Over the last decade there has been a great deal of interest in variation of mate preferences for sexual attractiveness. Such variation is important as it has consequences for the rate and direction of sexual selection (Turner & Burrows 1995). Population level variation has been used extensively to examine the evolutionary history of female mate preference and its coevolution with male traits (e.g. Wilkinson et al. 1998). In contrast, variation in preference between individuals is less well studied. Yet individual preferences can provide important insights into the selective forces that shape mating decisions, since preference is predicted to be highly sensitive to both the costs of choice and the benefits derived from it (Pomiankowski 1987; Houle & Kondrashov 2002). Variation in preference among individuals can also be used to investigate the mechanisms underlying preference by measuring associations with other traits (e.g. Hingle et al. 2001a).
Most studies of female mate preference have focused on variation at the population or group level (reviewed in Jennions & Petrie 1997; Wagner 1998). However, extrapolation of such findings to the level of the individual can prove misleading, as individual preferences may differ widely in shape or form (Wagner et al. 1995; Wagner 1998). Selection may generate adaptive variation in individual preference if a female benefits from having preferences different from the population mean, for instance when the optimal strength of preference is dependent on the context of mate choice or the qualities of potential mates (Qvarnström 2001; Badyaev & Qvarnström 2002). In addition, selection can generate variance via the quality of the choser, if, for example, function preferences or choosiness are condition-dependent (Tomlinson & O’Donald 1996; Fawcett & Johnstone 2003). Despite its importance, only a few studies have successfully investigated individual preference variation (e.g. Wagner et al. 1995), while others have suffered from deficiencies in experimental design and choice of preference measure (reviewed in Wagner 1998).
Preference should be distinguished, both conceptually and empirically, from choice. Preference comprises the sensory and behavioural components that influence to mate differentially with certain phenotypes, whereas choice is the pattern of mating that is influenced not only by preference, but also other factors such as availability and the costs of choice (Jennions & Petrie 1997). This limits the utility of typical experiments that assess ‘preference’ when individuals are given a choice between simultaneously presented stimuli. This design forces to choose, which may misrepresent how individuals respond to the full range of phenotypes (Wagner et al. 1995; Wagner 1998).
If we adopt a ‘no-choice’ design, individual preference functions are derived from their responses to sequentially presented stimuli which vary in ornament value (Wagner 1998; Shackleton et al. 2005). In order to be exploited fully, no-choice tests need to assay choosers with a range of natural phenotypes. Studies that simply employ a few stimulus (typically syntetic composites and/or with extreme values of ornamentation) cannot accurately measure the strength of directional selection or detect stabilizing (e.g. Gerhardt 1991; Ritchie 1996; Hunt et al. 2005) or disruptive preference functions (e.g. Sappington & Taylor 1990; Greene et al. 2000). They also have limited power to resolve differences in the preference of individuals (Wagner 1998). It is also clear that the accuracy of preference functions will increase with the number of levels of a given ornament for which a chooser’ response is measured. However, care needs to be taken in repeated sampling of mating decisions to assess changes in receptivity and/or preference through time.
Ideal framework for studying preference are therefore those which perform easily distinguishable specific behaviours that indicate mating intent, such as exposing single people to solicitations of courtship/dating or active rejection of unwanted suitors.
2. DEFINITION OF MATE CHOICE.
Distinguishing preference and choosiness
A commonly adopted framework for describing mate choice is to distinguish preference from choosiness (e.g., Cotton et al. 2006 ; Jennions and Petrie 1997 ; Widemo and Sæther 1999 ). This demarcation is important as it differentiates innate tendencies toward specific mates (i.e., preferences – choices that would be made if cost was no object) and the actual mating bias that results depending on the amount of effort that a choosing individual is willing or able to invest in mate choice (i.e., choosiness; Cotton et al. 2006 ).
Jennions and Petrie (1997) provide a detailed description of terminology in which mate choice is defined as the pattern of mating that arises from “mating preferences.” “Mating preferences” are further divided into “preference functions” and “choosiness.” “Preference functions” are defined as the order in which an individual ranks potential mates. In contrast, “choosiness” is the amount of resources invested into choice, principally mate search effort and mate assessment effort. As mate search effort is coupled with variation in acceptance thresholds (i.e., individuals with higher thresholds will need to expend more effort searching for mates; see Acceptance thresholds above), variation in acceptance thresholds is positively correlated with “choosiness” in this definition ( Jennions and Petrie 1997 ).
In an approach echoing that of Jennions and Petrie (1997) , Cotton et al. (2006) expanded the description of preference functions to further define the “form” and “strength” of a preference function. The “form” of a preference function can take many shapes, for example, directional, stabilizing, or disruptive ( Figure 1a–e ) and is thus analogous to the ranking of potential mates. For a positive directional preference function larger trait values are ranked highest (and vice versa), for a stabilizing preference function intermediate trait values are ranked highest, and for a disruptive preference function extreme trait values are ranked highest. In addition, Cotton et al. (2006) defined the “strength” of a preference function as the rate of change, or slope, of the preference function, that is, how much higher an individual is ranked for a given phenotypic difference (e.g., Figure 1f–i ).
3. MATE CHOICE PARAMETERS:
3.1. Preference functions
Preference function: The order in which potential mates are ranked. The relationship between a phenotypic trait in potential mates ( x axis) and the reproductive resources invested in a mate ( y axis).
Hypothetical preferences of individual females. Despite substantial differences in the forms of the preferences, in all cases females would prefer trait value B to trait value A.
The concept of preference functions, and the term “preference function” have been widely and consistently used to describe patterns of mate choice (e.g., Basolo 1998 ; Gerhardt et al. 2000 ; Jennions and Petrie 1997 ; Ritchie 1996 ; Wagner 1998 ; Figure 1 ).
Preference functions have been particularly influential in the study of mate choice as the concept spans both empirical and theoretical approaches. The mathematical interpretation of preference functions has been widely used to model variation in choice, of which many examples can be traced back to the influential work of figures such as Lande (e.g., Lande 1981).
On the y axis of a preference function is a variable describing the will of invest in reproduction with each potential mate ( Bonduriansky 2001 ). On the x axis of a preference function is a phenotypic trait expressed by potential mates.
Furthermore, phenotypic variation in potential mates can often be complex in nature, and this can be depicted in multivariate preference functions (e.g., Backwell and Passmore 1996 ; Brooks et al. 2005 ; Candolin 2003
In summary, preference functions are incredibly useful for the description of choice because a wide range of preferred traits and expressions of choice can be depicted within the same framework. When viewing preference functions, a preference is shown whenever variation in liking/willingness to mate ( y axis) is dependent upon phenotypic variation in potential mates/attractiveness ( x axis).
Multivariate preference functions
We could considered only the effect of manipulating one visual character at a time. In reality, selection seldom operates on a single trait independently of other traits, and combinations of traits could have effects on individual fitness that cannot be predicted from consideration of the effect of varying a single trait in an experimental study (Lande and Arnold 1983).
Indeed, nonlinear selection analysis (Lande and Arnold 1983; Phillips and Arnold 1989) has formally shown that combinations of traits can have multiplicative effects on fitness via the action of correlational selection (e.g., Brodie 1992; Blows et al. 2003; LeBas et al. 2003).
The resulting pattern of selection operating on a suite of traits can thus be complex (e.g., Blows et al. 2003; Blais et al. 2004; McGlothlin et al. 2005) and impossible to predict from univariate analyses alone. Interestingly, correlational selection, in which two or more traits components influence attractiveness multiplicatively, has been invoked as a possible cause of directional, concave sexual selection (LeBas et al. 2003; McGlothlin et al. 2005).
The shape of a preference function:
Types of mate preference functions.
Preference strength/choosiness: Variation in the slope of a preference function—in general reference.
3.4. Acceptance thresholds
A step preference function in which potential mates with trait values greater than the threshold are accepted and all others are rejected.
Responsiveness (an aspect of choosiness).
Receptivity: Average response to potential mate.
Discrimination (an aspect of choosiness). Another approach to the description of “choosiness” has been to describe choosiness as an outcome of “responsiveness” and “discrimination” ( Bailey 2008 , 2011 ; Brooks 2002 ; Brooks and Endler 2001 ; Ritchie et al. 2005 ).
“Responsiveness” can be defined as a measure of motivation to mate, or the mean response of a focal individual to potential mates, and can thus be represented by a wide variety of traits, for example, courtship intensity, response latency, or association time. Variation in responsiveness can be depicted as a vertical shift in the position of the preference function ( Figure 1o–r and see also Bailey 2008 ).
In contrast, “discrimination” has been defined as the variation in response to different individuals. This can be calculated, for example, as the standard deviation of all responses ( Brooks and Endler 2001 ) or the difference between a response to the most preferred stimulus and the average response to all stimuli ( Gray and Cade 1999 ), though either method can yield similar results ( Bailey 2008 ).
- Linear preference function:
- Stabilizing preference function
Preference function components. Mate preference functions are shown for two populations or individuals (black dotted and blue solid lines). Female preference functions can be described by three aspects. Peak: Which male trait values are preferred, indicated by the trait value that elicits maximal response. Choosiness: How much deviation from the peak in male trait is tolerated, indicated by the width of the function. Responsiveness: The degree to which females respond to the trait, indicated by the height of the function. Choosiness and responsiveness combine to determine preference strength. Of note is that some ecological factors will change preference function width (strength) and others will change what’s preferred (peak). When both occur, this can result in clustering.
“Responsiveness” and “discrimination” can vary when there is no choice. Each of the 8 preference functions depicts motivation to mate ( y axis) versus phenotypic variation in potential mates ( x axis). Arrows indicate mean motivation to mate, that is “responsiveness” sensu Brooks and Endler (2001) . The distance between dashed lines indicates variation in responsiveness to mates, that is, “discrimination” sensu Brooks and Endler (2001) . Plots in the left panel illustrate how variation in responsiveness (a vs. b), discrimination (a vs. c), and both responsiveness and discrimination (a vs. d) can reflect variation in mate choice. Corresponding plots in the right panel (e, f, g, and h) illustrate analogous variation in these traits without any expression of choice.
- Selectivity: Variation in the response to potential mates
- Permissiveness: A response to a signal that is normally unattractive.
3.7. The slope of a preference function
The slope of a preference function can be defined as the difference in reproductive resources invested, including the likelihood of mating, per unit change in trait value of potential mates (e.g., Murphy and Gerhardt 2000 ). The slope of a preference function can be described for both linear and non-linear preferences through regression coefficients (e.g., Basolo 1998 ; Hunt et al. 2005 ; Murphy and Gerhardt 2000 ; Wagner et al. 1995 ; Figure 1g–i ).
The slope of a preference function can also be described for categorical traits as the difference in resources invested in potential mates belonging to different classes (e.g., Fisher and Rosenthal 2006 ; Qvarnström et al. 2000 ; Tinghitella et al. 2013 ; Figure 1f ).
There are two aspects of variation in a preference function (i.e., slope and horizontal position) each describe a distinct aspect of choice. In layman terms, the horizontal position of a preference function can be viewed as describing “what” is preferred. For find out choosiness/preference threshold subjects must make dichotomous decisions about whether an opposite-sex individual is attractive or not.
For example, the location of a peak of a stabilizing preference function is the most preferred trait value and the acceptance threshold describes the range of potential mates that will be accepted. For a categorical trait, variation in the horizontal position of a preference function is equivalent to varying the category of trait that is preferred without varying the slope or vertical position of the function (Figure 1k).
In contrast, the slope of a preference function can be viewed as describing “by how much” something is preferred, that is, by how much are preferred mate types favored relative to other mate types. Whereby subjects are asked to make judgements (i.e. scoring a likert scale) about a the same sequence of randomly varying stimuli.
These 2 pieces of information that describe mate choice echo previous approaches to the definition of mate choice that differentiated innate biases an individual might have toward certain mates from the manifestation of those biases (e.g., Cotton et al. 2006 ; Jennions and Petrie 1997 ). Consequently, the most appropriate term for referring to any variation in the horizontal position of a preference function would be “preference,” that is, which type of mate does the individual have a “preference” for. This term has previously been used in the same context when describing the location of a peak of a stabilizing preference function (e.g., Lande 1981 ; Rodriguez et al. 2013b ). “Choosiness” might then be used to refer to variation in the slope of a preference function (e.g., Ratterman et al. 2014 ). An important corollary of defining “choosiness” as the slope of a preference function is that, irrespective of taxa and the shape of a preference function, the absence of choice could universally be described as whenever “choosiness” is equal to zero. In summary, an individual could be described as exhibiting a “preference” for certain mate types (the horizontal position of a preference function), but the extent of those preferences will depend on “choosiness” (the slope of a preference function).
4. CHOICE DESIGNS
a) Simultaneous or sequential
There are simultaneous stimulus presentations. Stimuli are presented to a focal individual at the same time. Methods can vary depending upon the type of animal taxa and the investigator conducting the work. The responses of females can be scored dichotomously (who’s preferred- yes or no for each stimulus) or on a continuous scale (How desirable/attractive is it; the score rated with each stimulus).
In sequential test, each stimulus is presented individually to females. For example, we might measure responses to potential mates of different physical appearance, where all stimuli are presented sequentially. Responses to a series of sequentially presented stimuli can then be measured, and preference functions can be derived by examining the relationship between female response and stimulus trait value (e.g. Basolo 1995).
b) Choice/No-choice test.
An important way in which experiments testing mate preferences can vary is in the number of options the subject is presented with during the test, which we refer to as the “choice paradigm” or “choice design.” Tests can use either no-choice or choice designs ( Wagner 1998 ).
- In a no-choice test each subject is presented with a single stimulus. Several no-choice trials may be performed using the same subject. these are referred to as sequential choice tests
- In contrast, in a choice test each subject is given a choice between multiple stimuli presented simultaneously.
On a binary test, this variable would be the probability of an individual mating and finding out aceptance threholds. By the other hand, “Likert” scales are also usefully represent actual investment/willingness to invest in mating with a mate. Research must have a method of ascribing quantitative value to each indivual sampled, to make it amenable to statistical analysis,where a numerical value is assigned to each potential mate on a 1 to 10 scale ( or 1-5 scale, 1-7 scale.)
c) Population level
Population-level preference functions are the most commonly used method to analyse mate preference. Typically, a group of females within a population are tested only once for their response to a range of male attractiveness levels and a statistical model is fitted relating female response as a function of male signal. For genetic analysis this approach has limited value because variation in preference within and among females is unaccounted for. Therefore, a population-level function can in fact contain multiple female preference phenotypes.
d) Individual level
Individual preference functions measure a female’s response to an array of male attractiveness levels. Typically, a single female is repeatedly tested for her response to a randomized series of attractiveness levels and preference functions are estimated using either polynomial regression or cubic spline methods.
It is important that females are tested multiple times for their response to the same stimulus to allow variances to be calculated around mean response levels and variation among individuals to be compared.
In a experimental design for the analysis of individual preference functions for attractiveness, each individual is tested for her response to sample of stimuli with n trials conducted for each stimulus. This example is an absolute preference function, as females were not given a choice between males.
In the example in part b, genetic variation in female preference functions. Each green line is the female preference function for an individual. Note that although many functions are overlapping, there is significant variation among genotypes within this population.
In the example in part b, genetic variation in female preference functions. Each green line is the female preference function for an individual. Note that although many functions are overlapping, there is significant variation among genotypes within this population.
5. Material and methods
a.1. Selection of stimuli:
They should be analyzed attractiveness assessments made to static images and dynamic video clips of the same persons:
- Photographs: ratings of static images.
- Video clips: ratings made to dynamic images. Which provide richer samples of appearance. Presence of much more information in the videos (multiple views, neutral and smiling expressions and speech-related movements)
– Recruitment of single participants. A great number (Ns) of male and female participants are recruited. They are photographed and videotaped, digitally processed. On a separate process independent judges (number Nr) rate the facial/body photographs and clips. Judges are asked to rate the attractiveness of Ns sample on a 1-10 numeric scale.
a.2) Another alternative would be to use a face/body database, which include standardized photographs of male and females subjets of varying ethnicity between a widre range of age. Extensive norming data are available for each individual model. These data include both physical attributes (e.g., face size) as well as subjective ratings by independent judges (e.g., attractiveness).
– Use of databases: If it is not feasible to recruit a large number of participants for our study, or the available databases do not contain such a large number of individuals, the set of stimuli may be replaced by non-standardized real-world profile pictures downloaded from the public domain. (e.g online dating webs).
– The use of profile pictures retrieved from online dating websites. Non-standardized images varying composition, face size and background cues.
a.2 Rating/sorting of stumuli.
Pictures/videos of participants are rated/sorted for attractiveness by judges recruited. Judges of both sex rate each image on a 10-point rating scale from ‘‘not at all’’ to ‘‘extremely’’ attractive.
1. Binary attractiveness rating. Probability function.
On a higher level, we seem to regard attractiveness in very binary terms(yes, or no).
A Np sample of subjects (unmated participants) interesed in meeting a potential partner, are recruited.
Study asks to make judgments about a sequence of randomly varying stimuli. It must adapt the non-choice sequence task to ask a pertinent question about mate selection, in which participants will rate each picture in a sequence, and will make also binary decisions (attractive or unattractive) in a simplified, real-world context.
It’s a design of binary task mimicking the selection interface (currently popular in online dating websites):
“Are you interested in dating this person? If you want to have a date with him/her, then tap on the “Yes” icon. You must swip left on “No” button if he/she is not attractive enough as potential partner.
When you are playing binary task, if you and another member both Like each other or slide each others photos into the ‘yes’ pile, then we’ll let both of you know. If either of you slide the photo to the ‘No’ pile, then nothing happens—if you don’t heart someone, they’re not hidden from you forever.
If you and another user like each other, we’ll let you both know by sending a notification when the experiment is finished. Moreover your contact info (i.e phone number) will be provided to each person into your mutual list.”
Participants are assayed for individual mate preference by scoring rejection or acceptance towards all oposite-sex stimuli (two males from each eyespan class). Once a response for a given stimuli had been determined, this one is removed and other subject is shown.
Participants must be assayed for individual mate preference by scoring rejection or acceptance of all the photographic stimuli. Once a response for a given picture had been determined, the picture is removed, and the next one is shown.
2) Attractiveness assessment. Preference Function.
A trial start with a picture is shown randomly from the set of opposite sex potential individuals. In an unspeeded task participants give a score on a 1-10 scale and then assest as attractive or not and the next face followed immediately.
Individual-level (left column) and population-level (right column) female preference functions for overall male attractiveness.
In systems where species encounter mates sequentially and the costs of lost mating opportunities may be high, n-choice tests may overestimate the strength of mate selection. But the converse may also be true; in lek breeding animals (e.g humans), individuals may seldom encounter a single courter. Thus, no-choice paradigms may underestimate the strength of selection in lek breeding species.
In the case of a no-choice paradigm, the receiver likely compares the signal it receives to some internal template (although this template may not even be fixed, e.g., Taylor and Ryan 2013). In a n-choice test, the receiver may compare several signals to an internal template or it could bypass an internal template and compare the signals directly to each other.
The employment of multiple approaches within a specie can provide a richer understanding of perceptual processing and the evolution of mating signals.
We should take great care when designing studies of mate choice if our goal is to project our conclusions to natural populations or to make quantitative predictions about how mate choice translates into selection on male traits.
If either is our aim, we need to rely on field studies or experimental studies conducted under settings that closely mimic those in the wild.
In turn, if a role for mate rejection costs is demonstrated, it will yield insights into the economics of mate choice. For example, perhaps the high cost of rejecting a mate in males, where variance in mating success can be so high, accounts for the tendency of male choice to vary little between choice and no choice formats
If we can understand the impact of some set of environmental factors on the strength of choice, we have arguably gained some level of understanding of the factors that shape the evolution of choice.
Experience of sexual signals can alter mate preferences and influence the course of sexual selection. It is likely that patterns of experience-mediated plasticity in mate preferences that can arise in response to variation in the composition of mates in the environment.
The precise characters conferring male attractiveness is a composite trait that cannot be totally captured by simple measurements of single characters. That is to say, even if individual traits that are subject to sexual selection are heritable, this need not imply attractiveness in total is heritable and can evolve.
We don’t know if humans have been selected to adjust preference selectivity according to the variability of potential mates in their social environment, as well as to the presence/absence of preferred mates.
However it seems like a dead end since it does not seem feasible to control the life-time experience prior to the measurement/evaluation process for human subjects.
It is likely that experience-mediated plasticity in mate preferences can influence the strength of selection on male signals and can result in evolutionary dynamics between variation in preferences and signals that either promote the maintenance of variation or facilitate rapid trait fixation.
If we could control for patterns there would test hypotheses about potential sources of selection favouring experience-mediated plasticity.
Manipulated signal experience on experiment with the following treatments:
-Absence of stimuli
– Exposure to unattractive stimuli.
– Exposure to attractive stimuli.
-A mixture of attractive and unattractive stimuli.
In some cases, a female’s mate preferences may depend on experiences with conspecifics in the social environment prior to making a mating decision, providing evidence for socially cued anticipatory plasticity (SCAP) in mating preferences.
Features that comprise the attractiveness complex are intricate parts of an n-dimensional feature space. This feature space is organized such that all features point in the same direction. Attractiveness thus follows the redundant signalling hypothesis.
We do not assume innate beauty detectors; we rather propose that the brain has an innate tendency and basic rules on how to create beauty templates, which then are filled up during ontogeny. When the media raise attractiveness standards by prototyping beauty, then unreal expectations to mate quality (beauty) will emerge. If the template created is more attractive than reality, no mate selection can occur on realistic grounds, leading to a high proportion of singles.
a)‘Norm-based coding’ (Rhodes, Brennan & Carey, 1987), where averaging a large number of faces in the brain derives the norm.
b) ‘Density alone hypothesis’, where the Gestalt is a point by-point representation in a multidimensional space.
c)‘Template’ hypothesis, which suggests that the brain analyses the single parts with templates and then reintegrates them.
Assessment of facial attractiveness could depend on a genetically specified template or it could reflect general pattern-learning mechanisms. Early visual experience appears critical to the normal processing of facial configuration later in life (Le Grand
et al. 2001). Exposure to faces and prototype abstraction during early life may also have long-lasting effects on face processing including judgements of facial attractiveness. Thus, differences in the types of face an individual is exposed to may lead to subtle differences in the facial prototypes extracted, which in turn may bias attractiveness judgements towards exposed facial characteristics.