See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/277478426

Partner choice decision making and the integration of multiple cues

Article in Evolution and Human Behavior · May 2015

DOI: 10.1016/j.evolhumbehav.2015.05.001

CITATIONS READS

17 344

2 authors, including:

Ryan Schacht
East Carolina University
44 PUBLICATIONS 733 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Embodied Capital View project

Study of Choyero Science in the Sierra de la Giganta, Baja California Sur, View project

All content following this page was uploaded by Ryan Schacht on 17 October 2017.

The user has requested enhancement of the downloaded file.

 Evolution and Human Behavior 36 (2015) 456–466

Contents lists available at ScienceDirect

Evolution and Human Behavior

journal homepage: www.ehbonline.org

Original Article

Partner choice decision making and the integration of multiple cues

Ryan Schacht a,⁎, Mark Grote b
a
Department of Anthropology, University of Utah, Salt Lake City, UT, 84112
b
Department of Anthropology, University of California-Davis, Davis, CA, 95616

a r t i c l e i n f o a b s t r a c t

Article history: Uncertainty about fitness-enhancing traits in a potential mate, as well as variability in social and ecological envi-
Initial receipt 28 June 2014 ronments, favors the use of multiple cues in selecting a partner. Though how individuals respond with adaptive
Final revision received 7 May 2015 mating preferences is an open question. Here we investigate mate choice decision making among the Makushi
of Guyana and compare two competing approaches: 1) a prioritized trait approach, in which preferences are
Keywords:
determined by the independent evaluation of relevant partner traits; and 2) an integrative approach, in which
Sexual selection
Mate choice
preferences are determined by reducing multiple, interrelated traits to a few latent dimensions. Within these
Sex ratio two approaches we measure the effects of several key factors — sex, adult sex ratio, and community-to-
Item response models community variability — thought to pattern preferences. We find support for cue integration and contextual
variability in preferences. Sex and adult sex ratio are weak predictors of preferences in the Makushi: preferences
are best explained by unstructured community effects. These findings highlight two key issues in mate choice
studies: 1) simple biologically-based models do not seem adequate to explain variation in preferences, either
within or among populations; and 2) while context, generally speaking, matters in determining preferences, we
lack theoretically-informed predictions about relevant contextual factors. The importance of cues, as well as
what they signal in a potential partner, is likely to vary with location-specific factors that are yet unexplored.
© 2015 Elsevier Inc. All rights reserved.

1. Introduction
In sexually reproducing species the choice of a mate is key to fitness
(Andersson, 1994; Bateson, 1983). Research on mate choice has primarily
focused on identifying preferred traits, insofar as these serve as sig-
nals of the potential benefits a mate may offer (reviewed in
Andersson, 1994), be they direct (e.g., parental investment) or indirect
(e.g., immunocompetence). Far from involving simple decisions, selec-
tion of a mate likely requires paying attention to multiple, potentially
competing, signals. It is an open question how individuals utilize infor-
mation from multiple cues to make an adaptive mate choice decision.
In general, signals are thought to be organized in one of two ways:
1) as a collection of independently relevant traits, each reflecting a
single property of a potential partner and varying in importance to
the individual making the choice (the "chooser"; reviewed in Candolin,
2003); or 2) as suites of interrelated characters, in which relationships
between traits are as important to the chooser as the traits themselves
(e.g., Jennions & Petrie, 1997). We call the former the “prioritized traits”
approach and the latter the “integrated traits” approach, acknowledging
that these terms of convenience are imperfect. A central aim of our
analysis is to learn whether individuals choosing mates seem to take a
prioritized, or alternatively a more integrated, approach when evaluat-
ing traits of potential partners.
⁎ Corresponding author.
E-mail address: r.schacht@utah.edu (R. Schacht).
Within the two approaches, patterns of mate choice are often
thought to rest on a long-standing model of sexual selection, which
links differential parental investment to sex-differences in optimal mat-
ing rates (Trivers, 1972). A newer model pays attention to evolutionary
feedbacks that can strongly influence sex roles and subsequent patterns
of sex-differentiated behavior (Kokko & Jennions, 2008). In this paper
we will explore mate choice decision-making by measuring the relative
empirical support for the two approaches (prioritized vs. integrated
traits) as a function of sex and population-level parameters.
1.1. Why multiple cues?
The emphasis, historically, within the study of mate choice was on
the experimental manipulation of a single cue in homogenous environ-
ments (reviewed in Gerhardt, 1992). This work, while usefully
highlighting the traits which could serve as cues of mate quality, poorly
reflected actual animal displays that consisted of multiple cues
(e.g., Dale & Slagsvold, 1996; Hill et al., 1999; Kodric-Brown & Nicoletto,
2001). Additionally, studies revealed that individual reproductive deci-
sions change by age, condition, experience and context (reviewed in
Miller & Svensson, 2014). For example, in a well-known study among
lark buntings, Chaine and Lyon (2008) found the traits associated with
male pairing success to be highly variable from year to year, highlight-
ing temporal flexibility in female choice. Thus, variation in the social
and ecological environment, as well as among individuals, influences
which traits are potentially fitness-enhancing, thereby favoring

http://dx.doi.org/10.1016/j.evolhumbehav.2015.05.001
1090-5138/© 2015 Elsevier Inc. All rights reserved.
 R. Schacht, M. Grote / Evolution and Human Behavior 36 (2015) 456–466
multiple signals and reinforcing the need to pay attention to multiple
traits (Bro-Jorgensen, 2010).
Additionally, organisms have to make decisions under uncertainty,
as individual traits may not be reliable signals of underlying quality or
the condition of a potential mate (Brunswik, 1955). Because of this
uncertainty, several traits, each related to underlying condition but
likely only offering partial information, may need to be attended to.
Thus variation in trait priorities, or in how preferences for different traits
are combined, may arise due to individual and contextual factors
(reviewed in Bro-Jorgensen, 2010).
Mate choice interactions are likely even more dynamic and sensitive
to individual and contextual-level variables in species with mutual mate
choice and biparental investment (Bergstrom & Real, 2000; Hooper &
Miller, 2008; Johnstone, Reynolds, & Deutsch, 1996; Kokko & Johnstone,
2002). For example, in humans, cues that may signal underlying male
quality may also be associated with lower levels of parental investment
(due to that male's attractiveness to a larger number of mates) causing
females to make trade-offs over multiple cues when selecting a mate
(Scheib, 2001).
1.2. Patterning of choice (prioritized traits vs. integrated traits)
Mate choice studies of humans have successfully documented many
important individual traits used in selecting partners (e.g., body mass
index, waist to hip ratio, physical attractiveness, social status, kindness,
and honesty; reviewed in Gangestad & Scheyd, 2005). The prioritized
trait approach is useful for understanding mate selection in terms of
the traits that are more or less important to the chooser. However,
relationships between the chooser's preferences across multiple traits
may be lost in trait-by-trait comparisons (Miller, 1997). In a classic
and approachable example of trait integration, Moller et al. (1998)
find, when looking at male traits of song-rate and tail length among
barn swallows (Hirundo rustica), that females do not simply have a
greater preference for one trait over the other (i.e. they do not find
support for a prioritized trait approach). Instead, the importance
females place on male song-rate depends on male tail length (see
Kodric-Brown & Nicoletto, 2001; Scheib, 2001 for similar findings in
guppies and humans respectively).
In general, studies of mate choice take the prioritized trait approach,
treating each cue as one of a list of independently relevant characteris-
tics in a potential partner (e.g., Lippa, 2007). However, there is increas-
ing interest in exploring how traits interact in a synergistic manner
(Jennions & Petrie, 1997), focusing on variation and covariation in trait
preferences. This has led to the integrated trait approach (Miller,
1997), which argues that relationships among traits are essential for
understanding mate choice.
1.3. Variables influencing mate choice
The study of reproductive decision-making typically relies on the
long-standing model of sexual selection developed by Trivers (1972).
This model links mate choice preferences directly to differential invest-
ment in young by males and females. In humans, because investment by
women is more obligatory (through gestation and lactation), they are
expected to pay close attention to a partner's ability to provide
resources. In contrast, because a woman's reproductive value declines
with age, men are expected to pay close attention to physical attractive-
ness, as it serves as a signal of fertility (Symons, 1979).
Variability in preferences driven by contextual factors and cultural
norms (e.g. partner chastity; Buss et al., 1990) has long been acknowl-
edged in studies of mate choice. However, and in line with predictions
from traditional sexual selection theory, findings of men's preferences
for physical attractiveness and women's preferences for resources are
quite robust across the literature (Buss, 1989; Buss et al., 1990; Li
et al., 2002; Schmitt, 2005; Shackelford, Schmitt, & Buss, 2005). So
while gender differences in preferences are not expected across all
457
traits, men's preferences for reproductive capacity and women's prefer-
ences for investment potential are generally treated as human univer-
sals. While the evidence is quite impressive, the methodology and
theory underlying these findings can be productively critiqued.
First, almost all of the work on human mate preferences has been
based on questionnaire responses from college undergraduates
(Aspendorf & Penke, 2005; Gray, Heaney, & Fairhall, 2003; Griffiths,
2001; Laland & Brown, 2011), generally from the US or western
Europe. These populations are relatively easy to sample from and are
worthy of study, however the generalizability of findings from such
samples can be questioned (Henrich, Heine, & Norenzayan, 2010;
Smith, Borgerhoff Mulder, & Hill, 2001). Additionally, some studies
that purport to be cross cultural (Buss, 1989; Schmitt, 2005) draw on
university students in developing nations, who may be even less repre-
sentative of their local populations (Beckerman, 2005). Studies that do
look outside the west generally find results counter to conventions:
for example, among the Shuar (Pillsworth, 2008) and Hadza
(Marlowe, 2004) there was little support for a difference between
men and women in preferences for physical attractiveness. Additional-
ly, a recent study using data from 12 societies, both industrialized and
non-industrialized, found no consistent gender differences in partner
preferences (Scott et al., 2014).
Second, a reformulated theory of sexual selection critiques the sim-
plistic labeling of reproductive roles by gender based on inherent sex
differences in parental investment (Kokko & Jennions, 2008). As in
nonhumans (Clutton-Brock, 2007), patterns of sexual selection on
men and women can be highly variable (Borgerhoff Mulder, 2009;
Brown, Laland, & Mulder, 2009; Scelza, 2011). Reproductive strategies
are not an invariant, species-specific characteristic, but rather faculta-
tive responses to individual- and population-level social and ecological
circumstances (e.g., Nettle, 2009; Owens & Thompson, 1994; Szekely,
Webb, & Cuthill, 2000) requiring conditional decision-making
(Gangestad & Simpson, 2000; Nettle, Coall, & Dickins, 2011).
To counter concerns of (1) western overrepresentation and
(2) simplistic parental investment models, we conduct our study
among the Makushi of Guyana, measuring the support for a sexual
selection framework in which evolutionary feedbacks are predicted to
influence sex roles and subsequent patterns of sex-differentiated
investment in mating effort (Kokko, Klug, & Jennions, 2012). A key
feature of this framework is its game-theoretical foundation (Kokko &
Jennions, 2008), in which sex roles are partly determined by the relative
scarcity of the sexes (e.g. Fromhage, Elgar, & Schneider, 2005). As a
consequence, sex-structured pay-offs depending on the adult sex ratio
(ASR) generate predictions of sex-differentiated behavior. For example,
when females are in surplus, males may be able to leverage their
relative scarcity, behaving promiscuously and offering little parental
investment yet still obtaining mating opportunities. In contrast, when
females are scarce, males may need to show a commitment to marriage
and family in order to secure mating opportunities. The adult sex ratio
is therefore expected to play an important role in the patterning of
preferences (Schacht, Rauch, & Mulder, 2014; see Table 1). While
some authors have explored sex ratio effects on reproductive decision
making (Pedersen, 1991; Schmitt, 2005), they nevertheless assume
that sex ratios will impact mating strategies as proposed by PI theory
Table 1
Traits of greatest importance to males and females choosing partners, according to tradi-
tional and reformulated sexual selection theories.
Traditional (sex) Reformulated (sex * ASR)
Males Physical attractiveness, Male-biased ASR: Female-biased ASR:
faithfulness faithfulness, good physical attractiveness,
parent health
Females Resources, social status Male-biased ASR: Female-biased ASR:
faithfulness, good physical attractiveness,
parent health
Desired partner traits are in italics.
458 R. Schacht, M. Grote / Evolution and Human Behavior 36 (2015) 456–466

Table 2
Models and their predictions.

Cognition underlying mate choice

Prioritized traits Integrated traits

Intervening variables (variables patterning the variation)

Community Differences in importance of individual Suites of traits variably
traits by context important by community
Sex Differences in importance of individual Suites of traits variably
traits by sex important by sex
ASR × Sex Differences in importance of individual Suites of traits variably
traits by sex and ASR important by sex and ASR

A community effect is included in all models, and the fixed effect for sex is included in
models containing the ASR × sex interaction. Thus, a nested sequence of increasingly
complex models is produced by moving from top to bottom in the table.

(Emlen & Oring, 1977), namely that men will invest more in mating
effort when they are more numerous than women.
The predictive variables and contrasting approaches for patterning
choice can be thought of as existing in distinct theoretical domains.
The traditional sex-based and newer evolutionary feedback frameworks
are both compatible with either the prioritized or integrated trait
approaches. For example, integrated traits may be strongly patterned
by sex alone, or prioritized traits may be best explained by both sex Fig. 1. Map of Guyana and field site locations.
and ASR.
We will measure the support for models of mate choice involving
nested combinations of a) community effects, b) sex, and c) ASR, in her family (although a father may take sons) but are expected to be
two ways (prioritized vs. integrated traits; Table 2). In doing so, we provided for by a stepfather if the mother remarries.
not only examine variables that influence mate choice but also tackle
the newer issue of how humans evaluate traits in potential partners: 2.2. Data collection
as separate pieces of information or as an integrated package.
We conducted the mate choice preference survey across eight
2. Methods Makushi communities (Fig. 1). We first conducted a full census to deter-
mine the community composition. We then randomly sampled a
2.1. Study population minimum of 30 individuals from each village for a total of 148 men
and 152 women aged 18–45 (Table 3). We asked respondents to answer
The Makushi inhabit the Rupununi savannas of south-western questions regarding the importance of partner traits operationalized
Guyana, region 9. Living along the border with Brazil, this ethnic through 10 items: financial resources, physical attractiveness, faith-
group shares many cultural traits with other groups from the Xingu fulness, parenting qualities, social status, health, desire for children,
Basin. These include shifting cultivation, a focus on bitter cassava, devotion, hardworking, and strength of family bonds. These traits
matrilocal marriage, the performance of bride-service before marriage, were selected based on their ubiquity in the mate-choice literature
and fairly egalitarian gender relationships (Schacht, 2013). While (see Buston & Emlen, 2003) and pilot-tested to ensure cultural appro-
premarital sex is not disapproved, and is an expected avenue to secure priateness. Respondents rated each item with an item score, using a
a partner (Myers, 1993), the Makushi generally marry monogamously five-point scale (1 = not at all important, through 5 = very important).
and extended families typically share one residential area (Forte, In order to minimize response bias and other potential data-quality
1996). Makushi marriages are generally endogamous, in that mates problems, our study protocol incorporated: a) a long (16 month) period
are usually selected from within the village community (Myers, 1993). of fieldwork during which community rapport could be built across
As elsewhere in Guyana, outmigration has led to considerable each of the villages, b) gender-matched interviewers and interviewees
between-community variation in ASR, as men and women search for and c) the use of a nonverbal response card method (Lindstrom et al.,
economic opportunities. Principle activities for men are mining, cattle 2010) to guarantee the privacy of the interviewee's response, even
ranching, agricultural work and logging, activities which occur mainly from the interviewer. Furthermore, we asked questions using regional
in the more remote areas of the Rupununi or in the forested regions at colloquial language, after pilot-testing the wording of questions for
the center of the country, whereas women are attracted to urban salience. In this way we tried to make sure respondents understood
areas (such as the capital of Roraima in neighboring Brazil) and the what was being asked of them.
larger interior Guyanese towns (such as Lethem) in search of shop
and domestic work (Gafar, 2004).
Table 3
Community-level ASR strongly structures marital options for endoga- Descriptive statistics for each community.
mous marriage, but men are still expected to perform bride-service
in order to marry. This traditionally involved a year of service by the pro- Community ASR Men:Women # Men # Women Total
(18–45 years) interviewed interviewed population
spective husband, in which he clears and farms fields for his in-laws
while building a new dwelling nearby for himself and his wife. Men A 0.93 125:135 29 29 745
H 1.11 70:63 15 15 415
and women typically marry only once or twice, and conventions are
F 1.13 70:62 15 15 407
similar across all marriages, with men providing bride-service and E 1.16 87:75 18 19 596
thereafter bearing considerable responsibility to provide for their wives G 1.22 73:60 20 20 432
and children (including stepchildren who are valuable helpers) through D 1.33 80:60 19 19 406
farming, fishing and various forms of wage labor (Myers, 1993). At B 1.35 27:20 15 15 162
C 1.43 57:40 17 20 310
divorce, children remain largely the responsibility of the mother and
 R. Schacht, M. Grote / Evolution and Human Behavior 36 (2015) 456–466
2.3. Statistical approach
In order to weigh the evidence in our sample for different notions
about the cognition underlying mate choice, we evaluate the relative
support for nine theoretically-informed models fitted to the observa-
tions. Here, and in the Supplementary Online Materials S1–S4 (available
on the journal's website at www.ehbonline.org), we describe our
approach to estimation and comparison of prioritized trait models
(I–III) and integrated trait models (IV–IX). All models are fitted by
computationally-intensive Bayesian methods. Bayesian methods are a
practical (more than an ideological) choice here, for two main reasons:
1) the models are relatively “large”, containing many parameters to
infer, yet our main interests lie in a small subset of these parameters
or in whole-model comparisons; 2) the integrated trait models
(IV–IX) present problems of identification (see S2), which can be
partly remedied by appropriate choices of prior distributions. Com-
putational Bayesian methods naturally accommodate these needs.
All models I–IX contain basic linear equations (see S1) of familiar
regression form.
2.3.1. Prioritized traits
Within the prioritized trait approach we consider three nested or-
dered logit models (community, sex, and ASR; Table 2; S1). Ordered
logit regression models are used when the responses — here, item
scores — are ordered categories. All items and individuals are included
in each model. The direction and magnitude of effects in these models
are allowed to vary freely and are unique to each item. For example,
we do not require sex-specific differences in the importance of social
status and financial resources to conform to expected patterns of strong
female preference (e.g., Buss, 1989; Shackelford et al., 2005; Stewart,
Stinnett, & Rosenfeld, 2000). However, if those preferences do exist in
the sample they will be reflected by the estimated effects. We take
Fig. 2. Item score distributions by sex: Side-by-side histograms of item score frequencies. The x-axis indicates the frequency in the sample of each response option by sex (males in light
gray and females in dark gray). Axis scaling varies by item in order to accommodate different levels of variation.
459
this approach in order to give models the greatest latitude to find sex
differences if they exist.
2.3.2. Integrated traits
Here we consider six models using a dimension reduction approach.
In using dimension reduction, we assume that an individual's response
profile expresses interrelated preferences across items that can be re-
duced to a smaller number of variables. These variables are unobserved,
and we refer to them as latent traits; they are theoretical constructs
analogous to intelligence or leadership ability (e.g., Gardner, 1983;
Stogdill & Coons, 1957). It is unclear whether dimension reduction of
interrelated preferences should aim for a unidimensional or multidi-
mensional trait. For example, is one axis ranging from low to high
adequate to explain variation in response profiles? Previous findings
suggest that this may be overly simplistic (Botwin, Buss, & Shackelford,
1997) and that there may be multiple axes in play (e.g., individuals may
be placed low on one axis yet high on another). Accordingly we examine
both one and two dimensional models. We return to the question of
additional dimensions in the Discussion.
A traditional dimension reduction approach used in psychometrics
is factor analysis. This is most suitable when the observed variables —
here items — have Gaussian distributions. However, a preliminary
check showed that ordinary factor analysis would be a poor choice for
our sample: the item score distributions are discrete and very skewed
(see Fig. 2); and the variance/covariance matrix computed from the
item scores had several negative eigenvalues. A slightly better option
is to build a factor model on the polychoric correlation matrix, which
treats the item scores as thresholded continuous variables (Grilli &
Rampichini, 2003). However, again the matrix was ill-conditioned,
and it was clear that an approach better suited to our ordinal variables
was needed.
460 R. Schacht, M. Grote / Evolution and Human Behavior 36 (2015) 456–466
Item response models were developed by quantitative psychologists
due to the need for appropriate dimension reduction techniques for
data sets sharing features of ours (Hambleton, Swaminathan, & Rogers,
1991). Like factor analysis, item response models contain a linear
prediction equation involving one or more latent variables, and aim to
connect variation and covariation in the observed variables to the latent
variables. However, unlike factor analysis, an appropriate link function
is required to transform the linear predictor to the scale of the manifest
variables. The item response model establishes the position of each
individual in a latent space, and derives probabilities of responses to
multiple observed manifest variables as a function of item parameters
and the individual's position in the space. Essential to the questions at
hand is the possibility to include fixed effects, such as sex and ASR, as
well as contextual (“random”) effects in the linear predictor (see
Table 2 and S1). This facility allows us to estimate the effects of commu-
nity, sex and ASR on the latent trait. Bayesian methods built on Markov
Chain Monte Carlo (MCMC) algorithms make it possible to fit and
compare these item response models (structured as in Table 2; Zhu &
Stone, 2012).
2.3.3. Community effects
All community predictors in Table 2 are random intercepts, included
to capture unmeasured community-to-community heterogeneity in
trait preferences. The intercepts are community-specific (as well as
trait-specific in the prioritized trait models), and act to adjust
community-average preferences to a theoretical population baseline.
These predictors are “unstructured”, in the sense that they are not
derived from direct measurements of community-level variables such
as ASR. They can be used interpretively to ask whether unobserved
community-level (“contextual”) factors may act as important sources
of variation in preferences. We understand the inclusion of community
random effects to be both facilitated, and required by, the multi-level
nature of our dataset (consisting of individuals nested within eight
communities). Merlo et al. (2005) motivate the use of community
random effects as tools for detecting contextual phenomena in datasets
of similar form.
2.3.4. Model comparisons
We perform posterior predictive calculations, consequently producing
evidence ratios and model weights, in order to compare the relative
support for models I–IX (see Ntzoufras, 2009 sections 10.4, 11.10). We
understand comparison of the different models — which themselves
mirror different notions about how mates are chosen — as an attractive
alternative to traditional hypothesis testing. This weight-of-evidence
strategy keeps all models in play, but highlights those that are better
supported by the observations. Gelfand, Dey, and Chang (1992) give a
rationale for the specific approach to model comparison used here,
and point to advantages provided by computational Bayesian estima-
tion. The basic quantity calculated is a conditional predictive ordinate
(CPO) for each observed person-and-item score (see S4). CPOs are
specific to each model, and can be combined across persons and items
to produce model-wide summaries broadly analogous to information
measures such as the deviance information criterion (DIC; Spiegelhalter
et al., 2002).
3. Results
3.1. Item score distributions
To begin our investigation we examine the item score distributions
by sex (Fig. 2). What is initially striking is that the modal response for
all items, except desire for children, is “very important”, the highest
category. A visual inspection suggests that preferences vary by sex
(males in light gray, females in dark gray) for at least some traits:
more men than women rate physical attractiveness as “very important”
and more women than men rate financial resources and strength of
family bonds as “very important”. The frequent use of the response
“very important” produces highly skewed score distributions for most
items and pre-empts the use of traditional analytic tools built on Gaussian
distributions (such as factor analysis). Sampling strategies involving
“forced choices”, in which respondents rank items in order of importance
or allocate points to each item out of a fixed total, may help to circumvent
such ceiling effects. The ordered logit and item response models we use
here are not troubled by ceiling effects, provided that item scores are at
least moderately variable.
Fig. 2 aggregates responses for each item across individuals by sex, in
effect breaking each person's multivariate response profile (consisting
of a sequence such as financial resources = “very important”, physical
attractiveness = “fairly important”, etc.) into univariate parts. Looking
at Fig. 2, it could be imagined that a small number of common response
profiles are shared by many individuals, but this is not the case: each of
the n = 300 response profiles in the sample is unique. We understand
this to mean that preferences do indeed vary across the sample, and
that individuals evaluated how their preferences differ across traits.
3.2. Model comparisons
In Table 4 we display pseudo Bayes model weights, which allow us
to compare the relative support for each model. Our principle finding
is that the prioritized trait models (I–III) and unidimensional (1-D) inte-
grated trait models (IV–VI) are not competitive when compared to the
two dimensional (2-D) integrated trait models (VII–IX; Table 4). Of
the 2-D models, model VII, with random effects for community, is the
best-supported model. Notably, our best model does not include the
respondent's sex as a predictor of preferences, even though sex has
been found important in many previous studies of mate choice. An
additional surprise is that models containing the adult sex ratio as a
predictor are relatively uncompetitive, in spite of recent findings
showing a relationship between mating effort and ASR in this popula-
tion (Schacht & Borgerhoff Mulder, 2015).
3.3. What about sex differences?
In mate-choice studies, sex is often argued to be a primary determi-
nant of reproductive strategies. Here we examine this claim, investigate
the relationship between sex and trait preferences in our sample and
offer an explanation as to why sex is not a predictor in our best model.
3.3.1. Prioritized trait models
According to the prioritized trait approach, each trait is one of a list of
independently relevant individual characteristics. Within this approach
we measure the effects of three variables (community, sex, and ASR;
Table 2). Based on logCPO values (Table 4), model II, with a unique
Table 4
Posterior predictive calculations and model weights.
LogCPO is calculated for each model as described in S4. Models are nested within classes; for
example, model IV is nested within models V and VI in the class of integrative 1-dimensional
models. Δlog CPO is the natural logarithm of the pseudo Bayes factor (a type of evidence
ratio) for each model compared to model VII, the best-supported model in the set. Pseudo
Bayes weights are calculated from Δlog CPO as described in S4. Model weights may
exaggerate the support for model VII in comparison to the other two-dimensional models
(VIII and IX). The logCPO scores of the 2-D models are very similar; however model VII is
the best model by a small amount.
 R. Schacht, M. Grote / Evolution and Human Behavior 36 (2015) 456–466
Fig. 3. Model II, prioritized traits by sex: posterior means and 95% credibility intervals for
the linear predictor (ζij) in model II, evaluated for male subjects. For each item the uncer-
tainty about the linear predictor has two components: 1) uncertainty about the sex effect
and 2) uncertainty about the community random effect. Here we distinguish these sources
of uncertainty: the dark gray segment shows the 95% credibility interval for the sex effect
and the lighter gray segment shows the 95% credibility interval for the community
random effect, averaged across communities. We recover sex differences for three
items: male subjects value physical attractiveness in a partner more than female subjects,
but the converse holds for strength of family bonds and financial resources. For these items
the credibility intervals for the sex effects are bounded away from zero.
sex effect for each item, is the best of the three prioritized trait models
(ordering the logCPOs of models I–III from most- to least-negative
produces an ordering of least-supported to best-supported models).
This is a reassuring finding given the apparent sex differences in several
item score distributions (Fig. 2). In Fig. 3 we display estimated linear
predictors for item scores in model II. For a given item the linear
predictor is the sum of a community effect and a sex effect (see S1).
Items rated higher by males tend to have positive linear predictors
whereas those rated higher by females tend to have negative linear
predictors. Although most of the variation in item scores is attributable
to community effects, we nonetheless recover sex differences for three
items: strength of family bonds, financial resources, and physical attrac-
tiveness (the first two have a higher female rating and the last has a
higher male rating). For each of these items, the sex effect (αi), an addi-
tive term in the linear predictor, is statistically well supported and
distinguishable from zero. We have thus confirmed, within the priori-
tized trait approach, two key predictions of PI theory following Buss
et al. (1990): females prefer partners who can provide resources, and
males prefer physically attractive, fertile partners. While high female
ratings for strength of family bonds are not suggested by PI theory,
they do fit quite neatly with a general reliance of Makushi women on
extended kin to meet household resource needs.
3.3.2. Integrated trait models: one latent dimension
Under the 1-D approach, item responses are functions of a single
latent variable capturing interrelated preferences. Among the 1-D
models, the model with community and sex effects (model V) is best
supported by logCPO (Table 4). Here the apparent utility of sex as a
predictor resembles our finding from the prioritized trait approach.
To continue our evaluation of sex and its effects on preferences, we
turn to Fig. 4. Here we show the response probability curves for the
item physical attractiveness in the top panel and densities of the latent
variable by sex and location in the bottom. Physical attractiveness was
selected because mate-choice studies typically find a robust sex
461
Fig. 4. Model V, 1-D sex: response probability curves for the item physical attractiveness
along with community- and sex-specific densities for the latent trait. For each value on
the horizontal axis, the curves of the upper panel give the probabilities of each response
option; these probabilities sum to 1. To obtain the lower panel we calculated the posterior
mean of the linear predictor for each individual (λj) from model V (individual positions
are represented by ticks along the axis, females in black and males in gray). We then
estimated kernel densities of the λjs for each sex in each community (16 densities in all,
females in black and males in gray) and superimposed these above the axis. In model V
there is a single axis and a single male:female contrast α, therefore the lower panel is
the same for all traits. The item response curves displayed in the upper panel will vary
in appearance across items, depending on discrimination (βi) and cut point parameters
(see S1).
difference in preferences for this trait, and because we found a sex effect
for this item using the prioritized trait approach. For each value on the
horizontal axis (lower panel), the curves of the upper panel give the
probabilities of each response option. Therefore a person with a position
close to 0 on the latent trait will respond “very important” with proba-
bility near 0.5, and “fairly important” with probability near 0.3. The
superimposed densities of the bottom panel show where male and
female respondents of each community lie on the axis. Left-to-right
shifts of the densities reflect sex and community differences in place-
ment along the axis. The response probability curves for other items
will vary depending on item-specific parameters; but because the latent
variable and the male–female contrast α are shared across items in the
1-D model, the lower panel is the same for all items.
We see considerable overlap in the distributions of males and
females across communities. If male–female differences were large
and communities were not important sources of variation, we would
see two sets of distributions, well separated by sex but indistinguishable
by community. Instead, in this sample demanding individuals of either
sex find physical attractiveness “very important” in a partner. Thus, if
the goal is to predict the importance of physical attractiveness, the sex
of the respondent is of limited value.
3.3.3. Integrated trait models: two latent dimensions
Fig. 5 shows response probability curves for physical attractiveness
and densities for the two latent dimensions of the 2-D model with sex.
As a feature of the model, the response probability curves for physical
attractiveness, as well as the densities, are different for dimensions
one and two. As for the 1-D model, within a dimension the latent
variable and the male–female contrast are shared across all items,
while the response probability curves vary across items as a function
of item-specific parameters. A slight separation between male and
female densities is present in dimension 1, but sex differences are
462 R. Schacht, M. Grote / Evolution and Human Behavior 36 (2015) 456–466

Fig. 5. Model VIII, 2-D sex: response probability curves for the item physical attractiveness along with community- and sex-specific densities for the latent variables on dimensions 1 and 2.
These graphs display model VIII, but details of their construction are as in Fig. 4.

difficult to detect in dimension 2. Physical attractiveness tends to be individuals in the four quadrants (indicated by light grey vertical and
“very important” to individuals positive on dimension 1; interestingly, horizontal lines). We label the horizontal dimension (D-1) “general
a few female respondents occupy the positive extremes of this dimen- demand” because positively placed individuals are more likely to use
sion. The polarity is reversed along dimension 2, where individuals in the response option “very important” when rating items. However,
the positive extremes appear to be relatively indifferent to physical individuals negatively placed in D-1 are not simply undemanding, but
attractiveness in a potential partner. As in the 1-D model, sex is of instead are selective in their preferences, rating some traits “very
limited value in predicting placement along the axes. important” but other traits notably less important. Therefore the two
dimensions should not be interpreted independently of each other.
3.4. The “best” model: 2-dimensions with community effect We label the vertical dimension (D-2), “natal tendencies”, with
individuals positive on the dimension rating the desire for children as
While we have focused on the predictor sex, our best model (model “very important”.
VII) does not include it. To understand this result graphically we turn to
Fig. 6. Here we plot individual placement along the two dimensions, for
the 2-D model including only random intercepts for community. We do
not see distinct groupings by sex, as would be expected if the sex of the
respondent were informative about their position in the latent space.
Instead the latent positions of males and females appear to be well
mixed across the two dimensions. We return to model VII interpretively
in the next section.

4. Discussion

Our findings address two questions outlined at the beginning: 1) How

can we better describe the decision-making process of individuals
choosing mates (prioritized vs. integrated traits)? and 2) Which variables
(community, sex, ASR or some combination thereof) best predict mate
preferences? In answer to question 1, and in line with researcher calls
for exploring the interaction of multiple cues (e.g., Miller & Svensson,
2014), we find strong support for preferences built upon trait integration.
Regarding question 2, we detect meaningful variation in preferences,
but find that neither the traditional nor reformulated models of sexual
selection persuasively explain the observations.
Our best model (model VII) includes community-level random
intercepts within two dimensions of preference. Fig. 7 gives a schematic Fig. 6. Model VII, 2-D community: individuals are plotted by posterior means of the latent
interpretation of model VII based on response profiles for virtual traits (λj, φj) and shaded by sex.
 R. Schacht, M. Grote / Evolution and Human Behavior 36 (2015) 456–466
Fig. 7. Preferences of virtual individuals in each of the four quadrants: the more positive an
individual is on D1, the more likely they are to find nearly all traits “very important”. The
more positive an individual is on D2, the more likely they are to find desire for children as a
“very important” trait in a partner.
In general, individuals positive on D-1 rate traits “very important”.
Here are our interpretations of these individuals based on their place-
ment along D-2:
(1, 1) Individuals positive on both dimensions find all items “very
important”. These are demanding individuals wanting a partner
who “has it all”.
(1, − 1) Individuals positive on D-1 but negative on D-2 find all
items “very important” except desire for children, which is “not at
all important”. A desire for a family is not driving these individuals'
mate choice decisions, partner quality is.
Individuals negative on D-1, are not demanding across all items.
However, this is not to say that they are generally unchoosy: devotion,
parenting qualities, health and hardworking are consistently rated as
“very important”. However the importance of other traits depends on
their placement on D-2:
(−1, 1) Individuals negative on D-1 but positive on D-2 find, in addi-
tion to the traits outlined above, desire for children and faithfulness in
a partner “very important”. We understand these individuals to prefer
mates exhibiting “Partner and Child Commitment”.
(− 1, − 1) Individuals negative on both dimensions, find family
bonds “very important” and desire for children “not at all impor-
tant”. We understand these individuals to prefer “Kin Support”.
What we uncover through an examination of the quadrants is not a
simple story about choosy vs. unchoosy individuals. Instead, and espe-
cially for those negative on D-1, we see mixed preferences, with some
key traits being “very important” while others are not.
It is possible that more than two dimensions are needed to fully
characterize mate preferences. A preliminary check in this sample
suggested, however, that very little could be learned by adding a third
dimension. In trial computing experiments, we found that person- and
item-specific parameters for a third dimension reproduced those of
the first or second dimension (bringing to mind a repeating mirror).
Although we cannot rule out the existence of additional dimensions of
mate preference (especially in other samples) we believe that two
dimensions are adequate for these data, and that three or more dimen-
sions would unduly strain the available variation.
463
In response to question 2, we find that community random effects
best explain patterning in preferences. Sex is often argued to be a
primary determinant of reproductive strategies due to potential differ-
ences in optimal mating rates, leading women to place a greater impor-
tance on cues that signal ability to acquire resources and men to place a
greater importance on cues that signal fertility and reproductive value
(e.g., Buss, 1989). Had we concluded our analysis after fitting only the
prioritized trait models, we would have confirmed yet again the impor-
tance of sex (model II). Sex differences in three traits align in the pre-
dicted direction (see Fig. 3).
Sex differences are not unimportant in Makushi reproductive strate-
gies, but they are likely not as straight-forward as conventionally
argued. The Makushi offer a useful comparison to “typical” studies of
mate choice focusing on college undergraduates in western industrial-
ized settings (Henrich et al., 2010). In the west, cultural institutions
create traditional sex roles where males control most of the resources
(Eagly & Wood, 1999). Unsurprisingly, this is also associated with
robust and consistent sex differences across mate choice studies.
The Makushi are quite different. Kinship is matrilineal, marriage is
matrilocal, and men must perform bride-service in order to marry
(Schacht, 2013). Makushi women generally provide most calories for
subsistence, and as a consequence sex roles are quite egalitarian, with
women having considerable power in household decision-making.
The relative economic autonomy of Makushi women may explain why
preferences do not diverge strongly across the genders; instead,
situationally-dependent preferences may be adopted by both males
and females (Fig. 7).
Similar work among the Shuar of Ecuador finds that resource
preferences are patterned by community market integration
(Pillsworth, 2008). Like the Makushi, the Shuar are traditionally matri-
lineal and matrilocal with subsistence centering on cassava propagation
and female food production and processing. Pillsworth found that in a
traditional village, where families depend on women's food production,
men expressed a greater preference for partner resource provisioning
than did women. In the market integrated village, where families
were heavily dependent on purchased goods, western-style ‘traditional’
preferences emerged. Therefore, put plainly, preferences for a resource-
provider do not seem to be rigid and simply based on sex, but instead
vary facultatively based on who provides the bulk of the resources in
a particular community.
Newer sexual selection models show as well that sex does not auto-
matically act as a predictor of behavior, because sex roles coevolve with
sex-structured pay-offs to mating strategies, time spent in parental care
and sex-biased mortality differentials (Kokko et al., 2012). We hypothe-
sized that sex might be an important predictor of preferences in interac-
tion with ASR, based on earlier work (Schacht & Borgerhoff Mulder,
2015) investigating mating effort among males. However, ASR contrib-
uted essentially nothing to the models here. It seems that partner
preference is less sensitive than mating effort to the availability of part-
ners. Why? This is an open question, but it may be that the traits of an
ideal mate are those associated with successful long-term relationships.
These traits may be related to ecological conditions that are possibly less
labile over time, thus unlikely to change with short-term fluctuations in
ASR. What we can say, however, is that predictions based only on sex
and partner availability are inadequate.
Social scientists should welcome this attention to variable sex roles
insofar as cross-cultural variability in human reproductive and mating
behavior seems to be the rule rather than the exception (Brown et al.,
2009). It has long been argued that sexual stereotyping arises from
models inappropriately linking men's and women's reproductive
strategies to the constraints of parental care (Eagly & Wood, 1999;
Gowaty, 1997; Hrdy, 1997). Several studies have tested alternative
models for mate preferences and have found that income inequality
(Eagly & Wood, 1999), cultural affiliation (Boyd & Silk, 2006), self-
ratings (Buston & Emlen, 2003), socioeconomic status (Buss, 1989)
and trait matching (McClintock, 2014) are often better predictors of
464 R. Schacht, M. Grote / Evolution and Human Behavior 36 (2015) 456–466
preferences than gender. Thus, the apparent robustness of gender
differences may be an artifact of the focus on gender at the expense of
other, possibly more meaningful, variables.
Subsequently to our main investigation, we explored three
individual-level variables: age, income, and relationship status (the
latter at the suggestion of a reviewer); and one community-level
variable: market integration. These post-hoc checks were not included
in our planned analyses, so we understand them to serve an interpre-
tive, rather than inferential, function.
Age and income did not appear to be related to placement in the
D-1/D-2 latent space, but some patterning with relationship status
can be seen (Fig. 8). Single individuals tend to be negative on D-1, there-
fore perhaps less demanding of a potential mate, having a smaller set of
traits that are deemed to be “very important”. Both Makushi men and
women desire to secure a long-term marriage partner (Myers, 1993;
Schacht, 2013). A stable partner is not only necessary to meet house-
hold needs according to the gender division of labor, but also to gain
respect from community members as a result of reaching adult status.
Unmarried individuals in our sample may have traits that make them
less attractive as marriage partners, in contrast to married individuals
who were desirable enough to secure a mate. Numerous studies
have found that individuals express preferences for, and pair with,
mates of similar value across traits (Buss, 1989; Buston & Emlen, 2003;
McClintock, 2014; Stevens, Owens, & Schaefer, 1990). Additionally,
individuals who score favorably on a particular trait are even more dis-
criminating in their preferences for that trait in a partner (Buss, 1989;
Buston & Emlen, 2003). Positive assortative mating (reviewed in Buller,
2005) for mate quality may help to explain why single individuals in our
sample appear to have less stringent preferences.
Next, we turned to a community level variable: market integration
(i.e. reliance on wage labor), measured as community distance to the
regional capital. A display of individuals, shaded by community
membership, reveals moderate but detectable clustering of light and
dark points (Fig. 9). In the lower left quadrant we see lighter shading:
these individuals live in communities farther away from wage labor
opportunities. They are also characterized by desiring kin support
(i.e. they rate strength of family bonds to be “very important”; Fig. 7).
The lack of available wage labor opportunities, which can provide a
buffer in times of resource shortage, may cause individuals in these
communities to be more reliant on their extended families to meet
household needs. Turning to the upper right quadrant, we find a greater
Fig. 8. Model VII, 2-D community: individuals are plotted by posterior means of the latent
traits (λj, φj) and shaded by relationship status.
Fig. 9. Model VII, 2-D community: individuals are plotted by posterior means of the latent
traits (λj, φj) and shaded by distance between their community of residence and the
regional capital. Darker circles indicate communities closer to the capital. Distances
between communities and the regional capital range from 3 km (darkest grey circles) to
100 km (lightest grey).
concentration of darker points (these individuals tend to be in closer
proximity to wage labor). These are also individuals who tend to rate
all items as “very important”. These preferences may be in response to
relatively recent market integration and the need to pay attention to
many traits because of uncertainty about the relationships between
traits and mate quality in novel environments (see Candolin, 2003).
Additionally, individuals in communities closer to the regional capital
are generally more pronatal. This is surprising and is counter to typical
predictions from both competitive market and demographic transition
models (e.g., Kaplan 1996). Generally, as population measures of market
integration increase, natal preferences decrease due to the importance of
education and rising costs to having multiple children. However, among
the Makushi, success is largely independent of education. Most jobs
available for men and women are low-skill, wage labor positions. These
jobs therefore are new-found sources of wealth that can be used to
support more children. However, over time, as success becomes more
tightly linked to education, predictions from competitive market and de-
mographic transition models may hold, as individuals begin investing
more in fewer children as opposed to many.
In sum, neither sex nor its interaction with ASR is included in our
best model; this appears to be because community-to-community
variation dominates other factors predicting mate preferences. While
the drivers of community-level variability are yet unknown, we do
uncover “types” of preferences from the two-dimensional model.
Some individuals make very few concessions when choosing a partner,
whereas others are selective, targeting differing combinations of the 10
traits. Additionally, or more precise, measures of individual- (e.g. mate
value) and population-level (e.g. market integration) variables may
help to more clearly explain trait preferences and their integration
into the patterns we observe.
5. Conclusion
Integrative trait models allow us to ask which variables predict
preferences as well as examine relationships between preferences for
different traits. We recover two underlying dimensions explaining
variation in preferences: a general-demand dimension and a pronatal
dimension. A model in which unstructured community-level effects
 R. Schacht, M. Grote / Evolution and Human Behavior 36 (2015) 456–466
predict individual placement on these dimensions is better-supported
than models incorporating more familiar predictors of trait preferences.
Changes in social and ecological environments, as well as uncertainty
about traits that accurately signal partner quality, may prompt individuals
choosing a mate to use different combinations of traits in different
contexts. These ephemeral sources of variation may contribute to the
low repeatability, or what may be labeled ‘noise’, in preferences that we
see across mate choice studies (Candolin, 2003; Jennions & Petrie,
1997). Our study highlights the need to incorporate unstructured varia-
tion (random effects) in predictive models along with more traditional in-
dividual and contextual covariates (fixed effects). Unstructured variation
could arise from community-level heterogeneity as here, or from unmea-
sured time-varying factors in a longitudinal study (Chaine & Lyon, 2008).
Optimal mating strategies are likely influenced by sex-specific pay-
offs to behaviors, but assortative mating and situationally-dependent
factors could result in an apparent lack of sex differences in preferences.
The robust literature surrounding sex differences largely makes use of a
prioritized trait approach that may miss much of the relevant complexity
underlying partner choice decision-making. We do not claim that sex
differences found in previous studies are ‘wrong’, as far as they go.
But, as we have shown here, these studies may be methodologically
incapable of revealing the rich tapestry of factors necessary to predict
situationally-dependent mating behavior.
We have investigated an alternative model for mate choice in
humans (the integrated trait model), and have compared it to a more
conventional prioritized trait model, using computational Bayesian
methods. In this sample the integrative model performs markedly
better than the prioritized trait model. We believe that the underlying
psychology shaping mate choice is unknown and open to investigation;
however, we offer the integrative approach as a promising way to
understand mate choice. The weak support for sex differences is not
unexpected in this sample, given the background ethnography and
previous findings for this population (Schacht & Borgerhoff Mulder,
2015). Perhaps most interesting is the fact that the patterning of mate
choice may be related to community membership — possibly more
specifically to market-economy access — suggesting a direction for
further investigation.
Supplementary materials
Supplementary data to this article can be found online at http://dx.
doi.org/10.1016/j.evolhumbehav.2015.05.001.
Acknowledgements
We are grateful to J. Schacht who provided fundamental contribu-
tions to the fieldwork. We thank M. Borgerhoff Mulder, J. Scheib, and
B. Winterhalder for comments on the manuscript, and the Human
Behavioral Ecology and Cultural Evolution lab groups at UC Davis for
advice. Lastly we thank the Makushi of south-western Guyana
who opened up their homes to us. We acknowledge financial support
from the National Science Foundation 0962440 and the Wenner-Gren
Foundation.
References
Andersson, M. (1994). Sexual selection. Princeton, NJ: Princeton University Press.
Aspendorf, J. B., & Penke, L. (2005). Sex differences: Empiricism, hypothesis testing, and
other virtues. Behavioral and Brain Sciences, 28, 275–276.
Bateson, P. (1983). Mate choice. New York: Cambridge University Press.
Beckerman, S. (2005). Sociosexual strategies in tribes and nations. Behavioral and Brain
Sciences, 28, 276–277.
Bergstrom, C. T., & Real, L. A. (2000). Towards a theory of mutual mate choice: Lessons
from two-sided matching. Evolutionary Ecology Research, 2(4), 493–508.
Borgerhoff Mulder, M. (2009). Serial monogamy as polygyny or polyandry? Marriage in
the Tanzanian Pimbwe. Human Nature, 20(2), 130–150.
465
Botwin, Michael D., Buss, David M., & Shackelford, Todd K. (1997). Personality and mate
preferences: Five factors in mate selection and marital satisfaction. Journal of
Personality, 65(1), 107–136.
Boyd, R., & Silk, J. (2006). How humans evolved. New York: W.W. Norton.
Bro-Jorgensen, J. (2010). Dynamics of multiple signalling systems: Animal communica-
tion in a world in flux. Trends in Ecology & Evolution, 25(5), 292–300.
Brown, Gillian R., Laland, Kevin N., & Mulder, Monique Borgerhoff (2009). Bateman's
principles and human sex roles. Trends in Ecology & Evolution, 24(6), 297–304.
Brunswik, E. (1955). Representative design and probabilistic theory in a functional
psychology. Psychological Review, 62(3), 193–217.
Buller, David J. (2005). Adapting minds: Evolutionary psychology and the persistent quest for
human nature. Cambridge, Mass: MIT Press.
Buss, D. M. (1989). Sex-differences in human mate preferences - Evolutionary hypothesis
tested in 37 cultures. Behavioral and Brain Sciences, 12(1), 1–14.
Buss, David M., et al. (1990). International preferences in selecting mates: A study of 37
cultures. Journal of Cross-Cultural Psychology, 21(1), 5–47.
Buston, P. M., & Emlen, S. T. (2003). Cognitive processes underlying human mate choice:
The relationship between self-perception and mate preference in western society.
Proceedings of the National Academy of Sciences of the United States of America,
100(15), 8805–8810.
Candolin, U. (2003). The use of multiple cues in mate choice. Biological Reviews, 78(4),
575–595.
Chaine, A. S., & Lyon, B. E. (2008). Adaptive plasticity in female mate choice dampens sexual
selection on male ornaments in the lark bunting. Science, 319(5862), 459–462.
Clutton-Brock, Tim (2007). Sexual selection in males and females. Science, 318(5858),
1882–1885.
Dale, S., & Slagsvold, T. (1996). Mate choice on multiple cues, decision rules and sampling
strategies in female pied flycatchers. Behaviour, 133, 903–944.
Eagly, A. H., & Wood, W. (1999). The origins of sex differences in human behavior -
Evolved dispositions versus social roles. American Psychologist, 54(6), 408–423.
Emlen, S. T., & Oring, L. W. (1977). Ecology, sexual selection, and the evolution of mating
systems. Science, 197(4300), 215–223.
Forte, J. A. (1996). About Guyanese Amerindians. Georgetown, Guyana: J. Forte.
Fromhage, Lutz, Elgar, Mark A., & Schneider, Jutta M. (2005). Faithful without care: The
evolution of monogyny. Evolution, 59(7), 1400–1405.
Gafar, J. (2004). The performance of the labour market in Guyana. Canadian Journal of
Development Studies-Revue Canadienne D Etudes Du Developpement, 25(4), 663–685.
Gangestad, S. W., & Scheyd, G. J. (2005). The evolution of human physical attractiveness.
Annual Review of Anthropology, 34, 523–548.
Gangestad, S. W., & Simpson, J. A. (2000). The evolution of human mating: Trade-offs and
strategic pluralism. Behavioral and Brain Sciences, 23(4), 573–587.
Gardner, H. (1983). Frames of mind: The theory of multiple intelligences. New York: Basic Books.
Gelfand, A. E., Dey, D. K., & Chang, H. (1992). Model determination using predictive distri-
butions with implementation via sampling-based methods. In J. M. Bernardo, J. O.
Berger, A. P. Dawid, & A. F. M. Smith (Eds.), Bayesian Statistics 4 (pp. 147–167). Oxford:
Oxford University Press.
Gerhardt, H. C. (1992). Conducting playback experiments and interpreting their results. In
M. P.K. (Ed.), Playbacks and Studies of Animal Communication. New York: Plenum Press.
Gowaty, P. A. (Ed.). (1997). Feminism and evolutionary biology - Boundaries, intersections
and frontiers. New York: Springer.
Gray, R. D., Heaney, M., & Fairhall, S. (2003). Evolutionary psychology and the challenge of
adaptive explanation. In K. Sterelny, & J. Fitness (Eds.), From Mating to Mentality:
Evaluating Evolutionary Psychology (pp. 247–268). New York: Psychology Press.
Griffiths, P. E. (2001). From adaptive heuristic to phylogenetic perspective: Some lessons
from the evolutionary psychology of emotion. In H. R. HolcombIII (Ed.), Conceptual
Challenges in Evolutionary Psychology, Innovative Research Strategies: Studies in
Cognitive Systems (pp. 309–325). Dordrecht, Netherlands: Kluwer Academic Publishers.
Grilli, L., & Rampichini, C. (2003). Alternative specifications of multivariate multilevel
probit ordinal response models. Journal of Educational and Behavioral Statistics,
28(1), 31–44.
Hambleton, R. K., Swaminathan, H., & Rogers, H. J. (1991). Fundamentals of item response
theory. Newbury Park, CA: Sage Press.
Henrich, Joseph, Heine, Steven J., & Norenzayan, Ara (2010). The weirdest people in the
world? Behavioral and Brain Sciences, 33(2–3), 61–83.
Hill, J. A., et al. (1999). Mate choice based on static versus dynamic secondary sexual traits
in the dark-eyed junco. Behavioral Ecology, 10(1), 91–96.
Hooper, Paul L., & Miller, Geoffrey F. (2008). Mutual mate choice can drive costly signaling
even under perfect monogamy. Adaptive Behavior, 16(1), 53–70.
Hrdy, Sarah Blaffer (1997). Raising Darwin’s consciousness. Human Nature, 8(1), 1–49.
Jennions, M. D., & Petrie, M. (1997). Variation in mate choice and mating preferences: A
review of causes and consequences. Biological Reviews of the Cambridge
Philosophical Society, 72(2), 283–327.
Johnstone, Rufus A., Reynolds, John D., & Deutsch, James C. (1996). Mutual mate choice
and sex differences in choosiness. Evolution, 50(4), 1382–1391.
Kodric-Brown, A., & Nicoletto, P. F. (2001). Female choice in the guppy (Poecilia
reticulata): The interaction between male color and display. Behavioral Ecology and
Sociobiology, 50(4), 346–351.
Kokko, Hanna, & Jennions, Michael D. (2008). Parental investment, sexual selection and
sex ratios. Journal of Evolutionary Biology, 21(4), 919–948.
Kokko, H., & Johnstone, R. A. (2002). Why is mutual mate choice not the norm? Opera-
tional sex ratios, sex roles and the evolution of sexually dimorphic and monomorphic
signalling. Philosophical Transactions of the Royal Society of London Series B-Biological
Sciences, 357(1419), 319–330.
Kokko, H., Klug, H., & Jennions, M. D. (2012). Unifying cornerstones of sexual selection:
Operational sex ratio, Bateman gradient and the scope for competitive investment.
Ecology Letters, 15(11), 1340–1351.
466 R. Schacht, M. Grote / Evolution and Human Behavior 36 (2015) 456–466
Laland, K. N., & Brown, G. R. (2011). Sense and nonsense: Evolutionary perspectives on
human behaviour. Oxford: Oxford University Press.
Li, Norman P., et al. (2002). The necessities and luxuries of mate preferences: Testing the
tradeoffs. Journal of Personality and Social Psychology, 82(6), 947–955.
Lindstrom, D. P., et al. (2010). Nonmarital sex and condom knowledge among Ethiopian
young people: Improved estimates using a nonverbal response card. Studies in
Family Planning, 41(4), 251–262.
Lippa, R. A. (2007). The preferred traits of mates in a cross-national study of heterosexual
and homosexual men and women: An examination of biological and cultural influ-
ences. Archives of Sexual Behavior, 36(2), 193–208.
Marlowe, Frank W. (2004). Mate preferences among Hadza hunter-gatherers. Human
Nature, 15(4), 365–376.
McClintock, Elizabeth Aura (2014). Beauty and status: The illusion of exchange in partner
selection? American Sociological Review.
Merlo, J., et al. (2005). A brief conceptual tutorial of multilevel analysis in social epidemiology:
Linking the statistical concept of clustering to the idea of contextual phenomenon.
Journal of Epidemiology and Community Health, 59(6), 443–449.
Miller, G. F. (1997). Mate choice: From sexual cues to cognitive adaptations. In C.
Foundation (Ed.), Characterizing Human Psychological Adaptations (pp. 71–82). Wiley.
Miller, Christine W., & Svensson, Erik I. (2014). Sexual selection in complex environments.
Annual Review of Entomology, 59(1), 427–445.
Moller, A. P., et al. (1998). Paternity and multiple signaling: Effects of a secondary sexual
character and song on paternity in the barn swallow. American Naturalist, 151(3),
236–242.
Myers, I. (1993). The Makushi of the Guiana-Brazilian Frontier in 1944: A study of culture
contact. Antropologica, 80, 3–99.
Nettle, D. (2009). Ecological influences on human behavioural diversity: A review of
recent findings. Trends in Ecology & Evolution, 24(11), 618–624.
Nettle, D., Coall, D. A., & Dickins, T. E. (2011). Early-life conditions and age at first pregnancy
in British women. Proceedings of the Royal Society of London. Series B-Biological Sciences,
278(1712), 1721–1727.
Ntzoufras, I. (2009). Bayesian modeling using WinBugs. New Jersey: John Wiley & Sons.
Owens, Ian P. F., & Thompson, Desmond B. A. (1994). Sex differences, sex ratios and sex
roles. Proceedings of the Royal Society of London. Series B-Biological Sciences,
258(1352), 93–99.
Pedersen, Frank (1991). Secular trends in human sex ratios. Human Nature, 2(3), 271–291.
Pillsworth, Elizabeth G. (2008). Mate preferences among the Shuar of Ecuador: Trait
rankings and peer evaluations. Evolution and Human Behavior, 29(4), 256–267.
Scelza, B. A. (2011). Female choice and extra-pair paternity in a traditional human popu-
lation. Biology Letters, 7(6), 889–891.
View publication stats
Schacht, R. (2013). Cassava and the Makushi: A shared history of resiliency and transfor-
mation. In H. Garth (Ed.), Food and Identity in the Caribbean (pp. 15–29). London: Berg
Publishers.
Schacht, R., & Borgerhoff Mulder, M. (2015). Sex ratio effects on reproductive strategies in
humans. Royal Society Open Science, 2.
Schacht, Ryan, Rauch, Kristin Liv, & Mulder, Monique Borgerhoff (2014). Too many men:
The violence problem? Trends in Ecology & Evolution, 29(4), 214–222.
Scheib, J. E. (2001). Context-specific mate choice criteria: Women's trade-offs in the
contexts of long-term and extra-pair mateships. Personal Relationships, 8(4),
371–389.
Schmitt, D. P. (2005). Sociosexuality from Argentina to Zimbabwe: A 48-nation study of
sex, culture, and strategies of human mating. Behavioral and Brain Sciences, 28(2),
247–311.
Scott, Isabel M., et al. (2014). Human preferences for sexually dimorphic faces may be
evolutionarily novel. Proceedings of the National Academy of Sciences, 111(40),
14388–14393.
Shackelford, T. K., Schmitt, D. P., & Buss, D. M. (2005). Universal dimensions of human
mate preferences. Personality and Individual Differences, 39(2), 447–458.
Smith, E. A., Borgerhoff Mulder, M., & Hill, K. (2001). Controversies in the evolu-
tionary social sciences: A guide for the perplexed. Trends in Ecology & Evolution,
16(3), 128–135.
Spiegelhalter, D. J., et al. (2002). Bayesian measures of model complexity and fit. Journal of
the Royal Statistical Society Series B-Statistical Methodology, 64, 583–616.
Stevens, Gillian, Owens, Dawn, & Schaefer, Eric C. (1990). Education and attractiveness in
marriage choices. Social Psychology Quarterly, 53(1), 62–70.
Stewart, S., Stinnett, H., & Rosenfeld, L. B. (2000). Sex differences in desired characteristics
of short-term and long-term relationship partners. Journal of Social and Personal
Relationships, 17(6), 843–853.
Stogdill, R. M., & Coons, A. E. (Eds.). (1957). Leader behavior: Its description and
measurement. Columbus: Ohio State University Press.
Symons, D. (1979). The evolution of human sexuality. New York: Oxford University Press.
Szekely, T., Webb, J. N., & Cuthill, I. C. (2000). Mating patterns, sexual selection
and parental care: An integrative approach. In M. Apollonio, M. Festa-Bianchet,
& D. Mainardi (Eds.), Vertebrate Mating Systems (pp. 194–223). London: World
Science Press.
Trivers, R. L. (1972). Parental investment and sexual selection. In B. Campbell (Ed.), Sexual
selection and the Descent of Man, 1871–1971 (pp. 136–179). Chicago: Aldine.
Zhu, X. W., & Stone, C. A. (2012). Bayesian comparison of alternative graded response
models for performance assessment applications. Educational and Psychological
Measurement, 72(5), 774–799.