972
Forstmeier et al. (2014) reviewed the EPP literature and urged
more research on “nonadaptive” hypotheses of female extrapair mating, instead of the traditional, possibly excessive, focus on adaptive
explanations of female choice. We join Forstmeier et al. in encouraging researchers to examine such alternative explanations for extrapair
behavior. In particular, future research into the male manipulation
hypothesis, taking into account changes with male age, seems promising. Finally, we wish to propose a new name for the alternative to
the good-genes hypothesis: the “sugar-free daddy” hypothesis. It is
increasingly apparent that extrapair males do not provide any direct,
and probably also no indirect, benefits to females and that these
males are likely, on average, to be older than the females.
Address correspondence to S. Nakagawa. E-mail: [email protected].
Received 13 March 2015; accepted 16 March 2015; Advance Access
­publication 11 April 2015.
doi:10.1093/beheco/arv041
Editor-in-Chief: Leigh Simmons
REFERENCES
Akçay E, Roughgarden J. 2007. Extra-pair paternity in birds: review of the
genetic benefits. Evol Ecol Res. 9:855–868.
Arct A, Drobniak SM, Cichoń M. 2015. Genetic similarity between mates
predicts extrapair paternity—a meta-analysis of bird studies. Behav Ecol.
26:959–968.
Cleasby IR, Nakagawa S. 2012. The influence of male age on within-pair
and extra-pair paternity in passerines. Ibis. 154:318–324.
Forstmeier W, Nakagawa S, Griffith SC, Kempenaers B. 2014. Female
extra-pair mating: adaptation or genetic constraint? Trends Ecol Evol.
29:456–464.
Hadfield JD, Richardson DS, Burke T. 2006. Towards unbiased parentage assignment: combining genetic, behavioural and spatial data in a
Bayesian framework. Mol Ecol. 15:3715–3730.
Hsu YH, Schroeder J, Winney I, Burke T, Nakagawa S. 2014. Costly infidelity: low lifetime fitness of extra-pair offspring in a passerine bird.
Evolution. 68:2873–2884.
Hsu Y-H, Schroeder J, Winney I, Burke T, Nakagawa S. 2015. Are extrapair males different from cuckolded males? A case study and a meta-analytic examination. Mol Ecol. 24:1558–1571.
Ihle M, Forstmeier W. 2013. Revisiting the evidence for inbreeding avoidance in zebra finches. Behav Ecol. 24:1356–1362.
Kingma SA, Hall ML, Peters A. 2013. Breeding synchronization
facilitates extrapair mating for inbreeding avoidance. Behav Ecol.
24:1390–1397.
Krause ET, Krüger O, Kohlmeier P, Caspers BA. 2012. Olfactory kin recognition in a songbird. Biol Lett. 8:327–329.
Lehtonen J, Kokko H. 2015. Why inclusive fitness can make it adaptive to
produce less fit extra-pair offspring. Proc R Soc Lond B. 282:20142716.
Løvlie H, Gillingham MA, Worley K, Pizzari T, Richardson DS. 2013.
Cryptic female choice favours sperm from major histocompatibility complex-dissimilar males. Proc R Soc Lond B. 280:20131296.
Nakagawa S, Waas JR. 2004. ‘O sibling, where art thou?’—a review of
avian sibling recognition with respect to the mammalian literature. Biol
Rev. 79:101–119.
Petrie M, Krupa A, Burke T. 1999. Peacocks lek with relatives even in the
absence of social and environmental cues. Nature. 401:155–157.
Schroeder J, Nakagawa S, Rees M, Mannarelli M-E, Burke T. 2015.
Reduced fitness in progeny from old parents in a natural population.
Proc Natl Acad Sci USA. doi: 10.1073/pnas.1422715112.
Szulkin M, Stopher KV, Pemberton JM, Reid JM. 2013. Inbreeding avoidance, tolerance, or preference in animals? Trends Ecol Evol. 28:205–211.
Westneat DF, Stewart IRK. 2003. Extra-pair paternity in birds: causes, correlates, and conflict. Annu Rev Ecol Evol Syst. 34:365–396.
Behavioral Ecology
Caution is needed when 90% of all
possible estimates remain unpublished: a
comment on Arct et al.
Wolfgang Forstmeier
Department of Behavioural Ecology and Evolutionary Genetics, Max
Planck Institute for Ornithology, Eberhard-Gwinner-Straße 7, 82319
Seewiesen, Germany
The meta-analysis by Arct et al. (2015) is an important and valuable
first step toward answering the question whether extrapair mating
evolved as an inbreeding avoidance strategy. Like most other metaanalyses, and for understandable reasons, the study is based on estimates from the published literature only. In this context, I would
like to raise a provocative question that concerns meta-analyses
in general: To what extent can we draw conclusions from a set of
data that has gone through several steps of filtering? This is not
at all meant as a critique of Arct et al. or of the many authors
who have published the original research, but rather as a critique
of our present publication system, which filters the research by its
outcome (Fanelli 2010) and thereby hinders an unbiased assessment
of research questions and ultimately scientific progress.
Imagine you tested the hypothesis that extrapair mating serves
inbreeding avoidance in a species where inbreeding is rather
uncommon and you obtained a null result. In the current publication environment, such an estimate would most likely remain
unpublished because referees would criticize the limited power and
the improbability of the hypothesis for your study system. A significant finding, in contrast, would likely get published irrespective of
the plausibility of the hypothesis for a given species.
For the study of Arct et al. (2015) we can in fact inspect this issue
more closely. A literature data base of our Department (Kempenaers
B, Valcu M, unpublished data) indicates that at least 521 estimates
of levels of extrapair paternity in birds have been published up until
2012 (the end of data collection in the Arct study). Of these, 472
detected extrapair paternity, which then should in principle allow testing the hypothesis of inbreeding avoidance because relatedness can be
estimated from the molecular markers. However, it appears that most
of these tests were either not carried out or not published because
only 43 estimates (9.1%) are included in the present meta-analysis.
With more than 90% of all possible estimates remaining unpublished, it appears dangerous to draw conclusions from the currently
published subset (comprising 31 nonsignificant effects, 11 significant
positive effects, and 1 significant negative effect) and to assume that
research findings did not influence the probability of publication.
Additional bias could result from so-called “researcher degrees
of freedom” (Simmons et al. 2011), meaning that researchers may
often (even unconsciously) prefer statistical tests that yield significance because they perceive them as more powerful. When faced
with arbitrary decisions (e.g., which statistic of relatedness to compute, whether or not to exclude molecular markers that contain
null alleles, whether or not to exclude females mated to polygynous
males), researchers may often have opted for choices that made
their findings more coherent with their expectations and hence
more easily publishable.
In the present case, publication bias and researcher degrees of
freedom would mean that large positive effects should be overrepresented in the sample. This would also lead to an inflation of
the heterogeneity of effect sizes, which was found to be large and
highly significant in the study by Arct et al. (2015). Attempts by
the authors of explaining the heterogeneity of effect sizes were all
unsuccessful, and although the article highlights the possible effect
Reid • Relatedness and extrapair paternity
of using different molecular methods for estimating relatedness,
I was unable to detect such an effect when applying an Anova to
the given effect sizes (F2,40 = 0.13, P = 0.88). Another factor contributing to the apparently high heterogeneity in effect sizes in this
meta-analysis is that confidence intervals were estimated from sample sizes rather than from standard errors in the original publications. According to Figure 2 in Arct et al. (2015), it appears that 19
out of 43 effect size estimates were significantly different from zero,
but examination of the original publications reveals that only 12 of
these 19 estimates are in fact significant.
To address the problem of publication bias, meta-analysts have
developed a range of tools to test for the presence of bias. In the Arct
et al. (2015) study, all these tests turned out nonsignificant, though
1 test showed a trend in the direction expected when bias is present. However, these tests should generally be regarded with caution
(Ioannidis and Trikalinos 2007), and it might be naïve to think that a
simple statistical test could fix all the bias caused by the current publication culture. I therefore believe that the only satisfactory way of
examining the representativeness of the published fraction would be
the retrieval of unpublished estimates (McAuley et al. 2000).
Hence, the most valuable avenue for future work would be to
gather and analyze unpublished data sets, while strictly adhering to
pre-defined analysis strategies, thereby avoiding “researcher degrees
of freedom” and the resulting inflated effect size estimates. Until
such ideal analyses become available, I prefer a cautious interpretation of the current study: Extra pair mating may have evolved as an
inbreeding avoidance strategy in some species (where the evidence is
indeed highly convincing), yet inbreeding avoidance does not represent a universal explanation for the occurrence of extrapair mating.
FUNDING
W.F. was supported by the Max Planck Society.
I am grateful to M. Valcu for providing the literature database, to A. Rutten
and K. Temnow for literature search, and to S. Nakagawa and T. Albrecht
for helpful comments on previous drafts of the manuscript.
Address correspondence to W. Forstmeier. E-mail: [email protected].
Received 16 March 2015; accepted 17 March 2015; Advance Access
­publication 11 April 2015.
doi:10.1093/beheco/arv044
Editor-in-Chief: Leigh Simmons
REFERENCES
Arct A, Drobniak SM, Cichon M. 2015. Genetic similarity between mates
predicts extra-pair paternity – a meta-analysis of bird studies. Behav Ecol.
26:959–968.
Fanelli D. 2010. Do pressures to publish increase scientists’ bias? An
empirical support from US States Data. PLoS One. 5:e10271.
Ioannidis JP, Trikalinos TA. 2007. The appropriateness of asymmetry tests for publication bias in meta-analyses: a large survey. CMAJ.
176:1091–1096.
McAuley L, Pham B, Tugwell P, Moher D. 2000. Does the inclusion of
grey literature influence estimates of intervention effectiveness reported
in meta-analyses? Lancet. 356:1228–1231.
Simmons JP, Nelson LD, Simonsohn U. 2011. False-positive psychology:
undisclosed flexibility in data collection and analysis allows presenting
anything as significant. Psychol Sci. 22:1359–1366.
973
Extrapair paternity and genetic
similarity—we are not quite there yet: a
response to comments on Arct et al.
Szymon M. Drobniak, Aneta Arct, and Mariusz Cichoń
Institute of Environmental Sciences, Jagiellonian University,
Gronostajowa 7, 30-348 Kraków, Poland
Sexual selection is an increasingly dynamic and fast-changing field
of research. The spectrum and breadth of comments written in
response to our meta-analysis of the relationship between genetic
relatedness and extrapair paternity in birds (Arct et al. 2015)
emphasizes the importance of studying mate choice in the context
of population-wide genetic parameters. The picture that emerges
from these commentaries is one of a complex and still not well
understood or defined phenomenon, and a field that also suffers
from a number of methodological shortcomings. We would like to
briefly summarize these controversies and offer a general view on
how extrapair paternity (EPP) research should develop in order to
make the most of the published studies, and contribute the most to
our understanding of the evolution of EPP.
The first and probably fundamental issue is the lack of consensus
on how to define and measure EPP. Although EPP has an apparently unambiguous definition, some studies tend to include, for
example, lek-breeding species (Akcay and Roughgarden 2007) on
the assumption that genetic relatedness should influence breeding
decisions in such species in way that is analogous to monogamous
species. The simplest solution would be of course the removal of
such studies, which does not invalidate our results as the overall
effect size remains positive and statistically significant. However,
as pointed out in Griffith (2015), our understanding of processes
relating female breeding decisions with genetic similarity will likely
benefit after considering other kinds of breeding behaviors in the
context of genetic relatedness (Du and Lu 2009; Ryder et al. 2010).
It is possible that mate choice in monogamous species, lek dynamics, and patterns of mating in collectively breeding animals all share
similar physiological and ecological mechanisms that can link them
to genetic similarity. Expanding syntheses to include such systems
would surely contribute greatly to our understanding of the evolution of EPP. Similarly, it is difficult to address questions relating specifically extrapair behavior and genetic similarity in the absence of
accurate theoretical models of inbreeding avoidance and its interplay with population-wide genetic parameters. Explicit models,
taking into account behavioral characteristics of birds (e.g., their
variable ability to recognize kin [Zelano and Edwards 2002; Arct
et al. 2010; Nakagawa, 2015]; the relationship between inbreeding and kin cooperation in bird societies [Rubenstein and Lovette
2007; Hatchwell 2009; Szulkin et al. 2013; Eliassen and Jorgensen
2014]; and the connection between physiological variables such as
age—and EPP behavior) are needed in order to generate testable
predictions. We expect that the field will benefit from not only identifying these new research directions but also formulating them in
a way that enables experimental tests: so far the vast majority of
studies have used correlational evidence which largely limits the discovery of causal links in the observed patterns.
Even the most intricate hypotheses are doomed to remain
untested if not enough care and attention is given to the way the
data on EPP is reported and analyzed. Förstmeier (2015) points
out that in most cases, the outcomes of meta-analyses rely on a
limited number of studies and thus should be treated with caution.