Received: 28 March 2018 | Revised: 10 May 2018 | Accepted: 15 May 2018

DOI: 10.1002/ece3.4272

ORIGINAL RESEARCH

Diversity of warning signal and social interaction influences the

evolution of imperfect mimicry

Renan Janke Bosque1 | J. P. Lawrence1 | Richard Buchholz1 |

Guarino R. Colli2 | Jessica Heppard1 | Brice Noonan1

1
The University of Mississippi, University,
Mississippi Abstract
2
Universidade de Brasília, Brasília, Brazil Mimicry, the resemblance of one species by another, is a complex phenomenon
where the mimic (Batesian mimicry) or the model and the mimic (Mullerian mimicry)
Correspondence
Renan Janke Bosque, The University of gain an advantage from this phenotypic convergence. Despite the expectation that
Mississippi, University, MS 38677.
mimics should closely resemble their models, many mimetic species appear to be
Email: rjbosque@go.olemiss.edu
poor mimics. This is particularly apparent in some systems in which there are multiple
Funding information
This study was supported by Conselho available models. However, the influence of model pattern diversity on the evolution
Nacional de Desenvolvimento Científico of mimetic systems remains poorly understood. We tested whether the number of
e Tecnológico, Coordenação de
Aperfeiçoamento de Pessoal de Nível model patterns a predator learns to associate with a negative consequence affects
Superior, Fundação de Apoio à Pesquisa their willingness to try imperfect, novel patterns. We exposed week-­old chickens to
do Distrito Federal and the United States
Agency for International Development. coral snake (Micrurus) color patterns representative of three South American areas
that differ in model pattern richness, and then tested their response to the putative
imperfect mimetic pattern of a widespread species of harmless colubrid snake
(Oxyrhopus rhombifer) in different social contexts. Our results indicate that chicks
have a great hesitation to attack when individually exposed to high model pattern
diversity and a greater hesitation to attack when exposed as a group to low model
pattern diversity. Individuals with a fast growth trajectory (measured by morphologi-
cal traits) were also less reluctant to attack. We suggest that the evolution of new
patterns could be favored by social learning in areas of low pattern diversity, while
individual learning can reduce predation pressure on recently evolved mimics in areas
of high model diversity. Our results could aid the development of ecological
­predictions about the evolution of imperfect mimicry and mimicry in general.

KEYWORDS
aposematism, Batesian, generalization, mimicry, Mullerian, social learning, warning signal

1 | I NTRO D U C TI O N the signal of a defended prey species (Ruxton, Sherratt, & Speed,
2004). Color combinations including red, yellow, white, and black
Mimicry is an evolutionary strategy often employed by organisms are broadly used as warning signals in many defended taxa, such as
to escape predation. Mimetic phenotypes can generally be classi- Hymenoptera (Hines & Williams, 2012), Coleoptera (Bocak & Yagi,
fied as either camouflage/masquerade, for example, insects mimick- 2010), Lepidoptera (Jiggins, Mallarino, Willmott, & Bermingham,
ing leaves (Skelhorn & Ruxton, 2010) or warning, that is, co-­opting 2006), Lissamphibia (Kraemer & Adams, 2014; Symula, Schulte, &

This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium,
provided the original work is properly cited.
© 2018 The Authors. Ecology and Evolution published by John Wiley & Sons Ltd.

7490 | www.ecolevol.org
 Ecology and Evolution. 2018;8:7490–7499.
BOSQUE et al. | 7491

Summers, 2001), and Squamata (Campbell & Lamar, 2004). These model species number, mimic species number, pattern and color-
warning colors can elicit aversion in a wide variety of visually ori- ation diversity (Figure 1), and extent of overlap between mimics and
ented predators (Ruxton et al., 2004). The aversion of conspicuous models (Bosque, Noonan, & Colli, 2016; Campbell & Lamar, 2004;
prey can even be socially transmitted (Thorogood, Kokko, & Mappes, Roze, 1996). Species of Micrurus transmit a clear warning signal to
2017), reducing the predation pressure on newly evolved signals. potential predators through varying combinations of contrasting
Aversion can also be affected by individual variation in personality red, black, yellow, and white rings (Brodie, 1993; Brodie & Janzen,
(Exnerová, Svádová, Fučíková, Drent, & Štys, 2010), which can be 1995; Smith, 1976). These same colors are also used by harmless
genetically inherited (Drent, Oers, & Noordwijk, 2003) and be ac- snakes, with varying fidelity in color and pattern to Micrurus models,
companied by differences in morphological and physiological traits making this one of most remarkable examples of mimetic interaction
(Goerlich, Nätt, Elfwing, Macdonald, & Jensen, 2012). Whether this (Savage & Slowinski, 1992).
aversion is innate, self-­
learned, or socially transmitted, warning Regional variation in the warning coloration of mimics could
signals are known to have a strong influence on how a predatory occur simply because different predators may interpret mimic-­
animal will explore and interact with prey (Aronsson & Gamberale-­ model resemblance using different sensory cues or cue components
Stille, 2012; Ham, Ihalainen, Lindstrom, & Mappes, 2006; Lindstrom, (Aubier & Sherratt, 2015; Pekar, Jarab, Fromhage, & Herberstein,
Alatalo, & Mappes, 1999; Rowe & Guilford, 2000). 2011). Further, different populations of a mimetic species may occur
At the community level, Batesian mimicry, where an unde- in areas with different predators, with local color variants emerging
fended mimic benefits from a resemblance to a harmful model, is by predation pressure. Nonetheless, even within a single predator
perhaps the most evolutionarily complex mimicry system (Bates, species, individual experience with model pattern richness (i.e., the
1862; Ruxton et al., 2004). Multiple predator species may co-­occur number of different prey patterns) by direct contact or via social ob-
with both multiple defended and multiple undefended prey species servation may also directly affect the evolution of mimetic lineages.
that employ a variety of warning colors and patterning, and the di- A particularly vexing problem in the macroevolutionary study of
mensionality of these components of the mimicry system can vary mimicry complexes that might benefit from a deeper understanding
geographically. For example, New World coral snakes (Micrurus) and of predator learning is that, despite a presumed selective pressure to
their mimics of the genus Oxyrhopus exhibit many combinations of attain perfect resemblance with their models, imperfect mimics are

High
Medium

High
Medium

Number of
color patterns
0
1
2 Low Low
3
4
5
6
7
8

F I G U R E 1 Map with one-­degree cells showing Micrurus color pattern richness. To the right are patterns used in the exposure phase. In
pink the distribution of Oxyrhopus rhombifer. Map based on data from Bosque et al., 2016
7492 | BOSQUE et al.

not uncommon in nature. The reasons for the maintenance of im- (John, 1994). As in most bird taxa, the left testis is usually larger in
perfect mimicry are still unclear but several authors have suggested mature phasianid birds such as the chicken (Calhim & Montgomerie,
plausible explanations (Kazemi, Gamberale-­Stille, Tullberg, & Leimar, 2015) and thus chicks with greater asymmetry in this direction can
2014; Kikuchi & Pfennig, 2013). One explanation focuses on the se- be assumed to be on a more rapid trajectory toward the adult form.
lective pressures on the mimic when many models exist in the same Directional asymmetry in adult testis size has been associated with
area. When multiple models are present within a mimic’s geographic male sexual ornamentation and mate quality in some birds (Møller,
distribution, mimics may be selected by predators to either resemble 1994). We predicted that chicks who invest more in organ matura-
only one model or, if the models are not sympatric with each other, tion would be more motivated to feed and thus less likely to avoid a
the mimics can adopt an intermediate phenotype (Edmunds, 2000; novel food item, despite having learned previously that similar cues
Sherratt, 2002). If just one model is present, selection is expected were aposematic.
to drive mimics toward signal identity with the defended model
(Ruxton et al., 2004). However, if several sympatric, defended mod-
2 | M E TH O DS
els vary in phenotype, predators in this area may be conservative
in the avoidance of harmless species with similar warning signals,
2.1 | Study subjects and housing
even if mimicry of the defended models is inexact (Edmunds, 2000).
Experimental evidence demonstrates that predators indeed gen- As model predators we used approximately 10-­day-­old, male do-
eralize a bad experience with one prey species to others (Hotová mestic chickens (Gallus gallus domesticus). The capacity of chickens
Svádová, Exnerová, Kopečková, & Štys, 2013). to discriminate between two objects based on their wavelength is
Model diversity may also drive generalization to novel patterns comparable to several bird species (Hart, 2001), which reinforces the
that are not even found in models (Ham et al., 2006; Kikuchi & adequacy of the species selected as model predator. Birds are com-
Pfennig, 2013). Historically, avoidance of novel prey has been at- monly used as model predators in warning coloration experiments
tributed to innate neophobia; the avoidance of a previously unen- because their color vision is well documented, and they are known
countered signal simply because it is new/unusual (Greenberg & to be the main predators of snakes, including coral snakes (Buasso,
Mettke-­Hofmann, 2001). Because neophobia may disappear with Leynaud, & Cruz, 2006; Hinman et al., 1997; Kikuchi & Pfennig,
exposure experience, the generalization and neophobia hypotheses 2010). Commercial chick feed (corn-­meal) was provided ad libitum
for explaining novel mimic-­like patterns make opposite predictions except for the 60 min immediately prior to exposure and testing ses-
about the outcome of predator learning as the number of models sions, so that the chicks were motivated to “attack.” Housing and
increases. More models provide predators more cues from which to testing conditions were approved by the University of Mississippi
generalize, making them cautious about new prey patterns, but also Institutional Animal Care and Use Committee (#15-­0 09). To repli-
increase the familiarity with novelty, thus fostering less neophobia cate the snake patterns found in nature, we painted Wild Harvest™
toward it. tube feeders with brown spray paint to represent brown snakes and
Previous researchers have demonstrated generalization of coral wrapped experimental feeders with colored electrical tape to repre-
snake warning patterns by free-­ranging avian predators. In these sent the coral snake color pattern(s) present in three regions of South
studies, the birds avoided a mimetic morph with a pattern that dif- America (Figure 1 and Supporting Information Figure S1) (Bosque
fered from the local model but with the same colors (Brodie & Janzen, et al., 2016). We filled the aposematic (henceforth, we use apose-
1995; Kikuchi & Pfennig, 2010). To investigate the evolution of more matic and warning signal interchangeably) feeders with chick feed
complex systems with multiple models and imperfect mimics, we that was previously sprayed with 10% chloroquine solution, making
tested whether the number of models that an avian predator experi- the feed distasteful but not harmful (Lindstrom, Alatalo, & Mappes,
ences affects the breadth of its avoidance generalization to a novel 1997; Ruxton et al., 2004); brown feeders had normal chick feed.
pattern. In this study, a “novel pattern” is also an imperfect mimic, These feeders were not meant to be exact replicas of coral snakes,
a pattern not seen previously by the subject, and yet incorporating but simply represent a variety of patterns from which the chicks had
features (colors and shapes) shared with the aposematic models. We to learn. To simulate natural encounters with aposematic prey, we
also exposed chickens to different contexts using social and individ- used two different approaches: group exposure and individual ex-
ual exposure as these may affect learned responses to distasteful posure. Using these two approaches, we could not only identify how
prey (Thorogood et al., 2017). In order to understand how differ- pattern richness affected generalization to a new pattern but also
ences in individual development of chicks could impact their will- the effect of social exposure versus individual exposure.
ingness to sample imperfect mimics, we investigated morphological
traits that may reveal ontogenetic growth trade-­offs between gen-
2.2 | Group exposure
eral investment in somatic growth (mass, tarsus and body condition)
and organ-­specific development associated with immune prepared- Chicks were housed in three groups of 43 in poultry brooder cages
ness (spleen mass) and sexual maturation (directional testis asym- during exposure to aposematic feeders. Each exposure group expe-
metry). The spleen is an important immune organ in birds, the size rienced only one of the pattern richness treatment levels (Figure 1):
of which reflects immune activity and possibly immunocompetence highest color pattern richness—H (8 patterns), intermediate color
BOSQUE et al. | 7493

pattern richness—M (4 patterns), or low color pattern richness—L (1 we removed the brown feeder and presented a random aposematic
pattern). feeder for up to 2 min. If the chick pecked the food, we allowed it
In addition to regular (trough-­style) chick feeders, chicks were to eat for up to a cumulative total of 10 s and then we removed the
exposed to brown feeders for 8 hr per day during the first 4 days. aposematic feeder. We repeated this procedure until all the 16 feed-
On the 5th day, 16 bird feeders (8 brown and 8 aposematic) were ers were presented according to each subject’s treatment group
positioned randomly along the perimeter of each enclosure for a (H: 16 feeders with 8 different aposematic patterns; L: 16 feeders
10 min exposure session. The feed in each feeder was weighed be- with 1 aposematic pattern—Supporting Information Figure S1) and
fore and after each exposure session. This procedure was repeated recorded the hesitation time, that is, time until the first peck. We
an additional five times over 2 days. A final (6th) exposure session did not record the quantity of feed eaten by chicks during individual
before testing lasted 1 hr, to ensure that chicks were completely training.
avoiding the aposematic feeders. Notably, our group exposure train-
ing procedure allows for social learning (Slagsvold & Wiebe, 2011)
2.5 | Individual testing
as the chicks in the same cage may learn from each other’s negative
reaction to the feed in aposematic feeders. The learned aversion After the exposure described above, we presented a feeder with
from conspecifics is still a theme that deserves investigation as con- an imperfect mimic (i.e., Oxyrhopus rhombifer) pattern alongside a
trasting results have been reported (Sherwin, Heyes, & Nicol, 2002; brown feeder in the testing arena. The arrangement (left or right)
Thorogood et al., 2017). of the feeders was randomized to avoid lateralization bias. We re-
corded the hesitation time and first feeder choice. To evaluate
whether morphological characteristics could explain individual vari-
2.3 | Group testing
ation in hesitation time, we took the following postmortem meas-
After the conclusion of group exposure, we individually tested chicks ures of each chick at the end of the experiment: tarsus length, body
for their reaction to a feeder featuring either the imperfect mimetic mass, directional testes length asymmetry, spleen mass and body
pattern of the false coral snake (Oxyrhopus rhombifer) or a brown condition. The entire length of each testis was measured, unless the
feeder. The testing arena consisted of a 60 cm × 60 cm wood box organ was not fully differentiated, in which case only the length of
containing a small wire cage with two chick companions to prevent portion consisting of white (as opposed to purple-­red) tissue was
isolation stress of the test chick. Each chick was tested only once. measured. Directional testis asymmetry was calculated as (left
Despite a broad geographic distribution, overlapping with many length–right length). Body condition was calculated as mass/tarsus
species of Micrurus, Oxyrhopus rhombifer has a tricolor pattern with length (Brown, 1996).
black saddles bordered by white on a red dorsum (Figure 1), a pattern
not found in any Micrurus species. A previous study using plasticine
2.6 | Statistical analysis
replicas has demonstrated that the Oxyrhopus rhombifer phenotype
does provide protection against free-­range predators (Buasso et al., We fitted Cox proportional hazards models to assess the depend-
2006), but the mechanisms of avoidance are still poorly understood. ency of hesitation time on predictor variables, using the survival
We recorded the reaction to feeder exposure as the hesitation package (Therneau, 2015) in R (R Core Team 2017). Survival analysis
time (time until the first peck). Each trial lasted up to five minutes models the time (i.e., survival time) it takes for a given event to occur
or until the first attack (peck). If we did not observe any attack after and the factors that affect it (Moore, 2016). For the group testing, we
five min, we stopped the trial. Before each trial, we offered small modeled hesitation time as a function of pattern richness exposure
pieces of dry mealworm (Tenebrio molitor) to ensure that chicks were (H, M, or L), feeder type (aposematic or brown), and their interaction.
hungry and willing to attack. All trials were recorded using a digital For the individual testing, we modeled hesitation time as a function
camera (videos available upon request). of pattern richness exposure (high or low), feeder type (aposematic
or brown), their interaction, and the postmortem morphological
variables (tarsus length, body mass, testis length asymmetry, spleen
2.4 | Individual exposure
mass and body condition). We used stepwise model selection based
In order to explore the impact of individual exposure to different on the Akaike information criterion (AIC) to assess predictor impor-
model community diversity we deprived 27 chicks of food for one tance. For each model we checked (a) the proportional hazards as-
hour. We then individually exposed 14 chicks to high color pattern sumption by examination of scaled Schoenfeld residuals using the
richness (Figure 1)—H (8 patterns) and 13 chicks to low color pat- cox.zph function of package survival; (b) the nonlinearity assump-
tern richness—L (1 pattern). Eight additional individuals were used as tion using Martingale residuals; and (c) the presence of influential
buddy chicks. The exposure (training) and testing arena consisted of observations using case deletion residuals (dfbetas) (Moore, 2016).
a cardboard box 38 cm × 30 cm with two buddy chicks inside a small In all cases, we found no violation of assumptions or any influential
wire cage. In each treatment, we started by presenting one brown observation. When needed, we performed pairwise comparisons of
feeder for up to 2 min. Starting after the first peck, we allowed them treatments using the log-­rank test as implemented by the function
to eat for a cumulative time of 10 s to prevent satiation. After that, pairwise_survdiff in package survminer (Kassambara, Kosinski, Biecek,
7494 | BOSQUE et al.

& Fabian, 2018), adjusting p-­values with the Benjamini–Hochberg’s between low and high pattern richness, based on pairwise com-
method (Benjamini & Yosef, 1995). parisons (Benjamin–Hochberg adjustment; high–low: p = 0.001;
high-medium: p = 0.081; low-medium: p = 0.293).

3 | R E S U LT S
3.2 | Individual exposure
3.1 | Group exposure
When presented individually, feeder pattern (brown or aposematic
Across the first five exposure sessions, mean consumption of imperfect) was not a part of our final model, showing that chicks
feed from the aposematic feeders was lower (H: 1.40 ± 1.44 g; M: had no preference for feeder type. The final model contained only
1.99 ± 2.68 g; L: 1.85 ± 3.13 g) than from the brown feeders (H: three predictors: pattern richness exposure (high vs. low), spleen
15.27 ± 8.42 g; M: 18.60 ± 8.36 g; L: 14.20 ± 7.43 g). This pattern mass and directional testes asymmetry (r2 = 0.445, Wald test = 13.3,
was found for all three cages in all exposure sessions (Figure 2). The df = 3, p = 0.004). Chicks exposed to low pattern richness were 3.63
last session (#6) demonstrated that the chicks were avoiding the times more likely to peck a feeder, regardless of color/pattern, than
aposematic patterns: brown feeders were nearly empty, whereas those exposed to high pattern richness (log hazard ratio for low
aposematic feeders were largely avoided (average of food left in- pattern richness exposure = 1.291, Z = 2.552, p = 0.011, Figure 4,
side the feeders during the #6 session H: aposematic: 77.4%, brown: Supporting Information Figure S3). Chicks with higher spleen mass
17.10%; M: aposematic: 85.67%, brown: 8.06%; L: aposematic: and higher testes asymmetry also had a much higher probability of
84.11%, brown: 27.22%). pecking a feeder than less developed chicks (log hazard ratio for
During the testing, we recorded a wide range of attack laten- spleen mass = 7.771, Z = 2.304, p = 0.021; log hazard ratio for tes-
cies from 1 s to 228 s. In 16 trials chicks never attacked the feeder, tes asymmetry = 3.916, Z = 2.437, p = 0.015, Figure 5, Supporting
and thus their trials were terminated at 5 min, and these data were Information Figure S4). Body condition, body mass and tarsus length
right-­censored in our survival analysis. The final model derived from did not contribute to our final model of factors influencing predation.
analysis of group exposure contained only one predictor: pattern
richness exposure (r 2 = 0.074, Wald test = 8.48, df = 2, p = 0.014).
Chicks exposed to low pattern richness had 0.47 times less risk of 4 | D I S CU S S I O N
pecking the novel aposematic feeder than chicks in the high pat-
tern richness treatment (log hazard ratio for low pattern richness The evolution of novel aposematic patterns in nature is a theme
exposure = −0.755, Z = −2.848, p = 0.004, Figure 3, Supporting of intense debate among evolutionary biologists (Lindstrom, 1999;
Information Figure S2). The birds in the medium richness treatment Mappes & Alatalo, 1997). If a novel aposematic pattern is not pro-
showed only a marginal difference from the high pattern richness tected by previous predator education from similar warning patterns
group in the risk of pecking the feeder (log hazard ratio for medium already extant in the region, the attention drawn to a bold, new pat-
pattern richness exposure = −0.47, Z = −1.898, p = 0.058, Figure 3 tern will subject it to a high degree of predator attack. Consequently,
Supporting Information Figure S2). Hesitation time differed only the intense predation on new patterns can slow or even inhibit their

30
High
Pattern richness
Medium High
25
Low
Mean mass consumed (g)

20 Medium
Brown

15

Low
10

5 –1.5 –1 –0.5 0 0.5

Aposematic
Log hazard ratio
0
1 2 3 4 5 F I G U R E 3 Survival analysis modeling hesitation time for chicks
Trial exposed as a group to different coral snake pattern richness to peck
at feeders painted with nonaposematic (brown) or aposematic-­
F I G U R E 2 Bird food mass eaten by chickens after 10 min imperfect patterns as a function of pattern richness. Graphs depict
(rounds 1–5) of exposure. Top lines show feeders with brown log hazard ratios estimated by a Cox proportional hazards model
coloration. Bottom lines show aposematic feeders (Micrurus having high color pattern richness as reference compared to log
patterns). High: eight aposematic patterns; medium: four hazard ratio of medium and low pattern richness; horizontal bar
aposematic patterns; low: one aposematic pattern represents 95% confidence interval
BOSQUE et al. | 7495

(a) 4.1 | Group exposure

High
Pattern richness Despite the low attack rate (food consumption) on aposematic feed-
ers during the exposure phase, we found no evidence of discrimi-
Low nation between novel aposematic and brown prey during testing;
whether previously exposed to low, medium or high color pattern
training. This outcome suggests that novel imperfect mimics will not
–0.5 0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
benefit from previous predator education on how to discriminate be-
Log hazard ratio
(b) 3 tween edible and aposematic prey. Instead, all prey under low pat-
tern richness benefit because socially trained predators are hesitant
Log hazard ratio

2
when facing any type of prey. In contrast, chicks exposed as a group
to more than one aposematic pattern were less cautious and, thus,
1
all prey patterns would be equally subjected to attack. This latter
0 outcome has several possible causes. Young chickens may not be up
to the cognitive task of integrating the many aposematic pattern fea-
–1 tures found in pattern-­rich environments. Similarly, because chicks
needed to navigate both social interactions and multiple patterns
0.15 0.20 0.25 0.30 0.35 0.40 0.45
during training sessions, they were distracted such that they were
Spleen mass
(c) 2 not conditioned to aposematic cues. Alternatively, chicks may have
indeed learned to avoid specific aposematic phenotypes, but also
1
Log hazard ratio

eventually learned from sampling so many feeders that there was

0 little consequence of testing new prey.
Our results suggest that social predators can encourage the evo-
–1
lution of imperfect mimicry in areas of low model pattern diversity
–2 as imperfect mimics receive a crucial time to escape a predation
attempt. However, once multiple color patterns are established in
–3
a particular area, the information overload received by social pred-
–0.2 –0.1 0.0 0.1 0.2 0.3 0.4 0.5
ators can hinder the evolution of imperfect mimics as predators
Testes asymmetry
promptly attack their prey.
F I G U R E 4 Survival analysis modeling hesitation time for chicks
individually exposed to different coral snake pattern richness
to peck on feeders painted with nonaposematic (brown) and 4.2 | Individual exposure
aposematic-­imperfect patterns as a function of pattern richness,
As with the socially exposed subjects, individually exposed subjects
spleen mass, and testes asymmetry. Graphs depict log hazard ratios
did not discriminate against the novel aposematic feeder. However,
estimated by a Cox proportional hazards model as a function of the
three predictors. (a) Log hazard ratio reference (high color pattern individuals exposed to multiple patterns had a higher hesitation to
richness) compared to log hazard ratio of low pattern richness; feed from either feeder during their test trials. In pattern-­diverse
horizontal bar represents 95% confidence interval. (b) Linear fit areas, the uncertainty about the dangerousness of prey can make
of the log hazard ratio as a function of spleen mass; dashed line solitary predators more reluctant to try new food items presented
represents 95% confidence interval. (c) Linear fit of the log hazard
to them. If so, in areas with many models and different aposematic
ratio as a function of testes asymmetry; dashed line represents 95%
confidence interval patterns imperfect mimics are better protected because nonsocial
predators will not immediately attack their prey, creating opportu-
nity for escape.
evolution (Turner, 1988), leaving scientists puzzled as to the selec- Our individual subjects varied greatly in their latency to attack
tive mechanisms by which new patterns can evolve. Our initial ex- suggesting that motivational factors other than those caused by the
pectation was that greater pattern diversity exposure would lead to treatments were at play. Difference in hunger is the most obvious ex-
greater hesitation time to attack imperfect phenotypes, as birds are planation for this variation, but this seems unlikely given that chicks
expected to transfer knowledge of diverse visual cues to new prey were fed ad libitum in their rearing brooder and each had equivalent
(Svádová et al., 2009). Instead, we found that the effect of multiple opportunities to feed during the exposure events. Importantly, chick
aposematic models is dependent on the opportunity for social learn- body condition did not explain latency to attack. Our results did,
ing. Chicks exposed as a group to several patterns were less cautious however, confirm our suspicion that the nutritional demands of al-
than chickens exposed to one aposematic pattern. In contrast, when ternative individual growth trajectories would contribute to explain-
exposed individually, chickens are more cautious with a novel pattern ing the variation in feeding hesitation by chicks. Although immune
when their previous aversive exposure involved multiple patterns. and reproductive development differs the most between strains of
7496 | BOSQUE et al.

Social Nonsocial
predators predators

H – +

L +
–
F I G U R E 5 Diagram showing the effect of social and nonsocial predators on the evolution of mimicry/color pattern diversity. In areas of
high model color diversity (H), new color patterns can be favored (+) by reduced predation pressure as a result of higher attack hesitation of
nonsocial predators and disfavored (−) by lower attack hesitation of social predators. In areas of low pattern diversity (L), new color patterns
can be favored (+) by reduced predation pressure as a result of higher attack hesitation of social predators and disfavored (−) by lower attack
hesitation of nonsocial predators

chickens, intrastrain differences among individuals in organ size or (English, Fawcett, Higginson, Trimmer, & Uller, 2016). Individuals
activity occur and can be found as early as day one (Apanius, 1998; with bold personalities often have a higher food intake rate (Biro
de Reviers & Williams, 1984). Rapid growth of the spleen and devel- & Stamps, 2008; Kurvers et al., 2010). Thus early differences in
opment of adult-­like asymmetry in the testes were associated with individual personality traits, such as boldness and the propensity
greater urgency to begin feeding in our study, independent of body to quickly explore space, may allow some chicks to begin feeding
condition. This result suggests that individual organ growth trajec- sooner and develop faster relative to individuals that are shy and
tories may create feeding motivations that are not reflected by ex- slow to explore. Consequently the weaker aversion to the novel im-
ternal morphological measurements, but affect the opportunity for perfect mimic by our more developed subjects may be the direct
the evolution of novel aposematic prey types. Individual variation on and independent result of the bold personality itself, rather than
the willingness to attack, also documented in other species like the simply a product of the growth trajectory initiated by their precocity
quail Coturnix japonica (Marples & Brakefield, 1995), can affect the at feeding. We did not measure personality traits in our subjects,
evolution of new aposematic prey (Speed, 2000). When individuals but in another bird, the great tit (Parus major), fast explorers showed
with rapid development are more prone to attack aposematic prey, shorter attack latency for an aposematic insect than slow individuals
this can enhance the risk of extinction of new conspicuous prey. On (Exnerová et al., 2010), a result similar to our chicks with advanced
the other hand, slow-­growing individuals could initially ease the se- organ development. Nevertheless, the physiological demands of a
lection on new aposematic prey. bold personality may still be the driving force for the eagerness of
Although we conclude that the individual variation in attack la- such chicks to peck at aposematic prey. Bold individuals often have
tency results from the motivation to feed imposed by the energetic a higher metabolic rate than shy ones (Biro & Stamps, 2008), are at
demands of different growth trajectories, growth and learning are greater risk of starvation (Lichtenstein et al., 2017), and thus may
not independent; feeding successfully results both in an increase need to be less catholic in their feeding, showing greater resistance
in body size and reinforces learning about how to feed effectively to learning to avoid noxious prey (Exnerová et al., 2010). Clearly, the
BOSQUE et al. | 7497

experimental disentanglement of predator personality, early devel- trajectory, and the interrelationship between social and nonsocial
opment and motivation to feed discriminately is both relevant to our predators on the evolution of imperfect mimicry will surely benefit
understanding of the evolution of mimicry and a complex challenge from further consideration.
worthy of further research effort.
We demonstrated that color pattern diversity and social trans-
AC K N OW L E D G M E N T S
mission of information might have an influence on the evolution
of imperfect mimicry and mimicry in general, which corroborates RJB thanks the Ciência sem Fronteiras program of Conselho
mathematical models (Thorogood et al., 2017). However, we are Nacional de Desenvolvimento Científico e Tecnológico for a doc-
aware that the evolution of imperfect mimicry may be facilitated by torate fellowship and the Biology department, of University of
other extrinsic factors like niche preferences, predators with differ- Mississippi. GRC thanks Coordenação de Aperfeiçoamento de
ent visual systems (i.e., mammals vs. birds), and biogeographic his- Pessoal de Nível Superior, Conselho Nacional de Desenvolvimento
tory in areas with elevated model color diversity, as is the case for Científico e Tecnológico, Fundação de Apoio à Pesquisa do
Micrurus in western Amazonia (Bosque et al., 2016). There are few Distrito Federal and the United States Agency for International
cases where predation of coral snakes has been observed in nature Development PEER program under cooperative agreement AID-
(DuVal, Greene, & Manno, 2006) but it has been reported that in OAA-A-11-00012 for financial support. The authors offer spe-
one specific site at least 90 species are potential predators of coral cial thanks to Marcella Gonçalves Santos for helping to build the
snakes (França, 2008). Predators of coral snakes have sufficient op- chicken’s feeders.
portunity for social learning, given the number of species in a par-
ticular area (interspecific leaning) and the various degree of sociality
C O N FL I C T O F I N T E R E S T S
of each species, ranging from less social species (red-­legged seriema
Cariama cristata), to highly social species (greater ani, Crotophaga We have no competing interests.
major).
Interestingly, this empirical demonstration of the effects of
AU T H O R S ’ C O N T R I B U T I O N S
model diversity and social interaction lends some insight into how
mimicry systems arise at all. In low model diversity systems, social Renan Janke Bosque was responsible for the conception, design,
predators facilitate the initial evolution of mimics while nonsocial acquisition, interpretation, and analysis and draft of the manu-
predators are an opposing force. After a single color pattern model script. J. P. Lawrence contributed to the acquisition of data, design,
is established in a particular area, mediated by selection of social interpretation, and revision of the manuscript. Richard Buchholz
predators, the number of models/color patterns can further increase contributed to the conception, design, interpretation, revision of
by selection of nonsocial predators (Figure 5). In this sense, in areas the manuscript, and dissection of the chickens. Guarino Rinaldi
with high model color diversity, nonsocial predators will favor re- Colli contributed to the conception, design, interpretation, anal-
cently evolved mimics. Personal experience is probably more com- ysis, and revision of the manuscript. Jessica Heppard aided with
mon than eavesdropped information, which might be another factor experimental design, data collection, and revision of the manu-
to explain why we find more mimics of coral snakes in areas of high script. Brice Noonan contributed to the conception, design, inter-
color diversity of models (Davis Rabosky et al., 2016). pretation, and revision of the manuscript. All authors gave final
approval for publication. All authors agreed to be accountable for
all aspects of the work in ensuring that questions related to the
5 | CO N C LU S I O N accuracy or integrity of any part of the work are appropriately in-
vestigated and resolved.
Newly evolved patterns can be favored by social learning in areas
of low pattern diversity and disfavored by individual learning.
E T H I C S S TAT E M E N T
These findings can shed light on the evolution of imperfect mimicry
(Kikuchi & Pfennig, 2013), which were not previously explored. Our Approval granted to carry out the experiment IACUC 15-­0 09.
findings indicate that this phenomenon can be favored in areas of
low and high model diversity by two distinct mechanisms. We sug-
DATA AC C E S S I B I L I T Y
gest that imperfect mimicry can be favored in areas of high model
diversity by reduced predation pressure as a result of attack hesita- No data deposition is applicable.
tion by nonsocial predators. In areas of low pattern diversity, imper-
fect mimics can be better protected because social predators are
not so cognitively overloaded that they become less prone to attack ORCID
prey. Individual growth trajectory determines how predators will in-
Renan Janke Bosque http://orcid.org/0000-0003-3729-9301
teract with their prey, making fast-­growing individuals less hesitant
to attack. Our understanding of how information overload, growth JP Lawrence http://orcid.org/0000-0001-8577-4870
7498 | BOSQUE et al.

Drent, P. J., Oers, K. V., & Noordwijk, A. J. V. (2003). Realized heritability

Richard Buchholz http://orcid.org/0000-0003-3336-2676
of personalities in the great tit (Parus major). Proceedings of the Royal
Guarino R. Colli http://orcid.org/0000-0002-2628-5652 Society B: Biological Sciences, 270, 45–51. https://doi.org/10.1098/
rspb.2002.2168
Jessica Heppard http://orcid.org/0000-0003-4022-9371
DuVal, E. H., Greene, H. W., & Manno, K. L. (2006). Laughing fal-
con (Herpetotheres cachinnans) predation on coral snakes
(Micrurus nigrocinctus). Biotropica, 38, 566–568. https://doi.
REFERENCES org/10.1111/j.1744-7429.2006.00162.x
Apanius, V. (1998). Ontogeny of immune function. In J. M. Starck, & R. Edmunds, M. (2000). Why are there good and poor mimics?
E. Ricklefs (Eds.), Avian growth and development (pp. 203–222). New Biological Journal of the Linnean Society, 70, 459–466. https://doi.
York, NY: Oxford University Press. org/10.1111/j.1095-8312.2000.tb01234.x
Aronsson, M., & Gamberale-Stille, G. (2012). Colour and pattern simi- English, S., Fawcett, T. W., Higginson, A. D., Trimmer, P. C., & Uller, T.
larity in mimicry: Evidence for a hierarchical discriminative learning (2016). Adaptive use of information during growth can explain long-­
of different components. Animal Behaviour, 84, 881–887. https://doi. term effects of early life experiences. The American Naturalist, 187,
org/10.1016/j.anbehav.2012.07.011 620–632. https://doi.org/10.1086/685644
Aubier, T. G., & Sherratt, T. N. (2015). Diversity in Müllerian mimicry: Exnerová, A., Svádová, K. H., Fučíková, E., Drent, P., & Štys, P. (2010).
The optimal predator sampling strategy explains both local and re- Personality matters: Individual variation in reactions of naive bird
gional polymorphism in prey. Evolution, 69, 2831–2845. https://doi. predators to aposematic prey. Proceedings of the Royal Society
org/10.1111/evo.12790 B: Biological Sciences, 277, 723–728. https://doi.org/10.1098/
Bates, H. W. (1862). Contributions to an insect fauna of the Amazon rspb.2009.1673
Valley (Lepidoptera: Heliconidae). Transactions of the Linnean Society França, F. G. R. (2008). O mimetismo das serpentes corais em ambien-
of London, 23, 495–566. https://doi.org/10.1111/j.1096-3642.1860. tes campestres, savânicos e florestais da América do Sul. Tese de
tb00146.x Doutorado. PhD, Universidade de Brasília.
Benjamini, Y., & Yosef, H. (1995). Controlling the false discovery rate: A Goerlich, V. C., Nätt, D., Elfwing, M., Macdonald, B., & Jensen, P. (2012).
practical and powerful approach to multiple testing. Journal of the Transgenerational effects of early experience on behavioral, hor-
Royal Statistical Society, 57, 289–300. monal and gene expression responses to acute stress in the pre-
Biro, P. A., & Stamps, J. A. (2008). Are animal personality traits linked to cocial chicken. Hormones and Behavior, 61, 711–718. https://doi.
life-­history productivity? Trends in Ecology & Evolution, 23, 361–368. org/10.1016/j.yhbeh.2012.03.006
https://doi.org/10.1016/j.tree.2008.04.003 Greenberg, R., & Mettke-Hofmann, C. (2001). Ecological aspects of neo-
Bocak, L., & Yagi, T. (2010). Evolution of mimicry patterns in phobia and neophilia in birds. In V. Nolan Jr, & C. Thompson (Eds.),
Metriorrhynchus (Coleoptera: Lycidae): The history of dispersal Current ornithology (pp. 119–178). New York, NY: Springer.
and speciation in Southeast Asia. Evolution, 64, 39–52. https://doi. Ham, A. D., Ihalainen, E., Lindstrom, L., & Mappes, J. (2006). Does co-
org/10.1111/j.1558-5646.2009.00812.x lour matter? The importance of colour in avoidance learning, mem-
Bosque, R. J., Noonan, B. P., & Colli, G. R. (2016). Geographical co- orability and generalisation. Behavioral Ecology and Sociobiology, 60,
incidence and mimicry between harmless snakes (Colubridae: 482–491. https://doi.org/10.1007/s00265-006-0190-4
Oxyrhopus) and harmful models (Elapidae: Micrurus). Global Ecology Hart, N. S. (2001). The visual ecology of avian photoreceptors. Progress
and Biogeography, 25, 218–226. https://doi.org/10.1111/geb.12401 in Retinal and Eye Research, 20, 675–703. https://doi.org/10.1016/
Brodie, E. D. (1993). Differential avoidance of coral snake banded pat- S1350-9462(01)00009-X
terns by free-­ranging avian predators in Costa Rica. Evolution, 47, Hines, H. M., & Williams, P. H. (2012). Mimetic colour pat-
227–235. https://doi.org/10.1111/j.1558-5646.1993.tb01212.x tern evolution in the highly polymorphic Bombus trifascia-
Brodie, E. D., & Janzen, F. J. (1995). Experimental studies of coral snake tus (Hymenoptera: Apidae) species complex and its comimics.
mimicry: Generalized avoidance of ringed snake patterns by free-­ Zoological Journal of the Linnean Society, 166, 805–826. https://doi.
ranging avian predators. Functional Ecology, 9, 186–190. org/10.1111/j.1096-3642.2012.00861.x
Brown, M. E. (1996). Assessing body condition in birds. In V. Nolan, & Hinman, K. E., Throop, H. L., Adams, K. L., Dake, A. J., McLauchlan,
E. D. Ketterson (Eds.), Current ornithology (pp. 67–135). Boston, MA: K. K., & McKone, M. J. (1997). Predation by free-­ r anging
Springer. https://doi.org/10.1007/978-1-4615-5881-1 birds on partial coral snake mimics: The importance of
Buasso, C. M., Leynaud, G. C., & Cruz, F. B. (2006). Predation on snakes ring width and color. Evolution, 51, 1011–1014. https://doi.
of Argentina: Effects of coloration and ring pattern on coral and false org/10.1111/j.1558-5646.1997.tb03684.x
coral snakes. Studies on Neotropical Fauna and Environment, 41, 183– Hotová Svádová, K., Exnerová, A., Kopečková, M., & Štys, P. (2013). How
188. https://doi.org/10.1080/01650520600630725 do predators learn to recognize a mimetic complex: Experiments
Calhim, S., & Montgomerie, R. (2015). Testis asymmetry in birds: The in- with naive Great Tits and aposematic Heteroptera. Ethology, 119,
fluences of sexual and natural selection. Journal of Avian Biology, 46, 814–830. https://doi.org/10.1111/eth.12121
175–185. https://doi.org/10.1111/jav.00503 Jiggins, C. D., Mallarino, R., Willmott, K. R., & Bermingham, E. (2006). The
Campbell, J. A., & Lamar, W. W. (2004). The venomous reptiles of the phylogenetic pattern of speciation and wing pattern change in neo-
Western Hemisphere. Volume 1. Ithaca, NY: Comstock Publishing tropical Ithomia butterflies (Lepidoptera: Nymphalidae). Evolution, 60,
Associates. 1454–1466. https://doi.org/10.1111/j.0014-3820.2006.tb01224.x
Davis Rabosky, A. R., Cox, C. L., Rabosky, D. L., Title, P. O., Holmes, I. A., John, J. L. (1994). The avian spleen: A neglected organ. Quarterly Review
Feldman, A., & McGuire, J. A. (2016). Coral snakes predict the evolu- of Biology, 69, 327–351. https://doi.org/10.1086/418649
tion of mimicry across New World snakes. Nature Communications, 7, Kassambara, A., Kosinski, M., Biecek, P., & Fabian, S. (2018). survminer:
11484. https://doi.org/10.1038/ncomms11484 drawing survival curves using ‘ggplot2’. R package version 0.4.2.
de Reviers, M., & Williams, J. B. (1984). Testis development and pro- https://CRAN.R-project.org/package=survminer.
duction of spermatozoa in the cockerel (Gallus domesticus). In F. J. Kazemi, B., Gamberale-Stille, G., Tullberg, Birgitta S., & Leimar, O. (2014).
Cunningham, P. E. Lake & D. Hewitt (Eds.), Reproductive biology of Stimulus salience as an explanation for imperfect mimicry. Current
poultry (pp. 183–202). Harlow, UK: British Poultry Science Ltd. Biology, 24, 965–969. https://doi.org/10.1016/j.cub.2014.02.061
BOSQUE et al. | 7499

Kikuchi, D. W., & Pfennig, D. W. (2010). Predator cognition permits im- UK: Oxford University Press. https://doi.org/10.1093/acprof:
perfect coral snake mimicry. The American Naturalist, 176, 830–834. oso/9780198528609.001.0001
https://doi.org/10.1086/657041 Savage, J. M., & Slowinski, J. B. (1992). The colouration of the venomous
Kikuchi, D. W., & Pfennig, D. W. (2013). Imperfect mimicry and the limits coral snakes (family Elapidae) and their mimics (families Aniliidae and
of natural selection. Quarterly Review of Biology, 88, 297–315. https:// Colubridae). Biological Journal of the Linnean Society, 45, 235–254.
doi.org/10.1086/673758 https://doi.org/10.1111/j.1095-8312.1992.tb00642.x
Kraemer, A. C., & Adams, D. C. (2014). Predator perception of Batesian Sherratt, T. N. (2002). The evolution of imperfect mimicry. Behavioral
mimicry and conspicuousness in a salamander. Evolution, 68, 1197– Ecology, 13, 821–826. https://doi.org/10.1093/beheco/13.6.821
1206. https://doi.org/10.1111/evo.12325 Sherwin, C. M., Heyes, C. M., & Nicol, C. J. (2002). Social learning in-
Kurvers, R. H. J. M., Prins, H. H. T., van Wieren, S. E., van Oers, K., Nolet, fluences the preferences of domestic hens for novel food. Animal
B. A., & Ydenberg, R. C. (2010). The effect of personality on social Behaviour, 63, 933–942. https://doi.org/10.1006/anbe.2002.2000
foraging: Shy barnacle geese scrounge more. Proceedings of the Royal Skelhorn, J., & Ruxton, G. D. (2010). Mimicking multiple models:
Society B: Biological Sciences, 277, 601–608. https://doi.org/10.1098/ Polyphenetic masqueraders gain additional benefits from crypsis.
rspb.2009.1474 Behavioral Ecology, 22, 60–65.
Lichtenstein, J. L. L., Wright, C. M., Luscuskie, L. P., Montgomery, G. A., Slagsvold, T., & Wiebe, K. L. (2011). Social learning in birds and its role
Pinter-Wollman, N., & Pruitt, J. N. (2017). Participation in coopera- in shaping a foraging niche. Philosophical Transactions of the Royal
tive prey capture and the benefits gained from it are associated with Society B: Biological Sciences, 366, 969–977. https://doi.org/10.1098/
individual personality. Current Zoology, 63, 561–567. rstb.2010.0343
Lindstrom, L. (1999). Experimental approaches to studying the initial Smith, S. M. (1976). Predatory behaviour of young turquoise-­browed
evolution of conspicuous aposematic signalling. Evolutionary Ecology, motmots, Eumomota superciliosa. Behaviour, 56, 309–320. https://
13, 605–618. https://doi.org/10.1023/A:1011004129607 doi.org/10.1163/156853976X00082
Lindstrom, L., Alatalo, R. V., & Mappes, J. (1997). Imperfect Batesian Speed, M. P. (2000). Warning signals, receiver psychology and predator
mimicry -­The effects of the frequency and the distastefulness of memory. Animal Behaviour, 60, 269–278. https://doi.org/10.1006/
the model. Proceedings of the Royal Society B: Biological Sciences, 264, anbe.2000.1430
149–153. https://doi.org/10.1098/rspb.1997.0022 Svádová, K., Exnerová, A., Štys, P., Landová, E., Valenta, J., Fučíková,
Lindstrom, L., Alatalo, R. V., & Mappes, J. (1999). Reactions of hand-­ A., & Socha, R. (2009). Role of different colours of aposematic in-
reared and wild-­c aught predators toward warningly colored, gregar- sects in learning, memory and generalization of naïve bird pred-
ious, and conspicuous prey. Behavioral Ecology, 10, 317–322. https:// ators. Animal Behaviour, 77, 327–336. https://doi.org/10.1016/j.
doi.org/10.1093/beheco/10.3.317 anbehav.2008.09.034
Mappes, J., & Alatalo, R. V. (1997). Batesian mimicry and signal accuracy. Symula, R., Schulte, R., & Summers, K. (2001). Molecular phylogenetic
Evolution, 51, 2050–2053. https://doi.org/10.1111/j.1558-5646.1997. evidence for a mimetic radiation in Peruvian poison frogs supports
tb05129.x a Müllerian mimicry hypothesis. Proceedings of the Royal Society
Marples, N. M., & Brakefield, P. M. (1995). Genetic variation for the B: Biological Sciences, 268, 2415–2421. https://doi.org/10.1098/
rate of recruitment of novel insect prey into the diet of a bird. rspb.2001.1812
Biological Journal of the Linnean Society, 55, 17–27. https://doi. Therneau, T. M. (2015). A package for survival analysis in S. version 2.38,
org/10.1016/0024-4066(95)90026-8 <URL:https://CRAN.R-project.org/package=survival.
Møller, A. P. (1994). Directional selection on directional asymmetry: Thorogood, R., Kokko, H., & Mappes, J. (2017). Social transmission of
Testes size and secondary sexual characters in birds. Proceedings avoidance among predators facilitates the spread of novel prey.
of the Royal Society B: Biological Sciences, 258, 147–151. https://doi. Nature Ecology & Evolution, 2, 254–261.
org/10.1098/rspb.1994.0155 Turner, J. R. G. (1988). The evolution of mimicry: A solution to the problem
Moore, D. F. (2016). Applied survival analysis using R, 1st ed. Basel, of punctuated equilibrium. The American Naturalist, 131, S42–S66.
Switzerland: Springer International Publishing. https://doi.
org/10.1007/978-3-319-31245-3
Pekar, S., Jarab, M., Fromhage, L., & Herberstein, M. E. (2011). Is the evo- S U P P O R T I N G I N FO R M AT I O N
lution of inaccurate mimicry a result of selection by a suite of pred-
ators? A case study using myrmecomorphic spiders. The American Additional supporting information may be found online in the
Naturalist, 178, 124–134. https://doi.org/10.1086/660287 Supporting Information section at the end of the article.    
R Core Team (2017). R: A language and environment for statistical comput-
ing. Vienna, Austria: R Foundation for Statistical Computing.
Rowe, C., & Guilford, T. (2000). Aposematism: To be red or dead. How to cite this article: Bosque RJ, Lawrence JP, Buchholz R,
Trends in Ecology & Evolution, 15, 261–262. https://doi.org/10.1016/ Colli GR, Heppard J, Noonan B. Diversity of warning signal and
S0169-5347(00)01897-8
social interaction influences the evolution of imperfect
Roze, J. A. (1996). Coral snakes of the Americas: Biology, identification, and
venoms. Malabar, FL: Krieger Publishing Company. mimicry. Ecol Evol. 2018;8:7490–7499. https://doi.
Ruxton, G. D., Sherratt, T. N., & Speed, M. P. (2004). Avoiding attack: The org/10.1002/ece3.4272
evolutionary ecology of crypsis, warning signals and mimicry. Oxford,