe- Theor. Biol. 117:651-664. male swordtails prefer heterospecific males. Science MEYER, A., J. MORRISSEY, AND M. SCHARTL. 1994. Mo- 236:595-597. lecular phytogeny of fishes of the genus Xiphophorus THORNHILL, R., AND J. ALCOCK. 1983. The evolution suggests repeated evolution of a sexually selected of insect mating systems. Harvard Univ. Press, trait. Nature 368:539-542. Cambridge, Massachusetts. MILLER, R. R., AND W. L. MINCKLEY. 1963. Xiphophorus TRTVERS, R. L. 1972. Parental investment and sexual gordoni, a new species of platyfish from Coahuila, selection. Pages 136-179 in Sexual selection and the Mexico. Copeia 1963:538-546. descent of man (B. Campbell, ed.). Aldine, Chicago. WAGNER, W. E. JR., A.-M. MURRAY, AND W. H. CADE. MORIZOT, D. C, AND M. J. SICILIANO. 1984. Gene mapping in fishes and other vertebrates. Pages 173- 1995. Phenotypic variation in the mating preferences of female field crickets (Gryllus integer). Anim. 228 in Evolutionary genetics of fishes (B. J. Turner, Behav. 49:1269-1281. ed.). Plenum, New York. WEST-EBERHARD, M. 1984. Sexual selection, competi- POMIANKOWSKI, A. 1988. The evolution of female tive communication and species-specific signals in mate preferences for male genetic quality. Oxf. Surv. insects. Pages 283-324 in Insect communication (T. Evol. Biol. 5:136-184. Lewis, ed.). Academic Press, Toronto. PROCTOR, H. C. 1992. Sensory exploitation and the WEST-EBERHARD, M., J. W. BRADBURY, N. B. DA VIES, P. evolution of male mating behavior: A cladistic test H. GOUYON, P. HAMMERSTEIN, B. KONIG, G. A. PAR- using water mites (Acari, Parasitengona). Anim. Be- KER, D. C. QUELLER, N. SACHSER, T. SLAGSVOLD, F. hav. 44:745-752. TRILLMICH, AND C. VOGEL. 1987. Conflicts between RAUCHENBERGER, M., K. D. KALLMAN, AND D. C. MOR- and within the sexes in sexual selection: Group re- IZOT. 1990. Monophyly and geography of the Pa- port. Pages 181-195 in Sexual selection: Testing the nuco Basin swordtails (genus Xiphophorus) with de- alternatives (J. W. Bradbury and M. B. Andersson, scriptions of four new species. Am. Mus. Novit. eds.). Wiley, New York. 2974:1-41. WIENS, J. J., AND M. R. MORRIS. 1996. Character def- REEVE, H. K., AND P. W. SHERMAN. 1993. Adaptation initions, sexual selection and the evolution of and the goals of evolutionary research. Q. Rev. Biol. swordtails. Am. Nat. 147:866-869. 68:1-31. WILLIAMS, G. C. 1966. Adaptation and natural selec- ROSEN, D. E. 1979. Fishes from the uplands and in- tion. Princeton Univ. Press, Princeton, New Jersey. termontane basins of Guatemala: Revisionary stud- ZAHAVI, A. 1977. Mate selection—A selection for a ies and comparative geography. Bull. Am. Mus. handicap. J. Theor. Biol. 53:205-214. Nat. Hist. 162:267-376. ZANDER, C. D., AND M. DZWILLO. 1969. Untersuchun- ROSEN, D. E., AND R. M. BAILEY. 1963. The poeciliid gen zur Entwicklung und Vererbung des Caudal- fishes (Cyprinodontiformes), their structure, zoo- fortsatzes der Xiphophorus-Arten (Pisces). Z. Wiss. geography and systematics. Bull. Am. Mus. Nat. Zool. 178:276-315. Hist. 126:1-176. Received 15 September 1995; accepted 1 May 1996 ROSEN, D. E., AND M. GORDON. 1953. Functional anat- Associate Editor: Jonathan Losos