Etolog(a. 3:209-218(1993)

Secondary sexual ornaments as signals: the handicap approach and three potential problems

A. Lotem

Dept. of Zoology, The University of British Columbia,6270 Univ. Boulevard,Vancouver, B.C. CanadaV6TIZ4, e-mail: [email protected]

ABSTRACf. Secondarysexual ornaments as signals:77tehandicap approachand three potential problems.- Secondary sexual ornaments have recently been discussedin the context of biological signals, and the handicapprinciple has been suggestedas a model explaining their . The handicap principle predicts that at equilibrium, sexual ornaments will be honest signals of the male's quality. This is becausethe cost of ornamentsto a potential cheater(a low quality male) will be greater than to an honest signaler (a high quality male), to an extent that makes cheating maladaptive. Accordingly, the cost of the ornament (the handicap) should be related to the quality it reveals. In the following, I discuss three problems with the handicap approach: i) It is difficult to determine the cost of the handicap and the quality it reveals. Nevertheless, I suggest that it is feasible and is worth doing. ii) It is not clear whether phylogenetic data can be used to distinguish between the handicap model and the sensory exploitation model. I suggestthat it can be usedonly underrestricted conditions. ill) Cheating (dishonest signalling) seemsto contradict the handicap model. I will try to show, however, that some forms of cheating can be explained by the handicapmodel.

KEY WORDS. , Signlas, Handicap principle, Sensoryexploitation, Cheating.

Introduction of secondarysexual ornamentsand deal with some related issues. I will first describethe handicap approachand the way I believe it should be used.I The handicap principle (Zahavi, 1975; 1977; will then discussthree problems likely to be fiĀ„::ed 1981; 1987)is a generalmodel for the evolution of when trying to use the handicapapproach. The first biological signals. Yet, it is most frequently problem is how to determinethe specific cost of a discussedin the specific context of mate choice handicap and the quality it reveals, and how to (Zahavi, 1975; Grafen, 1990a; Maynard Smith, determinethat a certain cost really doesfunction as a 1991). According to the handicap principle, handicap. The second problem is whether secondarysexual ornamentsare extravagantsignals phylogeneticdata can be usedto reject the handicap developedby animalsto attract the oppositesex, ard modeland to supportthe sensoryexploitation model like other signals, their evolution may be explained instead.Finally, I will try to show how cheating by the handicapmechanism. In the following, I (dishonest signalling) can be explained by the shall discussthe handicapapproach to the evolution handicapmodel.

209

Presentedin a plenary sessionon the evolution of secondarysexual ornaments at the XXIII InternationalEthological Conference, Torremolinos, Spain. Lotem

The handicap principle: how it not have to be perfectly reliable (perfectly con'elated works with male quality), but only reliable enough (on the average)to benefit femaleswho use them (i.e. to make the response to the signals better than a The reasons that brought many scientists to randomchoice). When femalesprefer reliable signals reject the handicap principle in its earlier days will and ignore unreliable ones, males gain no benefit not be discussed here (see Grafen, 1991, m:rl from using unreliable signals. Most males still Collins, 1993 for discussion on that subject). benefit, however, from using reliable signals to Recently, however, there is an increasing agreement show that they are still better than some other among theoreticians that the handicap principle males. In conclusion, at equilibrium, both males could work, and this has been demonstrated in and females are expectedto use reliable signals. several models (Nur & Hasson, 1984; KO

210 Etolog{a, Vol. 3, 1993

birds, behavioralecologists do not argue that great tits lay an averageclutch size of eight eggs simply becausethis is the optimal clutch size. Instead,they +" ~ use optimization models (with a methodological II c: assumption about equilibrium and optimality) in II .0 orderto explore trade-offsand constraintsthat affect ... 0 clutch size in birds. The question being askedin +" (/I 0 most of thesestudies is not whetherclutch size is u optimal, but why a certain clutch size might be more optimal than others.I believethat in the same way, the handicapprinciple should not be usedonly as an explanation for the existenceof extravagant ornaments,but rather as a model to explore their FIGURE 1. Optimal tai11ength for a low quality male meaning. Obviously, if we assume evolutionary (LQ) and for a high quality male (HQ)when a long tail equilibrium, any extravagant ornament used for is a handicap (see explanations in the text). Cost and communicationmay be a handicap.This is because benefit are measuredusing the same units of fitness the handicapmechanism is the equilibrium model (e.g. numberof offspring). This allows to illustrate the handicap mechanismas in an additive model. It should for the evolution of signals. It would be interesting, be noted, however, that the actual functions creating of course,to know whetherthe ornamentis indeeda the cost and the benefit curves might be multiplicative handicap,as well as whetherclutch size in birds is rather than additive (see, for example, Nur & Hasson, really optimal. However, using these models as a 1984). researchprogram might be more productive than [Longitud 6ptima de cola para un macho ~ baja calidad (LQ) y otro de alta (HQ) cuandouna cola trying to prove that they are completely realistic larga es un handicap.] (see Maynard Smith, 1978). The idea that an ornamentis possibly a handicap,may lead to series This condition createsthe linkage betweenthe cost of questions and hypotheses with testable of the handicap and the quality it reveals as predictions;If an ornamentis a handicap,what type suggestedby Zahavi (1987). of handicapis it? What is the cost of this handicap? What is the quality it reveals?In other words, what is the messageof the signal and why is it so The handicap approach and how to importantfor females? use it Thereare many possible costs for handicapsarxI many qualities they can reveal (Zahavi, 1987); For example, if elaboratedornaments handicap a male's As I have mentioned above, although the ability to escapepredation, the quality they revealis handicap principle is a unique model for the that an ornamentedmale is relatively good in evolution of signals, it is an equilibrium model am escapingpredators. Ornaments that are energetically is based on the optimization approach.Accepting costly to carry may indicate that their owners are this notion, might imply that similarly to the physically strong. Growing extravagantornaments optimization approach(Maynard Smith, 1978), this may be costly becauseit requiresthe investment of principle should be used not only as a post-hoc resourcesthat could be used for other maintenance explanation,but mainly as a modelto pursuefurther activities. In such a case,well developedornaments research.For example,when studying clutch size in indicatethat their ownerswere in good condition at

21 Lotem

the time they grew them, and were able to recruit function of a dietary intake and is correlatedwith the neededresources. These kind of handicaps male over winter survival and parentalcare activity. provide females with information on the male's However, although it appearsthat ornamentation history, which might be important if variations in was correlatedwith the males' general quality, the male quality are mostly pronouncedunder the harsh specific cost which createsthis correlation is not environmental conditions that precedethe breeding clear. Is it the cost of foraging for the required season.Another possible cost of ornamentsis their pigments, the cost of defendingfood resourcesthat vulnerability to parasites(Hamilton & Zuk, 1982). are rich in pigments, or the cost of expressingthe In sucha case,ornamented male may advertisetheir pigment (that may increase risk, or relative resistanceto parasites.The cost of having amplify information about certainqualities). long ornaments may also be that they are more An additional indirect evidencefor the cost of likely to be damagedduring fights betweenmales. If secondarysexual ornaments comes from a recent so, only good fighters can afford to grow long comparativestudy (promislow et al., 1992), which ornaments.The cost of a handicapneed not be a indicatesthat mortality cost may limit the evolution very direct one. Ornamentsor color patternmay not of male brightnessin passerinebirds. However,rere be costly by themselvesbut may clarify or amplify again the specific cost of brightnessthat eventually information about quality (Zahavi, 1979; 1987; affects mortality is unknown. Hasson, 1989; 1991). The cost of the ornamentin To test the specificcost of a handicapdirectly, an this caseis due to the information it provides.Low experimentalmanipulation is required.In numerous quality males may loose from showing their exact studiesmale ornamentswere manipulated to test for quality. femalepreferences. However, in only a few of these The examples above are probably not the only has the cost of such a manipulation been measured. possible ones. With a little bit of imagination mrl M~ller (1989) has shown that male barn swallows with some understandingof the animal's , with experimentally elongated tail ornaments other good hypothesescan be generated.The first captured smaller, less profitable prey than those problem, however,is how to test thesehypotheses. with shortenedtails, and were more likely to have fault bars in their feathers(an indication of food deficiency during molt). Evans and his colleagues Problem I: How to determine the went even one step further. In a first experiment cost of a handicap Evans& Hatchwell (1992) showedthat the long tail of a male scarlet-tuftedmalachite sunbird handicaps its ability to fly and its efficiency at catching ~al The fact that some sexually selectedornaments insects.To test the cost of the handicapeven more appear to be condition dependent(Endler, 1980, specifically, a secondexperiment was conducted.By 1983; Evans, 1991; Hill, 1991; 1992)suggests that adding extra mass on legs of some males aID they are indeed costly, becauseotherwise, males manipulating tail length of others, Evans & should grow them, or express them, under any Thomas (1992) were able to show that the circumstances.The problem, however, is that this aerodynamiceffect of a long tail, not its extra mass, evidenceis only indirect, and not very specificas to is the cost of the handicap. what the exact cost might be. For example, Elegant experiments of this type demonstrate Geoffery Hill (1991; 1992) showed that female that the specific cost of a handicapcan be tracked. house finches prefer to mate with brightly colored But, do they really show that the observedcost is males. Male plumage brightness,in this case, is a the one that functions as a handicap(the one which

212 Etolog(a, Vol. 3, 1993

createthe correlationbetween the ornamentand the quality)?To confirm that this is the case,one hasto show quantitatively that the observedcost is high enough to make cheating maladaptive.To explain ... ~ this point, let us imagine a case, illustrated by ., c figure 2, in which carrying a long tail has been ., .Q ... found to be energeticallycostly, and, as expected, 0 this cost is relatively high for low quality males. ... ~ However, when comparing the energeticcost to the tJ benefit gained by mating success,it is found that the observedtail lengths in the natural population fall far below the optimal tail length predictedunder the assumptionthat the cost of the tail is energetic. The conclusion in such a caseis that energeticcost alone cannot accountfor the honestyof the signal, and additional costs should be searchedfor. The FIGURE2. Expectedoptimal tail length of low quality (LQ) and high quality (HQ) males, based on the generalconclusion here is that although a cost of assumption that the cost of the tail is energetic, in any type is likely to be higher for low quality relation to the observed tail length of these males in individuals, it can be regardedas a handicaponly if .The hypothetical case illustrated here suggests it is high enoughto limit ornamentsize. that energetic cost alone cannot account for the A major difficulty when attemptingto determine honesty of long tails as a signal. To simplify the illustration, the energetic cost, and the benefit of a tail the cost of a handicapis to resolve casesin which are transelated into the same units of fitness (e.g. no apparentcost is found. The problem with such number of offspring), allowing to present the case as negativeresults is that it is virtually impossible to in an additive model (seecomment in legend to Fig. 1). prove that there is no cost (i.e. to prove that [Longitud 6ptima de cola esperada para something does not exist). One can always argue machos de baja (LQ) yalta (HQ) calidad, basado en la premisa de que el costo de la cola es energetico, en that the researchershave simply failed to find the relaci6n a la longitud de la cola de esos machos right cost, but perhaps one will be found in the observadaen la naturaleza.] future. If so, is there any way to rejectthe handicap model and to support an alternative explanation exploit this pre-existingpreference of females.This instead? model is commonly known as the "sensory exploitation model", although pre-existing preferencesmay not necessarilybe basedon sensory Problem II: Can phylogenetic data biases (A. Basolo personal communication). be used to reject the handicap According to the sensoryexploitation model, male principle and to support the sensory ornamentsmay not be a handicapand may carry no exploitation model instead? message about a male quality. The ornaments evolved simply because females preferred ornamentedmales regardless of their quality. Basolo (1990) and Ryan et al. (1990) have There aretwo aspectsto the sensoryexploitation suggestedthat in somecases the evolution of female model. The fIrst is the model itself, and the second preferencepreceded the evolution of the male trait is the use of phylogeny to test it (Basolo, 1990; (ornament),and that the male trait has evolved to Ryan et al., 1990; Ryan & Keddy-Hector,1992).

213 Lotem

The model itself is a good alternative scenariofor the evolution of male ornaments.The difficulties p - p- PT PT PT arise when trying to distinguish between the \ \ alternativemodels. The first problem, as pointedout \. by Balmfordand Read (1991), is that a trait that first evolved to exploit a pre-existing preferencemay be exaggeratedlater on, becomecostly, andturn into a handicap(see also Basolo, 1990; Ryan et al., 1990). Accordingly, even if we acceptthat in the past the Genet~ change b Genetk: change trait was not a handicap, we cannot exclude the a. Is rare 0 Is frequent possibility that it doesfunction as a handicapat the present. The second problem, which I discuss below, is whetherthe use of phylogeny can really FIGURE3. Alternative models to explain a phylogeny distinguishbetween the alternativemodels. in which females of the least derived species have a Using reconstructedphylogeny and experiments preference"P" for a male trait "T" which occurs only in the more derived species. a) Assuming that genetic on female behavior, Basolo (1990) and Ryan et al. changeis rare (i.e. parsimony), the preferenceevolved (1990) showedtwo clades(one of fish and a second before the trait. b) Assuming that genetic change is of frogs) in which females of the least derived frequent, the preference and the trait may have speciesprefer a male trait which occursin the more coevolved but the trait has subsequently been lost, derivedspecies but not in their own species(see fig. probably due to selection (see explanations in the text). 3). They suggestedthat the most parsimonious [Modelos alternativos para explicar una explanation for such a phylogeny is that the filogenia en la quelas hembrasde las especiesmenos preferenceevolved beforethe trait (fig. 3a). Basedon derivadastienen una preferencia"P" por una rasgo00 parsimony, the alternativepossibility, that the trait macho"Y' queocurre solamente en las especiesmas and the preferencecoevolved together but that the derivadas.] trait has been lost, should be rejectedbecause it and lessparsimonious scenario, is more likely to be requires more genetic changes(fig. 3b). There is the correct one. Many break downs have probably evidence, however, that in some cases male occurredduring the years,but the car has beenfixed. ornamentshave beenlost while femaleretained their Using parsimony in this case is improper because preference(McPhail, 1969; Basolo, 1991; Hill in the concept of parsimony (Felsenstein, 1983) press.). This might be expectedif the trait was a assumesthat changesin the system are rare, while handicap and became too costly under some in fact, we know that many "changes"are expected circumstances(Schluter & Price, 1993; Hill in to occur in a car within a period of 20 years. In press). Accordingly, the alternative scenario (Fig. conclusion, the critical information required to 3b) is possible, but it is only less parsimonious. distinguish between the two alternatives is the The use of parsimonyis thereforethe sole criterion frequency of changes. Because the concept of to prefer one model over the other. parsimony assumesalready that this rate is low, it Although the concept of parsimony is widely is not an independentcriterion that can test which of accepted,inadequate use of parsimony may lead to the alternativesis correct. unrealistic conclusions. For example, based on Let us return now to sexualselection and the two parsimony, one can arguethat a 20 yearold car can alternative models illustrated by figure 3. If we still run properly becauseit has neverbroken down. assume that the genetic change (the mutation) It is quite obvious, however, that the alternative, requiredfor creating a male trait is very rare (fig.

214 Etolog(a, Vol. 3, 1993

3a.), then it is likely that it OCCUlTedonly once, arKi that the males of the least derived species are Cost experiencing a long evolutionary lag: the trait is old-LQ potentially adaptive (females prefer it) but it has not LQ :HQ yet evolved. If, on the other hand, we have reasons ~ v Benefit c to believe that these mutations are relatively v ~ common, then, it seems incongruent that during a .. 0 long period of time this potentially adaptive ~ mutation occurred frequently but never spread in the C.I f"1 population. In this case the simplest explanation is that the trait is too costly for the males of that species, and was selected against (fig. 3b). In other ~~~ words, the absence of male trait is adaptive arKi represents an equilibrium state. Similar to the FIGURE4. "Cheating" by an old male: If the cost of a example above, the critical information needed to given tail length decreasesin old age, the optimal tail distinguish between the alternatives is the frequency length for an old low quality male (Old LQ) may be of changes. The concept of parsimony already higher than that of a low quality young male (LQ) and assumes that this frequency is low, but this may not similar to that of a high quality young male (HQ). See text for more explanations. To simplify the necessarily be the case. illustration, cost and benefit are measuredusing the In contrast with the concept of parsimony that same units of fitness (e.g. number of offspring), minimizes the rate of genetic change (Felsenstein, allowing to present the case as in an additive model 1983), optimization could occur only if there is (seecomment in the legend to fig. 1). ["Trampa" de un macho viejo: si el costo 00 sufficient genetic variance (Maynard Smith, 1978). una longitud dadade cola decreceal aumentarla edad,la Hence, these two legitimate approachesare based on longitud 6ptima para un macho viejo de mala calidad opposite assumptions. Considering this, it is not (Old LQ) puede ser superior a la de uno j6ven de baja surprising that parsimony supports the evolutionary calidad(LQ) y similar a uno j6ven de alta calidad (HQ).] lag scenario (fig. 3a) and rejects the equilibrium scenario (fig. 3b). The handicap model which is over the handicapmodel, should be basedon some based on equilibrium, is therefore doomed to be knowledgeof the rate of geneticchange, rather than rejected by parsimony under many circumstances. on parsimonyand its assumptionof low rates. Parsimony cannot be used as an independent criterion to decide between equilibrium models arKi models that assume an evolutionary lag. It can Problem III: How cheating can be only be used in cases where the rate of genetic explained by the handicap model change is known to be a limiting factor, and in such cases, equilibrium models are questionable anyhow. It is important to note that the sensory When explaining the handicapprinciple earlierin exploitation model itself does not dependon a low this paper, I emphasizedthat the cost of the rate of genetic change. Rather, it is the phylogenetic handicapmakes cheating maladaptive.If so, how trees used to support the model that requires can we explain evidence for cheating? In the parsimony. I am not therefore arguing against the following, I will try to show that accordingto the sensory exploitation model itself. I am, however, handicapprinciple cheatingmay evolve and may be suggesting that the use of phylogeny to support it stable, but only under certain circumstances.The

215 Lotern

concept is applicable for any signalling system The important conclusionsfrom this exampleare basedon handicaps,but will be discussedhere only that asymmetriesin the cost of the handicapallow in the context of secondarysexual ornaments. for somelevel of cheatingin a system of handicaps. In caseswhere the male ornamentis a long tail and that this asymmetry determines which (fig. 1), cheats would grow tails longer than the individuals can be cheatersand which are honest averagetail length of malesof the samequality. In signalers. general, this will not be possible becauseall the males of a certain quality are affectedby the same cost-benefit balance. Yet, cheating might be Conclusions possible if thereis an asymmetryin the cost-benefit balance imposed on different males of the same quality. A possible examplefor such a casehas been The handicapprinciple is suggestedfor use not recently suggestedby Oren Hasson(in press) and I only as an explanation for the evolution of will further discussit here. secondarysexual ornaments,but mostly as a model It is assumedthat one componentof the cost of a to explore their meaning. Hypothesesregarding the long tail is reducingmale probability to survive to specific cost of a handicapand the quality it reveals the next breedingseason. This component is less can be tested experimentally. However, a important for old malesin their last breedingseason quantitative cost-benefit analysis is required to (becausein the suggestedscenario the chancesof old confirm that an observedcost really doesfunction as males surviving to the next year are very small). a handicap.Using phylogeny to reject the handicap Hence, there is an asymmetry of costs between model and to support the sensory exploitation young and old males. In this casethe optimal tail model, can be doneonly when there are reasonsto length of an old low quality male can be similar or believe that the rate of genetic changeis low. In evenequal to that of a high quality young male (fig. such a case, however. the handicap model that 4). If females chose males based on tail length assumes equilibrium is questionable anyhow. alone, a femalethat picked up an old male, chosea Finally, even at equilibrium, asymmetries in the male of lower quality than the averagequality of cost of the handicapbetween individuals of the same males with similar tail lengths. We can then argue quality allow for somelevel of cheatingin a system that this femalehas beencheated. It shouldbe noted of handicaps. that the averagequality of males with a given tail length, and therefore the degree of cheating, is affectedby the relative proportion of old and young Resumen males in the population. This kind of cheatingwill be stable if: a) It is too costly for females to discriminate betweenold and young males (if they Ornamentossexuales secundarioscomo selial: fa can discriminate,cheaters will be selectedagainst). aproximaci6n del handicap y Ires problemas b) Femalesstill benefit on averagefrom choosing potenciales. males based on tail length, despite occasional Los ornamentos sexuales secundarios se han cheating(otherwise they should not usethe signal at discutido recientementeen el contexto de sefiales all). c) Long tails are the best signal available to biol6gicas, y se ha sugerido el principio reI distinguish between males of different quality (If handicap como un modelo explicativo de su more reliable signals are available, females should evoluci6n. El principio del handicappredice que en usethem instead). el equilibrio, los ornamentossexuales seran sefiales

216 Etolog(a, Vol. 3, 1993

honestas de la calidad del macho. Esto es debido a preference.Science, 253:1426-1427. que el costo de los ornarilentos para un tramposo Collins, S., 1993. Is there only one type of male potencial (un macho de baja calidad) sera mayor ~ handicap?Proc.R. Soc. Lond. B,252:193-197. para un senalizador honesto (un macho de aIta Endler, J.A., 1980. Natural and sexual selection on calidad), hasta el punto de que haga la tampa color patterns in Poecilia reticulata. Evolution, maladaptativa. De acuerdo con esto, el costo reI 34:76-91. ornamento (el handicap) deberfa estar relacionado con Endler, J.A., 1983. Natural and sexual selection on la calidad que manifiesta. Se discuten tres problemas color patternsin poeciliid fishes. Environ. Biol. con la aproximaci6n del handicap: i) Es diffcil Fishes,9: 173-190. determinar el costo del handicap y la cali dad ~ Evans,M.R., 1991. The size of adornmentsof male revela. No obstante, se sugiere que es factible y scarlet-tufted malachite sunbirds varies with merece la pena que se haga. ii) No esta claro si los environmental conditions, as predicted by datos filogeneticos se pueden usaf para distinguir handicaptheories. Anim. Behav., 42:797-803. entre los model os del handicap y de la explotaci6n Evans, M.R. & Hatchwell B.J., 1992. An sensorial. Se sugiere que s6lo se puede hacer bajo experimental study of male adornment in the condiciones restrictivas. iii) La trampa (senalizaci6n scarlet-tuftedmalachite sunbird: II. The role of deshonesta) parece contradecir el modelo reI the elongated tail in mate choice and handicap. Se tratara de mostrar, sin embargo, ~ experimental evidence for a handicap. Behav. algunas formas de engano pueden seTexplicadas POT Ecol. Sociobiol., 29:421-427. el modelo del handicap. Evans, M.R. & Thomas A.L.R., 1992. The aerodynamicand mechanicaleffect of elongated tails in the scarlet-tufted malachite sunbird: Acknowledgments measuringthe cost of a handicap.Anim. Behav., 43:337-347. Fe1senstein,J., 1983. Parsimony in systematics: I thank my mentor, , for teaching biological andstatistical Issues. Ann. Rev. Ecol. me the handicap principle, P. Abrams, M. Syst., 14:313-33. Adamson, G. Hill, J. Hoglund, I. Jones, and J. Godfray, H.C.J., 1991. Signalling of need by Smith for suggestions,and A. Basolo, O. Hasson, offspring to their parents.Nature, 352:328-330. D. Westcott, D. Schluter, and A. Zahavi for Grafen, A., 1990a. Sexual selection unhandicapped commentson the manuscript. by the Fisher process.J. Theor. Biol., 144:473- 516. Grafen,A., 1990b. Biological signals as handicaps. References J. Theor. Biol., 144:517-546. Grafen, A., 1991. Modelling in behavioural ecology. In: Behavioural Ecology: aJ Balmford, A. & Read, A.F., 1991. Testing evolutionaryapproach: 5-31 (J.R. Krebs & N.B. alternative models of sexual selection through Davies,Eds.). Oxford: Blackwell, femalechoice. Tree,6:274-276. Hamilton, W.O. & Zuk, M., 1982. Heritable true Basolo, A.L., 1990. Female preferencepredates the fitness and bright birds: A role for parasites? evolution of the sward in swordtails. Science, Science,218:384-387. 250: 808-81 O. Hasson, 0., 1989. Amplifiers and the handicap Basolo, A.L.. 1991. Male swards and female principle in sexual selection: a different

217 emphasis.Proc. R. Soc. Lond. B, 235:383-406. Pomiankowski, A.N., 1987. Sexual selection: the Hasson, 0., 1991. Sexual displays as amplifiers: handicap principle does work - sometimes.Proc. practical exampleswith an emphasison feather R. Soc. Lond. B, 231:123-145. decorations.Behav. Ecol., 2:189-197. Pomiankowski, A.N., Iwasa, Y. & Nee, S., 1991. Hasson, 0., (in press). Cheating signals. J. Theor. The evolution of costly mate preferences: I. Bioi. Fisher and biased mutations. Evolution, Hill, G.E., 1991. Plumage coloration is a sexually 45:1422-1431. selected indicator of male quality. Nature, Promislow, D.E.L., Montgomerie, R. & Martin, 350:337-339. T .E., 1992. Mortality costs of sexual Hill, G.E., 1992. The proximate basis of variation dimorphism in birds. Proc. R. Soc. Lond. B, in carotenoid pigmentation in male house 250:143-150. finches.Auk, 109:1-12. Ryan, M.J., Fox, J.H., Wilczynski, W. & Rand, Hill, G.E., (in press). Geographicvariation in male S.A., 1990. Sexual selection for sensory ornamentationand female mate preferencein the exploitation in the frog Physalaemus pustulosus. house finch: a comparative test of models of Nature, 343:66-67. sexualselection. Behav. Ecol. Ryan, M.J. & Keddy-Hector, A., 1992. Directional Iwasa, Y., Pomiankowski, A. & Nee, S., 1991. patterns of female mate choice and the role of The evolution of costly matepreferences: II. The sensory biases. Am. Nat., 139:S4-S35. handicapprinciple. Evolution, 45:1431-1442. Schluter, D. & Price, T., 1993. Honesty, Jennions, M.D., 1993. Female choice in birds and perception, and population divergence in sexually the cost of long tails. Tree,8:230-232. selected traits. Proc. R. Soc. Lond. B, 253:117- Johnstone, R.A. & Grafen, A., 1992. Error-prone 122. signalling. Proc. R. Soc. Lond. B, 248, 229- Zahavi, A., 1975. Mate selection - a selection for a 233. handicap. J. Theor. Bioi., 53:205-214. Kodric-Brown, A. & Brown, J.H., 1984. Truth in Zahavi, A., 1977. The cost of honesty (further advertising: the kind of traits favoredby sexual remarks on the handicap principle). J. Theor. selection.Am. Nat., 124:309-323. Bioi., 6:146-151. McPhail, J.D., 1969. Predationand the evolution of Zahavi, A., 1978. Decorative patterns and the a stickleback (Gasterosteus).J. Fish. Res. Bd. evolution of art. New Scientist, 19:182-184. Cant:r:/a26:3183-3208. Zahavi, A., 1981. Natural selection, sexual Maynard Smith, J., 1978. Optimization theory in selection and the selection of signals. In: evolution. Ann. Rev. Ecol. Syst., 9:31-56. Evolution today. Proc. of the second Inter. Maynard Smith, J., 1991. Theories of sexual Congo of Syst. and Evol. (Scudder, G.G.E. & selection.Tree, 6: 146-151. Reveal, J.L., Eds.). Pittsburgh: Carnegie-Mellon M~lIer, A.P., 1989. Viability costs of male tail Univ. ornamentsin a swallow. Nature,339:132-135. Zahavi, A., 1987. The theory of signal selection arxI Nur, N. & Hasson,0., 1984. Phenotypicplasticity some of its implications. In: International and the handicapprinciple. J. Theor. Bioi., 110, Symposium of Biological Evolution, (U.P. 275-297. Delfino, Ed.). Bari: Adriatica Editrice.

(Recibido:enero 1994)