Behavioral Ecology Vol. 11 No. 4: 405-410
© 2000 International Society for Behavioral Ecology
The evolution of courtship rituals in monogamous species
Department of Zoology, University of Stockholm, S-106 91 Stockholm, Sweden
Address correspondence to C.-A. Wachtmeister. E-mail : carladam{at} zool.su.se .
Received 8 February 1999; revised 26 October 1999; accepted 10 November 1999.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONSimulation modelRESULTSDISCUSSIONREFERENCES |
|---|
In this paper we propose an alternative explanation for the evolution of courtship rituals in monogamous species. We demonstrate, using computer simulations, how male courtship might develop as males exploit response biases in females to manipulate the female into starting reproduction before she has been able to assess the male's intentions. In our coevolutionary simulations, a recurrent, artificial neural network is used to model the female recognition mechanism, while the displaying male is represented by a sequence of signals. Our particular model situation is just one example of how a reproductive conflict could result in the evolution of ritualized displays in monogamous species. Since reproductive conflicts occur even after pair formations, the explanation we propose may also apply to rituals that occur after pair formation.
Key words: artificial neural networks, courtship, ritualization, coyness, manipulation, mate choice, monogamy, reproductive conflict, receiver bias, sexual selection.
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONSimulation modelRESULTSDISCUSSIONREFERENCES |
|---|
Avariety of signals and behaviors are used in interactions with the opposite sex. In monogamous species these rituals may occur over an extended period of time. Many of these interactions are identified as courtship and are today often regarded as means of assessing partner quality (Andersson, 1994
;
Trivers, 1972). For instance,
partner quality may be revealed because high-quality males can afford to use
more elaborate displays or because a direct relationship exists between
quality and performance (see, e.g.,
Andersson, 1994). Older
ethological explanations of courtship rituals have stressed the importance of
stimulation and cooperation, rather than choice of partner. For instance,
Huxley (1914 : 516) proposed
that these ritualized displays serve "to keep the two birds of a pair
together, and to keep them constant to each other" (see also
Armstrong, 1963). A similar
idea is that courtship rituals provide necessary stimulation and that they
coordinate the reproductive physiology of the male and the female
(Bastock, 1967 ;
Lehrman, 1959,
1964). The elaboration of
displays was considered to reduce ambiguity
(Cullen, 1966).
In this paper we describe a new hypothesis for the evolution of rituals
within a monogamous pair. Here the courtship display is not informative to the
female apart from indicating the presence of a male of her own species. Our
hypothesis accords with classical ethology (theories of ritualization ; see,
e.g., Eibl-Eibesfeldt, 1975)
insofar as the emphasis is on stimulation, but it rests on reproductive
conflict and manipulation rather than on cooperation. In monogamous species
there are a variety of such reproductive conflicts, many concerning parental
duties. To illustrate how this can lead to evolution of elaborated display and
manipulations, we developed a model. Consider a female that has to find a
partner in an environment of both faithful and philandering males. These males
differ in the amount of parental care that they provide. Males are eager to
court any female and start reproduction as soon as they meet a female.
Females, on the other hand, might benefit by staying coy and evaluating the
intention of a potential partner before actual reproduction, even if this is
associated with some costs (Wachtmeister
and Enquist, 1999). Our suggestion is that complex rituals may
evolve as a male attempts to manipulate the female's decision process to start
reproduction earlier than is optimal for her.
For manipulation to be possible, biases must exist in receiver's behavior
so that it is possible for a sender to elaborate on existing stimuli and
thereby elicit responses more favorable to the sender. There is ample evidence
that receivers are biased in this way ; the classical evidence is from studies
of supernormal stimuli in ethology (e.g.,
Tinbergen, 1951) and peak
shift in learning psychology (e.g.,
Mackintosh, 1974). More
systematic studies have recently been introduced studying biases in related
species and taking phylogeny into consideration
(Basolo, 1990 ;
Ryan et al., 1990). The latter
studies have shown that the same bias may exist in related species, although
it has only been exploited in some of them. That receiver biases can drive
evolution has been suggested many times
(Dawkins and Krebs, 1978,
Enquist and Arak, 1993 ;
Holland and Rice, 1998 ;
Krakauer and Johnstone, 1995 ;
Krebs and Dawkins, 1984 ;
Leimar et al., 1986 ;
O'Donald, 1977 ;
Ryan, 1998 ;
Staddon, 1975).
It is difficult to account for the possibility of manipulation with traditional modeling techniques such as game theory because the reaction to novel stimuli is problematic. The reason for this is that such reactions are not shaped by selection with any precision. Instead, reactions to novel stimuli are determined largely by the mechanism that has been established to deal with familiar stimuli. Thus, to study manipulation we need knowledge about how real biological recognition systems operate, and we need suitable models of such systems.
In this study we used an artificial neural network to model the female's
behavioral mechanism (see Enquist and
Arak, 1998 ; Ghirlanda and
Enquist, 1998, for information about artificial neural networks as
models of stimulus control). The displaying male is represented by a sequence
of visual signals. In one set of simulations the male can use different
signals in the sequence, and in another set he is limited to repeating a
single signal. Neural network models have been used earlier to study the
evolution of signals (e.g., Enquist and
Arak, 1993, 1998)
and have revealed some basic evolutionary relationships between conflicts of
interests, ornamentation, and manipulation. The current study is technically
different from our earlier simulations in that the network now has internal
feedback loops and thereby can react to a sequence of stimuli.
|
Simulation model |
|---|
|
TOPABSTRACTINTRODUCTIONSimulation modelRESULTSDISCUSSIONREFERENCES |
|---|
In this section we introduce the female and male player and continue with an explanation of the game. We finish with a description of the evolutionary algorithm.
An artificial neural network models the female recognition mechanism. This
network has three layers and is partially recurrent
(Hutchinson, 1994). The input
layer consists of seven cells, the hidden layer of four cells, and the output
layer of four cells (Figure 1). Three feedback loops from three cells in the output layer provide input to a
corresponding number of cells in the input layer. These input layer cells
receiving feedback are referred to as context unit cells. These context cells
provide the network with a memory of the past and recurrent networks of this
architecture can thus be used to recognize signal sequences
(Hertz et al., 1991).
|
The different displays used by a courting male are represented by a sequence of signals (one signal for each time step). Each signal in the sequence consists of four components. Each component activates a single input layer cell in the network with a value between 0 and 1 (Figure 1). Two types of sequences are considered. In one set of simulations we allow the male to use different signals, referred to as a variable sequence. In the other set males can only repeat a single signal, referred to as a uniform sequence. A male uses the same signal sequence independent of whether he is faithful or philandering.
The network receives input from a sequence of signals. Four of the input
cells are activated by the signal. The three others are each activated by one
of the three output layer cells involved in the feedback system. Each cell in
the middle layer receives input from each cell in the first layer, and each
output cell receives input from each middle layer cell. Any middle or output
layer cell, j, produces an output, s, according to the
sigmoid function
 |
In all, the probability of response depends on the properties of the
signal, the state of the network, and on an additional motivational factor.
This motivational factor varies according to a normal distribution with mean 0
and a standard deviation of 0.01. If the sum of the network's output and the
internal factor is greater than a threshold of 0.5, the receiver reacts. The
varying internal factor means that a particular stimulus and network state
produce slightly different responses at different occasions. For more
information about how artificial neural networks can be used for modeling
receivers, see, for example, Enquist and Arak
(1993,
1998).
A single female regularly meets potential mates that are either faithful or philandering. The faithful male remains with the female after reproduction. The philandering male stays with the female until she either starts to reproduce or she exposes and rejects him. A female that rejects a philandering male will search for a new mate.
Let x be the duration of female coyness ; that is, the time that a
female spends with a particular male before she starts reproducing. The time
that has passed from the start of the season is denoted t. Both
x and t are measured in discrete time units. After the male
and female pair, the male stimulates the female, using his particular
courtship sequence. A female can only reproduce once in a season, and her
reproductive success declines exponentially (rate r) with time of
onset of reproduction in the season. Female reproductive success also depends
on whether she reproduces with a faithful or a philandering male. If the male
is a philanderer, female reproductive success is devalued with a factor
.
A female may be unpaired, paired, or reproducing. An unpaired female finds
a male with probability (Pc + Pd)
during a time unit, where Pc and Pd
are the probabilities of meeting a faithful or philandering male,
respectively. An unpaired female always pairs with an encountered male. If the
male is a philanderer, she will detect this with a probability q each
time step. How is it possible for a female to detect a philanderer ? First, a
philanderer might expose himself by not being able to spend as much time with
the female as a faithful male (e.g.,
Stenmark et al., 1988).
Second, the primary female of an already mated male that philanders might
intervene or be detected (Breiehagen and
Slagsvold, 1988). Once the female has detected a philanderer, she
rejects the male and starts searching for a new one. The longer the female is
paired with a male, the more certain she is of his intentions. The probability
of being paired to a faithful male increases with the time the female spends
with the male without rejecting him. Details depend on the possibility to
expose philanders (i.e., the detection rate, q) as well as on the
proportion of philanderers in the male population,
Pd/(Pc + Pd).
Every time step that a female is unpaired, she is stimulated by a uniform
sequence in which each component is set to 0. This stimulus indicates that no
conspecific male is present.
In all coevolutionary simulations, both the network population and the
signaler population is fixed at 1,000,000 individuals each, and the parameters
are set to q = 0.2, r = 0.03, = 0.15,
(Pc + Pd) = 0.6, and
Pd/(Pc + Pd) =
0.8. In each population there can be a maximum of 20 types of individuals. All
simulations start out with a male and a female population, each consisting of
a single type referred to as the starting male type and the starting female
type, respectively. Independent of whether the simulation allows males to use
a variable or a uniform sequence, the starting male type is a uniform signal
sequence with all signal components set at 0.5. The starting female type is a
network that responds optimally to this uniform male type. This starting
female type was produced by letting the network evolve in a simulation where
the starting male signal was kept unchanged until the optimal response was
reached. During the simulation each type in the population contributes to the
next generation in proportion to their frequency and their fitness (see
below). Each simulation consists of 60,000 generations.
In each generation, provided that there are fewer than 20 types, a maximum of 2 individuals in the population is mutated. Each mutated individual becomes a new type initially represented by a single individual.
When a male with a variable sequence mutates, there is a probability of 0.1
that each signal in the sequence is modified. A signal chosen for modification
can either be exchanged for another signal in the sequence (probability 0.99)
or be slightly altered (probability 0.01). In the latter case each signal
component is changed with a probability of 0.25 by adding a normally
distributed increment (µ = 0, 
= 0.1) given that the component
remains between 0 and 1. In males with a uniform sequence, the signal was
mutated by letting each component change with a probability of 0.25 by adding
a normally distributed increment (µ = 0, = 0.1), given that the
component remains between 0 and 1. Females are mutated by modifying each
connection weight with probability of 0.045 by adding a normally distributed
increment (µ = 0, = 0.05)
To compute the average coy response of a female of a particular type
i, we first calculate the distribution of coy responses,
vij, toward a given male type j. We then sum over
all male types weighting with their frequencies, mj, in
the male population.
 |
The fitness of a female type depends on whether the male is faithful or
philandering and when reproduction starts in the season and is calculated as
 |
(0 < < 1). Reproducing without any male renders no fitness to
the female. The value of reproducing declines exponentially with time in
season at rate r (see above).
For males we used a fitness function that declines with duration of
coyness. Using the distributions of coy responses from above and the
distribution of female types f, we define male fitness as
 |
This expression of fitness is somewhat simplified but serves the purpose of favoring in evolution male types that can persuade females to start reproducing earlier.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONSimulation modelRESULTSDISCUSSIONREFERENCES |
|---|
Before starting any simulations, we checked whether a female (i.e., the network) is able to evolve an optimal coy response to a courting male when the male is not allowed to change the signal sequence. As shown in Figure 2, the female always evolves the optimal response, as compared with results from an analytical model (Wachtmeister and Enquist, 1999
). In both simulation types, new, more stimulating male
displays repeatedly emerged, causing females to start reproducing earlier than
optimal. In response, the female continuously evolved a reduced sensitivity to
the new successful displays.
|
During all five simulations with a variable male sequence (Figure 3), females typically stayed coy between five and six time steps, although optimal response is eight time steps (Figure 4a). The average coyness is 5.6 (excluding the first turbulent 5000 generations). Studying the results in closer detail, we notice that shorter or longer intervals of no or small changes of stability appear between the more turbulent periods (Figure 4b). This indicates that it takes the evolutionary processes some time to come up with better solutions both in males and females. We also see in Figure 4b that sometimes females exist that are coyer than the optimal duration of coyness (i.e., super coy females). These females have replaced less fit sub-coy females. Excluding the first 5000 generations, the average rate of change in males is 2.1 and in females 1.9. Rate of change is measured as the rate of replacement of the most common type in a 5000-generation interval.
|
|
Limiting the males to use only a uniform sequence (Figure 3) results in a drastically reduced success at manipulating a female's reproductive decision. In five simulations of the co-evolutionary process between females and males with uniform sequences, it is only at the start of the simulations that the males have some success (Figure 5a), when the females still are naive. In the subsequent intervals the females continually respond with near optimal coyness. The average coyness is 8.0 (excluding the first turbulent 5000 generations), clearly higher compared with the simulations in which males use a variable signal sequence (Mann-Whitney U test, U = 0, p =.0040, exact one-tailed probability). The rate of change is also lower. A uniform male signaler is obviously limited in its ability to design a more stimulating sequence (compare Figure 5b and Figure 4b). Excluding the first 5000 generations, the average rate of change is 0.4 for both males and females.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONSimulation modelRESULTSDISCUSSIONREFERENCES |
|---|
Here we have given one example of how a reproductive conflict might give rise to behavior rituals and attempts at manipulating the female. The display itself does not provide any information about the female apart from indicating the presence of a conspecific male. These results agree with other theoretical studies of coevolution between senders and receivers in conflict (Arak and Enquist, 1995
;
Enquist and Arak, 1998). An
important result is that coevolution between senders and receivers never
settles at equilibrium with respect to signal form and the receivers'
reactions to the signal (Arak and Enquist,
1995). The effect of this coevolutionary race on the senders has
been investigated in a number of studies
(Arak and Enquist, 1993,
1995 ;
Enquist and Arak, 1993,
1994,
1998 ;
Enquist and Johnstone, 1997 ;
Hurd et al., 1995 ;
Phelps and Ryan, 1998). It is
shown how increasing conflict between senders and receivers results in more
elaborate and costly displays (Arak and
Enquist, 1995).
How the coevolution affects the receiver has received less attention (but
see Johnstone, 1994 ;
Phelps and Ryan, 1998).
Dawkins and Krebs (1978,
Krebs and Dawkins, 1984), in
their pioneering work on sender-receiver coevolution, argued that receivers
might be manipulated. Manipulation occurs when the receiver (in our model the
female) makes a suboptimal decision due to the display of the sender (in our
model the male). Models of behavioral mechanisms such as artificial neural
networks provide us with a tool to study manipulation (e.g.,
Enquist and Arak, 1998 ;
Ghirlanda and Enquist, 1998).
Our results show how females can be manipulated by displaying males that
exploit biases in recognition systems. Once manipulation becomes costly for
the female, selection will favor reduced sensitivity to the current display.
However, any change in the females' memory will create new biases
(Enquist and Arak, 1998), and
selection will favor males that can exploit these biases, leading to a
never-ending race.
Our simulations gave us two types of results. When male signals were variable, females started reproduction earlier than was optimal. When male signals were uniform, the manipulation was transient. Our results suggest that manipulation will be more common when signals are free to vary. This is, of course, because this freedom permits a more exhaustive search of receiver biases by the evolving signal. Conversely, the evolution of manipulation also depends on how rapidly receivers can evolve resistance.
To what degree are females manipulated in nature ? This is a difficult
problem, and to our knowledge no information exists on manipulation in natural
interactions between monogamous males and females. Our results show that both
successful (variable) and unsuccessful (uniform) males develop conspicuous
signals. A consequence of this is that by observing display alone, it is hard
to infer whether females really are manipulated to any significant degree.
However, the hypothesis we propose here for the evolution of behavioral
rituals in monogamous species has the advantage of being valid in a wider
context than theories building on mate choice (e.g., sexual selection). For
instance, there are a lot of ornamented rituals performed between the two
members of a monogamous pair that continue far beyond the act of pair
formation (Armstrong, 1963,
1965 ;
Huxley, 1923). These can be
understood if we recognize that the conflict between the male and the female
continue after the choice of a partner. These conflicts might concern
investments in parental duties such as nest building, incubation, feeding of
young, territory defense, and whether or not to have another clutch.
Our results depend partly on the artificial neural network we use as the
model of the females' recognition mechanism. These models are, of course,
crude in comparison with real nervous systems and do not closely imitate the
structure and complexity of a biological neural network
(Dawkins and Guilford, 1995).
Despite this, these models can still capture the most basic properties of
stimulus control. These simple networks are capable of generalization and
develop response biases as a by-product of discrimination and have
successfully repeated results from experimental psychology
(Enquist and Arak, 1998 ;
Ghirlanda and Enquist, 1998).
Today, there are no good alternatives to artificial neural networks when
modeling how stimulus controls response. Also, all other models of stimulus
control are actually much simpler than network models. Furthermore, this is
one of the first studies using networks that considers time as a parameter.
For another such successful attempt, see Phelps and Ryan
(1998).
Our simulations contain some deliberate simplifications. For instance, the
probability of meeting a male, as well as the proportion of philanderers in
the male population, remains constant throughout the season. This and the fact
that reproductive success declines exponentially with time in season results
in a constant female coy response
(Wachtmeister and Enquist,
1999).
An interesting development of our model would be to let the time that the
male stays in the presence of the female influence her reproductive decision.
This would directly model the female assessment process. Philanders may be
less attentive because they also are involved with other females. This would
be one way to investigate the old pair-bond hypothesis
(Armstrong, 1963,
1965 ; Huxley, 1923). A further
possibility is to limit the number of males and to evolve their tendency to
philander.
To summarize, we argue that courtship rituals evolve as a sender exploits
biases in the receiver to influence the outcome of sender-receiver conflicts.
According to our model, courtship is mainly manipulation and not an exchange
of detailed information. It is important to notice that manipulation always is
possible, although the degree of manipulation observed at any given time
depends on how we choose to set the parameters. We may distinguish between two
types of model parameters. The first type (q, Pd,
Pc, r, and ) determines what is best for the female to
do, thereby influencing the degree of conflict between males and females. When
it pays the female to be more coy, the conflict increases as males always
benefit from reproducing immediately. With growing conflict, the incentive for
manipulation increases, which tends to result in more manipulation. The second
type of parameters regulates the mutation processes, such as the number of new
mutants per generation and the nature of these mutations. Of key importance is
the relative ability of males and females to adapt to each other. If mutation
parameters are chosen so that favorable mutations appear at a faster rate in
males than in females, we observe more manipulation than if favorable
mutations appear faster in females. Note, however, as long as there is a
conflict, new manipulative male signals will regularly evolve
(Arak and Enquist, 1995). What
varies is how quickly females counteradapt to these innovations. It is, of
course, a weakness of our model that little is known about the relative
evolutionary plasticity of senders and receivers in reality. Empirical
observations of receivers range from those that seem completely fooled (e.g.,
foster parents that raise cuckoo chicks) to those that respond appropriately
to subtle differences. In our simulation mutation parameters were chosen so
that males and females evolved at similar rates.
A problem with testing our results is that some of our predictions coincide
with those of models considering courtship as quality advertisement
(Andersson, 1994 ;
Zahavi, 1975,
1977). Both models predict a
development of exaggerated signals in males and a preference for these signals
in females. However, "good genes" models require that signals are
costly and that only males of high quality can afford to develop the most
extreme and the most preferred signals
(Grafen, 1990 ; see also
Krakauer and Johnstone, 1995).
Our model, on the other hand, makes these predictions whether signals are
costly or not and whether there is an important difference in quality or not
(Enquist and Arak, 1998). In
conclusion, our model offers an alternative explanation to male courtship
sequences in monogamous species in which reproductive conflicts and the
females' recognition mechanisms are key factors.
|
ACKNOWLEDGEMENTS |
|---|
We thank Stefano Ghirlanda, Risa Rosenberg, Innes Cuthill, and two anonymous referees for suggestions and comments. This research was supported by grants from the Swedish Natural Science Research Council, from Marianne och Marcus Wallenbergs Stiftelse, from Riddarhuset, and from Wachtmeisterska släktföreningen.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONSimulation modelRESULTSDISCUSSIONREFERENCES |
|---|
Andersson M, 1994. Sexual selection. Princeton, New Jersey : Princeton University Press.
Arak A, Enquist M, 1993. Hidden preferences and the evolution of signals. Phil Trans R Soc Lond B 340 : 207-213.
Arak A, Enquist M, 1995. Conflict receiver bias and the evolution of signal form. Phil Trans R Soc Lond B 349 : 337-344.[Web of Science][Medline]
Armstrong EA, 1963. A study of bird song. London : Oxford University Press.
Armstrong EA, 1965. Bird display and behaviour. New York : Dover.
Basolo A, 1990. Female preference predates the
evolution of the sword in swordtail fish. Science
250: 808-810.
Bastock M, 1967. Courtship : an ethological study. Chicago : Aldine.
Breiehagen T, Slagsvold T, 1988. Male polyterrotoriality and female-female aggression in pied flycatcher, Ficedula hypoleuca. Anim Behav 36: 604-606.
Cullen JM, 1966. Reduction of ambiguity through ritualization. Phil Trans R Soc Lond B 251: 363-374.
Dawkins MS, Guilford T, 1995. An exaggerated preference for simple neural network models of signal evolution? Proc R Soc Lond B 261: 357-360.[Medline]
Dawkins R, Krebs JR, 1978. Animal signals : information or manipulation ? In : Behavioural ecology. An evolutionary approach (Krebs JR, Davies NB, eds). Oxford : Blackwell; 282-309.
Eibl-Eibesfeldt I, 1975. Ethology. New York : Holt, Rinehart & Winston.
Enquist M, Arak A, 1993. Selection of exaggerated male traits by female aesthetic senses. Nature 361: 446-448.[Medline]
Enquist M, Arak A, 1994. Symmetry, beauty and evolution. Nature 372: 169-172.[Medline]
Enquist M, Arak A, 1998. Neural representation and the evolution of signal form. In : Cognitive ecology. The evolutionary ecology of information processing and decision making (Dukas R, ed). Chicago : University of Chicago Press; 21-87.
Enquist M, Johnstone R, 1997. Generalization and the
evolution of symmetry preferences. Proc R Soc Lond B
264: 1345-1348.
Ghirlanda S, Enquist M, 1998. Artificial neural networks as models of stimulus control. Anim Behav 56: 1383-1389.[Web of Science][Medline]
Grafen A, 1990. Biological signals as handicaps. J Theor Biol 144: 517-546.[Web of Science][Medline]
Hertz J, Krogh A, Palmer RG, 1991. Introduction to the theory of neural computation. Redwood City, California : Addison-Wesley.
Holland B, Rice W, 1998. Perspective chase-away sexual selection : antagonistic seduction versus resistance. Evolution 52: 1-7.
Hurd PL, Wachtmeister C-A, Enquist M, 1995. Darwin's
principle of antithesis revisited : a role for perceptual biases in the
evolution of intraspecific signals. Proc R Soc Lond B
259: 201-205.
Hutchinson A, 1994. Algorithmic learning. Oxford : Clarendon Press.
Huxley JS, 1914. The courtship-habits of the great crested grebe (Podiceps cristatus); with an addition to the theory of sexual selection. Proc Zool Soc Lond 35: 491-562.
Huxley JS, 1923. Courtship activities in the Red Throated Diver (Colymbus stellatus Pontopp.); together with a discussion of the evolution of courtship in birds. J Linn Soc (Zool) 35: 253-292.
Johnstone R, 1994. Female preferences for symmetrical males as a byproduct of selection for mate recognition. Nature 372: 172-175.[Medline]
Kraukauer DG, Johnstone R, 1995. The evolution of exploitation and honesty in animal communication : A model using artificial neural networks. Proc R Soc Lond B 348: 355-361.
Krebs JR, Dawkins R, 1984. Animal signals : Mind-reading and manipulation. In : Behavioural ecology. An evolutionary approach (Krebs JR, Davies NB, eds). Sunderland, Massachusetts : Sinauer Associates; 380-420.
Lehrman DS, 1959. Hormonal responses to external stimuli in birds. Ibis 101: 478-497.
Lehrman DS, 1964. The reproductive behaviour of ring doves. Sci Am 211: 48-54.
Leimar O, Enquist M, Sillén-Tullberg B, 1986. The evolutionary stability of aposematic coloration and prey unprofitability : a theoretical model. Am Nat 128: 469-490.[Web of Science]
Mackintosh NJ, 1974. The psychology of animal learning. London : Academic Press.
O'Donald P, 1977. Theoretical aspects of sexual selection. Theor Popul Biol 12: 298-334.[Web of Science][Medline]
Phelps M, Ryan MJ, 1998. Neural networks predictresponse biases of female túngara frogs. Proc R Soc Lond B 265: 279-285.[Medline]
Ryan MJ, Fox JH, Wilczynski W, Rand AS, 1990. Sexual selection for sensory exploitation in the frog Physalaemus pustulosus. Nature 343: 66-67.[Medline]
Ryan MJ, 1998. Sexual selection, receiver bias, and
the evolution of sex differences. Science
281: 1999-2003.
Staddon JER, 1975. A note on the evolutionary significance of "supernormal" stimuli. Am Nat 109: 541-545.[Web of Science]
Stenmark G, Slagsvold T, Lifjeld JT, 1988. Polygyny in pied flycatcher, Ficedula hypoleuca : a test of the deception hypothesis. Anim Behav 36: 1646-1657.
Tinbergen N, 1951. The study of instinct. Oxford : Clarendon Press.
Trivers RL, 1972. Parental investment and sexual selection. In : Sexual selection and the descent of man, 1871-1971 (Campbell B, ed). London : Heinemann; 136-179.
Wachtmeister CA, Enquist M, 1999. The evolution of female coyness-trading time for information. Ethology 105: 983-992.[Web of Science]
Zahavi A, 1975. Mate selection-a selection for a handicap. J Theor Biol 53: 205-214.[Web of Science][Medline]
Zahavi A, 1977. The cost of honesty (further remarks on the handicap principle). J Theor Biol 67: 603-605.[Web of Science][Medline]
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  J. M McNamara, L. Fromhage, Z. Barta, and A. I Houston The optimal coyness game Proc R Soc B, March 7, 2009; 276(1658): 953 - 960. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. Gavrilets and D. Waxman Sympatric speciation by sexual conflict PNAS, August 6, 2002; 99(16): 10533 - 10538. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||