A systematic review and meta-analysis of eyespot anti-predator mechanisms

Abstract

Eyespot patterns have evolved in many prey species. These patterns were traditionally explained by the eye mimicry hypothesis, which proposes that eyespots resembling vertebrate eyes function as predator avoidance. However, it is possible that eyespots do not mimic eyes: according to the conspicuousness hypothesis, eyespots are just one form of vivid signals where only conspicuousness matters. They might work simply through neophobia or unfamiliarity, without necessarily implying aposematism or the unprofitability to potential predators. To test these hypotheses and explore factors influencing predators’ responses, we conducted a meta-analysis with 33 empirical papers that focused on bird responses to both real lepidopterans and artificial targets with conspicuous patterns (i.e. eyespots and non-eyespots). Supporting the latter hypothesis, the results showed no clear difference in predator avoidance efficacy between eyespots and non-eyespots. When comparing geometric pattern characteristics, bigger pattern sizes and smaller numbers of patterns were more effective in preventing avian predation. This finding indicates that single concentric patterns have stronger deterring effects than paired ones. Taken together, our study supports the conspicuousness hypothesis more than the eye mimicry hypothesis. Due to the number and species coverage of published studies so far, the generalisability of our conclusion may be limited. The findings highlight that pattern conspicuousness is key to eliciting avian avoidance responses, shedding a different light on this classic example of signal evolution.

Full text

Mizuno etal. eLife 2024;13:RP96338. DOI: https://doi.org/10.7554/eLife.96338 1 of 23
A systematic review and meta- analysis of
eyespot anti- predatormechanisms
Ayumi Mizuno1,2,3*, Malgorzata Lagisz2, Pietro Pollo2†, Yefeng Yang2†,
Masayo Soma1, Shinichi Nakagawa2,3,4
1Department of Biology, Faculty of Science, Hokkaido University, Sapporo, Japan;
2Evolution & Ecology Research Centre, School of Biological, Earth and Environmental
Sciences, The University of New South Wales, Sydney, Australia; 3Department of
Biological Sciences, Faculty of Science, The University of Alberta, Edmonton, Canada;
4Theoretical Sciences Visiting Program, Okinawa Institute of Science and Technology
Graduate University, Onna, Japan
eLife assessment
This meta- analysis presents valuable findings that reexamine the function of butterfly eyespots in
predator avoidance and report for conspicuousness over mimicry. The analysis is robust, but the
evidence supporting the importance of conspicuousness is incomplete due to the limitations of the
literature, and this debate would benefit from additional experiments that would strengthen these
claims. This paper is of interest to evolutionary biologists and ecologists working on the evolution of
morphology and predator- prey interactions.
Abstract Eyespot patterns have evolved in many prey species. These patterns were traditionally
explained by the eye mimicry hypothesis, which proposes that eyespots resembling vertebrate eyes
function as predator avoidance. However, it is possible that eyespots do not mimic eyes: according
to the conspicuousness hypothesis, eyespots are just one form of vivid signals where only conspic-
uousness matters. They might work simply through neophobia or unfamiliarity, without necessarily
implying aposematism or the unprofitability to potential predators. To test these hypotheses and
explore factors influencing predators’ responses, we conducted a meta- analysis with 33 empirical
papers that focused on bird responses to both real lepidopterans and artificial targets with conspic-
uous patterns (i.e. eyespots and non- eyespots). Supporting the latter hypothesis, the results showed
no clear difference in predator avoidance efficacy between eyespots and non- eyespots. When
comparing geometric pattern characteristics, bigger pattern sizes and smaller numbers of patterns
were more effective in preventing avian predation. This finding indicates that single concentric
patterns have stronger deterring effects than paired ones. Taken together, our study supports the
conspicuousness hypothesis more than the eye mimicry hypothesis. Due to the number and species
coverage of published studies so far, the generalisability of our conclusion may be limited. The find-
ings highlight that pattern conspicuousness is key to eliciting avian avoidance responses, shedding a
different light on this classic example of signal evolution.
Introduction
Naturalists have long pondered the evolution and function of the many signals and cues animals use
to communicate (Endler, 1992; Endler, 1993; Andersson, 1994; Johnstone, 1996; Martin Schaefer
et al., 2004; Johansson and Jones, 2007; Hill, 2009; Jones and Ratterman, 2009; Rose etal.,
2022). Visual signals, such as vibrant colours and contrasting patterns, have attracted more interest
RESEARCH ARTICLE
*For correspondence:
ayumi.mizuno5@gmail.com
†These authors contributed
equally to this work
Competing interest: The authors
declare that no competing
interests exist.
Funding: See page 17
Preprint posted
22 January 2024
Sent for Review
12 February 2024
Reviewed preprint posted
30 July 2024
Reviewed preprint revised
09 September 2024
Version of Record published
12 December 2024
Reviewing Editor: George
H Perry, Pennsylvania State
University, United States
Copyright Mizuno etal. This
article is distributed under the
terms of the Creative Commons
Attribution License, which
permits unrestricted use and
redistribution provided that the
original author and source are
credited.

Research article Evolutionary Biology
Mizuno etal. eLife 2024;13:RP96338. DOI: https://doi.org/10.7554/eLife.96338 2 of 23
from researchers than other signals, likely because our species is visually oriented (Endler, 1992;
Kelber et al., 2003; Endler et al., 2005). Eyespot patterns, characterised by concentric rings of
different colours with a light outer ring and a dark centre (Stevens, 2005), are well- known patterns
believed to reduce predation. Although eyespots have been researched for a long time (Stevens,
2005; Kodandaramaiah, 2011; Stevens and Ruxton, 2014; Drinkwater etal., 2022), researchers
continue to debate why eyespots might deter predation.
Three hypotheses have been proposed to explain why eyespot patterns can contribute to prey
survival (reviewed in Stevens, 2005; Kodandaramaiah, 2011; Stevens and Ruxton, 2014; Figure1).
First, the eye mimicry hypothesis suggests that eyespots play a role in deterring predators from
attacking prey and reducing predation risks by mimicking the eyes of vertebrates (Blest, 1957a; Vallin
etal., 2005; Kjernsmo and Merilaita, 2017). This hypothesis predicts that if the pattern has specific
Figure 1. A visual summary of three hypotheses that explain the predation avoidance function of eyespot patterns and the predictions that can
be derived from these two hypotheses. The resemblance of eyespots to actual eyes is discussed through the predator mimicry hypothesis and the
conspicuous signal hypothesis. The table shows the predictions derived from these two hypotheses. The references of the examples illustrated in the
gure: cuckoos and hawks (Davies and Welbergen, 2008; Ma etal., 2018); moths and spiders (Rota and Wagner, 2006); poison frogs (Saporito
etal., 2007); ladybugs (María Arenas etal., 2015); plovers (de Framond etal., 2022); lizards (Bateman and Fleming, 2009).

Research article Evolutionary Biology
Mizuno etal. eLife 2024;13:RP96338. DOI: https://doi.org/10.7554/eLife.96338 3 of 23
characteristics (e.g. eye- like shape) and is presented as a pair, predation avoidance will increase,
assuming eyespots imitate potential predators. Second, the conspicuousness hypothesis posits that
eyespots are simply conspicuous patterns that prevent attacks due to negative predator responses
caused by sensory bias, neophobia, or sensory overload (Stevens, 2005; Stevens and Ruxton, 2014).
The hypothesis states that the eye- like shape and patterns arranged in pairs do not necessarily deter
predators. Rather, it is their conspicuous appearance that makes them effective predator deterrents,
and any resemblance to eyes is coincidental. Eyespots can act as an aposematic signal for potential
predators. For example, if the size of the pattern (one of the measures of conspicuousness) increases,
the avoidance effect will also increase. Third, the deflection hypothesis suggests that predator attacks
should be directed toward eyespots to avoid damage to vital body parts (Hill and Vaca, 2004;
Olofsson etal., 2010; Kodandaramaiah etal., 2013; Olofsson etal., 2013a; Merilaita etal., 2017).
The eye mimicry and conspicuousness hypotheses are usually applied to explain large eyespots, while
the deflection hypothesis is used to interpret the function of small ones (Stevens, 2005; Kodandara-
maiah, 2011; Stevens and Ruxton, 2014). The first two of these hypotheses focus on how eyespots
prevent predators from attacking, specifically whether it is because they resemble eyes or are conspic-
uous. The third hypothesis focuses on whether eyespots divert a predator’s attack away from vital
body parts by drawing the predator’s attention to them. Thus, in this third hypothesis, whether the
eyespots resemble eyes or are conspicuous is not the central issue (Stevens, 2005; Kodandaramaiah,
2011; Stevens and Ruxton, 2014). Although there seems to be little disagreement in the deflection
hypothesis (Lyytinen etal., 2004; Pinheiro etal., 2014; Ho etal., 2016, but see also Lyytinen etal.,
2003), why large eyespots can intimidate avian predators has been controversial (Stevens, 2005;
Stevens and Ruxton, 2014). This is because while the eye mimicry and conspicuousness hypoth-
eses are not mutually exclusive, the key mechanism that explains why predators react negatively to
eyespots is clearly different.
Lepidopterans, such as butterflies and moths, have been the leading models for testing the eye
mimicry and conspicuousness hypotheses. A typical empirical study has adult individuals, caterpillars,
or their models as prey, with birds as predators (reviewed in Stevens, 2005; Stevens and Ruxton,
2014; Kodandaramaiah, 2011). According to the eye mimicry hypothesis, avian predators perceive
the eyespots as the eyes of a potential enemy. For example, great tits (Parus major) showed more
aversive responses to animated butterflies with a pair of large eyespots than those without, and such
eyespots were more effective than modified, less mimetic, but equally contrasting patterns (De Bona
etal., 2015). Although several studies have supported the eye mimicry hypothesis (e.g. Blest, 1957a;
Merilaita etal., 2011; De Bona etal., 2015), many conspicuous patterns other than eyespots, such
as dots and stripes, likely deter attacks from predators as well (Stevens etal., 2008a; Stevens etal.,
2009a; Dell’aglio etal., 2016; Ximenes and Gawryszewski, 2020). Some field experiments with
artificial prey have supported the conspicuousness hypothesis, demonstrating survival rates for both
conspicuous (eyespots and non- eyespots) pattern prey stimuli were higher than control prey stimuli
(Stevens etal., 2007b; Stevens etal., 2008a; Stevens etal., 2009a). Such discrepancies might have
arisen from differences in experimental design between studies, such as the size, number, and shape
of the presented pattern stimuli or the bird species used as subjects in the experiments (Stevens,
2005; Stevens, 2007a). However, there has been no systematic attempt to synthesise and compare
earlier studies quantitatively.
Here, we conduct a systematic review with meta- analysis to synthesise empirical evidence on the
intimidating effects of eyespots and the factors that contribute to predator avoidance responses
towards them. To examine the two hypotheses above, we ask three interrelated questions. First, we
examine whether conspicuous patterns, namely eyespots and non- eyespot patterns (i.e. conspicuous
patterns other than eyespots), influence bird responses or prey survival in a manner that increases the
success of predator avoidance. Second, we test whether pattern resemblance to eyes (eye- like shape)
is the key to predator avoidance (which differentiates the eye mimicry hypothesis from the conspicu-
ousness hypothesis). For the first and second questions, we use (phylogenetic) multilevel meta- analytic
models. Third, we examine what factors promote bird response and increase prey survival by testing
eight moderators (treatment stimulus pattern types, namely eyespots vs. non- eyespots, pattern area,
the number of pattern shapes, prey material type, maximum pattern diameter/length, total pattern
area, total area of prey surface, and prey shape type) (Figure2bc). For the third question, we apply
meta- regression models to evaluate how these moderators influence predator avoidance. We assess

Research article Evolutionary Biology
Mizuno etal. eLife 2024;13:RP96338. DOI: https://doi.org/10.7554/eLife.96338 4 of 23
?
Figure 2. Overview of the dataset. (a)Preferred Reporting Items for Systematic Reviews and Meta- Analyses
(PRISMA)- like owchart of the systematic literature search for the meta- analysis. (b) and (c)Details of the main
moderators examined in the meta- analysis. (d)The phylogenetic tree of bird species included in the meta- analysis,
together with the sample sizes and number of effect sizes per species.

Research article Evolutionary Biology
Mizuno etal. eLife 2024;13:RP96338. DOI: https://doi.org/10.7554/eLife.96338 5 of 23
publication bias to check the robustness of our findings. Throughout our review and analysis process,
we adhere to PRISMA (Preferred Reporting Items for Systematic Reviews and Meta- Analyses; Moher
et al., 2009) and PRISMA- EcoEvo (Preferred Reporting Items for Systematic reviews and Meta-
analyses in Ecology and Evolutionary biology; O’Dea etal., 2021) guidelines and report this study
(Figure2a; Supplementary file 1; for detailed methods, see Materials and methods).
Results
Screening outcomes and dataset characteristics
We obtained 270 effect sizes from 33 studies (164 experiments) for our analysis (Blest, 1957a; Jones,
1980; Inglis et al., 1983; Wourms and Wasserman, 1985; Lyytinen etal., 2003; Forsman and
Herrström, 2004; Lyytinen etal., 2004; Vallin etal., 2005; Stevens etal., 2007b; Stevens etal.,
2008a; Stevens etal., 2008b; Stevens et al., 2009a; Stevens etal., 2009b; Brilot et al., 2009;
Kodandaramaiah etal., 2009; Vallin etal., 2010; Merilaita etal., 2011; Vallin etal., 2011; Blut
etal., 2012; Hossie and Sherratt, 2012; Wert, 2012; Hossie and Sherratt, 2013; Olofsson etal.,
2013a; Olofsson etal., 2013b; Stevens etal., 2013; Skelhorn et al., 2014; Hossie etal., 2015;
Mukherjee and Kodandaramaiah, 2015; Olofsson etal., 2015; Ho etal., 2016; Skelhorn et al.,
2016; Postema, 2022). The screening process and reasons for exclusion at the full- text screening
stage are summarised in the PRISMA- like flowchart (Figure2a), with additional details available in
Supplementary file 2, which comprises a list of included/excluded studies. Of the dataset, 68.9%
of effect sizes came from eyespot presentation experiments (Figure2b). The remaining 31.1% of
effect sizes came from non- eyespot pattern presentation experiments (Figure2b). The latter cate-
gory encompassed various shapes, including circles (71.4%), rectangles (16.7%), diamonds (6.0%),
complex patterns (combinations of circles and diamonds; 4.8%), and stripes (1.1%); 93.7% of the
control stimuli used in these experiments involved the removal of the pattern used in the treatment
stimuli; the remaining stimuli were camouflage patterns (6.3%). Prey shape type used for stimulus
presentation varied from real or imitation of a particular butterfly (24.4%) to simply a piece of paper
(21.5%) (Figure2b). The number of pattern shapes varied between studies from one to 11, but in
artificial
real
non-eyespots
eyespots
overall
effect
0.00 2.50
k = 270 (33)
k = 186 (26)
k = 84 (11)
k = 66 (16)
k = 204 (17)
precision (1/se) 10 20 30
effect sizes (log response ratio: lnRR)
ns
ns
(a)
(b)
(c)
Figure 3. Mean effect sizes of (a)overall for conspicuous patterns (eyespots and non- eyespots), (b)effects split by experiments with eyespot versus
non- eyespot patterns, and (c)two prey types used in the experiments. Thick horizontal lines represent 95% condence intervals, and thin horizontal lines
represent 95% prediction intervals. The points in the centre of each thick line indicate the average effect size. k is the number of effect sizes used to
estimate the statistics, followed by the number of studies in the brackets.

Research article Evolutionary Biology
Mizuno etal. eLife 2024;13:RP96338. DOI: https://doi.org/10.7554/eLife.96338 6 of 23
most experiments, they were two (i.e. a pair of shapes; Figure2c). Additionally, we found that the size
of these patterns, both area and maximum diameter/length, exhibited considerable variation across
studies (Figure2c). The total area of the patterns and stimulus also varied widely (Figure 2c). The
studies reported responses to conspicuous pattern stimuli by seven bird species (Figure2d). Chickens
(Gallus gallus) and common starlings (Sturnus vulgaris) were the most studied birds in our dataset.
Apart from chickens (eight studies) and Eurasian blue tits (Cyanistes caeruleus; five studies), effect
sizes were available from just one or two studies per species. Six of the seven species were omnivores,
and one (yellow bunding; Emberiza sulphurata) was a granivore (Tobias etal., 2022).
Does the presence of conspicuous patterns affect predator avoidance?
The overall mean effect size, calculated as the natural logarithm of the response ratio (lnRR) in this
study, was statistically significant (for details on effect size calculation, see Materials and methods).
This showed a 21.86% (the percentage value is the back- transformed values of lnRR) increase in the
probability of predator avoidance, such as higher prey survival rates or eliciting fewer attacks from
birds (estimate=0.20, 95%CI = [0.08, 0.31], t[df = 268] = 3.40, p = 0.0008), in prey with conspicuous
patterns than in prey without such patterns (Figure3a). Total heterogeneity across effect sizes was
high (I2=96.50%); more specifically, observation ID (representing the within- study effect) accounted
for the most heterogeneity, 79.88%, with study ID (representing between- study effect) accounting for
the remaining 16.61%.
Is there a difference in predator avoidance between eyespots and non-
eyespot patterns?
There was no statistically significant difference between the effects of eyespots and non- eyespot
patterns (F[df1 = 1, df2 = 268] = 0.33, p = 0.57, R2 = 0.27%; Figure3b). On average, eyespot patterns resulted
in 24.37% (estimate = 0.22, 95%CI = [0.08, 0.35], t[df = 268] = 3.17, p = 0.002) and non- eyespot patterns
in 17.11% (estimate = 0.16, 95%CI = [–0.02, 0.34], t[df = 268] = 1.71, p = 0.09) increases in predator
avoidance compared with control stimuli, although this trend was not statistically significant for non-
eyespots (Figure3b).
What factors promote predator avoidance?
Our uni- moderator meta- regression model with pattern area (individual shape area) showed that larger
patterns were associated with an increase in predator avoidance (estimate = 0.11, 95%CI = [0.03,
0.19], t[df = 268] = 2.71, p = 0.007, R2 = 8.56%; Figure4a). The total pattern area also promoted predator
avoidance (estimate = 0.09, 95%CI = [0.004, 0.17], t[df = 268] = 2.07, p = 0.04, R2 = 5.18%; Figure5a).
Similarly, the maximum diameter/length of the pattern positively influenced predator avoidance (esti-
mate = 0.19, 95%CI = [0.04, 0.35], t[df = 268] = 2.46, p = 0.01, R2 = 6.62%; Figure5b). In contrast, an
increased number of pattern shapes significantly reduced the effect of predator avoidance (estimate
= –0.06, 95%CI = [-0.11, –0.008], t[df = 268] = –2.29, p = 0.02, R2 = 2.46%; Figure4b). We found no
significant effects of total prey surface area on predator avoidance (estimate = –0.03, 95%CI = [–0.15,
0.09], t[df = 268] = –0.48, p = 0.63, R2 = 0.42%; Figure5c). Predator avoidance was not statistically signifi-
cantly affected by differences in whether the presented prey looked like a real lepidopteran species
(F[df1 = 1, df2 = 268] = 0.12, p = 0.72, R2 = 0.13%). Both types of prey material (real/imitation and abstract
butterfly) had similar positive trends (Figure3c), with the former increasing predator avoidance by
25.55% (estimate = 0.23, 95%CI = [0.03, 0.43], t[df = 268] = 2.24, p = 0.03) and the latter by 20.07%
(estimate = 0.18, 95%CI =[0.04, 0.33], t[df = 268] = 2.44, p = 0.02). Furthermore, when also considering
prey type (Figure6), abstract and real butterflies significantly exhibited increased predator avoidance
by 37.98% (estimate = 0.32, 95%CI = [0.11, 0.53], t[df = 268] = 3.04, p = 0.003) and by 25.40% (estimate
= 0.23, 95%CI = [0.03, 0.42], t[df = 268] = 2.25, p = 0.03), respectively, but artificial abstract caterpillars
(estimate = 0.07, 95%CI = [–0.18, 0.31], t[df = 266] = 0.53, p = 0.60) and artificial abstract prey (estimate
= 0.01, 95%CI = [–0.35, 0.37], t[df = 266] = 0.06, p = 0.95) did not, respectively. When comparing each
prey type (e.g. abstract butterfly vs. real butterfly), none of the differences was statistically significant
(Figure6).
The multi- moderator (full) regression model showed that only pattern area positively affected pred-
ator avoidance (estimate = 0.10, 95%CI = [0.009, 0.18], t[df = 266] = 2.16, p = 0.03; Supplementary file
3). Contrary to the uni- moderator regression model, the number of patterns showed no significant

Research article Evolutionary Biology
Mizuno etal. eLife 2024;13:RP96338. DOI: https://doi.org/10.7554/eLife.96338 7 of 23
effects on predator avoidance, although the consistent trend remained (estimate = –0.05, 95%CI =
[–0.11, 0.004], t[df = 266] = –1.84, p = 0.07; Supplementary file 3). The full model accounted for 8.33%
of the variation in the dataset. The complete output of the multi- moderator model is displayed in
Supplementary file 3.
Publication bias
The funnel plot showed no visual sign of funnel asymmetry (Figure7a). The meta- regression anal-
ysis, which included the square root of the inverse of the effective sample size, further supported
this observation by showing that the effective sample size did not significantly predict the effect size
values (estimate = –0.09, 95%CI = [–0.83, 0.65], t[df = 266] = –0.24, p = 0.81; Figure7b). There was
−2.00−1.00 0.00 1.00 2.00 3.00 4.00
1.50
(4.48 mm2)
3.00 4.50 6.00
(403.48 mm2)
log−transformed area (mm2)
(a)
246810
number of patterns
precision (1/se) 10 20 30
(b)
k = 270
k = 270
precision (1/se) 10 20 30
−2.00−1.00 0.00 1.00 2.003.004.00
effect sizes (lnRR)
effect sizes (lnRR)
Figure 4. The relationships between (a)prey conspicuous pattern area (log- transformed) and effect sizes and (b)number of prey conspicuous patterns
and effect sizes. Circle sizes are scaled according to precision, k represents the number of effect sizes. Each tted regression line is shown as a coloured
straight line, and 95% condence and prediction intervals are shown as dashed and dotted coloured lines, respectively.

Research article Evolutionary Biology
Mizuno etal. eLife 2024;13:RP96338. DOI: https://doi.org/10.7554/eLife.96338 8 of 23
no detectable trend suggesting that more recent publications consistently showed lower or higher
effect size values, which would have indicated the presence of time- lag publication bias (estimate =
−0.0008; 95%CI = [−0.01, 0.01], t[df = 266] = –0.12, p = 0.90; Figure7c). We obtained the same trends
from multi- moderator meta- regressions (Figure8).
Discussion
Eyespots and non- eyespot patterns did not differ significantly in the magnitude of deterring effects
(Figure 3b). Avian predators showed similar avoidance responses to the conspicuous patterns
compared to control ones (Figure3a). Specifically, larger pattern sizes played a crucial role in eliciting
negative responses from birds (Figure4a). Furthermore, negative responses from birds showed the
tendency to decline with increasing pattern number: single patterns were likely more intimidating
2.003.00 4.005.006.0
07
.00
log−transformed total pattern area
effect sizes (lnRR)
precision (1/se) 10 20 30
(a)
1.00 2.00 3.00
log−transformed diameter
precision (1/se) 10 20 30
(b)
−2.000.002.004.00
5.00 6.007.00 8.009.00 10.00
log−transformed total prey surface area
precision (1/se) 10 20 30
(c)
k = 270
k = 270
k = 270
−2.000.002.004.00−2.00 0.00 2.00 4.00
effect sizes (lnRR)effect sizes (lnRR)
Figure 5. The relationships between (a)total pattern area, (b)pattern maximum diameter/length, and (c)total prey surface area and effect sizes. k shows
the number of effect sizes. Each tted regression line is shown as a solid straight line, and 95% condence and prediction intervals are shown as dashed
and dotted lines, respectively.

Research article Evolutionary Biology
Mizuno etal. eLife 2024;13:RP96338. DOI: https://doi.org/10.7554/eLife.96338 9 of 23
than a group of patterns (Figure4b). Taken together, our results support the conspicuousness hypoth-
esis rather than the eye mimicry hypothesis.
Eye mimicry or conspicuousness hypothesis?
Overall, our meta- analysis showed that conspicuous patterns could increase predator avoidance
by over 20%. Specifically, our results indicate that conspicuousness per se can be advantageous in
avoiding bird predation (Figures3ab and 4). The evidence favouring the conspicuousness hypothesis
comes mainly from a series of field experiments by Stevens and his colleagues (Stevens etal., 2007b;
Stevens etal., 2008a; Stevens etal., 2009a). They showed that both eyespots and non- eyespots
improved the prey survival similarly compared to non- conspicuous patterns (Stevens etal., 2007b;
Stevens etal., 2008a; Stevens etal., 2009a). In addition, their research showed prey with more
conspicuous patterns (i.e. large- size patterns) tended to survive more than others (Stevens et al.,
2007b; Stevens etal., 2008a; Stevens etal., 2009a), and eye resemblance (e.g. number or pattern
shapes) did not significantly affect the prey’s survival (Stevens etal., 2007b; Stevens etal., 2008a;
Stevens etal., 2009a). Given that these pattern stimuli used in the experiments are rarely or never
found in natural environments (Stevens etal., 2007b), the most parsimonious explanation for these
results is neophobia or dietary conservatism in birds (Ord etal., 2021; Marples etal., 1998; Marples
and Kelly, 1999). Both phenomena appear to diminish with habituation and/or learning. A few studies
investigated such factors for intimidating effects, and they showed that repeated encounters made
birds more habituated to eyespot patterns (Blest, 1957a; Inglis etal., 1983; Skelhorn etal., 2014).
We need more systematic tests of bird habituation to vividly- or aposematic- coloured patterns to
better understand the evolution and function of such patterns in Lepidoptera.
While our meta- analytic results favour the conspicuousness hypothesis, several empirical studies
support the eye mimicry hypothesis. For example, De Bona etal., 2015 found that a pair of eyespots
abstract
butterfly
abstract
caterpillar
abstract
prey
real
butterfly
0.00
2.50
precision (1/se) 10 20 30
k = 66 (16)
k = 68 (3)
k = 68 (6)
k = 78 (8)
effect sizes (log response ratio: lnRR)
Figure 6. Mean effect sizes of total prey shape types. Thick horizontal lines represent 95% condence intervals, and thin horizontal lines represent
prediction intervals. The points in the centre of each thick line indicate the average effect size. k shows the number of effect sizes.

Research article Evolutionary Biology
Mizuno etal. eLife 2024;13:RP96338. DOI: https://doi.org/10.7554/eLife.96338 10 of 23
Figure 7. Funnel plot and relationships between effect Sizes, effective sample size, and publication year. (a) Funnel
plot using effect size and its inverse standard error. The relationship between effect sizes and (b) the square root
of the inverse of effective sample size and (c) publication year. In (b) and (c), circle sizes are scaled accordingly to
precision, and k represents the number of effect sizes. Each tted regression line is shown as a straight line, and
95% condence and prediction intervals are shown as dashed and dotted lines, respectively.

Research article Evolutionary Biology
Mizuno etal. eLife 2024;13:RP96338. DOI: https://doi.org/10.7554/eLife.96338 11 of 23
of Caligo martia was as effective as true owl eyes and more efficient in eliciting predator avoidance
responses than less mimetic but equally contrasting circles. Blut and Lunau, 2015 created artificial
eye- spotted prey with different similarities to the vertebrate eyes and checked their survival rates in a
field experiment. They revealed that the prey with the most mimetic pattern had the highest survival
rate (Blut and Lunau, 2015). Although studies on Lepidoptera larvae are relatively limited, caterpillar
eyespots are considered part of snake mimicry (Stevens and Ruxton, 2014). Some research exam-
ined the benefit of eyespots by presenting artificial caterpillars (marked with eyespots and control)
made from dyed pastry to wild birds and showed that eyespots improved survival (Hossie and Sher-
ratt, 2012; Hossie and Sherratt, 2013; Skelhorn etal., 2014). Despite these convincing pieces of
empirical evidence, our meta- analytic results showed that eye resemblance did not improve predator
−2.000.00 2.00 4.00
0.20 0.40 0.60
square root of inverse of effective sample size
precision (1/se) 10 20 30
(a)
−2.000.00 2.00 4.00
1960 1980 2000 2020
year of publication
precision (1/se) 10 20 30
(b)
effect sizes (lnRR)
k = 270
k = 270
effect sizes (lnRR)
Figure 8. The relationship between (a) effect sizes and the square root of the inverse of effective sample size and (b) relationship between effect sizes
and publication year. Both plots were based on the multi- moderator model. k shows the number of effect sizes. Each tted regression line is shown as a
solid straight line, and 95% condence intervals and prediction intervals are shown as dashed and dotted lines, respectively.

Research article Evolutionary Biology
Mizuno etal. eLife 2024;13:RP96338. DOI: https://doi.org/10.7554/eLife.96338 12 of 23
avoidance. If the eye mimic hypothesis was true, we would have seen a clear difference between
studies investigating eyespots and non- eyespots.
However, we observed little heterogeneity among studies, despite finding high heterogeneity
within individual studies. This finding implies that if each study followed similar experimental proce-
dures within studies, our main result on predator avoidance would be more generalisable. The high
within- study heterogeneity can be caused by varying stimulus characteristics contributing to the effect
size variations, even in the same studies. Bird phylogenetic relatedness explained little heterogeneity
in our predator dataset, but this may have occurred because a limited number of subject bird species
(i.e. chickens, common starlings, Eurasian blue tits) dominated our dataset (Figure 2d). While we
cannot exclude the possibility of species differences in birds’ responses to the conspicuous patterns,
our analysis indicated that bird species identity did not explain the observed variation in predator
avoidance.
We also note that conspicuous patterns can also be important for conspecific communication in
butterflies, not just for avoiding predation (Stevens, 2005; Crees etal., 2021). For example, eyespots
on Bicyclus anynana are known to function as sexual signals. For example, males choose females
depending on eyespot size and reflectance (Robertson and Monteiro, 2005). Regarding the non-
eyespot patterns, males of Heliconius cydno and H. pachinus can recognise conspecific females by
the bright colour of wing patches (Kronforst etal., 2006; Finkbeiner etal., 2014). Conspicuous
patterns can also act as social signals in other taxa (e.g. birds: Mason and Bowie, 2020), but this
function remains unclear in butterflies. Therefore, the diversity of patterns on wings could be shaped
by intra- specific and inter- specific communication. We should simultaneously consider the influence of
anti- predator and sexual/social signalling functions on the evolution of butterfly conspicuous patterns
(Robertson and Monteiro, 2005; Ng etal., 2017; Huq etal., 2019).
What factors explain the observed heterogeneity?
The indicators of pattern size, including each pattern area (Figure4a), total pattern area (Figure5a),
and maximum diameter/length (Figure 5b), were the most important moderators of effect sizes,
overall indicating that large patterns could promote predator avoidance. Notably, these size metrics
were correlated, so they are not independent of each other. Several studies suggested that the
pattern size difference is related to the difference in prey survival (Stevens etal., 2008a; Kodan-
daramaiah etal., 2013; Ho etal., 2016). For example, eyespots larger than 6.0mm may have a
strong deterrent effect with increasing size (Ho etal., 2016), but such patterns may increase the
visibility of lepidopterans, and their presence may increase predation rates as well (Lindström etal.,
2001). Indeed, small conspicuous patterns tend to attract predators' attention, as explained by the
deflection hypothesis (Stevens, 2005; Humphreys and Ruxton, 2018). The effect may contribute
to the observed negative overall effect sizes (Figures3 and 4). Considering studies on B. anynana
with eyespots with a deflecting effect (maximum diameter is about 5.0mm; Supplementary file 5),
a size of at least 6.0mm is required to avoid predator approach. However, it is uncertain whether the
effect would linearly increase with size or whether an optimal size exists. Although eyespot sizes on
actual Lepidoptera may be restricted by their body or wing size (e.g. Hossie etal., 2015, but see also
Kodandaramaiah etal., 2013), it would be interesting to find a maximum threshold for patterns that
promote predator avoidance responses in birds.
Among other moderators tested (prey material type, total pattern area, and prey shape type),
the only moderator that seemed to explain heterogeneity was the number of patterns (Figure4b;
yet it is likely inconclusive; see Supplementary file 3). Previous studies predominantly employed
a single pattern or a pair of patterns, leading to limited variations. Nonetheless, our findings indi-
cate that a single eyespot is equally or more effective than a pair of eyespots. Consequently, the
resemblance to a pair of eyes, a crucial aspect of the eye mimicry hypothesis, may be optional for
effective predator avoidance. Indeed, we should note that the presence of both eyes is unnecessary
for birds to recognise their predators because birds may often see only one eye of their predators.
To disentangle the two hypotheses, we recommend conducting the following experiments with two
key features (Stevens, 2007a; Stevens etal., 2008a; Stevens etal., 2009b): a set of stimuli that (1)
have the same size (area or diameter/maximum length of each pattern or total pattern area) but with
different numbers of patterns ranging from a few usually found in Lepidoptera to numerous patterns
unlike those seen in them, and (2) are presented with the same number of patterns and the same size

Research article Evolutionary Biology
Mizuno etal. eLife 2024;13:RP96338. DOI: https://doi.org/10.7554/eLife.96338 13 of 23
but different pattern shapes. Results from these experiments could deepen our current knowledge,
allowing us to inch toward a more definitive answer.
Knowledge gaps and future opportunities
Along with other conspicuous patterns, eyespots are believed to deter bird predation, and our meta-
analysis supports this function. However, five major gaps remain in the current literature and our
knowledge. First, birds and humans likely perceive eye- like shapes differently based on the interspe-
cific diversity of bird vision (Martin, 2017). For example, most bird species can detect ultraviolet light,
which is invisible to humans, and the ultraviolet reflection of the butterflies' eyespots may contribute
to predator avoidance (e.g. Olofsson etal., 2010, Olofsson etal., 2013a). In addition, researchers
can quantify and objectively evaluate conspicuousness, such as size and number, but the assessment
of 'eye mimicry' remains subjective. Thus, it could be premature to conclude that eyespots on Lepi-
doptera resemble vertebrate eyes universally.
Second, some lepidopterans present conspicuous patterns to potential predators in combina-
tion with other elements, such as sounds and movements (Blest, 1957a; Blest, 1957b; Bura etal.,
2016; Vallin etal., 2005; Drinkwater etal., 2022), presumably to emphasise the conspicuousness of
the patterns. Most of the current literature does not take these effects into account in experiments,
although some studies argue in favour or against their importance (e.g. Blest, 1957a; Vallin etal.,
2005). We should also consider how factors other than those constituting the pattern (e.g. colour,
number, and size) are involved in the predator avoidance function of eyespots. The location of the
butterfly’s eyespot patterns varies from species to species as well; eyespots exist on the wings' ventral,
dorsal, or both sides. Not only the dorsal eyespot patterns, which were used in most studies, but also
the ventral eyespot patterns should be explored. In addition, we need to avoid presenting patterns
unnaturally when using real butterflies in experiments. For example, many owl butterflies (family
Caligo) have a pair of eyespot patterns on the ventral side. Their eyespots are usually visible to birds
when the wings are closed and would not present side by side as in the eyes of the owl’s frontal face.
Third, recent studies have shown that birds are sensitive to the gaze of other individuals and may
respond more aversively when their gazes are directed at them (e.g. Carter etal., 2008; Clucas etal.,
2013; Davidson etal., 2015). Skelhorn and Rowland, 2022 showed that the anti- predation effect
may be further enhanced if the inner circle of the eyespot is in a more gazing- like position for subject
birds. However, further research is needed to investigate the importance of the position of the inner
circle.
Fourth, as mentioned above, studies focusing on caterpillar eyespots are much more scarce
compared to butterflies; Hossie and Sherratt, 2014 have shown similarities between caterpillars and
snakes, but the response of birds to actual caterpillars has not been experimentally tested. Conversely,
in butterflies, similarities between the eyespot patterns on wings and the eyes of birds of prey have
not been investigated.
Finally, birds are generally considered as potential predators of butterflies and caterpillars.
Although other taxa species, such as invertebrates (Sang and Teder, 2011; Prudic etal., 2015; Chan
etal., 2021), lizards (Lyytinen etal., 2003; Vlieger and Brakefield, 2007; Halali etal., 2019), and
rats (Wiklund etal., 2008; Olofsson etal., 2011; Olofsson et al., 2012; Postema, 2022), are also
known to prey on lepidopterans, there are much fewer studies using non- avian species as preda-
tors. The effectiveness of eye mimicry versus being conspicuous may vary depending on the pred-
ator, and either one may be more effective depending on specific predator species. Therefore, we
should expand the range of taxa used for experiments to get a better and more generalisable under-
standing of the eyespots’ function and evolution in butterflies and caterpillars. Additionally, much of
the research has been conducted in Europe and North America. Of the studies we included, only two
were from other regions (India Mukherjee and Kodandaramaiah, 2015 and Singapore Ho etal.,
2016). The empirical results may differ in areas with many species of lepidopterans with eyespot
patterns (e.g. Ord etal., 2021).
Knowing the effects of conspicuous patterns may contribute to creating a world where birds and
humans can live more harmoniously. Both eyespots and non- eyespot patterns have already been used
to control birds, particularly in agriculture, although their effectiveness has been questioned (e.g.
Avery etal., 1988; Nakamura etal., 1995). Such uncertainty may reflect our limited understanding
of why birds avoid eyespots and non- eyespots. Nevertheless, visual stimuli are less likely to harm birds

Research article Evolutionary Biology
Mizuno etal. eLife 2024;13:RP96338. DOI: https://doi.org/10.7554/eLife.96338 14 of 23
or affect the natural environment than others (e.g. nest/egg destructions or toxic chemicals; reviewed
in Linz et al., 2015). Therefore, when proven effective, they could be used for better pest control,
population management, and conservation (McLennan etal., 1995).
Conclusion
We have shed light on a traditional but controversial research topic that has fascinated behavioural
ecologists for decades. Our findings provide a better understanding of the evolution of signal designs,
but also show that more work is needed to understand the function of the eyespot patterns in Lepi-
doptera, such as whether eyespot patterns evolved due to mimicry or conspicuousness.
Materials and methods
We preregistered our methods and planned analyses before data extraction and analysis in Open
Science Framework (https://osf.io/ymwvb; Mizuno etal., 2023).
Search protocols
We used the PICO (Population, Intervention, Comparator, Outcome; Table1) framework (Foo etal.,
2021) to specify the scope of our research questions and to inform our literature searching and
screening. We conducted a comprehensive literature search across multiple databases, including
Scopus, ISI Web of Science, Google Scholar (for non- English studies), and Bielefeld Academic Search
Engine (for unpublished theses; i.e. grey literature). We designed the search strings (see Supplemen-
tary file 4) to identify studies that used experimental methods to examine the effects of eyespot
patterns on birds' predation behaviours. We did not set any temporal restrictions on the database
searches. Additionally, we conducted backward and forward reference searches within the Scopus
database using four key publications (Stevens, 2005; Kodandaramaiah, 2011; Stevens and Ruxton,
2014; Drinkwater et al., 2022). The strings were translated for searches in non- English languages,
and search results were assessed by reviewers with expertise in the respective languages: AM for
Japanese, ML for Polish and Russian, PP for Portuguese and Spanish, and YY for Simplified and Tradi-
tional Chinese. We limited Google Scholar searches to the top 100 results in each language, sorted
by relevance. In cases of disagreement between the reviewers, discrepancies were discussed and
resolved to reach a consensus. The screening process and results are shown in the PRISMA- like flow-
chart (Figure2a).
Eligibility criteria
We set specific criteria for including studies in our meta- analysis (according to our pre- registered
protocol). Initial screening, including titles, abstracts, and keyword assessment for English- language
bibliographic records, was conducted by AM and ML using Rayyan (https://www.rayyan.ai; Ouzzani
etal., 2016) following predefined inclusion criteria. Subsequently, AM and PP independently screened
the full texts of studies that passed the initial screening. To be eligible, a study had to conduct exper-
iments and provide data on bird behavioural responses or prey survival/attacked rates. We excluded
studies solely involving non- avian predators, such as fish, insects, mammals, or other species.
However, studies that included a mix of species from different taxonomic groups were allowed if the
primary focus was on avian predation. In our analysis, we only considered research that presented
both conspicuous and control (non- conspicuous) patterns as stimuli. We omitted studies using actual
Table 1. Descriptions of the population, intervention, comparator, and outcome (PICO) were used to
define the scope of this study.
PICO Description
Population Birds as predators and butteries, moths, caterpillars, and their models as prey
Intervention Presenting eyespot or conspicuous pattern stimulus to birds
Comparator Presenting stimulus that is neither eyespot nor conspicuous patterns
Outcome
Avian behavioural responses to eyespot or conspicuous pattern stimuli
The probability of prey surviving or being attacked (for the stimuli)

Research article Evolutionary Biology
Mizuno etal. eLife 2024;13:RP96338. DOI: https://doi.org/10.7554/eLife.96338 15 of 23
predator or human eyes as stimuli since we focused on understanding how eyespot patterns in butter-
flies and caterpillars, which are unlikely to resemble specific bird or vertebrate species eyes, affect
predation avoidance (Janzen etal., 2010). We also excluded studies that used bright and contrasting
patterns as control stimuli because such stimuli would prevent comparison with eyespots or non-
eyespot patterns. Furthermore, we focused only on studies that used real or artificial butterflies,
moths, caterpillars, or a piece of paper as prey or presented stimuli. We also did not consider research
that only investigated avian physiological responses to conspicuous patterns. In addition, we did not
include studies that only assessed whether prey with eyespots or conspicuous patterns were less likely
to be attacked by birds, based on wing or body damage alone, without including control stimuli. This
is because it was not possible to quantitatively assess the effect of eyespots or non- eyespot patterns
on predation avoidance without control stimuli.
Data collection
We extracted four types of information from each study. First, we collected citation information, such
as title, author name, and publication year. Second, we gathered the details of the presented stimuli
used in each experiment within studies: type of control pattern (plain neutral- coloured or camou-
flaged), type of treatment pattern (eyespots or non- eyespot patterns), pattern area (mm2: area per
shape comprising the pattern), total pattern area (mm2: when multiple patterns exist on the presented
stimulus, it denotes the total area of all patterns; for stimuli with single eyespot or distinct pattern,
the value equals the pattern area), linear size of the pattern (mm: e.g. maximum diameter or length
of pattern), number of shapes in pattern, total area of prey surface (mm2: e.g. butterfly wings and
caterpillar bodies), prey material type (i.e. whether a real butterfly or a complete imitation of a partic-
ular butterfly was used as prey), and prey shape type (a further subdivision of the former). For non-
eyespot patterns, we also noted pattern shapes (e.g. circles, stripes, and triangles). In each study,
bird responses to control and treatment pattern stimuli and prey survival/attacked rates when these
patterns were present were reported. Bird responses contained a variety of measures, including the
number of attacks and escape behaviours, latency to attack, latency to approach, and the proportion
of birds attacking the presented stimuli. Henceforth, we refer to these measures and responses as
‘predator avoidance.’ Third, we obtained data for calculating effect sizes (e.g. mean, standard devi-
ation or standard error, and sample size of control and treatment group) from plots using WebPlot-
Digitizer 4.6.0 (https://automeris.io/WebPlotDigitizer), detailed tables, texts, or raw data. In survival
analysis plots, we extracted data at the point in time when the difference between the ‘survival’ or
‘attacked’ rates of the intervention and comparison groups was greatest as outcomes. Study design
(i.e. whether experiments were done independently or dependently between the control and treat-
ment group) was also recorded. Fourth, we gathered predator and prey information, specifically, the
study species (common English name and scientific name) and predator diet type. In some cases,
studies did not use a specific bird species as a predator or a specific lepidopteran species as prey. We
contacted authors when such information was ambiguous or missing. When the paper did not report
the pattern area and diameter of the treatment stimulus or the presented stimulus surface area, AM
calculated or measured them from available images using ImageJ v.1.53i (Abramoff and Ram, 2004).
The dataset was originally divided into two parts. The first part involved the data from presenting
eyespot patterns to avian predators and directly observing their responses (predator dataset). The
sample size or unit of analysis in this part was based on the number of individual avian predators. The
second part involved the data from using real or artificial abstract butterflies, moths, or caterpillars
with eyespots or non- eyespot patterns as stimuli or prey, and observing their survival/attacked prob-
abilities in the field (prey dataset). The sample size or unit of analysis in this part was based on the
number of real or artificial abstract prey. However, we also used the combined dataset that included
both predator and prey datasets, as detailed in the ‘Meta- analysis and meta- regressions’ and ‘Publi-
cation bias’ sections.
Effect size calculation
To obtain the effect size point estimates and sampling variances, we used lnRR (the natural logarithm
of the response ratio) between the means of the treatment and the treatment control stimulus groups
(Hedges etal., 1999; Lajeunessei, 2011; Senior etal., 2020). Positive lnRR values indicate height-
ened aversion in birds and enhanced prey survival, while negative lnRR values signify diminished bird

Research article Evolutionary Biology
Mizuno etal. eLife 2024;13:RP96338. DOI: https://doi.org/10.7554/eLife.96338 16 of 23
aversion and increased prey mortality. The point estimate and sampling variance (var) of lnRR can be
then calculated in:
lnRR
=ln
(M
T
MC)
(1)
var (
lnRR
)
=SD
2
T
N
T
M2
T
+SD
2
C
N
C
M2
C
−2r
√
SD
2
T
N
T
M2
T√
SD
2
C
N
C
M2
C
(2)
where
MT
and
MC
are mean responses of treatment and control groups (e.g. total frequency of
attacking prey, latency of approach, or prey survivability), respectively.
SD
and
N
are (sample) standard
deviations and sample size, respectively. The term, r is the correlation coefficient between responses
of the two groups. Some of our eligible studies used the paired (dependent) study design where
treatment and control samples originated from the same individuals, and sample sizes between the
two groups were the same. None of these studies provided an estimate of
r
. Thus, when calculating
our effect sizes, we assumed that this correlation was 0.5, which is conservative (Noble etal., 2017).
For the other studies that used independent study design, we set
r=0
.
We note that our dataset included proportion (percentage) data (e.g. predator attack rate or prey
survival probability), which are bounded at 0 (0%) and 1 (100%). Therefore, we transformed group
means (
M
) and group standard deviations (
SD
) for proportion data using Equations (3) and (4) before
applying (1) and (2) to calculate lnRR and the sampling variance:
f(M)=arcsine (√M)
(3)
SD
(
f
(
M
))
=
√
SD
2
4
M
(1−
M
)
(4)
where
f
indicates a function, in our case, the arcsine transformation. The standard deviation (SD)
related to this transformation was derived using the delta method before calculating lnRR and the
sampling variance (Macartney etal., 2022). We have also assumed that the standard deviation was
SD (f(M))= 1/√8
if SD was not available.
Meta-analysis and meta-regressions
We used the rma. mv function from the package meta for v.4.4.0 (Viechtbauer, 2010) in R v.4.3.1 (R
Development Core Team, 2023) for our analyses. We started by fitting multilevel, mixed- effect meta-
analytic models to the predator and prey datasets. These meta- analytic models explicitly incorporated
random factors, Study ID, Cohort ID (groups of the same subjects), and Shared control ID (indicating
effect sizes sharing control groups) (Nakagawa etal., 2023b) along with Observation ID, fitted by the
above function (Viechtbauer, 2010). The model for the predator dataset included Species ID and a
correlation matrix related to phylogenetic relatedness for the species as random factors (Nakagawa
and Santos, 2012). This is because we had data on the bird species used in the experiment in the
predator dataset, and we needed to control for phylogenetic relationships between birds. We also
quantified the total I2 (a measure of heterogeneity not attributed to sampling error: Higgins etal.,
2003) and how much each random factor was explained (partial I²), calculated by the i2_ml func-
tion from the package orchaRd v.2.0.0 (Nakagawa etal., 2023a). After running both meta- analytical
models, we found that phylogeny and Species ID did not need to be controlled for in the predator
dataset, as their partial I² were zero (I²=0.00%). That is, these factors explained little heterogeneity
between effect sizes.
Therefore, we merged predator and prey datasets (i.e. full dataset) without considering phylo-
genetic information and used them for the following models. We had, as random effects, Study ID,
Cohort ID, Shared control ID, and Observation ID for our meta- analytic model using the full dataset.
The Cohort ID and Shared control ID were removed from our subsequent meta- regressions because
they both explained little heterogeneity (both partial I²<0.001%). This intercept- only (meta- analytic)
model tested the conspicuous patterns (eyespots and non- eyespots) that affected predator avoidance
(i.e. our first question).
Next, we tested whether eyespots and non- eyespot patterns differ in the magnitude and direc-
tion of the effect of elicited bird predator avoidance and what factors contribute to the deterring

Research article Evolutionary Biology
Mizuno etal. eLife 2024;13:RP96338. DOI: https://doi.org/10.7554/eLife.96338 17 of 23
effects of conspicuous patterns. We performed uni- moderator meta- regression models with each of
eight moderators: treatment stimulus pattern types, pattern area, the number of pattern shapes, prey
material type, maximum pattern diameter/length, total pattern area, total area of prey surface, and
prey shape type (Figure2 bc). We also ran a multi- moderator meta- regression model, including the
first four of the eight variables mentioned in the uni- moderators, due to moderator correlations. We
used log- transformed data for pattern area, total pattern area, total area of prey surface, and pattern
maximum diameter/length in our analysis to normalise these moderators. We created all result plots
in the orchard_plot and bubble_plot functions from the package orchaRd (Nakagawa etal., 2023a).
Publication bias
We used three approaches to assess the presence of publication bias in our study. First, we visually
assessed the funnel plot asymmetry by examining the residuals from a meta- analytic model, which
included all the random factors utilised in our study. These residuals were plotted against the precision
of the effect sizes. Second, we performed an alternative method to Egger’s regression. This method
used the inverse of the effective sample size as a moderator within a multilevel meta- analytic model
(Nakagawa etal., 2022). Third, we examined the possibility of time- lag bias by including publication
year as a moderator in our multilevel meta- analytic model. Uni- moderator models were run for each
inverse of the effective sample size and publication year, and a multi- moderator model was carried
out with the full model including both inverse of the effective sample size and publication year as
moderators.
Additions and deviations
We made two changes to the pre- registration: the addition of four new moderators and the removal
of two moderators. The new moderators were pattern area, total pattern area, total area of prey
surface, and prey shape types, although similar moderators were in the pre- registration such as the
number of eyespots (patterns) and diameter of an eyespot (a pattern). These post- hoc decisions were
taken to refine our initial moderators. We subsequently used them in our meta- regression analyses.
We originally intended to include the broad outcome categories of predator avoidance measure as a
moderator in the models, but the diversity of reported results made categorisation impossible. There-
fore, we did not include it as a moderator. We also collected information on bird diet but decided
not to include it. This decision was because six of the seven bird species in our study were omnivores,
resulting in a lack of variability needed to detect diet effects in our data (for more details, please see
Results).
Acknowledgements
The authors thank Martin Stevens and Ben Brilot for sharing data. This work was supported by Japan
Society for the Promotion of Science Research Fellowship for Young Scientists [JP22KJ0076], Japan
Society for the Promotion of Science Overseas Challenge Program for Young Researchers [202280247]
to Ayumi Mizuno; ARC [DP210100812, DP230101248] to Malgorzata Lagisz and Shinichi Nakagawa;
Japan Society for the Promotion of Science Grant- in- Aid for Scientific Research [20K06809] to Masayo
Soma.
Additional information
Funding
Funder Grant reference number Author
Japan Society for the
Promotion of Science
JP22KJ0076 Ayumi Mizuno
Japan Society for the
Promotion of Science
202280247 Ayumi Mizuno
Australian Research
Council
DP210100812 Malgorzata Lagisz
Shinichi Nakagawa

Research article Evolutionary Biology
Mizuno etal. eLife 2024;13:RP96338. DOI: https://doi.org/10.7554/eLife.96338 18 of 23
Funder Grant reference number Author
Australian Research
Council
DP230101248 Malgorzata Lagisz
Shinichi Nakagawa
Japan Society for the
Promotion of Science
20K06809 Masayo Soma
The funders had no role in study design, data collection and interpretation, or the
decision to submit the work for publication.
Author contributions
Ayumi Mizuno, Conceptualization, Data curation, Software, Formal analysis, Supervision, Funding
acquisition, Investigation, Visualization, Methodology, Writing - original draft, Project administration,
Writing – review and editing; Malgorzata Lagisz, Data curation, Funding acquisition, Investigation,
Methodology, Writing – review and editing; Pietro Pollo, Yefeng Yang, Data curation, Investigation,
Writing – review and editing; Masayo Soma, Conceptualization, Funding acquisition, Visualization,
Writing – review and editing; Shinichi Nakagawa, Conceptualization, Software, Supervision, Funding
acquisition, Methodology, Project administration, Writing – review and editing
Author ORCIDs
Ayumi Mizuno
https://orcid.org/0000-0003-0822-5637
Malgorzata Lagisz
https://orcid.org/0000-0002-3993-6127
Pietro Pollo
http://orcid.org/0000-0001-6555-5400
Yefeng Yang
https://orcid.org/0000-0002-8610-4016
Masayo Soma
http://orcid.org/0000-0002-8596-1956
Shinichi Nakagawa
https://orcid.org/0000-0002-7765-5182
Peer review material
Reviewer #1 (Public Review): https://doi.org/10.7554/eLife.96338.3.sa1
Author response https://doi.org/10.7554/eLife.96338.3.sa2
Additional files
Supplementary files
• Supplementary file 1. Preferred Reporting Items for Systematic Reviews and Meta- Analyses
(PRISMA)- EcoEvo Checklist.
• Supplementary file 2. List of (a) included and (b) excluded studies at the full- text screening stage
with exclusion reasons.
• Supplementary file 3. Summary of a multi- moderator model including all moderators. The
bold typeface is used when a 95% confidence interval (CI) does not contain zero; thus, it can be
interpreted as an existing significant effect in predator avoidance.
• Supplementary file 4. Search strings used for each database. We accessed Scopus, ISI Web of
Science core collection, Google Scholar (Japanese, Polish, Portuguese, Russian, Spanish, Simplified
Chinese, and Traditional Chinese) on 08/06/2023, and Bielefeld Academic Search Engine (BASE) on
26/06/2023. BASE was used as a source of grey literature. We conducted backward and forward
reference searches for key review articles using Scopus on 19/06/2023. We modified search strings
to collect studies to capture studies examining the effects of eyespot patterns on birds using
experimental methods. Search strings were adapted to the structure of each database.
• Supplementary file 5. Average maximum diameter of eyespots on Bicyclus anynana. AM obtained
the pictures from lepdata.org/photos/animals/ and https://data.nhm.ac.uk/ and measured the
eyespot diameters. Raw data is available here: https://ayumi-495.github.io/eyespot/ and on GitHub
(copy archived at Mizuno, 2024) and Zenodo.
• MDAR checklist
Data availability
Raw data, analysis script and supplementary materials are available at https://ayumi-495.github.io/
eyespot/ and GitHub (copy archived at Mizuno, 2024) and Zenodo.

Research article Evolutionary Biology
Mizuno etal. eLife 2024;13:RP96338. DOI: https://doi.org/10.7554/eLife.96338 19 of 23
The following dataset was generated:
Author(s) Year Dataset title Dataset URL Database and Identifier
Mizuno A, Nakagawa
S
2024 A systematic review and
meta- analysis of eyespot
anti- predator mechanisms
https:// doi. org/
10. 5281/ zenodo.
13147019
Zenodo, 10.5281/
zenodo.13147019
References
Abramoff MP, Ram SJ. 2004. Image processing with ImageJ. Biophotonics Int 11:36–42.
Andersson MB. 1994. Sexual Selection. Princeton, NJ: Princeton University Press.
Avery ML, Daneke DE, Decker DG, Lefebvre PW, Matteson RE, Nelms CO. 1988. Flight pen evaluations of
eyespot balloons to protect citrus from bird depredations. Proc Vertebr Pest Conf 13:277–280.
Bateman PW, Fleming PA. 2009. To cut a long tail short: a review of lizard caudal autotomy studies carried
out over the last 20 years. Journal of Zoology 277:1–14. DOI: https://doi.org/10.1111/j.1469-7998.2008.
00484.x
Blest AD. 1957a. The function of eyespot patterns in the lepidoptera. Behaviour 11:209–258. DOI: https://doi.
org/10.1163/156853956X00048
Blest AD. 1957b. The evolution of protective displays in the saturnioidea and sphingidae (lepidoptera).
Behaviour 11:257–309. DOI: https://doi.org/10.1163/156853957X00146
Blut C, Wilbrandt J, Fels D, Girgel EI, Lunau K. 2012. The ‘sparkle’ in fake eyes – the protective effect of mimic
eyespots in lepidoptera. Entomologia Experimentalis et Applicata 143:231–244. DOI: https://doi.org/10.1111/
j.1570-7458.2012.01260.x
Blut C, Lunau K. 2015. Effects of lepidopteran eyespot components on the deterrence of predatory birds.
Behaviour 152:1481–1505. DOI: https://doi.org/10.1093/jisesa/iez123
Brilot BO, Normandale CL, Parkin A, Bateson M. 2009. Can we use starlings’ aversion to eyespots as the basis
for a novel ‘cognitive bias’ task? Applied Animal Behaviour Science 118:182–190. DOI: https://doi.org/10.
1016/j.applanim.2009.02.015
Bura VL, Kawahara AY, Yack JE. 2016. A comparative analysis of sonic defences in bombycoidea caterpillars.
Scientific Reports 6:31469. DOI: https://doi.org/10.1038/srep31469, PMID: 27510510
Carter J, Lyons NJ, Cole HL, Goldsmith AR. 2008. Subtle cues of predation risk: starlings respond to a predator’s
direction of eye- gaze. Proceedings. Biological Sciences 275:1709–1715. DOI: https://doi.org/10.1098/rspb.
2008.0095, PMID: 18445559
Chan IZW, Ngan ZC, Naing L, Lee Y, Gowri V, Monteiro A. 2021. Predation favours Bicyclus anynana butterflies
with fewer forewing eyespots. Proceedings. Biological Sciences 288:20202840. DOI: https://doi.org/10.1098/
rspb.2020.2840, PMID: 34034526
Clucas B, Marzluff JM, Mackovjak D, Palmquist I. 2013. Do American crows pay attention to human gaze and
facial expressions. Ethology : Formerly Zeitschrift Fur Tierpsychologie 119:296–302. DOI: https://doi.org/10.
1111/eth.12064
Crees LD, DeVries P, Penz CM. 2021. Do hind wing eyespots of Caligo butterflies function in both mating
behavior and antipredator defense? (Lepidoptera, Nymphalidae) . Annals of the Entomological Society of
America 114:329–337. DOI: https://doi.org/10.1093/aesa/saaa050
Davidson GL, Clayton NS, Thornton A. 2015. Wild jackdaws, Corvus monedula , recognize individual humans
and may respond to gaze direction with defensive behaviour. Animal Behaviour 108:17–24. DOI: https://doi.
org/10.1016/j.anbehav.2015.07.010
Davies NB, Welbergen JA. 2008. Cuckoo–hawk mimicry? An experimental test. Proc R Soc B 275:1817–1822.
DOI: https://doi.org/10.1098/rspb.2008.0331
De Bona S, Valkonen JK, López- Sepulcre A, Mappes J. 2015. Predator mimicry, not conspicuousness, explains
the efficacy of butterfly eyespots. Proceedings. Biological Sciences 282:20150202. DOI: https://doi.org/10.
1098/rspb.2015.0202, PMID: 25854889
de Framond L, Brumm H, Thompson WI, Drabing SM, Francis CD. 2022. The broken- wing display across birds
and the conditions for its evolution. Proceedings. Biological Sciences 289:20220058. DOI: https://doi.org/10.
1098/rspb.2022.0058, PMID: 35350855
Dell’aglio DD, Stevens M, Jiggins CD. 2016. Avoidance of an aposematically coloured butterfly by wild birds in a
tropical forest. Ecological Entomology 41:627–632. DOI: https://doi.org/10.1111/een.12335, PMID: 27708481
Drinkwater E, Allen WL, Endler JA, Hanlon RT, Holmes G, Homziak NT, Kang C, Leavell BC, Lehtonen J,
Loeffler- Henry K, Ratcliffe JM, Rowe C, Ruxton GD, Sherratt TN, Skelhorn J, Skojec C, Smart HR, White TE,
Yack JE, Young CM, etal. 2022. A synthesis of deimatic behaviour. Biological Reviews of the Cambridge
Philosophical Society 97:2237–2267. DOI: https://doi.org/10.1111/brv.12891, PMID: 36336882
Endler JA. 1992. Signals, signal conditions, and the direction of evolution. The American Naturalist 139:S125–
S153. DOI: https://doi.org/10.1086/285308
Endler JA. 1993. Some general comments on the evolution and design of animal communication systems.
Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences 340:215–225. DOI:
https://doi.org/10.1098/rstb.1993.0060, PMID: 8101656

Research article Evolutionary Biology
Mizuno etal. eLife 2024;13:RP96338. DOI: https://doi.org/10.7554/eLife.96338 20 of 23
Endler JA, Westcott DA, Madden JR, Robson T. 2005. Animal visual systems and the evolution of color patterns:
sensory processing illuminates signal evolution. Evolution; International Journal of Organic Evolution 59:1795–
1818. DOI: https://doi.org/10.1554/04-669.1, PMID: 16329248
Finkbeiner SD, Briscoe AD, Reed RD. 2014. Warning signals are seductive: relative contributions of color and
pattern to predator avoidance and mate attraction in Heliconius butterflies. Evolution; International Journal of
Organic Evolution 68:3410–3420. DOI: https://doi.org/10.1111/evo.12524, PMID: 25200939
Foo YZ, O’Dea RE, Koricheva J, Nakagawa S, Lagisz M. 2021. A practical guide to question formation,
systematic searching and study screening for literature reviews in ecology and evolution. Methods in Ecology
and Evolution 12:1705–1720. DOI: https://doi.org/10.1111/2041-210X.13654
Forsman A, Herrström J. 2004. Asymmetry in size, shape, and color impairs the protective value of conspicuous
color patterns. Behavioral Ecology 15:141–147. DOI: https://doi.org/10.1093/beheco/arg092
Halali D, Krishna A, Kodandaramaiah U, Molleman F. 2019. Lizards as predators of butterflies: shape of wing
damage and effects of eyespots. The Journal of the Lepidopterists’ Society 73:78. DOI: https://doi.org/10.
18473/lepi.73i2.a2
Hedges LV, Gurevitch J, Curtis PS. 1999. The meta- analysis of response ratios in experimental ecology. Ecology
80:1150–1156. DOI: https://doi.org/10.1890/0012-9658(1999)080[1150:TMAORR]2.0.CO;2
Higgins JPT, Thompson SG, Deeks JJ, Altman DG. 2003. Measuring inconsistency in meta- analyses. BMJ
327:557–560. DOI: https://doi.org/10.1136/bmj.327.7414.557, PMID: 12958120
Hill RI, Vaca JF. 2004. Differential wing strength in pierella butterflies (nymphalidae, satyrinae) supports the
deflection hypothesis. Biotropica 36:362–370. DOI: https://doi.org/10.1111/j.1744-7429.2004.tb00328.x
Hill PSM. 2009. How do animals use substrate- borne vibrations as an information source? Die
Naturwissenschaften 96:1355–1371. DOI: https://doi.org/10.1007/s00114-009-0588-8, PMID: 19593539
Ho S, Schachat SR, Piel WH, Monteiro A. 2016. Attack risk for butterflies changes with eyespot number and size.
Royal Society Open Science 3:150614. DOI: https://doi.org/10.1098/rsos.150614, PMID: 26909190
Hossie TJ, Sherratt TN. 2012. Eyespots interact with body colour to protect caterpillar- like prey from avian
predators. Animal Behaviour 84:167–173. DOI: https://doi.org/10.1016/j.anbehav.2012.04.027
Hossie TJ, Sherratt TN. 2013. Defensive posture and eyespots deter avian predators from attacking caterpillar
models. Animal Behaviour 86:383–389. DOI: https://doi.org/10.1016/j.anbehav.2013.05.029
Hossie TJ, Sherratt TN. 2014. Does defensive posture increase mimetic fidelity of caterpillars with eyespots to
their putative snake models? Current Zoology 60:76–89. DOI: https://doi.org/10.1093/czoolo/60.1.76
Hossie TJ, Skelhorn J, Breinholt JW, Kawahara AY, Sherratt TN. 2015. Body size affects the evolution of
eyespots in caterpillars. PNAS 112:6664–6669. DOI: https://doi.org/10.1073/pnas.1415121112, PMID:
25964333
Humphreys RK, Ruxton GD. 2018. What is known and what is not yet known about deflection of the point of a
predator’s attack. Biological Journal of the Linnean Society 123:483–495. DOI: https://doi.org/10.1093/
biolinnean/blx164
Huq M, Bhardwaj S, Monteiro A. 2019. Male bicyclus anynana butterflies choose females on the basis of their
ventral UV- reflective eyespot centers. Journal of Insect Science 19:25. DOI: https://doi.org/10.1093/jisesa/
iez014, PMID: 30794728
Inglis IR, Huson LW, Marshall MB, Neville PA. 1983. The feeding behaviour of starlings (sturnus vulgaris) in the
presence of ‘Eyes’. Zeitschrift Für Tierpsychologie 62:181–208. DOI: https://doi.org/10.1111/j.1439-0310.1983.
tb02151.x
Janzen DH, Hallwachs W, Burns JM. 2010. A tropical horde of counterfeit predator eyes. PNAS 107:11659–
11665. DOI: https://doi.org/10.1073/pnas.0912122107, PMID: 20547863
Johansson BG, Jones TM. 2007. The role of chemical communication in mate choice. Biological Reviews of the
Cambridge Philosophical Society 82:265–289. DOI: https://doi.org/10.1111/j.1469-185X.2007.00009.x, PMID:
17437561
Johnstone RA. 1996. Multiple displays in animal communication: ‘backup signals’ and ‘multiple messages’.
Philosophical Transactions of the Royal Society of London. Series B 351:329–338. DOI: https://doi.org/10.1098/
rstb.1996.0026
Jones RB. 1980. Reactions of male domestic chicks to two- dimensional eye- like shapes. Animal Behaviour
28:212–218. DOI: https://doi.org/10.1016/S0003-3472(80)80025-X
Jones AG, Ratterman NL. 2009. Mate choice and sexual selection: what have we learned since Darwin? PNAS
106 Suppl 1:10001–10008. DOI: https://doi.org/10.1073/pnas.0901129106, PMID: 19528643
Kelber A, Vorobyev M, Osorio D. 2003. Animal colour vision--behavioural tests and physiological concepts.
Biological Reviews of the Cambridge Philosophical Society 78:81–118. DOI: https://doi.org/10.1017/
s1464793102005985, PMID: 12620062
Kjernsmo K, Merilaita S. 2017. Resemblance to the enemy’s eyes underlies the intimidating effect of eyespots.
The American Naturalist 190:594–600. DOI: https://doi.org/10.1086/693473, PMID: 28937816
Kodandaramaiah U, Vallin A, Wiklund C. 2009. Fixed eyespot display in a butterfly thwarts attacking birds.
Animal Behaviour 77:1415–1419. DOI: https://doi.org/10.1016/j.anbehav.2009.02.018
Kodandaramaiah U. 2011. The evolutionary significance of butterfly eyespots. Behavioral Ecology 22:1264–
1271. DOI: https://doi.org/10.1093/beheco/arr123
Kodandaramaiah U, Lindenfors P, Tullberg BS. 2013. Deflective and intimidating eyespots: a comparative study
of eyespot size and position in Junonia butterflies. Ecology and Evolution 3:4518–4524. DOI: https://doi.org/
10.1002/ece3.831, PMID: 24340191

Research article Evolutionary Biology
Mizuno etal. eLife 2024;13:RP96338. DOI: https://doi.org/10.7554/eLife.96338 21 of 23
Kronforst MR, Young LG, Kapan DD, McNeely C, O’Neill RJ, Gilbert LE. 2006. Linkage of butterfly mate
preference and wing color preference cue at the genomic location of wingless. PNAS 103:6575–6580. DOI:
https://doi.org/10.1073/pnas.0509685103, PMID: 16611733
Lajeunessei MJ. 2011. On the meta- analysis of response ratios for studies with correlated and multi- group
designs. Ecology 92:2049–2055. DOI: https://doi.org/10.1890/11-0423.1, PMID: 22164829
Lindström L, Alatalo RV, Lyytinen A, Mappes J. 2001. Predator experience on cryptic prey affects the survival of
conspicuous aposematic prey. Proceedings. Biological Sciences 268:357–361. DOI: https://doi.org/10.1098/
rspb.2000.1377, PMID: 11270431
Linz GM, Bucher EH, Canavelli SB, Rodriguez E, Avery ML. 2015. Limitations of population suppression for
protecting crops from bird depredation: a review. Crop Protection 76:46–52. DOI: https://doi.org/10.1016/j.
cropro.2015.06.005
Lyytinen A, Brakefield PM, Mappes J. 2003. Significance of butterfly eyespots as an anti‐predator device in
ground‐based and aerial attacks. Oikos 100:373–379. DOI: https://doi.org/10.1034/j.1600-0706.2003.11935.x
Lyytinen A, Brakefield PM, Lindström L, Mappes J. 2004. Does predation maintain eyespot plasticity in Bicyclus
anynana? Proceedings. Biological Sciences 271:279–283. DOI: https://doi.org/10.1098/rspb.2003.2571, PMID:
15058439
Ma L, Yang C, Liang W. 2018. Hawk mimicry does not reduce attacks of cuckoos by highly aggressive hosts.
Avian Research 9:1–7. DOI: https://doi.org/10.1186/s40657-018-0127-4
Macartney EL, Lagisz M, Nakagawa S. 2022. The relative benefits of environmental enrichment on learning
and memory are greater when stressed: a meta- analysis of interactions in rodents. Neuroscience and
Biobehavioral Reviews 135:104554. DOI: https://doi.org/10.1016/j.neubiorev.2022.104554, PMID:
35149103
María Arenas L, Walter D, Stevens M. 2015. Signal honesty and predation risk among a closely related group of
aposematic species. Scientific Reports 5:11021. DOI: https://doi.org/10.1038/srep11021, PMID: 26046332
Marples NM, Roper TJ, Harper DGC. 1998. Responses of wild birds to novel prey: evidence of dietary
conservatism. Oikos 83:161. DOI: https://doi.org/10.2307/3546557
Marples NM, Kelly DJ. 1999. Neophobia and dietary conservatism:two distinct processes? Evolutionary Ecology
13:641–653. DOI: https://doi.org/10.1023/A:1011077731153
Martin GR. 2017. The Sensory Ecology of Birds. Clarendon, Oxford University press.
Martin Schaefer H, Schaefer V, Levey DJ. 2004. How plant–animal interactions signal new insights in
communication. Trends Ecol Evol 19:577–584. DOI: https://doi.org/10.1016/j.tree.2004.08.003, PMID:
16701323
Mason NA, Bowie RCK. 2020. Plumage patterns: Ecological functions, evolutionary origins, and advances in
quantification. The Auk 137:ukaa060. DOI: https://doi.org/10.1093/auk/ukaa060
McLennan JA, Langham NPE, Porter RER. 1995. Deterrent effect of eye‐spot balls on birds. New Zealand
Journal of Crop and Horticultural Science 23:139–144. DOI: https://doi.org/10.1080/01140671.1995.9513880
Merilaita S, Vallin A, Kodandaramaiah U, Dimitrova M, Ruuskanen S, Laaksonen T. 2011. Number of eyespots
and their intimidating effect on naïve predators in the peacock butterfly. Behavioral Ecology 22:1326–1331.
DOI: https://doi.org/10.1093/beheco/arr135
Merilaita S, Scott- Samuel NE, Cuthill IC. 2017. How camouflage works. Philosophical Transactions of the Royal
Society of London. Series B, Biological Sciences 372:2016034. DOI: https://doi.org/10.1098/rstb.2016.0341,
PMID: 28533458
Mizuno A, Lagisz M, Pollo P, Yang Y, Soma M, Nakagawa S. 2023. Meta- analysis of fear responses to eyespots on
butterfly wings in birds: a protocol for a systematic review with meta- analysis. Meta- analysis of fear responses
to eyespots on butterfly wings in birds: a protocol for a systematic review with meta- analysis. Open Science
Framework. DOI: https://doi.org/10.17605/OSF.IO/YMWVB
Mizuno A. 2024. Eyespot. swh:1:rev:eaa380be4df5ad8d3991e7373fa2d55174394933. Software Heritage.
https://archive.softwareheritage.org/swh:1:dir:67935a0da33c157ae7dc5cc6b4f8495cb007c4b2;origin=https://
github.com/Ayumi-495/eyespot;visit=swh:1:snp:ba79e9024e4af226d96e5486c31e8b21ee871cef;anchor=swh:
1:rev:eaa380be4df5ad8d3991e7373fa2d55174394933
Moher D, Liberati A, Tetzlaff J, Altman DG, PRISMA Group. 2009. Preferred reporting items for systematic
reviews and meta- analyses: the PRISMA statement. PLOS Medicine 6:e1000097. DOI: https://doi.org/10.1371/
journal.pmed.1000097, PMID: 19621072
Mukherjee R, Kodandaramaiah U. 2015. What makes eyespots intimidating- the importance of pairedness. BMC
Evolutionary Biology 15:34. DOI: https://doi.org/10.1186/s12862-015-0307-3, PMID: 25880640
Nakagawa S, Santos ESA. 2012. Methodological issues and advances in biological meta- analysis. Evolutionary
Ecology 26:1253–1274. DOI: https://doi.org/10.1007/s10682-012-9555-5
Nakagawa S, Lagisz M, Jennions MD, Koricheva J, Noble DWA, Parker TH, Sánchez‐Tójar A, Yang Y, O’Dea RE.
2022. Methods for testing publication bias in ecological and evolutionary meta‐analyses. Methods in Ecology
and Evolution 13:4–21. DOI: https://doi.org/10.1111/2041-210X.13724
Nakagawa S, Lagisz M, O’Dea RE, Pottier P, Rutkowska J, Senior AM, Yang Y, Noble DWA. 2023a. orchaRd 2.0:
An R package for visualising meta‐analyses with orchard plots. Methods in Ecology and Evolution 14:2003–
2010. DOI: https://doi.org/10.1111/2041-210X.14152
Nakagawa S, Yang Y, Macartney EL, Spake R, Lagisz M. 2023b. Quantitative evidence synthesis: a practical guide
on meta- analysis, meta- regression, and publication bias tests for environmental sciences. Environmental
Evidence 12:8. DOI: https://doi.org/10.1186/s13750-023-00301-6, PMID: 39294795

Research article Evolutionary Biology
Mizuno etal. eLife 2024;13:RP96338. DOI: https://doi.org/10.7554/eLife.96338 22 of 23
Nakamura K, Shirota Y, Kaneko T, Matsuoka S. 1995. Scaring effectiveness of eyespot balloons on the rufous
turtle dove, streptopelia orientalis (LATHAM), in a flight cage. Applied Entomology and Zoology 30:383–392.
DOI: https://doi.org/10.1303/aez.30.383
Ng SY, Bhardwaj S, Monteiro A. 2017. Males become choosier in response to manipulations of female wing
ornaments in dry season bicyclus anynana butterflies. Journal of Insect Science 17:81. DOI: https://doi.org/10.
1093/jisesa/iex053, PMID: 28973485
Noble DWA, Lagisz M, O’dea RE, Nakagawa S. 2017. Nonindependence and sensitivity analyses in ecological
and evolutionary meta- analyses. Molecular Ecology 26:2410–2425. DOI: https://doi.org/10.1111/mec.14031,
PMID: 28133832
O’Dea RE, Lagisz M, Jennions MD, Koricheva J, Noble DWA, Parker TH, Gurevitch J, Page MJ, Stewart G,
Moher D, Nakagawa S. 2021. Preferred reporting items for systematic reviews and meta- analyses in ecology
and evolutionary biology: a PRISMA extension. Biological Reviews of the Cambridge Philosophical Society
96:1695–1722. DOI: https://doi.org/10.1111/brv.12721, PMID: 33960637
Olofsson M, Vallin A, Jakobsson S, Wiklund C. 2010. Marginal eyespots on butterfly wings deflect bird attacks
under low light intensities with UV wavelengths. PLOS ONE 5:e10798. DOI: https://doi.org/10.1371/journal.
pone.0010798, PMID: 20520736
Olofsson M, Vallin A, Jakobsson S, Wiklund C. 2011. Winter predation on two species of hibernating butterflies:
monitoring rodent attacks with infrared cameras. Animal Behaviour 81:529–534. DOI: https://doi.org/10.1016/
j.anbehav.2010.12.012
Olofsson M, Jakobsson S, Wiklund C. 2012. Auditory defence in the peacock butterfly (Inachis io) against mice
(Apodemus flavicollis and A. sylvaticus). Behavioral Ecology and Sociobiology 66:209–215. DOI: https://doi.
org/10.1007/s00265-011-1268-1
Olofsson M, Jakobsson S, Wiklund C. 2013a. Bird attacks on a butterfly with marginal eyespots and the role of
prey concealment against the background. Biological Journal of the Linnean Society 109:290–297. DOI:
https://doi.org/10.1111/bij.12063
Olofsson M, Løvlie H, Tibblin J, Jakobsson S, Wiklund C. 2013b. Eyespot display in the peacock butterfly
triggers antipredator behaviors in naïve adult fowl. Behavioral Ecology 24:305–310. DOI: https://doi.org/10.
1093/beheco/ars167, PMID: 23243378
Olofsson M, Wiklund C, Favati A. 2015. On the deterring effect of a butterfly’s eyespot in juvenile and sub- adult
chickens. Current Zoology 61:749–757. DOI: https://doi.org/10.1093/czoolo/61.4.749
Ord TJ, Blazek K, White TE, Das I. 2021. Conspicuous animal signals avoid the cost of predation by being
intermittent or novel: confirmation in the wild using hundreds of robotic prey. Proceedings. Biological Sciences
288:20210706. DOI: https://doi.org/10.1098/rspb.2021.0706, PMID: 34102889
Ouzzani M, Hammady H, Fedorowicz Z, Elmagarmid A. 2016. Rayyan- a web and mobile app for systematic
reviews. Systematic Reviews 5:210. DOI: https://doi.org/10.1186/s13643-016-0384-4, PMID: 27919275
Pinheiro CEG, Antezana MA, Machado LP. 2014. Evidence for the deflective function of eyespots in wild junonia
evarete cramer (Lepidoptera, Nymphalidae). Neotropical Entomology 43:39–47. DOI: https://doi.org/10.1007/
s13744-013-0176-7, PMID: 27193402
Postema EG. 2022. The effectiveness of eyespots and masquerade in protecting artificial prey across
ontogenetic and seasonal shifts. Current Zoology 68:451–458. DOI: https://doi.org/10.1093/cz/zoab082,
PMID: 36090146
Prudic KL, Stoehr AM, Wasik BR, Monteiro A. 2015. Eyespots deflect predator attack increasing fitness and
promoting the evolution of phenotypic plasticity. Proceedings. Biological Sciences 282:20141531. DOI: https://
doi.org/10.1098/rspb.2014.1531, PMID: 25392465
R Development Core Team. 2023. R: A language and environment for statistical computing. R Foundation for
Statistical Computing. Vienna, Austria. https://www.r-project.org
Robertson KA, Monteiro A. 2005. Female bicyclus anynana butterflies choose males on the basis of their dorsal
UV- reflective eyespot pupils. Proceedings. Biological Sciences 272:1541–1546. DOI: https://doi.org/10.1098/
rspb.2005.3142, PMID: 16048768
Rose EM, Prior NH, Ball GF. 2022. The singing question: re- conceptualizing birdsong. Biological Reviews of the
Cambridge Philosophical Society 97:326–342. DOI: https://doi.org/10.1111/brv.12800, PMID: 34609054
Rota J, Wagner DL. 2006. Predator mimicry: metalmark moths mimic their jumping spider predators. PLOS ONE
1:e45. DOI: https://doi.org/10.1371/journal.pone.0000045, PMID: 17183674
Sang A, Teder T. 2011. Dragonflies cause spatial and temporal heterogeneity in habitat quality for butterflies.
Insect Conservation and Diversity 4:257–264. DOI: https://doi.org/10.1111/j.1752-4598.2011.00134.x
Saporito RA, Zuercher R, Roberts M, Gerow KG, Donnelly MA. 2007. Experimental evidence for aposematism in
the dendrobatid poison frog oophaga pumilio. Copeia 2007:1006–1011. DOI: https://doi.org/10.1643/0045-
8511(2007)7[1006:EEFAIT]2.0.CO;2
Senior AM, Viechtbauer W, Nakagawa S. 2020. Revisiting and expanding the meta- analysis of variation: the log
coefficient of variation ratio. Research Synthesis Methods 11:553–567. DOI: https://doi.org/10.1002/jrsm.1423,
PMID: 32431099
Skelhorn J, Dorrington G, Hossie TJ, Sherratt TN. 2014. The position of eyespots and thickened segments
influence their protective value to caterpillars. Behavioral Ecology 25:1417–1422. DOI: https://doi.org/10.1093/
beheco/aru154
Skelhorn J, Holmes GG, Hossie TJ, Sherratt TN. 2016. Multicomponent deceptive signals reduce the speed at
which predators learn that prey are profitable. Behavioral Ecology 27:141–147. DOI: https://doi.org/10.1093/
beheco/arv135

Research article Evolutionary Biology
Mizuno etal. eLife 2024;13:RP96338. DOI: https://doi.org/10.7554/eLife.96338 23 of 23
Skelhorn J, Rowland HM. 2022. Eyespot configuration and predator approach direction affect the antipredator
efficacy of eyespots. Frontiers in Ecology and Evolution 10:951967. DOI: https://doi.org/10.3389/fevo.2022.
951967
Stevens M. 2005. The role of eyespots as anti‐predator mechanisms, principally demonstrated in the
Lepidoptera. Biological Reviews 80:573–588. DOI: https://doi.org/10.1017/S1464793105006810
Stevens M. 2007a. Predator perception and the interrelation between different forms of protective coloration.
Proc R Soc B 274:1457–1464. DOI: https://doi.org/10.1098/rspb.2007.0220
Stevens M, Hopkins E, Hinde W, Adcock A, Connolly Y, Troscianko T, Cuthill IC. 2007b. Field experiments on the
effectiveness of ‘eyespots’ as predator deterrents. Animal Behaviour 74:1215–1227. DOI: https://doi.org/10.
1016/j.anbehav.2007.01.031
Stevens M, Hardman CJ, Stubbins CL. 2008a. Conspicuousness, not eye mimicry, makes “eyespots” effective
antipredator signals. Behavioral Ecology 19:525–531. DOI: https://doi.org/10.1093/beheco/arm162
Stevens M, Stubbins CL, Hardman CJ. 2008b. The anti- predator function of ‘eyespots’ on camouflaged and
conspicuous prey. Behavioral Ecology and Sociobiology 62:1787–1793. DOI: https://doi.org/10.1007/s00265-
008-0607-3
Stevens M, Cantor A, Graham J, Winney IS. 2009a. The function of animal ‘eyespots’: Conspicuousness but not
eye mimicry is key. Current Zoology 55:319–326. DOI: https://doi.org/10.1093/czoolo/55.5.319
Stevens M, Castor- Perry SA, Price JRF. 2009b. The protective value of conspicuous signals is not impaired by
shape, size, or position asymmetry. Behavioral Ecology 20:96–102. DOI: https://doi.org/10.1093/beheco/
arn119
Stevens M, Marshall KLA, Troscianko J, Finlay S, Burnand D, Chadwick SL. 2013. Revealed by conspicuousness:
distractive markings reduce camouflage. Behavioral Ecology 24:213–222. DOI: https://doi.org/10.1093/
beheco/ars156
Stevens M, Ruxton GD. 2014. Do animal eyespots really mimic eyes? Current Zoology 60:26–36. DOI: https://
doi.org/10.1093/czoolo/60.1.26
Tobias JA, Sheard C, Pigot AL, Devenish AJM, Yang J, Sayol F, Neate- Clegg MHC, Alioravainen N, Weeks TL,
Barber RA, Walkden PA, MacGregor HEA, Jones SEI, Vincent C, Phillips AG, Marples NM,
Montaño- Centellas FA, Leandro- Silva V, Claramunt S, Darski B, etal. 2022. AVONET: morphological, ecological
and geographical data for all birds. Ecology Letters 25:581–597. DOI: https://doi.org/10.1111/ele.13898,
PMID: 35199922
Vallin A, Jakobsson S, Lind J, Wiklund C. 2005. Prey survival by predator intimidation: an experimental study of
peacock butterfly defence against blue tits. Proceedings. Biological Sciences 272:1203–1207. DOI: https://doi.
org/10.1098/rspb.2004.3034, PMID: 16024383
Vallin A, Jakobsson S, Wiklund CG. 2010. Constant eyespot display as a primary defense - survival of male and
female emperor moths when attacked by blue tits. The Journal of Research on the Lepidoptera 43:9–17. DOI:
https://doi.org/10.5962/p.266504
Vallin A, Dimitrova M, Kodandaramaiah U, Merilaita S. 2011. Deflective effect and the effect of prey detectability
on anti- predator function of eyespots. Behavioral Ecology and Sociobiology 65:1629–1636. DOI: https://doi.
org/10.1007/s00265-011-1173-7
Viechtbauer W. 2010. Conducting meta- analyses in r with the metafor package. Journal of Statistical Software
36:1–48. DOI: https://doi.org/10.18637/jss.v036.i03
Vlieger L, Brakefield PM. 2007. The deflection hypothesis: eyespots on the margins of butterfly wings do not
influence predation by lizards. Biological Journal of the Linnean Society 92:661–667. DOI: https://doi.org/10.
1111/j.1095-8312.2007.00863.x
Wert L. 2012. Anti- predator adaptations and strategies in the Lepidoptera (Doctoral dissertation, University of
Glasgow). https://theses.gla.ac.uk/3541/ [Accessed August 15, 2012].
Wiklund C, Vallin A, Friberg M, Jakobsson S. 2008. Rodent predation on hibernating peacock and small
tortoiseshell butterflies. Behavioral Ecology and Sociobiology 62:379–389. DOI: https://doi.org/10.1007/
s00265-007-0465-4
Wourms MK, Wasserman FE. 1985. Butterfly wing markings are more advantageous during handling than during
the initial strike of an avian predator. Evolution; International Journal of Organic Evolution 39:845–851. DOI:
https://doi.org/10.1111/j.1558-5646.1985.tb00426.x, PMID: 28561349
Ximenes NG, Gawryszewski FM. 2020. Conspicuous colours in a polymorphic orb- web spider: evidence of
predator avoidance but not prey attraction. Animal Behaviour 169:35–43. DOI: https://doi.org/10.1016/j.
anbehav.2020.08.022