Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

Biol. Rev. (2019), 94, pp. 1430–1442.

1430
doi: 10.1111/brv.12509

All by myself? Meta-analysis of animal


contests shows stronger support for self than
for mutual assessment models
Nelson S. Pinto1 , Alexandre V. Palaoro2 and Paulo E. C. Peixoto3,∗
1
Graduate Program in Ecology, Universidade Federal da Bahia, Salvador, BA 40110-909, Brazil
2
LAGE do Departamento de Ecologia, Universidade de São Paulo, São Paulo, SP 05508-090, Brazil
3
Instituto de Ciências Biológicas, Departamento de Biologia Geral, Universidade Federal de Minas Gerais, Belo Horizonte, MG 31270-901,
Brazil

ABSTRACT

Since the 1970’s, models based on evolutionary game theory, such as war of attrition (WOA), energetic war of attrition
(E-WOA), cumulative assessment model (CAM) and sequential assessment model (SAM), have been widely applied to
understand how animals settle contests. Despite the important theoretical advances provided by these models, empirical
evidence indicates that rules adopted by animals to settle contests vary among species. This stimulated recent discussions
about the generality and applicability of models of contest. A meta-analysis may be helpful to answer questions such
as: (i) is there a common contest rule to settle contests; (ii) do contest characteristics, such as the occurrence of physical
contact during the fight, influence the use of specific contest rules; and (iii) is there a phylogenetic signal behind contest
rules? To answer these questions, we gathered information on the relationship between contest duration and traits
linked to contestants’ resource holding potential (RHP) for randomly paired rivals and RHP-matched rivals. We also
gathered behavioural data about contest escalation and RHP asymmetry. In contests between randomly paired rivals,
we found a positive relationship between contest duration and loser RHP but did not find any pattern for winners. We
also found a low phylogenetic signal and a similar response for species that fight with and without physical contact. In
RHP-matched rivals, we found a positive relationship between contest duration and the mean RHP of the pair. Finally,
we found a negative relation between contest escalation and RHP asymmetry, even though it was more variable than
the other results. Our results thus indicate that rivals settle contests following the rules predicted by WOA and E-WOA
in most species. However, we also found inconsistencies between the behaviours exhibited during contests and the
assumptions of WOA models in most species. We discuss additional (and relatively untested) theoretical possibilities that
may be explored to resolve the existing inconsistencies.

Key words: war of attrition, sequential assessment model, cumulative assessment model, meta-regression, resource-holding
potential, territoriality, agonistic interaction.

CONTENTS
I. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1431
II. Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1434
(1) Eligibility criteria and information sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1434
(2) Search, study selection and data collection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1434
(3) Data items and summary measures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1435
(4) Data analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1435
(a) Relationship between contest duration and traits linked to RHP in randomly paired rivals . . . . . . . 1435
(b) Relationship between contest duration and mean trait values in closely matched rivals . . . . . . . . . . 1435

* Address for correspondence (Tel: +55 (31) 99919-0958; Fax: +55 031 3409-2567; E-mail: pauloenrique@gmail.com).

Biological Reviews 94 (2019) 1430–1442 © 2019 Cambridge Philosophical Society


Meta-analysis of contest resolution rules 1431

(c) Relationship between contest escalation and asymmetry in traits linked to RHP in randomly paired
rivals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1435
(5) Risk of bias across studies and synthesis of results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1436
III. Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1436
(1) Relationship between contest duration and traits linked to RHP in randomly paired rivals . . . . . . . . . . 1436
(a) Analyses of the data set with all variables related to RHP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1436
(b) Analyses of the data set with the best predictors of RHP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1437
(2) Relationship between contest duration and mean trait values in RHP-matched rivals . . . . . . . . . . . . . . . 1437
(3) Relationship between contest escalation and RHP asymmetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1437
IV. Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1438
V. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1439
VI. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1440
VII. References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1440
VIII. Supporting Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1442

I. INTRODUCTION Briffa, 2013; Palaoro & Briffa, 2017). The four assessment
models assume that assymetries in RHP are important in
Animal contests over limited resources are ubiquitous in determining who wins the contest (Briffa & Hardy, 2013).
nature (Hardy & Briffa, 2013). Such contests are frequently However, they also assume differences on how animals gather
initiated when an intruder approaches a resource that is information about RHP during contests.
defended by another individual, and end when one of the According to WOA, E-WOA and CAM, individuals follow
individuals flees. Thus, it is fairly easy to determine when a self-assessment strategy that assumes that individuals do not
contests begin or end. However, what happens between obtain information about the rival’s RHP (Arnott & Elwood,
the beginning and ending of contests, namely the contest 2009). Consequently, each individual should give up when
dynamics, is highly variable among species. For example, its own cost accrual reaches a threshold determined by its
animal contests may consist of stereotyped behaviours own RHP (Arnott & Elwood, 2009). According to WOA, the
with distinct behavioural phases (Green & Patek, 2018), main cost should be the time spent in the interaction, while
endurance displays without physical contact (Peixoto & according to E-WOA, energy expenditure should be the main
Benson, 2011) or even disputes characterized by continuous cost. However, for most situations WOA and E-WOA are
pushing and biting between rivals (Painting & Holwell, 2014; empirically indistinguishable (e.g. when individual size is used
Palaoro et al., 2014). Interestingly, despite these frequent tugs as a proxy of RHP), and for this reason, we will refer to both
of war, contests rarely involve the death of an individual. models as (E)WOA. The last self-assessment strategy, CAM,
Rivals typically terminate the contest before any damage is assumes that costs also accrue from injuries caused by the
done (Lane & Briffa, 2017; Palaoro & Briffa, 2017). Since rival during physical contact (Payne, 1998). Hence, the key
individuals are contesting valuable resources and exposing difference between (E)WOA and CAM is that, in (E)WOA
themselves to damage, the rarity of death indicates that cost accrual is unaffected by the rival’s actions, while in CAM
individuals adopt some sort of criterion to decide when injuries inflicted by rivals affect individual persistence in the
to give up a contest before reaching extreme costs. The contest (Payne, 1998). For this reason, according to (E)WOA,
decision of when to give up on a contest has been modelled the loser’s persistence should be strongly correlated with its
using evolutionary game theory since the 1970’s (Briffa own RHP (such as size or energy reserves) and be unaffected
& Elwood, 2009). However, despite the large number of by the winners’ actions. Under CAM, however, the loser’s
models (Mesterton-Gibbons, Marden & Dugatkin, 1996; persistence is again strongly correlated with its own RHP,
Briffa & Sneddon, 2010; Kokko, 2013; Mesterton-Gibbons & but also negatively correlated to the winner’s RHP because
Heap, 2014), only four of them have been thoroughly tested stronger rivals cause losers to reach their cost threshold faster
empirically: war of attrition (WOA) (Mesterton-Gibbons and, therefore, persist for less time in a fight.
et al., 1996), energetic war of attrition (E-WOA) (Payne & In contrast to the self-assessment strategies, SAM assumes
Pagel, 1996), cumulative assessment model (CAM) (Payne, that rivals are capable of assessing their rivals’ RHP, making
1998) and sequential assessment model (SAM) (Enquist & a decision based on relative RHP, rather than absolute RHP
Leimar, 1983). (Enquist & Leimar, 1983). Thus, each individual decides to
One of the central concepts of all assessment models (i.e. give up when it considers itself to be the weaker rival (and not
WOA, E-WOA, CAM and SAM) is the resource holding when reaching a specific cost threshold). If both individuals
potential (or resource holding power, RHP sensu Parker, have similar RHPs, they should escalate the contest to
1974). The RHP is related to the persistence capacity during costlier behaviours that provide more accurate information
a contest and/or the ability to inflict costs on rivals (Vieira on the rival’s RHP, until one individual identifies itself
& Peixoto, 2013), and is often correlated with morphological as the weaker rival and gives up. Under SAM then, the
and physiological attributes of the individual (Hardy & loser’s persistence will depend on the winner’s RHP and

Biological Reviews 94 (2019) 1430–1442 © 2019 Cambridge Philosophical Society


1432 Nelson S. Pinto and others

its own RHP. This entails a positive relationship between the same variables for winners (Fig. 1). It is important to
loser persistence and loser RHP because, as the loser’s RHP note, however, that for SAM, if individuals have better
increases, it becomes more difficult for the loser to detect information about their own fighting ability than the fighting
that it is weaker than the winner – increasing the length of ability of their rivals, the relationship between fight duration
the contest. Furthermore, there is also a negative relation and individual RHP should be stronger for the losers than
between persistence and the winner’s RHP because, as the for the winners (Prenter, Elwood & Taylor, 2006).
winner’s RHP increases, the easier it will be for the loser The distinction between CAM and SAM is more difficult
to detect that it is weaker than the winner. In summary, to make and cannot be accomplished using the correlations
theory expects that under (E)WOA contest duration should between contest duration and traits linked to RHP (Briffa &
be mainly related to traits linked to the losers RHP. Under Elwood, 2009). At least two additional sources of information
CAM and SAM, however, information about winners (SAM) are needed to distinguish between CAM and SAM. First,
or damage received from them (CAM) also play a role it is necessary to use information about contest duration
in determining when losers should give up. Therefore, and traits linked to RHP for rivals with a high degree of
according to CAM and SAM contest duration should be similarity between these traits (i.e. RHP-matched rivals).
related to both loser and winner RHP. In a hypothetical scenario where it is possible to arrange
Empirical tests of (E)WOA, CAM and SAM have been contests between RHP-matched rivals across all ranges of
developed since 1990 (e.g. Marden & Waage, 1990; Leimar, RHPs in the population, different relations between contest
Austad & Enquist, 1991). However, detailed evaluations duration and the mean RHP of the pair would emerge
of these models were carried out only after the analytical if animals follow CAM or SAM. Since the relative RHP
framework proposed by Taylor & Elwood (2003) (e.g. Kelly, remains the same, under CAM (and also under (E)WOA),
2006; Stuart-Fox, 2006; Peixoto & Benson, 2011). Before this we expect a positive relation between contest duration and
analytical framework, most studies correlated fight duration the mean RHP of the pair, while under SAM, there should
with the difference in trait values between winners and be no relationship between contest duration and the mean
losers. A negative relationship between fight duration and RHP of the pair (Arnott & Elwood, 2009, Fig. 1). Second,
the difference in trait values was considered evidence in the assessment models were built based on differences in
support of SAM, and, in fact, many tests based on such escalation patterns, and for this reason, it is necessary to
correlations found support for SAM (e.g. Enquist et al., 1990; evaluate how individuals are escalating and de-escalating
Leimar et al., 1991; Hack, 1997). However, Taylor & Elwood their contests (Briffa, 2015). According to SAM, the most
(2003) showed that this approach led to biased conclusions. escalated interactions (e.g. change from displays to physical
When individuals are randomly paired, losers with low RHP contact during the contest, Painting & Holwell, 2014) should
have a high chance of fighting against much stronger winners. occur only between individuals with similar RHP (Enquist
This occurs because there is a large set of the population & Leimar, 1983). Therefore, if animals behave according
with individuals with greater RHP values than losers with to SAM, there should be a negative relationship between
low RHP. Losers with high RHP, on the other hand, would escalation probability and the difference in trait values for the
come from a small set of the population that is constituted by pair. However, under CAM, there should be no relationship
large RHP individuals. To lose, these high-RHP individuals between contest escalation and the difference in trait values
would have fought against individuals with even higher between rivals (Payne, 1998, Fig. 1).
RHP. Since there are few individuals at this extreme of the Despite methodological advances in how predictions from
population, losers with high RHP have a high probability of contest models are tested, few studies have found unequivocal
fighting against similar-RHP rivals. In this situation, even support for (E)WOA, CAM or SAM (e.g. Prenter et al.,
if individuals follow a self-assessment strategy, negative 2006; Briffa, 2008; Junior & Peixoto, 2013; Schnell et al.,
correlations should be obtained when regressing contest 2015). Most have found partial support for some of the
duration against trait differences between winners and losers. models, support for more than one model, or patterns that
To avoid finding negative correlations between contest were not predicted by any model (e.g. Palaoro et al., 2014;
duration and trait differences under self-assessment strate- Guillermo-Ferreira et al., 2015). These findings have sparked
gies, Taylor & Elwood (2003) suggested that assessment discussions about the applicability and generality of the
models should be tested using linear regressions between assessment models (Briffa & Sneddon, 2010; Lane & Briffa,
contest duration and winner and loser RHP separately 2017; Palaoro & Briffa, 2017). One largely ignored aspect,
(Taylor & Elwood, 2003; but see Briffa & Elwood, 2009). for instance, is that some assessement rules might be more
According to their framework, under (E)WOA, there should likely to evolve in some groups than in others. For example,
be a positive relationship between contest duration and in species that fight with physical contact, individuals may
traits linked to RHP for losers and no relationship (or be able to cause injuries during the fight. Consequently,
a weak positive one) between contest duration and traits fights settled according to CAM or SAM may be more likely
linked to RHP for winners. According to CAM and SAM, to evolve in species that fight with physical contact, while
on the other hand, there should be a positive relationship (E)WOA may be more common in species that fight without
between contest duration and traits linked to RHP for losers physical contact (although SAM may also explain assessment
and a negative relationship of similar magnitude between rules in some species that fight without physical contact – e.g.

Biological Reviews 94 (2019) 1430–1442 © 2019 Cambridge Philosophical Society


Meta-analysis of contest resolution rules 1433

Fig. 1. Legend on next Page.

Biological Reviews 94 (2019) 1430–1442 © 2019 Cambridge Philosophical Society


1434 Nelson S. Pinto and others

Vilela et al., 2017). Phylogenetic history may also be probability of escalation (i.e. change from less costly to costlier
important in explaining how individuals within a given clade behaviours) during the contest in relation to differences in
decide when to give up. In particular, closely related species traits linked to RHP between rivals. We conducted the
might make contest decisions similarly due to common literature search in ISI Web of Knowledge, Scopus and Google
ancestry. Hence, a systematic evaluation of the empirical tests Scholar between July 2014 and January 2019.
of these models is essential to: (i) show the level of empirical
support for the different assumptions and predictions of the (2) Search, study selection and data collection
assessment models; and (ii) determine if most species tend
to fight according to a specific model or whether different We performed two distinct types of searches. First, we
groups of species follow the rules of different models. conducted a search for articles that provided information
Here, we conducted a meta-analysis to evaluate the on the relationship between contest duration and traits
empirical support for (E)WOA, CAM and SAM. We first linked to RHP for losers and winners using the following
evaluated the empirical support distinguishing (E)WOA from key words: ‘animal contests’, ‘RHP’, ‘resource holding
CAM and SAM. For this, we investigated the relationship potential’, ‘contest resolution’, ‘SAM’, ‘sequential assessment
between contest duration and traits linked to RHP for model’, ‘WOA’, ‘war of attrition’, ‘E-WOA’, ‘energetic war
winners and losers in randomly matched pairs (Fig. 1). of attrition’, ‘CAM’, ‘cumulative assessment model’ and
We then investigated the empirical support for CAM ‘contest duration’. We performed the search using each key
and SAM by evaluating the relationship between contest word separately.
duration and traits linked to RHP for RHP-matched rivals For each selected paper, we first read the title and the
(Fig. 1). Additionally, we investigated the support for SAM abstract searching for information indicating that the study
through the relationship between escalation patterns and investigated contests between pairs of rivals. For studies that
RHP asymmetry (Fig. 1). For all tests we also measured provided this information, we searched the methods for
the phylogenetic signal in the observed patterns and, when evidence that the study evaluated contests according to the
possible, tested if the observed patterns differed between protocol of Taylor & Elwood (2003). If the study used this
species that fight with and without physical contact. protocol, we continued to the results to collect information on
the statistics for the relationships between traits linked to RHP
and contest duration for winners and losers (we considered
II. METHODS as a trait linked to RHP any morphological, physiological or
behavioural trait that was stated in the original article as an
attribute that indicated individual RHP). We also recorded
if the study used randomly paired rivals or RHP-matched
(1) Eligibility criteria and information sources rivals and if physical contact occurs during the contests for
We followed the PRISMA protocol because it provides a each species.
standardized method for reporting meta-analyses (Moher Second, we conducted a search for articles that provided
et al., 2009). Our eligibility criteria for study inclusion results regarding relationships between differences in traits
required that the study presented information on the linked to RHP of pairs of rivals and the probability of
relationships between contest duration and morphological, escalation during the contest. To perform this search,
physiological or behavioural traits linked to RHP for winners we used the following key words separately: ‘animal
and losers in dyadic contests (Taylor & Elwood, 2003). These contests’, ‘RHP’, ‘resource holding potential’, ‘contest
comparisons could involve contests between randomly paired resolution’, ‘SAM’, ‘sequential assessment model’, ‘CAM’,
individuals or individuals systematically paired to minimize ‘cumulative assessment model’ and ‘contest escalation’.
differences in traits linked to RHP (i.e. RHP-matched rivals). Contest escalation may involve escalation both between
We also included papers that provided information about the (i.e. a progressive use of costlier behaviours) and within

Fig. 1. Summary of the responses expected under different assessment models for the relationships between: (i) contest duration and
traits linked to resource holding potential (RHP) for winners and losers (often tested through linear regressions – top row); (ii) contest
duration and mean RHP of the fighting pair (often tested through linear regressions – middle row); and (iii) escalation patterns (i.e.
changes from less costly to costlier behaviours) and trait difference between rivals (often tested through logistic regressions – bottom
row). For each predicted relationship, we depict the expected pattern in the original articles (top figures in each row) and the
expected pattern in our meta-analyses (bottom figure in each row). The expected pattern in our meta-analysis is based on mean
values calculated from data recorded in the original studies and converted to Fisher’s z values (Zr), which represents the effect size
we used. A mean Zr value of zero (represented by dashed lines) indicates that there is no consistent relationship among species
between contest duration or escalation and winners/losers’ traits. (E)WOA represents the expectation for both War of attrition and
Energetic War of Attrition models, CAM represents the expectations for the Cumulative Assessment Model and SAM represent
the expectation for the Sequential Assessment Model. Randomly paired rivals occurred in experiments in which individuals were
randomly paired to induce contests, while RHP-matched pairs occurred in experiments in which rivals were paired to minimize the
difference between the traits linked to RHP.

Biological Reviews 94 (2019) 1430–1442 © 2019 Cambridge Philosophical Society


Meta-analysis of contest resolution rules 1435

(increase in the number or frequency of the same behaviour) most important determinant of RHP. Our second data set,
phases. However, because SAM assumes escalation between hence, only included one trait related to RHP per species.
phases, we only included articles with suitable data to test For each data set, we built a model considering the Zr
this assumption of SAM. In general, escalation between values as the response variable, and status (winner or loser),
phases was represented by the transition from non-contact occurrence of physical contact and the interaction between
behaviours to physical contact and grappling. status and occurrence of physical contact as moderator
As in the search for studies involving the relationship explanatory variables. We included study identity and the
between traits linked to RHP and contest duration, we read type of experiment (laboratory or field) as random variables
the title and the abstract searching for information indicating to account for variation among multiple effect sizes obtained
that the study investigated dyadic agonistic interactions. For for the same study and to account for variation among
studies that provided this information, we proceeded to the studies conducted in laboratory and field conditions. We
methods section to search for information indicating that the also included a variance/covariance matrix based on the
study measured the relationship between contest escalation phylogenetic relationships among species as a random
and traits linked to RHP for rivals. We then continued to variable to account for variation among groups of related
the results section to collect information about statistics for species (Chamberlain et al., 2012). We weighted each Zr
the relationships between the probability of escalation and value by the inverse of the variance of Zr (Borenstein et al.,
differences in values of traits linked to RHP for rivals. 2009). Therefore, in these models we tested if the mean Zr
differed between winners and losers for species that fight
with or without physical contact (Fig. 1). We calculated
(3) Data items and summary measures
the significance tests for moderator variables using the QM
We used Fisher’s z scale (Zr) as a measure of effect size for statistic (Borenstein et al., 2009). If (E)WOA was the best
the relationship between contest duration and traits linked model to explain animal contests, we expected that mean
to RHP. We also used the same effect size for the relation Zr values would be greater than zero for losers, but not for
between contest escalation and traits linked to RHP. Fisher’s winners. If SAM or CAM is the best explanation for animal
z is a normalization of the Pearson correlation coefficient contests, we expected that mean Zr values would be greater
(r) and allows comparisons to be made among studies with than zero for losers, but smaller than zero for winners (Fig. 1).
different sample efforts (Borenstein et al., 2009). To obtain Zr
values, we first obtained the Pearson correlation coefficients (b) Relationship between contest duration and mean trait values in
and then transformed them to Zr (Borenstein et al., 2009). closely matched rivals
When only the coefficient of determination was reported,
we transformed it to r using the square root of the observed To test if there was a relationship between contest duration
value (Nakagawa & Cuthill, 2007). When only F , t, χ 2 and mean RHP for RHP-matched rivals, we used a
values and degrees of freedom were reported, we performed random-effects meta-analysis (Koricheva et al., 2013). In this
transformations to Zr using the package compute.es (Del Re, model, we considered Zr values as the response variable and
2013). When a study provided information in graphs, we inserted the inverse of the variance of each Zr value as the
used the webplotdigitizer software (Rohatgi, 2012) to estimate weight – no moderator (explanatory) variables were used. In
raw values and then calculated the required information. this data set, all species had fights that can involve physical
When some information was unavailable in a paper, either contact. Therefore, for this analysis we only tested if the
explicitly or in graphs, we requested it from the authors. We mean Zr value differed from zero (Fig. 1). As in the previous
retained the signs of all original statistics (indicating positive analysis, we included study identity, type of experiment
or negative relationships) after all the transformations. (laboratory or field) and a variance/covariance matrix based
on the phylogenetic relationships among species as random
variables. Since we did not have moderator explanatory
(4) Data analyses variables, we tested if the mean Zr value differed from zero
using a z-test (Viechtbauer, 2010). According to CAM,
(a) Relationship between contest duration and traits linked to RHP in the mean Zr value should be greater than zero, while
randomly paired rivals according to SAM this value should not differ from zero
(Fig. 1).
To evaluate if contest duration is related to traits linked to
RHP of winners and losers, we performed a mixed-model
meta-regression (Koricheva, Gurevitch & Mengersen, 2013). (c) Relationship between contest escalation and asymmetry in traits
linked to RHP in randomly paired rivals
We performed the meta-regressions for two different data
sets. The first data set contained all effect sizes provided by To evaluate if the probability of escalation is related to
each study. However, since some studies measured more differences in traits linked to RHP between rivals, we used
than one trait related to RHP, we scanned the papers for a random effects meta-analysis (Koricheva et al., 2013).
any analysis or author commentaries stating which was the We considered Zr values as the response variable and
most important RHP trait for that study. We then pruned inserted the inverse of the variance of each Zr value as
our first data set and left only the effect sizes related to the the weight – again no moderators were used. We tested

Biological Reviews 94 (2019) 1430–1442 © 2019 Cambridge Philosophical Society


1436 Nelson S. Pinto and others

if the mean Zr values differed from zero because all Table 1. Heterogeneity I 2 and variance (σ 2 ; V ) measures for
species in this data set also fight with physical contact. the random variables included in each meta-analytical model
As random variables, we included the type of experiment used herein. I 2 values represents the percentage of variability
(laboratory or field) and a variance/covariance matrix based in estimates of effect sizes due to data heterogeneity rather than
on the phylogenetic relationships among species (we did sampling error. I 2 = 25, 50 and 75% indicates low, moderate
not include study identity because only one study provided and high heterogeneity, respectively (Higgins & Thompson,
2002; Higgins et al., 2003)
more than one datum). If rivals fight according to SAM,
we expected the mean Zr value to be smaller than zero
(a) Model 1.1 – randomly matched pairs: duration versus RHP
(Fig. 1). traits
I 2 (95% CI) V (95% CI)
(5) Risk of bias across studies and synthesis of
results Study ID 22.48 (20.56–24.4) 0.05 (0.03–0.09)
Experiment <0.001 (0.0–1.9) <0.001 (0.0–0.02)
To evaluate if there was publication bias in our data set, Phylogeny <0.001 (0.0–1.9) <0.001 (0.0–0.02)
we conducted a modified version of Egger’s test (Egger et al., Total 22.48 (20.56–24.4) —
1997), in which we regressed the residuals of each model
(b) Model 1.2 – randomly matched pairs: duration versus RHP
against the variance of the Zr values (Nakagawa & Santos, traits considering only the most important measure for each
2012). A significant intercept suggests that the publications species
used in our meta-analyses are concentrated toward positive
or negative effect sizes. We estimated the heterogeneity I 2 (95% CI) V (95% CI)
among effect sizes using I 2 and its corresponding 95% Study ID 27.24 (25.32–29.16) 0.06 (0.02–0.12)
confidence intervals for multivariate studies. The I 2 statistic Experiment <0.001 (0.0–1.9) <0.001 (0.0–0.34)
varies between 0 and 100% and can be interpreted as the Phylogeny <0.001 (0.0–1.9) <0.001 (0.0–0.13)
percentage of variability in estimates of effect sizes due to Total 27.24 (25.32–29.16) —
data heterogeneity rather than sampling error (Higgins & (c) Model 2 – RHP-matched rivals: duration versus mean RHP
Thompson, 2002; Higgins et al., 2003). We also calculated
the H 2 metric that estimates the phylogenetic signal in each I 2 (95% CI) V (95% CI)
test (Nakagawa & Santos, 2012). Values of H 2 near 100% Study ID 10.97 (9.05–12.89) 0.02 (0.0–0.09)
indicate a strong phylogenetic signal in the responses we Experiment <0.001 (0.0–1.9) <0.001 (0.0–0.43)
measured. Phylogeny <0.001 (0.0–1.9) <0.001 (0.0–0.10)
We conducted all analyses in R software (R Development Total 10.97 (9.05–12.89) —
Core Team, 2015). We used the metafor package (d) Model 3 – Escalation versus RHP difference including the
(Viechtbauer, 2010) to build and test all models. We used outlier study
the rotl package (Michonneau, Brown & Winter, 2016) to
I 2 (95% CI) V (95% CI)
build a phylogenetic tree with the species present in our
data set [the information to build this tree was obtained Experiment <0.001 (0.0–1.9) <0.001 (0.0–1.03)
from the Tree of Life Project (http://tolweb.org/tree/)]. We Phylogeny 25.83 (23.91–27.75) 0.05 (0.01–0.22)
then used the ape package to convert the topological tree to Total 25.83 (23.91–27.75) —
an ultrametric tree (see online Supporting information, Figs (e) Model 3 – Escalation versus RHP difference excluding one
S1–S3) with simulated branch lengths (Paradis, Claude & outlier study
Strimmer, 2004) in order to include a matrix of phylogenetic
I 2 (95% CI) V (95% CI)
correlations in our models. We present the results as mean
Zr values and 95% confidence intervals. Experiment 8.30 (6.38–10.22) 0.01 (0.0–2.23)
Phylogeny 0.38 (0.0–2.30) <0.001 (0.0–0.07)
Total 8.68 (6.76–10.60) —

III. RESULTS RHP, resource holding potential.

(1) Relationship between contest duration and estimate = −0.06; 95% CI = −0.15 to 0.02; P = 0.15)
traits linked to RHP in randomly paired rivals and a low level of heterogeneity (Table 1a). The variance
components and heterogeneity due to type of experiment
and to phylogeny were low; most of the variance was due
(a) Analyses of the data set with all variables related to RHP
to study ID (Table 1a). The phylogenetic signal was also low
For the data set with all variables related to RHP, we (H 2 < 0.001%).
obtained 174 effect sizes (88 for winners and 86 for losers) for The relationship between contest duration and traits
36 species (Table S1 and Fig. S1, Supporting information). linked to RHP was not influenced by the interaction
There was no evidence for publication bias (Egger’s test: between status and the occurrence of physical contact

Biological Reviews 94 (2019) 1430–1442 © 2019 Cambridge Philosophical Society


Meta-analysis of contest resolution rules 1437

Fig. 2. Mean Fisher’s z (Zr) values and their 95% confidence intervals for the relationship between contest duration and traits linked
to resource holding potential (RHP) for losers and winners of all contests (with and without physical contact). (A) Results from the
data set with all trait values used to estimate individual RHP in the original articles. (B) Results from the data set including only
the original authors’ best predictor of RHP for each species. A Zr value of zero (dashed line) indicates that contest duration was
unrelated to trait values.

(QM status*physical contact = 0.60; df = 1; P = 0.44) and did (2) Relationship between contest duration and
not differ between species that fight with and without mean trait values in RHP-matched rivals
physical contact (QM physical contact = 1.30; df = 1; P = 0.25).
We obtained 21 effect sizes from 16 species (Table S2 and
However, the relationship between contest duration and
Fig. S2, Supporting information). There was no evidence for
traits linked to RHP differed between winners and losers
publication bias (Egger’s test: estimate = 0.04; CI = −0.11
(QM status = 187.5; df = 1; P < 0.001). Contest duration
to 0.18; P = 0.59) and a low level of heterogeneity (Table 1c).
increased with loser RHP, but there was no relationship
The variance components and heterogeneity due to type of
between contest duration and winner RHP (Fig. 2A). These
experiment and phylogeny were low. Most of the variance
results thus provide support for the (E)WOA model regardless was concentrated on variation among studies (Table 1c). The
of the presence of physical contact. phylogenetic signal was also low (H 2 = 0.02%).
The mean effect size was greater than zero (z-score = 4.77;
(b) Analyses of the data set with the best predictors of RHP P < 0.001; Fig. 3), indicating that contest duration increases
For the second data set with only one trait linked to RHP with mean RHP of the pair. This result provides support for
per species, we obtained 68 effect sizes (34 for winners the self-assessment strategies-(E)WOA and CAM.
and 34 for losers) for 36 species (Table S1, Supporting
information). There was no evidence for publication bias (3) Relationship between contest escalation and
RHP asymmetry
(Egger’s test: estimate = 0.03; 95% CI = −0.13 to 0.18;
P = 0.74) and a low level of heterogeneity (Table 1b). The We obtained 20 effect sizes for 20 species (Table S3 and Fig.
variance components and heterogeneity due to type of S3, Supporting information). There was no evidence for pub-
experiment and to phylogeny were low; most of the variance lication bias (Egger’s test: estimate = −0.04; CI = −0.17 to
was due to study ID (Table 1b). The phylogenetic signal was 0.09; P = 0.55) and a low level of heterogeneity (Table 1d).
also low (H 2 < 0.001%). The variance component and heterogeneity due to type of
Similar to the results considering all traits linked to experiment was low; but in this case, most of the variance
RHP, the relationship between contest duration and traits was concentrated on phylogeny (H 2 = 99%, Table 1d).
linked to RHP was also not influenced by the interaction We found that the mean effect size did not differ from
between status and the occurrence of physical contact zero (Zr = −0.16; CI = −0.41 to 0.09; z-score = 1.23;
(QM status*physical contact = 0.67; df = 1; P = 0.41) and did P = 0.22), indicating that there is no relationship between
not differ between species that fight with and without contest escalation and RHP asymmetry. However, this
physical contact (QM physical contact = 2.72; df = 1; P = 0.10). result was strongly affected by one study on the crab
We also found a positive relation between contest duration Austruca annulipes (Bolton, Backwell & Jennions, 2013).
and loser RHP, while winner RHP was not correlated After removing this study, the variance component and
to contest duration (QM status = 170.7; df = 1; P < 0.001; heterogeneity due to type of experiment increased, while
Fig. 2B). Thus, regardless of the number of measured the variance explained by phylogeny decreased (Table 1e).
traits linked to RHP, the data suggest (E)WOA models We found that the mean effect size was smaller than zero
as the best explanation for how individuals give up on (z-score = 3.10; P = 0.002; Fig. 4), indicating that for the
contests. remaining 19 species, individuals were less prone to escalate

Biological Reviews 94 (2019) 1430–1442 © 2019 Cambridge Philosophical Society


1438 Nelson S. Pinto and others

which the losers do not gather information about the winners’


RHP – they give up in response to the costs they accrue
based on their own actions, regardless of the occurrence of
physical contact. When considering RHP-matched rivals, we
found that contest duration increased with mean trait values
of the fighting pair, which is consistent with self-assessment
strategies such as (E)WOA and CAM (Fig. 1). Therefore,
analyses based on the framework proposed by Taylor &
Elwood (2003) support self-assessment strategies as the
models that best describe the decision to give up on a contest.
On the other hand, when we consider the predictions made
by the models themselves regarding escalation patterns, the
results were more variable. When we considered all species,
we found no relation between probability of escalation
and RHP asymmetry. However, when we removed an
outlier study, we found a negative relation between the
Fig. 3. Mean Fisher’s z (Zr) value and its 95% confidence probability of escalation and RHP asymmetry (Fig. 4) – a
interval for the relationship between contest duration and mean pattern that provides support for SAM. It is important to
trait value of the fighting pair for contests in which rivals with note that non-significant results do not necessarily indicate
similar resource holding potential (RHP) values were paired that the species used in our analyses do not follow any
(RHP-matched rivals). A Zr value of zero (dashed line) indicates rule – non-significant results may suggest that species vary
that contest duration was unrelated to mean trait values.
widely in the rules they adopt to settle contests. Hence,
significant results indicate that different species may follow
a similar assessment model which decreases variation. For
this reason, although some species may settle their contests
based on mutual assessment (e.g. Junior & Peixoto, 2013;
Schnell et al., 2015; Green & Patek, 2018), our results for the
relationships between contest duration and traits linked to
RHP suggest that the proportion of species following mutual
assessment is low and, therefore, self-assessment seems to be
the most frequent contest resolution rule adopted by species.
It is surprising that (E)WOA models are the best candidates
to explain contest resolution for most animal species for
two reasons. First, these models were not developed to
explain physical contact – they were developed to explore
bouts of displays between rivals (Bishop & Cannings, 1978).
Second, in (E)WOA, contests do not escalate to different
phases. Individuals only increase the intensity of displays.
Fig. 4. Mean Fisher’s z (Zr) value and its 95% confidence Therefore, most studies do not agree with some or most of
interval for the relationship between contest escalation (i.e. the assumptions of (E)WOA. For instance, in the dragonfly
change from less costly to costlier behaviours during the contest) Diastatops obscura, males defend territories in which females
and the difference in the traits related to resource holding oviposit by performing bouts of aerial displays (Junior &
potential (RHP) of the fighting pair (excluding one outlier Peixoto, 2013). However, males can also escalate to physical
study). A Zr value of zero (dashed line) indicates that contest
contact when they have similar RHP, which violates the
escalation was unrelated to trait difference between rivals.
assumptions of (E)WOA. In fact, from the 36 species for
which we gathered information, only two could follow
as RHP asymmetry increases. The phylogenetic signal was
(E)WOA without violating any assumption (Hamadryas and
low, but it was the highest among all analyses (H 2 = 4.4%).
Hermeuptychia butterflies).
Perhaps the evidence in support of (E)WOA occurs due to
other factors that are not considered in the original models.
IV. DISCUSSION For example, recent studies suggest that damage capacity
(Palaoro & Briffa, 2017) and self-inflicted damage (Lane &
In this study, we showed that contest duration increased Briffa, 2017) may affect contest dynamics and, consequently,
with loser RHP, but not with winner RHP when individuals influence the relationship between contest duration and
are randomly matched for RHP (regardless of the number winners’ and losers’ RHP. If the importance of damage
of traits linked to RHP used in the analyses). This result is varies among species, we may find species in which the
consistent with predictions of the (E)WOA models (Fig. 1) in relationship between contest duration and winner RHP is

Biological Reviews 94 (2019) 1430–1442 © 2019 Cambridge Philosophical Society


Meta-analysis of contest resolution rules 1439

negative (e.g. species that bear weapons) and species in de-escalated behaviours, while winners use escalation to
which this relationship is absent (e.g. species that do not bear signal their willingness to persist in the contest (Briffa, Elwood
weapons). This would explain why isolated studies do not & Dick, 1998). Signals may also be used at the beginning
provide evidence for (E)WOA, while meta-analytic evidence of the interaction to evaluate the rival. Once the decision
indicates that (E)WOA explains contest decisions for most to engage or to continue is made, the contest may proceed
species. Thus, we argue that a self-assessment strategy is based on self-assessment. For example, in the fiddler crab,
indeed the general rule used to resolve contests, but it is Austruca mjoebergi, individuals seem to be able to decide when
a more flexible form of (E)WOA. Such flexibility in the to initiate a contest based on the difference between their size
assumptions would come from adding that contests are not and that of their rival (Morrell, Backwell & Metcalfe, 2005).
a single behavioural phase, that most species use physical A third possibility is that signals might also be used as threat
contact during contests, and that damage might play an displays to decrease the motivation of the rival to engage in
important role in some contests, but not in others. Alternative a contest (Számadó, 2003, 2008). Hence, signals might be
experimental approaches not based on correlations between used for reasons not necessarily related to the evaluation of
contest duration and individual traits are essential to clarify the rival. Ultimately, we need first to identify the function of
this picture (Arnott & Elwood, 2008, 2009). For example, by each phase in a contest and only then proceed to investigate
manipulating the type of visual cues exhibited by tethered the role of signalling.
rivals in the damselfly Mnesarette pudica, it was possible to show Based on the variation in, and inconsistencies between,
that, unlike the results involving correlations between fight model assumptions and empirical data, we suggest that
duration and individual traits, stronger males perform mutual the classical models such as (E)WOA, SAM, and CAM
assessment while weaker males perform self-assessment do not robustly explain how animals decide to give up
(Guillermo-Ferreira et al., 2015). on a contest. We have a considerable body of evidence
Self-assessment was advocated as the most probable suggesting that several factors might affect how the decision
contest resolution rule among animals because it is less to give up is made. Factors such as asymmetries in per-
cognitively complex than mutual assessment (Elwood & ceived resource value (Mesterton-Gibbons & Heap, 2014;
Arnott, 2012; but see Fawcett & Mowles, 2013). Since Mesterton-Gibbons & Sherratt, 2016), previous experience
most species follow self-assessment rules, it could be that (i.e. winner and loser effects; Hsu & Wolf, 2001), personality
self-assessment represents the basal assessment rule. Based (Briffa, Sneddon & Wilson, 2015), damage capacity (Palaoro
on this argument, mutual assessment would only be found in & Briffa, 2017), self-inflicted injuries (Lane & Briffa, 2017),
a few closely related groups – a hallmark of a phylogenetic and even individual fighting skill (Briffa & Fortescue, 2017)
signal. However, we found no evidence of phylogenetic are some of the variables that can potentially influence
signal in assessment models, indicating that both self- and how animals make decisions during contests. Among
mutual assessment occur scattered across the phylogeny these variables, only asymmetries in resource value have
regardless of the phylogenetic relationships among species. been included in contest theory (Mesterton-Gibbons &
One possible explanation for this pattern is the lack of Heap, 2014; Mesterton-Gibbons & Sherratt, 2016). We
studies in closely related species. We found only two studies also have increasing empirical evidence for variation
for congeneric fiddler crabs (Austruca spp.), one study for in assessment strategies during contests (Jennings et al.,
three species of stalk-eyed flies (Cyrtodiopsis spp.), and another 2005; Hsu et al., 2008; Lobregat et al., 2019), and among
study for two species of crayfish (Parastacus spp.). All other individuals (Guillermo-Ferreira et al., 2015; Edmonds &
studies were from distantly related species, which could Briffa, 2016). Such evidence has the potential to change
decrease the importance of phylogenetic history in our data. how we think about contests because assessment strategies
To highlight this issue, the strongest phylogenetic signal in do not need to be fixed – individuals might vary according
to their genotype, or according to the resource value
our analyses indicated that only 4.4% of the variation in
(Mesterton-Gibbons & Heap, 2014; Mesterton-Gibbons &
escalation patterns was explained by phylogeny. Clearly, we
Sherratt, 2016). But we still need to incorporate those ideas
need more studies on closely related species to understand
into the core of contest theory. In conclusion, we highlight
the basis of contest evolution.
that the classical models have been extremely important to
It is intriguing that SAM is not a general contest resolution
provide advances in the way we understand how individuals
rule for species that signal their RHP (Jennings et al., 2004;
settle contests. However, it may be time to move forward
Briffa, 2008; Painting & Holwell, 2014; Schnell et al., 2015),
and focus on new possibilities derived from such models.
species that have high energetic expenditure during the fight
(e.g. Copeland et al., 2011) or species that fight with high
chances of injury (e.g. Rudin & Briffa, 2011). However,
signals may not necessarily be associated with mutual V. CONCLUSIONS
evaluations of RHP. In some species individuals might signal
to show their willingness to engage or to remain in a contest (1) We tested if there was consistency among species in the
(Payne & Pagel, 1997; Jennings et al., 2012). For instance, rules used by rivals to decide the winner of animal contests.
in contests of the hermit crab Pagurus bernhardus, losers seem Although isolated papers show that some species settle their
to signal that they are giving up by using within-phase contests based on the mutual assessment predicted by SAM,

Biological Reviews 94 (2019) 1430–1442 © 2019 Cambridge Philosophical Society


1440 Nelson S. Pinto and others

we found that the majority of species fight according to Arnott, G. & Elwood, R. W. (2009). Assessment of fighting ability in animal
contests. Animal Behaviour 77, 991–1004.
self-assessment models. There was a positive relationship *Benítez, M. E., Pappano, D. J., Beehner, J. C. & Bergman, T. J. (2017). Evidence
between contest duration and loser RHP, but no relationship for mutual assessment in a wild primate. Scientific Reports 7, 1–11.
for contest duration and winner RHP when rivals where Bishop, D. T. & Cannings, C. (1978). A generalized war of attrition. Journal of
Theoretical Biology 70, 85–124.
randomly paired. For RHP-matched rivals, contest duration Bolton, J., Backwell, P. R. Y. & Jennions, M. D. (2013). Density dependence
seems to increase with mean RHP of the pair. Therefore, and fighting in species with indeterminate growth: a test in a fiddler crab. Animal
our results indicate that most species fight according to Behaviour 85, 1367–1376.
Borenstein, M., Hedges, L. V., Higgins, J. P. T. & Rothstein, H. R. (2009).
(E)WOA models. Introduction to Meta-analysis. John Wiley and Sons, New Jersey.
(2) We found a negligible phylogenetic signal in the *Brandt, Y. & Swallow, J. G. (2009). Do the elongated eye stalks of Diopsid flies
relationship between contest duration and loser and winner facilitate rival assessment? Behavioral Ecology and Sociobiology 63, 1243–1246.
Briffa, M. (2008). Decisions during fights in the house cricket, Acheta domesticus: mutual
RHP. This evidence indicates that closely related species do or self assessment of energy, weapons and size? Animal Behaviour 75, 1053–1062.
not necessarily make similar decisions on how to give up on Briffa, M. (2015). Agonistic signals: integrating analysis of functions and mechanisms.
a contest. Therefore, self- or mutual assessment are scattered In Animal Signaling and Function: An Integrative Approach (eds D. J. Irschick, M. Briffa
and J. Podos), pp. 141–173. JohnWiley & Sons, Inc., Hoboken, New Jersey.
across the phylogeny without a clear pattern. However, the Briffa, M. & Elwood, R. W. (2009). Difficulties remain in distinguishing between
lack of studies on closely related species might bias this result. mutual and self-assessment in animal contests. Animal Behaviour 77, 759–762.
(3) Although relationships between contest duration and Briffa, M., Elwood, R. W. & Dick, J. T. A. (1998). Analysis of repeated signals
during shell fights in the hermit crab Pagurus bernhardus. Proceedings of the Royal Society
traits linked to RHP provided support for self-assessment of London 265, 1467–1474.
models, other behavioural traits exhibited by species (such Briffa, M. & Fortescue, K. J. (2017). Motor pattern during fights in the hermit crab
as the occurrence of physical contact and predictable Pagurus bernhardus: evidence for the role of skill in animal contests. Animal Behaviour
128, 13–20.
behavioural phases during the fight) do not match Briffa, M. & Hardy, I. C. W. (2013). Introduction to animal contests. In Animal
all self-assessment assumptions (particularly for (E)WOA Contests (eds I. C. W. Hardy and M. Briffa), pp. 1–4. Cambridge University press,
New York.
models). In addition, we found that for a subset of species,
Briffa, M. & Sneddon, L. U. (2010). Contest behavior. In Evolutionary Behavioral
escalation patterns were related to RHP asymmetry. Ecology (eds D. F. Westneat and C. W. Fox), pp. 246–265. Oxford University
(4) We highlight that there is a considerable body of Press, New York.
Briffa, M., Sneddon, L. U. & Wilson, A. J. (2015). Animal personality as a cause
evidence that could be incorporated into the core of contest and consequence of contest behaviour. Biology Letters 11, 20141007.
theory. The incorporation of new ideas and alternative *Camerlink, I., Turner, S. P., Farish, M. & Arnott, G. (2015). Aggressiveness as
experimental approaches that are not based on correlations a component of fighting ability in pigs using a game-theoretical framework. Animal
Behaviour 108, 183–191.
between contest duration and traits linked to RHP might Chamberlain, S. A., Hovick, S. M., Dibble, C. J., Rasmussen, N. L., Van
solve the inconsistencies we identified herein. We know that Allen, B. G., Maitner, B. S., Ahern, J. R., Bell-Dereske, L. P., Roy, C. L.,
the classical models were important to understanding how Meza-Lopez, M., Carrillo, J., Siemann, E., Lajeunesse, M. J. & Whitney,
K. D. (2012). Does phylogeny matter? Assessing the impact of phylogenetic
animals decide to give up on contests, but we now need information in ecological meta-analysis. Ecology Letters 15, 627–636.
to move forward with empirical tests of new models that *Constant, N., Valbuena, D. & Rittschof, C. C. (2011). Male contest investment
incorporate new experiments and new evidence. changes with male body size but not female quality in the spider Nephila clavipes.
Behavioural Processes 87, 218–223.
Copeland, D. L., Levay, B., Sivaraman, B., Beebe-Fugloni, C. & Earley, R. L.
(2011). Metabolic costs of fighting are driven by contest performance in male convict
cichlid fish. Animal Behaviour 82, 271–280.
VI. ACKNOWLEDGEMENTS *Costa, J. R., Dalosto, M. M., Palaoro, A. V. & Santos, S. (2016). Contest
duration and dynamics are affected by body size in a potentially subsocial crayfish
(Crustacea: Decapoda). Ethology 122, 502–512.
We thank all authors that provided access to their *Dalosto, M. M., Palaoro, A. V., Costa, J. R. & Santos, S. (2013). Aggressiveness
original data. We also thank Christina Painting, Mark and life underground: the case of burrowing crayfish. Behaviour 150, 3–22.
Del Re, A.C. (2013). compute.es: compute effect sizes. R package, version 0.2-2.
Briffa, Robert Elwood, Alastair Wilson and an anonymous Electronic file available at http://cran.r-project.org/web/packages/compute.es
referee for helpful comments and suggestions. N.S.P. Edmonds, E. & Briffa, M. (2016). Weak rappers rock more: hermit crabs assess their
thanks CAPES for providing a PhD scholarship. own agonistic behaviour. Biology Letters 12, 20150884.
Egger, M., Davey Smith, G., Schneider, M. & Minder, C. (1997). Bias in
A.V.P. has a post-doctoral grant from Fundação de meta-analysis detected by a simple, graphical test. British Medical Journal 315,
Amparo à Pesquisa do Estado de São Paulo (FAPESP 629–634.
process number 2016/22679-3). P.E.C.P. has research *Elias, D. O., Kasumovic, M. M., Punzalan, D., Andrade, M. C. B. & Mason,
A. C. (2008). Assessment during aggressive contests between male jumping spiders.
grants from Conselho Nacional de Desenvolvimento Animal Behaviour 76, 901–910.
Científico e Tecnológico (CNPq produtividade em pesquisa: Elwood, R. W. & Arnott, G. (2012). Understanding how animals fight with Lloyd
305561/2014-6 and 311212/2018-2; CNPq Edital Universal Morgan’s canon. Animal Behaviour 84, 1095–1102.
*Emberts, Z., St. Mary, C. M., Herrington, T. J. & Miller, C. W. (2018). Males
2014: 443977/2014-3) and Fundação de Amparo à Pesquisa missing their sexually selected weapon have decreased fighting ability and mating
do Estado da Bahia (Fapesb: JCB0034/2013). success in a competitive environment. Behavioral Ecology and Sociobiology 78, 81.
Enquist, M. & Leimar, O. (1983). Evolution of fighting behaviour: decision rules
and assessment of relative strength. Journal of Theoretical Biology 102, 387–410.
Enquist, M., Leimar, O., Ljungberg, T., Mallner, Y. & Segerdahl, N. (1990).
VII. REFERENCES A test of the sequential assessment game: fighting in the ciclid fish Nannacara anomala.
Animal Behaviour 40, 1–14.
Fawcett, T. W. & Mowles, S. L. (2013). Assessments of fighting ability need not be
References marked with an asterisk are cited only within the supporting information. cognitively complex. Animal Behaviour 86, e1–e7.
Arnott, G. & Elwood, R. W. (2008). Information gathering and decision making *Fea, M. & Holwell, G. (2018). Combat in a cave-dwelling wētā (Orthoptera:
about resource value in animal contests. Animal Behaviour 76, 529–542. Rhaphidophoridae) with exaggerated weaponry. Animal Behaviour 138, 85–92.

Biological Reviews 94 (2019) 1430–1442 © 2019 Cambridge Philosophical Society


Meta-analysis of contest resolution rules 1441

Green, P. A. & Patek, S. N. (2018). Mutual assessment during ritualized fighting in Nakagawa, S. & Santos, E. S. A. (2012). Methodological issues and advances in
mantis shrimp (Stomatopoda). Proceedings of the Royal Society of London Series B 285, biological meta-analysis. Evolutionary Ecology 26, 1253–1274.
20172542. Painting, C. J. & Holwell, G. I. (2014). Exaggerated rostra as weapons and
Guillermo-Ferreira, R., Gorb, S. N., Appel, E., Kovalev, A. & Bispo, P. C. the competitive assessment strategy of male giraffe weevils. Behavioral Ecology 25,
(2015). Variable assessment of wing colouration in aerial contests of the red-winged 1223–1232.
damselfly Mnesarete pudica (Zygoptera, Calopterygidae). The Science of Nature 102, 13. Palaoro, A. V. & Briffa, M. (2017). Weaponry and defenses in fighting animals: how
Hack, M. A. (1997). Assessment strategies in the contests of male crickets, Acheta allometry can alter predictions from contest theory. Behavioral Ecology 28, 328–336.
domesticus. Animal Behaviour 53, 733–747. Palaoro, A. V., Dalosto, M. M., Costa, J. R. & Santos, S. (2014). Freshwater
Hardy, I. C. W. & Briffa, M. (2013). Animal Contests. Cambridge University press, decapod (Aegla longirostri) uses a mixed assessment strategy to resolve contests. Animal
New York. Behaviour 95, 71–79.
Higgins, J. P. T. & Thompson, S. G. (2002). Quantifying heterogeneity in a *Palaoro, A. V., Velasque, M., Santos, S. & Briffa, M. (2017). How does
meta-analysis. Statistics in Medicine 21, 1539–1558. environment influence fighting? The effects of tidal flow on resource value and
Higgins, J. P. T., Thompson, S. G., Deeks, J. J. & Altman, D. (2003). Measuring fighting costs in sea anemones. Biology Letters 13, 20170011.
inconsistency in meta-analyses. BMJ 327, 557–560. Paradis, E., Claude, J. & Strimmer, K. (2004). APE: analyses of phylogenetics and
Hsu, Y., Lee, S. P. P., Chen, M. H. H., Yang, S. Y. Y. & Cheng, K. C. C. evolution in R language. Bioinformatics 20, 289–290.
(2008). Switching assessment strategy during a contest: fighting in killifish Kryptolebias Parker, G. A. (1974). Assessment strategy and the evolution of fighting behaviour.
marmoratus. Animal Behaviour 75, 1641–1649. Journal of Theoretical Biology 47, 223–243.
Hsu, Y. & Wolf, L. L. (2001). The winner and loser effect: what fighting behaviours Payne, R. (1998). Gradually escalating fights and displays: the cumulative assessment
are influenced? Animal Behaviour 61, 777–786. model. Animal Behaviour 56, 651–662.
Jennings, D. J., Elwood, R. W., Carlin, C. M., Hayden, T. J. & Gammell, M. P. Payne, R. J. H. & Pagel, M. (1996). Escalation and time costs in displays of endurance.
(2012). Vocal rate as an assessment process during fallow deer contests. Behavioural Journal of Theoretical Biology 183, 185–193.
Processes 91, 152–158. Payne, R. J. H. & Pagel, M. (1997). Why do animals repeat displays? Animal Behaviour
Jennings, D. J., Gammell, M. P., Carlin, C. M. & Hayden, T. J. (2004). Effect 54, 109–119.
of body weight, antler length, resource value and experience on fight duration and Peixoto, P. E. C. & Benson, W. W. (2011). Fat and body mass predict residency
intensity in fallow deer. Animal Behaviour 68, 213–221. status in two tropical Satyrine butterflies. Ethology 117, 722–730.
Jennings, D. J., Gammell, M. P., Payne, R. J. H. & Hayden, T. J. (2005). An *Peixoto, P. E. C. & Benson, W. W. (2012). Influence of previous residency and
investigation of assessment games during fallow deer fights. Ethology 111, 511–525. body mass in the territorial contests of the butterfly Hermeuptychia fallax (Lepidoptera:
Junior, R. S. L. & Peixoto, P. E. C. (2013). Males of the dragonfly Diastatops Satyrinae). Journal of Ethology 30, 61–68.
obscura fight according to predictions from game theory models. Animal Behaviour 85, Prenter, J., Elwood, R. W. & Taylor, P. W. (2006). Self-assessment by males
663–669. during energetically costly contests over precopula females in amphipods. Animal
Kelly, C. D. (2006). Fighting for harems: assessment strategies during male–male Behaviour 72, 861–868.
contests in the sexually dimorphic Wellington tree weta. Animal Behaviour 72, *Prenter, J., Taylor, P. W. & Elwood, R. W. (2008). Large body size for winning
and large swords for winning quickly in swordtail males, Xiphophorus helleri. Animal
727–736.
Behaviour 75, 1981–1987.
Kokko, H. (2013). Dyadic contests: modelling fights between two individuals. In Animal
R Development Core Team (2015). R: A Language and Environment for Statistical Computing.
Contests (eds I. C. W. Hardy and M. Briffa), pp. 5–32. Cambridge University
R Foundation for Statistical Computing, Vienna Electronic file available at http://
press, New York.
www.R-project.org.
Koricheva, J., Gurevitch, J. & Mengersen, K. (2013). Handbook of Meta-analysis in
Reaney, L. T., Drayton, J. M. & Jennions, M. D. (2011). The role of body size
Ecology and Evolution. Princeton University Press, Princeton.
and fighting experience in predicting contest behaviour in the black field cricket,
Lane, S. M. & Briffa, M. (2017). The price of attack: rethinking damage costs in
Teleogryllus commodus. Behavioral Ecology and Sociobiology 65, 217–225.
animal contests. Animal Behaviour 126, 23–29.
*Reddon, A. R., Voisin, M. R., Menon, N., Marsh-Rollo, S. E., Wong,
Leimar, O., Austad, S. N. & Enquist, M. (1991). A test of the sequential assessment
M. Y. L. L. & Balshine, S. (2011). Rules of engagement for resource contests in a
game: fighting in the bowl and doily spider Frontinella pyramitela. Evolution 45, 862– 874.
social fish. Animal Behaviour 82, 93–99.
*Lim, M. L. M. & Li, D. (2013). UV-Green iridescence predicts male quality during
*Reichert, M. S. & Gerhardt, H. C. (2011). The role of body size on the outcome,
Jumping Spider contests. PLoS One 8, e59774.
escalation and duration of contests in the grey treefrog, Hyla versicolor. Animal Behaviour
Lobregat, G., Kloss, T. G., Peixoto, P. E. C. & Sperber, C. F. (2019). Fighting in
82, 1357–1366.
rounds: males of a neotropical cricket switch assessment strategies during contests. Rohatgi, A. (2012). WebPlotDigitalizer: HTML5 based online tool to
Behavioral Ecologyarz005. https://doi.org/10.1093/beheco/arz005. extractnumerical data from plot images. Electronic file available at http://arohatgi
*Lourenço, V. T. (2015). Defesa de territórios de acasalamento por Machos da estaladeira .info/WebPlotDigitizer/app/ Accessed 01.01.2019.
vermelha, Hamadryas amphinome (Lepidoptera: Nymphalidae), uma borboleta neotropical. Msc Rudin, F. S. & Briffa, M. (2011). The logical polyp: assessments and decisions during
Thesis: Unicamp, São Paulo. contests in the beadlet anemone Actinia equina. Behavioral Ecology 22, 1278–1285.
Marden, J. H. & Waage, J. K. (1990). Escalated damselfly territorial contests are Schnell, A. K., Smith, C. L., Hanlon, R. T. & Harcourt, R. (2015). Giant
energetic wars of attrition. Animal Behaviour 39, 954–959. Australian cuttlefish use mutual assessment to resolve male-male contests. Animal
*McGinley, R. H., Prenter, J. & Taylor, P. W. (2015). Assessment strategies Behaviour 107, 31–40.
and decision making in male-male contests of Servaea incana jumping spiders. Animal Stuart-Fox, D. (2006). Testing game theory models: fighting ability and decision
Behaviour 101, 89–95. rules in chameleon contests. Proceedings of the Royal Society of London Series B 273,
*McLean, C. A. & Stuart-Fox, D. (2014). Rival assessment and comparison of 1555–1561.
morphological and performance-based predictors of fighting ability in Lake Eyre Számadó, S. (2003). Threat displays are not handicaps. Journal of Theoretical Biology
dragon lizards, Ctenophorus maculosus. Behavioral Ecology and Sociobiology 69, 523–531. 221, 327–348.
Mesterton-Gibbons, M. & Heap, S. M. (2014). Variation between self- and mutual Számadó, S. (2008). How threat displays work: species-specific fighting techniques,
assessment in animal contests. American Naturalist 183, 199–213. weaponry and proximity risk. Animal Behaviour 76, 1455–1463.
Mesterton-Gibbons, M., Marden, J. H. & Dugatkin, L. A. (1996). On wars of Taylor, P. W. & Elwood, R. W. (2003). The mismeasure of animal contests. Animal
attrition without assessment. Journal of Theoretical Biology 181, 65–83. Behaviour 65, 1195–1202.
Mesterton-Gibbons, M. & Sherratt, T. N. (2016). How residency duration *Tsai, Y. J. J., Barrows, E. M. & Weiss, M. R. (2014). Pure self-assessment of size
affects the outcome of a territorial contest: complementary game-theoretic models. during male-male contests in the parasitoid wasp Nasonia vitripennis. Ethology 120,
Journal of Theoretical Biology 394, 137–148. 816–824.
Michonneau, F., Brown, J. W. & Winter, D. J. (2016). Rotl: an R package Viechtbauer, W. (2010). Conducting meta-analyses in R with the metafor package.
to interact with the Open Tree of Life data. Methods in Ecology and Evolution 7, Journal of Statistical Software 36, 1–48.
1476–1481. Vieira, M. C. & Peixoto, P. E. C. (2013). Winners and losers: a meta-analysis of
Moher, D., Liberati, A., Tetzlaff, J., Altman, D. G. & The PRISMA Group functional determinants of fighting ability in arthropod contests. Functional Ecology
(2009). Preferred reporting items for systematic reviews and meta-analyses: the 27, 305–313.
PRISMA statement. PLoS Medicine 6, e1000097. Vilela, D. S., Tosta, T. A. A., Rodrigues, R. R., Del-Claro, K. &
*Moore, J. C., Obbard, D. J., Reuter, C., West, S. A. & Cook, J. M. (2008). Guillermo-Ferreira, R. (2017). Colours of war: visual signals may influence
Fighting strategies in two species of fig wasp. Animal Behaviour 76, 315–322. the outcome of territorial contests in the tiger damselfly, Tigriagrion aurantinigrum.
Morrell, L. J., Backwell, P. R. Y. & Metcalfe, N. B. (2005). Fighting in fiddler Biological Journal of the Linnean Society 121, 786–795.
crabs Uca mjoebergi: what determines duration? Animal Behaviour 70, 653–662. *Walker, L. A. & Holwell, G. I. (2018). The role of exaggerated male chelicerae
Nakagawa, S. & Cuthill, I. C. (2007). Effect size, confidence interval and statistical in male–male contests in New Zealand sheet-web spiders. Animal Behaviour 139,
significance: a practical guide for biologists. Biological Reviews 82, 591–605. 29–36.

Biological Reviews 94 (2019) 1430–1442 © 2019 Cambridge Philosophical Society


1442 Nelson S. Pinto and others

*Wofford, S. J., Earley, R. L. & Moore, P. A. (2015). Evidence for assessment between contest duration and mean trait value of the fighting
disappears in mixed-sex contests of the crayfish, Orconectes virilis. Behaviour 152,
995–1018.
pair for contests in which rivals were paired in order to
*Xu, M. & Fincke, O. M. (2015). Ultraviolet wing signal affects territorial contest minimize their difference in resource holding potential (RHP)
outcome in a sexually dimorphic damselfly. Animal Behaviour 101, 67–74. (RHP-matched rivals).
*Yasuda, C., Takeshita, F. & Wada, S. (2012). Assessment strategy in male-male
contests of the hermit crab Pagurus middendorffii. Animal Behaviour 84, 385–390. Fig. S3. Phylogeny used to calculate the variance–
*Yasuda, C. I. & Koga, T. (2016). Importance of weapon size in all stages of covariance matrix for analyses involving the relationship
male-male contests in the hermit crab Pagurus minutus. Behavioral Ecology and Sociobiology between contest escalation (i.e. change from less costly to
70, 2175–2183.
costlier behaviours during the contest) and the difference in
traits related to the resource holding potential (RHP) of the
fighting pair.
Table S1. Effect sizes [Fisher’s z (Zr)] and their variances (V )
calculated for all studies used in the test of the relationship
VIII. SUPPORTING INFORMATION between contest duration and traits linked to resource holding
potential (RHP) for winners and losers in randomly paired
Additional supporting information may be found online in rivals.
the Supporting Information section at the end of the article. Table S2. Effect sizes [Fisher’s z (Zr)] and their variances
Fig. S1. Phylogeny used to calculate the variance– (V ) calculated for studies used in the test of the relationship
covariance matrix for meta-regression models involving the between contest duration and mean trait values of the fighting
relationship between contest duration and traits linked to pair in resource holding potential (RHP)-matched rivals.
resource holding potential (RHP) for winners and losers in Table S3. Effect sizes [Fisher’s z (Zr)] and their variances
randomly paired trials. (V ) calculated for studies used in the test of the relationship
Fig. S2. Phylogeny used to calculate the variance– between contest escalation and trait differences between
covariance matrix for analyses involving the relationship winners and losers.

(Received 1 August 2018; revised 4 March 2019; accepted 6 March 2019; published online 27 March 2019)

Biological Reviews 94 (2019) 1430–1442 © 2019 Cambridge Philosophical Society

You might also like