ecology letters, (2010) 13: 1494–1502 doi: 10.1111/j.1461...

L E T T E RMimicry between unequally defended prey can be

parasitic: evidence for quasi-Batesian mimicry

Hannah M. Rowland,1* Johanna

Mappes,2 Graeme D. Ruxton3 and

Michael P. Speed1

1Department of Evolution,

Ecology and Behaviour, Institute

of Integrative Biology, Faculty

of Health and Life Sciences,

Biosciences Building, University

of Liverpool, Crown Street,

Liverpool, L69 7ZB, UK2Centre of Excellence in

Evolutionary Research,

Department of Biological and

Environmental Science, P.O. Box

35, 40014 University of

Jyvaskyla, Finland3College of Medical, Veterinary

and Life Sciences, Institute of

Biodiversity, Animal Health and

Comparative medicine, Graham

Kerr Building, University of

Glasgow, Glasgow, G12 8QQ,

UK

*Correspondence: E-mail:

[email protected]

Abstract

The nature of signal mimicry between defended prey (known as Mullerian mimicry) is

controversial. Some authors assert that it is always mutualistic and beneficial, whilst

others speculate that less well defended prey may be parasitic and degrade the protection

of their better defended co-mimics (quasi-Batesian mimicry). Using great tits (Parus

major) as predators of artificial prey, we show that mimicry between unequally defended

co-mimics is not mutualistic, and can be parasitic and quasi-Batesian. We presented a

fixed abundance of a highly defended model and a moderately defended dimorphic

(mimic and distinct non-mimetic) species, and varied the relative frequency of the two

forms of the moderately defended prey. As the mimic form increased in abundance,

per capita predation on the model–mimic pair increased. Furthermore, when mimics were

rare they gained protection from predation but imposed no co-evolutionary pressure on

models. We found that the feeding decisions of the birds were affected by their

individual toxic burdens, consistent with the idea that predators make foraging decisions

which trade-off toxicity and nutrition. This result suggests that many prey species that

are currently assumed to be in a simple mutualistic mimetic relationship with their

co-mimic species may actually be engaged in an antagonistic co-evolutionary process.

Keywords

Aposematism, co-evolutionary arms race, Mullerian mimicry, Parus major, quasi-Batesian

mimicry.

Ecology Letters (2010) 13: 1494–1502

I N T R O D U C T I O N

In defensive mimicry, prey species share a visual signal that

serves to deter predation. Such mimicry is well documented

in a variety of taxa (Sbordoni et al. 1979; Waldbauer &

Cowan 1985; Dumbacher & Fleischer 2001; Kapan 2001;

Symula et al. 2001; Sanders et al. 2006) and is conventionally

separated into two distinct categories. In Batesian mimicry

(after Bates 1862), a palatable species evolves to resemble an

unpalatable species, known as a model. Batesian mimics are

likely to degrade the protection of their models, therefore

the evolutionary dynamics are antagonistic, characterized by

co-evolutionary arms races, with strong mimetic resem-

blances and divergence of mimetic signals (Nur 1970). In

Mullerian mimicry (after Muller 1878, 1879), unpalatable

species have a common warning signal which protects them

from predation by allowing the costs of predator education

to be shared. Mullerian co-mimics are considered mutual-

istic and are often expected to be monomorphic (Turner

1977), though this is not always observed (Mallet 1999,

2010).

Although these two forms of mimicry are widely

accepted, it has long been recognized that there is a

palatability spectrum in mimetic prey populations, so that

some species are more inedible than others (Wallace 1882;

Brower et al. 1968, 1972; Sargent 1995). Muller�s original

theory of mutually beneficial mimicry was poorly equipped

to deal with this fact, as he explicitly modelled mimicry

between equally defended prey (Muller 1879). The question

of whether unequally defended mimetic prey are always

mutualistic is controversial and remains an open empirical

question (Marshall 1908; Dixey 1909; Benson 1977; Turner

et al. 1984; Speed 1993a; Joron & Mallet 1998; Mallet 1999;

Sherratt et al. 2004; Rowland et al. 2007; Honma et al. 2008).

Perhaps the first documented suggestion that less well

defended species may be (parasitic) Batesian-like mimics,

rather than (mutualistic) Mullerian mimics was made by

Marshall (1908). Marshall argued that species of predator

Ecology Letters, (2010) 13: 1494–1502 doi: 10.1111/j.1461-0248.2010.01539.x

� 2010 Blackwell Publishing Ltd/CNRS

vary in their tolerance of prey toxins, therefore some would

feed on toxic prey, especially those with low toxin levels.

In addition, he argued that predators were often sufficiently

hungry that they would deliberately choose to eat some

unpalatable butterflies. Modern treatments represent this

argument as a trade-off between toxicity and nutrition

(Speed 1993a; Speed et al. 2000; Kokko et al. 2003; Sherratt

2008). Less well defended prey that act in a Batesian manner

have been termed quasi-Batesian mimics. The existence of

such mimics could explain the observed polymorphisms in

species that are assumed to show Mullerian mimicry (Speed

1993b).

Marshall�s idea has subsequently been re-invented in

theoretical models a number of times (see Nicholson 1927;

Huheey 1976; Owen & Owen 1984; Speed 1993b, 1999a;

Kokko et al. 2003; Sherratt et al. 2004; Balogh et al. 2008;

Mikaberidze & Haque 2009). In these models, a central

assumption is that per encounter attack probabilities are

higher on moderately defended than highly defended prey;

when there is mimicry between such prey, the per encounter

attack probability lies somewhere between these values.

In most theoretical models of quasi-Batesian mimicry, less

well defended prey may in some circumstances benefit their

better defended co-mimics, particularly if the less well

defended co-mimic is sufficiently abundant that it decreases

the per capita encounter rate on the better defended co-

mimic, such that any increases in per encounter attack risk

are offset (Speed 1993b, 1999a). Even if the attack

probability per encounter is increased by the abundance of

the moderately defended mimic, the per capita probability

that any individual is encountered will decline and offsets

this – provided that the mimic�s abundance is sufficiently

large (see Speed 1993b, 1999a,b; Kokko et al. 2003; Sherratt

et al. 2004; Rowland et al. 2007).

To date the empirical basis of quasi-Batesian mimicry is

equivocal. One field experiment shows good support (Speed

et al. 2000), whereas laboratory investigations provide varied

results including mutualistic relationships (Rowland et al.

2007) or inconsistent effects of unequal unpalatability levels

on mimicry (only sometimes parasitic depending for

instance on alternative prey or the strength of signals:

Lindstrom et al. 2006; Ihalainen et al. 2007, 2008). Two

important unanswered questions are: (1) can quasi-Batesian

mimicry be demonstrated under controlled laboratory

conditions? and (2) what kinds of ecological scenarios, if

any, will tend to make mimicry between defended prey

parasitic in nature?

To answer these questions, we examined the foraging

behaviour of naıve wild-caught captive great tits (Parus

major) which served as visually hunting predators. The

predators met three �species� of artificial prey: a cryptic

edible prey and two aposematic species. One of the

aposematic species was highly defended, of constant

abundance and monomorphic; this served as a model

species. The other aposematic species was moderately

defended and also has constant total abundance, but this

species had two visual forms, one of which was a mimic of

the highly defended model, whilst the other form was

distinctive and non-mimetic (Rowland et al. 2010 demon-

strate that the signals are equally visible and salient, and are

not generalized by great tits). Therefore, the total abundance

of members of the moderately defended species remained

constant but the relative frequencies of the mimic and non-

mimic form varied between zero and one by increments of

0.25 (see Table 1). This setup was designed to represent an

ancestral non-mimetic form of a species competing with an

alternative, mimetic form. After the learning trials, we

assessed whether the physiological state of the predators

determined their feeding responses in a manner consistent

with state-dependent explanations of quasi-Batesian mim-

icry. To establish whether mimicry was parasitic or

mutualistic, we had to demonstrate that predators recog-

nized the different levels of defence within our prey, and

that they could not discriminate models from mimics prior

Table 1 Experimental setup. An alternative edible prey was cryptic (a cross) was presented in all treatments, which matched the crosses on

the aviary background where prey was presented

Perfect mimic

density 0%

Perfect mimic

density 25%

Perfect mimic

density 50%

Perfect mimic

density 75%

Perfect mimic

density 100%

Signal Type Signal Type Signal Type Signal Type Signal Type

Cryptic 60 Cryptic 60 Cryptic 60 Cryptic 60 Cryptic 60

Model 60 Model 60 Model 60 Model 60 Model 60

Perfect mimic 0 Perfect mimic 15 Perfect mimic 30 Perfect mimic 45 Perfect mimic 60

Distinct mimic 60 Distinct mimic 45 Distinct mimic 30 Distinct mimic 15 Distinct mimic 0

For half of the birds models were always a black square signal and were highly unpalatable, for the other half models were circles and highly

unpalatable. Mimics were either squares or circles, depending on the model signal and were moderately defended. Distinct non-mimic forms

of the moderately defended species were the opposite signal to model, hence if model was square, distinct non-mimic was circle, and vice versa.

Numbers in the columns correspond to the number of each prey type presented at the start of a trial.

Letter Evidence for quasi-Batesian mimicry 1495


to attack. We have structured our Results section to reflect

this.

M A T E R I A L A N D M E T H O D S

Predators and housing

Eighty-one wild great tits (P. major) were caught at feeding

stations at Konnevesi Research Station, Finland, and served

as predators. The experiment was conducted between

October and December 2009, with permission number

KSU-2007-L-687 ⁄ 254 from the Central Finland Regional

Environmental Center and Animal Experiment Board ⁄ State

Provincial Office of Southern Finland, number ESLH-2007-

09311 ⁄ Ym-23. Two birds died for reasons unrelated to the

experiment; three birds failed to fulfil one or more stages of

the training programme and were released; and four birds

were deemed unfit to participate and were rehabilitated

without any involvement in the experiment (or any other

experiments) before being released.

Each bird was housed separately in indoor plywood

cages, with a daily light period of 11.5 h and ad libitum

supply of tallow, sunflower seed and water, except for the

days of training and trials when the birds were deprived for

c. 1 h to promote food-searching motivation. Following

testing, birds were ringed for identification and released at

the trapping site.

Artificial prey

The artificial prey items used in the experiment were small

pieces (c. 0.1 g) of almond glued (UHU non-toxic glue)

between two 8 mm · 8 mm pieces of white paper. Onto

the paper was printed a black signal (cross, circle or square;

see Table 1). Unpalatable prey (circles and squares) were

made either highly or moderately unpalatable by soaking the

almonds in a chloroquine phosphate solution for 1 h (2 g of

chloroquine dissolved in 30 mL of water or 0.25 g of

chloroquine dissolved in 30 mL of water). Cryptic edible

(cross) prey contained untreated almond.

Training

Opening prey items

All birds were trained in their home cages to open the

artificial prey and eat the almond inside in four stages

(following the procedure detailed by Ihalainen et al. 2007).

Each bird was required to eat all prey items at each stage

before moving to the next stage.

Foraging in the aviaries

Next, the birds were trained to find prey items on artificial

cross backgrounds on the aviary floor (in the small aviary),

and to use the perches in the small aviary (following the

procedure detailed by Ihalainen et al. 2007). We also

familiarized the birds with the large aviary by placing birds

to overnight and feed inside the aviary. The artificial

background was replaced by transparent plastic covered

with peanuts and sunflower seeds.

Aviaries

The experiments were conducted in three aviaries (one large:

57.7 m2 · 3.5 m height, used for the learning phase; and

two small: 13.5 m2 · 2.4 m height, used for the state-

dependent foraging trial). The large aviary�s floor was

covered by a background of white A3 size paper sheets

(297 mm · 420 mm), glued together and covered with

adhesive plastic to form a grid of 15 rows and 22 columns.

Between rows were wooden dividers (c. 6 cm wide boards)

to facilitate prey handling and movement of the birds.

Printed onto each A3 sheet were 70 crosses and 10 fake

cryptic prey items (8 mm · 8 mm pieces of cardboard with

printed crosses glued on the top) glued in random positions

on each sheet. The fake prey items made the background

three-dimensional, which aided camouflage of the cryptic

prey. In the large aviary, there were eight perches positioned

along three of the four walls and interspersed in the arena at

a height of 0.5 m to allow prey handling.

The small aviary floors were covered with A3 sized paper

sheets with crosses printed onto them and with fakes. The

small aviarys� background consisted of eight rows of 10

paper sheets. Each of the smaller aviaries had two perches.

Only one prey item was placed onto each A3 sheet, to aid

the observer (HMR) in identifying the attacked items. Birds

were observed through a one-way mirrored window. Fresh

water was always available in the aviaries.

Learning experiment

The signal of the model was a black square or a circle (see

Table 1). Our design was counterbalanced, so that half of

the birds were presented with square models and the

remainder received circles as models. When the model took

the form of a black square, the mimic form of the

moderately defended species also had that appearance,

whereas the non-mimetic form used the visually distinctive

circle. There is good evidence that the birds do not

generalize between the square and circle symbols (Rowland

et al. 2010). In all treatments, there were always 180 prey

presented (60 edible prey with a cross pattern, 60 of the

highly defended model species and 60 of the moderately

defended species, split between a mimic and a distinct non-

mimic form).

We released each bird into the large aviary, and recorded

the number and type of prey attacked (pecked) and �killed�.

1496 H. M. Rowland et al. Letter


A prey item was recorded as killed when the bird opened the

paper shell and took a bite or ate the almond contained

within the shell. The position of each prey type was

determined by randomly generated maps produced prior to

each test, such that no two birds experienced the same

configuration. Visually indistinguishable prey were denoted

as either models or mimics on the map for the observer to

track. Each bird was required to �kill� 50 prey items before

we ended the trial, at which point the remaining prey items

were collected, the aviary was swept, vacuum-cleaned and

reset for the next bird.

State-dependent foraging trial

On the two consecutive days following the main learning

trial the birds participated in a state-dependent foraging trial

that was run in the small aviary. Here, we aimed to

investigate whether recent experience with chloroquine

affected the birds� decision making in relation to the

mimetic prey. Chloroquine has been shown to have post-

ingestional effects (Skelhorn & Rowe 2007), and at very high

doses appears to be emetic to birds (Alcock 1970). Prior to

each test, the toxin burden of each subject was manipulated

by pre-feeding with two mealworms that were injected with

either 0.02 mL of water or 0.02 mL of 1% chloroquinine

phosphate solution (following Skelhorn & Rowe 2007). The

mealworms had no associated colour or taste cues by which

water or chloroquinine could be detected upon attack, and

birds showed no hesitation in eating all the mealworms

offered prior to the start of each trial. Only one bird refused

to eat either water-injected or chloroquine-injected meal-

worms; this individual did not participate in the state-

dependent trial further, and was released. Our design was

counterbalanced, so that half of the birds received water

mealworms first, whilst the remainder received chloroquine-

injected mealworms first.

State-dependent foraging trials were performed in the

small aviaries. The number of prey presented and the

number that the birds were required to attack were reduced.

Prey were presented in the same ratio as the learning trials

(i.e. 20 cryptic and 20 model prey items plus 20 mimic prey

in the appropriate distribution of mimic and distinct forms).

All prey items were palatable.

Statistical analyses

We determined the absolute numbers of prey killed by each

bird, and calculated the per capita mortality of each prey type

by dividing the total number of each type of prey killed

during the trial by the number presented. Signal type (square

or circle) did not have significant effects on the mortality of

the model (F1,52 = 0.024, P = 0.877) or either form of the

moderately defended species (mimic form, F1,40 = 0.012,

P = 0.89; distinct non-mimic form, F1,42 = 0.018, P =

0.894), and hence we pooled the data for the analyses.

The data generally satisfied the requirements of parametric

statistics and did not require transformation. We applied

a Greenhouse-Geisser correction for sphericity where

required. In the case where data could not be normalized

for a one-sample t-test, we also conducted and report

the results of a nonparametric Wilcoxon one-sample test.

We analysed the data using repeated measures and univariate

GLM in SPSS v.16.0 (SPSS, Inc., Chicago, IL, USA).

R E S U L T S

We consider the results as answers to five questions about

the edibility of the prey and the effects of signal mimicry on

survival.

Is there a spectrum of acceptability in the prey?

We first needed to demonstrate that the predators treated

the prey of different edibility levels differently. We examined

the condition in which 60 models exist with 60 distinct

non-mimic forms and 60 edible prey (crosses) to see if prey

types that we intended to have different edibility were

attacked at different rates. Edible crosses were attacked

more than moderately defended distinctive non-mimics,

which in turn were attacked more than the highly defended

model prey (edible, mean number attacked = 26.25,

SE = 2.125; moderately defended distinctive non-mim-

ics = 14.50, SE = 1.288; and highly defended mod-

els = 9.25, SE = 1.37). The edible cryptic prey items were

therefore attacked at a higher rate than the defended forms,

even though they were less conspicuous. To verify this

interpretation, we conducted a repeated measures ANOVA

(with Greenhouse-Geisser correction) which reported a

highly significant effect of edibility on mean attack rates of

the three prey (F1.363,14.993 = 18.81, P < 0.0001). Pairwise

Bonferroni corrected comparisons showed that per capita

attack levels were higher in the edible prey than the model

species (P = 0.001) and moderately defended distinct non-

mimic form (P = 0.012). Attack rates were also lower on

models than on the distinct non-mimic form (P = 0.022).

Do model and the mimic form share a mortality rate?

We next excluded the possibility that birds may have used

some cue (UV, odour, etc.) to discriminate between models

and mimics. We evaluated the per capita mortality levels of

model and the mimic form using repeated measures ANOVA,

with prey type as a within-subjects factor and mimic

frequency as a between-subjects factor (excluding the case

of mimic frequency of zero). We found no effect of prey

type on per capita mortality rates (F1,36 = 1.77, P = 0.19),



nor did prey type interact with the frequency of the mimic

form (F3,36 = 2.04, P = 0.126), hence there is no evidence

that the birds could discriminate models and mimics before

attacking them.

How does the frequency of the mimic affect per capitaattack rate on model and the mimic?

Given that the model and its mimic have attack rates that

are not significantly different, we calculated the overall

mortality of models plus mimics for this analysis (i.e. sum of

model and mimic attacked ⁄ total abundance of model and

mimic). To examine the effect of mimicry on attack rates,

we used a one-way ANOVA on the per capita mortality rate for

the model and its mimic. Where the mimic did not exist, we

used per capita rate for the model alone.

Mimic frequency (relative to the whole of the moderately

defended species) had a highly significant effect on per

capita attack rates on the co-mimetic prey (model and

mimic; F4,52 = 5.208, P = 0.001). Figure 1 shows that

attack rates were higher when the mimic form was common

in the moderately defended species. Thus, mimicry did not

reduce attack probability on the model–mimic pair, but

rather tended to increase it. Planned contrasts, comparing

mortality rates in the condition of mimic absent (fre-

quency = 0) to the remaining mimic frequencies, showed

that mean mortality was significantly higher when the mimic

was most common (i.e. 60 or 45 mimics; P = 0.025,

P = 0.012 respectively), but not when the mimic was rarer

(i.e. 30 or 15 mimics; P = 0.227, P = 0.221 respectively).

The same conclusions would be drawn if we analysed

model�s mortality separately – see Supporting information,

part 1).

The increased attack rates on the models and mimics

could be explainable by incomplete avoidance learning by

the birds. To assess this, we conducted two further sets of

analyses (see Supporting information, part 2). Birds killed

significantly fewer models by the end of the learning trial,

and the number of models �killed� in the last 10 prey (41–50)

did not differ to the penultimate 10 prey (31–40). Therefore,

we can discount the possibility that high mortality was due

to incomplete predator learning.

Do mimetic prey have higher survival rates thannon-mimetic forms?

We next considered whether the mimic form was always

superior in its survival to the distinct non-mimic form of the

moderately defended species. We calculated the ratio of per

capita mortality of the mimic form ⁄ per capita mortality of the

distinct non-mimic form for each bird for each frequency of

the mimic. Values lower than one indicate that the mimic

has lower mortality than the distinct non-mimic (and vice

versa). The ratio was calculated whenever two forms of the

moderately defended prey coexisted (i.e. mimic frequencies

of 0.25, 0.5, 0.75). For the condition mimic frequency = 0,

we took the value of per capita mortality for model to

represent the mortality value of a vanishingly rare mimetic

mutant in a population of ancestral distinctive non-mimics.

In this case, the mimic�s attack rate converges on that of the

model.

The data could not be normalized by transformation.

We conducted both a parametric one-sample t-test (as this

test is reasonably robust to deviations from normality; Zar

1999) and nonparametric Wilcoxon one-sample tests (see

Supporting information, part 3). The mimic tended to be

attacked less than the distinct non-mimic form, except in

the case that the mimic was very common (Fig. 2; one-

sample t-tests of the null hypothesis of a mean value of one).

Mimic [PM(0), P = <0.001; PM(0.25), P = 0.519; PM(0.5),

P = 0.006; PM(0.75), P = 0.095; Wilcoxon results equiva-

lent]. Note that the margins of error are relatively small for

the two conditions in which we estimated survival rate based

on attacks on relatively large numbers of prey (mimic = 0.5;

and 0, in which case its mortality rate is estimated from the

relatively abundant model).

In the cases where the numbers of one form of the

moderately defended species were small (i.e. mimic fre-

quency 0.25 or the distinctive non-mimic 0.25), the

variances were larger, because each attack makes a greater

increment to per capita mortality rates in small populations

than in large ones. It is not surprising then that these results

are non-significant. Variance in levels of selection is an

important evolutionary point which we consider in more

detail in the Discussion. Finally, we found that a mono-

morphic population of mimics has no survival advantage

Mimic density604530150

Mea

n t

ota

l mo

rtal

ity

of

mo

del

s +

mim

ics

(+/–

1 S

E)

0.30

0.25

0.20

0.15

0.10

0.05

0.00

Figure 1 Per capita attack rates on model and its mimic for a range

of mimic densities (error bars, 1 SE).



over a monomorphic population of distinctive non-mimics

(t20 = 0.989, P = 0.33), indicating that mimicry loses its

advantages completely as it becomes very common in the

population.

Is there evidence that physiological state of the predatorsaffects the mimetic relationships?

Several authors have hypothesized that the nutritional state

of a predator can account for the occurrence of quasi-

Batesian mimicry (Speed 1993a; Kokko et al. 2003; Sherratt

2003; Sherratt et al. 2004). In fact there is evidence to

support the idea that predators make foraging decisions

which trade-off toxicity and nutrition (Barnett et al. 2007;

Skelhorn & Rowe 2007). We therefore performed two

further tests after the learning trials in which the predators

encountered prey in the same ratios as their learning trials.

All prey items in these tests were palatable. Prior to each

test, the toxin burden of each bird was manipulated by pre-

feeding with two mealworms that were injected with either

0.02 mL of water or 0.02 mL of 1% chloroquinine

phosphate solution (following methods detailed in Skelhorn

& Rowe 2007).

In the first trial, if the birds were given chloroquine, attack

rates were higher when the mimic was common than when it

was rare or absent. Planned simple contrasts, comparing

mortality rates in the condition of mimic (frequency = 0) to

the remaining mimic frequencies, showed that mean mortal-

ity was significantly higher when the mimic was most

common (i.e. 60 mimics; P = 0.019), but not when the

mimic was rarer (i.e. 45 mimics; P = 0.052; 30 mimics;

P = 0.097; or 15 mimics; P = 0.771; see Fig. 3a). When the

birds were given water mealworms instead of chloroquine

mealworms on day 1, mimic frequency did not have a

significant effect on per capita attack rates on the model prey

(Fig. 3b; all planned contrasts P > 0.05).

On the second day, when birds were administered the

opposite treatment to day 1, mimic frequency did not have a

significant effect on per capita attack rates on the model

prey for either chloroquine or water (Fig. 3c,d;

7550250

Mea

n r

atio

of

mim

ic t

o d

isti

nct

no

n-m

imic

mo

rtal

ity

(+/–

1 S

E)

2.50

2.00

1.50

1.00

0.50

0.00

Figure 2 Mean of ratio of mortalities of perfect ⁄ distinct non-

mimic form for a range of mimic frequencies. Note: for perfect

mimic = 0, we take the model mortality rate, thereby estimating

survival of a vanishingly rare, novel mutant form.

Mimic density Mimic density

Mimic density Mimic density

604530150

Mea

n m

ort

alit

y o

f m

od

els

(+/–

1 S

E)

0.40

0.30

0.20

0.10

0.00

604530150

Mea

n m

ort

alit

y o

f m

od

els

(+/–

1 S

E)

0.40

0.30

0.20

0.10

0.00

604530150Mea

n m

ort

alit

y o

f m

od

el d

ay 2

(+/

– 1

SE

)

0.40

0.30

0.20

0.10

0.00

604530150Mea

n m

ort

alit

y o

f m

od

el d

ay 2

(+/

– 1

SE

)

0.40

0.30

0.20

0.10

0.00

(a) (b)

(c) (d)

Figure 3 Per capita attack rates on the model

prey during state-dependent foraging trial

when birds were pre-fed mealworms con-

taining either chloroquinine or water: (a) day

1 pre-fed chloroquine; (b) day 1 pre-fed

water; (c) day 2 pre-fed water (birds given

chloroquine on day 1); (d) day 2 pre-fed

cholorquine (birds given water on day 1).



F4,26 = 1.344, P = 0.287; F4,26 = 0.958, P = 0.451). These

results show that a predator�s toxin burden can have

significant influence on foraging decisions, and conse-

quently on model–mimic dynamics.

D I S C U S S I O N

This is, to our knowledge, the first purpose-designed

laboratory study which unequivocally supports the view that,

when there is an inequality in defence levels, mimicry

between two species can be parasitic and Batesian-like (or

quasi-Batesian) and not necessarily mutualistic and Mullerian.

As the mimic form increased in frequency to high levels

(compared to the alternative distinct non-mimic form of the

moderately defended prey), per capita predation rates on the

model–mimic pair increased, indicating that mimicry was

acting in a Batesian manner. In terms of current understand-

ing of the evolution of mimicry this is an important finding,

as quasi-Batesian mimics may be prone to signal diversifi-

cation and to cause antagonistic co-evolution with their

models. Furthermore, quasi-Batesian mimics may face much

stronger selection for signal accuracy than might Mullerian

co-mimics, leading to a higher quality of mimicry itself.

Implications for the evolution of mimicry

The �model� species in our experiment did not benefit from

co-existence with a visually indistinguishable moderately

defended mimic, and lost protection when the mimic form

was abundant. It is notable, however, that mimicry had no

effect on the survival of the model when it was compar-

atively rare (25 or 50% of the �mimic� population).

Consequently, we have a combination of parasitic mimicry

(when the mimic form is very common) and what has been

called �effectively neutral mimicry� (Turner et al. 1984), in

which there would be no co-evolutionary effect of a mimic

on its model. In this situation, the mimic gained protection

from predation, but would have imposed no co-evolutionary

force on its model.

Survival of the mimic form was generally higher than the

alternative distinctive non-mimetic form of the moderately

defended species. However, and in keeping with the view

that the moderately defended mimic form can be para-

sitic, the mimetic form also lost protection at high

abundance levels such that members of a monomorphic

population of mimics had no survival benefit compared to a

monomorphic population of the distinct non-mimic form.

This result demonstrates that rarity relative to non-mimetic

forms is advantageous in the mimetic form, and therefore

supports the idea that selection could favour diversification

in quasi-Batesian mimics (see Speed 1993a).

An important finding is that selection (measured as the

ratio of per capita survival rates between the mimic and

distinct non-mimic) was highly variable when either the

mimic or the distinct non-mimic form was rare (see also

Lindstrom et al. 2006). When there are relatively few

individuals of a given type to attack, comparatively small

levels of variance in absolute numbers of prey killed have a

large effect on the per capita levels of attack. For example,

suppose that the mean number killed is 5 prey (± 4). From a

large population of 100 individuals, the variance in per capita

attack frequency is small (based on ± 0.04 of the popula-

tion). In contrast, the variance from a small population of 10

individuals is large (based on ± 0.4 of the population). This

causes an increased level of variance in our measure of

fitness when the mimic form is rare (25%, see Fig. 2).

However, when the distinct non-mimic form was rare,

variance in the ratio of survival values was extremely high

(mimic frequency 0.75, Fig. 2) and this is explained by an

additional factor, that the rarity of the prey form itself

caused raised variance in predation rates. Some predators

tended to ignore the very rare distinct non-mimic, whilst

others attacked at relatively high rates. This did not happen

to the mimic when it was rare, because it shared the same

signal as the model and so this signal was always abundant.

This is an important finding since it predicts that random

drift in frequencies of mimetic and non-mimetic forms is:

(1) likely to be very important when prey forms are rare and

(2) asymmetrical in its effects, disturbing morph frequencies

most when a non-mimetic prey form is very rare locally (see

also Mallet 2010).

Evidence for quasi-Batesian mimicry in other experiments

The present experiment and that of Speed et al. (2000) offer

unequivocal support for quasi-Batesian mimicry. These have

in common an experimental design in which the mimic

species took two forms: one mimetic and the other

distinctive. This was intended to mimic the population

genetic scenario in which a mimic competes against an

ancestral non-mimetic form. Hence, the absolute number of

prey presented in these two experiments remained constant,

but the relative frequencies of the mimic vs. the non-mimic

form varied inversely with respect to each other.

In a related experiment, Rowland et al. (2007) simulated a

situation in which the mimetic species was monomorphic,

and increased in absolute abundance. Here, mimicry

between unequally defended prey was mutualistic, because

any increase in per encounter attack probability was more

than offset by a dilution effect, decreasing per capita

mortality rates. One conclusion is that if there is no dilution

effect (i.e. population density is not high), as in our present

experiment, then the parasitic nature of quasi-Batesian

mimicry may be seen much more readily than if changes in

mimic abundance increase the absolute number of prey

encountered.



A number of other experiments, using great tit predators

and prey with black and white symbols, have offered less

consistent support, either for the view that mimicry between

unpalatable prey is parasitic and quasi-Batesian, or mutual-

istic and Mullerian, supporting the view that dynamics

between co-mimics depend on ecological conditions.

In some cases, the data from these experiments can be

re-interpreted, so that they can be directly compared with the

experiment in this paper (see Supporting information, part 4

for reconsideration of data in Lindstrom et al. 2006). These

other experiments provide equivocal support for quasi-

Batesian mimicry, and are subject to the effects of inequalities

in signal salience (Lindstrom et al. 2004; Ihalainen et al. 2008)

and generalization of model ⁄ mimic signals (Ihalainen et al.

2007). In the work we present here, we build on these earlier

papers by designing signals of equal salience that were not

generalized. Hence, quasi-Batesian mimicry may be easy to

find when signal stimuli are simple and distinctive, but less

easy to find in more complex signalling environments.

State-dependent explanations for quasi-Batesian mimicry

Several authors have hypothesized that the nutritional state

of a predator can explain the occurrence of quasi-Batesian

mimicry (Speed 1993a; Kokko et al. 2003; Sherratt 2003;

Sherratt et al. 2004). In fact there is some evidence to

support the idea that predators make foraging decision to

trade-off toxicity and nutrition (Barnett et al. 2007; Skelhorn

& Rowe 2007). We found that manipulation of the state of

our predators, adding chloroquine or water to their diets,

could maintain or remove benefits from mimicry in entirely

edible prey. Birds fed chloroquine before the first state-

dependent foraging trial treated the edible prey as if they

contained quinine, whereas those fed with prey containing

water did not. In effect, the predators used their experience

from the learning trials, and their physiological state

(chloroquine fed or water fed) to determine attack rates.

These results match the assumptions of state-dependent

foraging models, in which predators reach attack decisions

based on information about hunger and toxicity (Speed

1993a; Kokko et al. 2003; Sherratt 2003, 2008). Therefore,

state-dependent foraging behaviour likely plays an important

role in mimetic relationships.

C O N C L U S I O N S : T H E E C O L O G Y O F Q U A S I -

B A T E S I A N M I M I C R Y

Our experiment demonstrates the conditions in which

mimicry between two unequally defended prey is quasi-

Batesian in nature, rather than mutualistic. Birds traded off the

benefits of nutrition against the costs of prey toxicity when

determining attack rates on prey (Speed 1993a). Our results

suggest some further questions of importance. First, as Dixey,

Marshall and others have suggested (Marshall 1908; Dixey

1909; Benson 1977), less well defended prey may operate as

(quasi-) Batesian mimics in some seasons (when food is

scarce), and Mullerian mimics in other periods (Speed 1993a;

Kokko et al. 2003; Sherratt et al. 2004). A consequence is that

there may be selection favouring strong mimetic resemblance,

and diversity in mimicry signals only in �Batesian� seasons. In

�Mullerian� seasons there may be relaxed selection on the

quality of mimicry but strong selection for mimetic mono-

morphisms. The combination of these alternating directions

of selection could explain why supposed Mullerian mimics

often have such strong resemblances to each other but tend to

lack Batesian-like mimetic diversity.

A C K N O W L E D G E M E N T S

We thank Helina Nisu and Lenka Trebatika for caring for

the birds and running the state-dependent foraging trials.

We also thank three anonymous referees for their comments

and suggestions. This work was funded by NERC grant

NE ⁄ D010667 ⁄ 1 and the Centre of excellence in evolution-

ary research at the University of Jyvaskyla.

R E F E R E N C E S

Alcock, J. (1970). Punishment levels and the response of black-

capped chickadees (Parus atricapillus) to three kinds of artificial

seeds. Anim. Behav., 18, 592–599.

Balogh, A., Gamberale-Stille, G. & Leimar, O. (2008). Learning

and the mimicry spectrum: from quasi-Bates to super-Muller.

Anim. Behav., 76, 1591–1599.

Barnett, C.A., Bateson, M. & Rowe, C. (2007). State-dependent

decision making: educated predators strategically trade off the

costs and benefits of consuming aposematic prey. Behav. Ecol.,

18, 645–651.

Bates, H.W. (1862). Contributions to an insect fauna of the

Amazon valley Lepidoptera: Heliconidae. Trans. Linn. Soc. Lond.,

23, 495–556.

Benson, W.W. (1977). On the supposed spectrum between

Batesian and Mullerian mimicry. Evolution, 31, 454–455.

Brower, L.P., Reyerson, W.N., Coppinger, L.L. & Glazier, S.C.

(1968). Ecological chemistry and the palatabilty spectrum. Science,

161, 1349–1351.

Brower, L.P., Williams, K.L., McEvoy, P.B. & Flannery, M.A.

(1972). Variation in cardiac glycoside content of monarch

butterflies from natural populations in eastern North-America.

Science, 177, 426–428.

Dixey, F.A. (1909). On Mullerian mimicry and diaposematism.

A reply to Mr GAK Marshall. Trans. R. Entomol. Soc. Lond.,

XXIII, 559–583.

Dumbacher, J.P. & Fleischer, R.C. (2001). Phylogenetic evidence

for colour pattern convergence in toxic pitohuis: Mullerian

mimicry in birds? Proc. R. Soc. Lond. B: Biol. Sci., 268, 1971–1976.

Honma, A., Takakura, K. & Nishida, T. (2008). Optimal-foraging

predator favors commensalistic Batesian mimicry. PLoS ONE, 3,

e3411.



Huheey, J.E. (1976). Studies in warning coloration and mimicry.

VII. Evolutionary consequences of a Batesian-Mullerian

spectrum: a model for Mullerian mimicry. Evolution, 30, 86–

93.

Ihalainen, E., Lindstrom, L. & Mappes, J. (2007). Investigating

Mullerian mimicry: predator learning and variation in prey

defences. J. Evol. Biol., 20, 780–791.

Ihalainen, E., Lindstrom, L., Mappes, J. & Puolakkainen, S. (2008).

Butterfly effects in mimicry? Combining signal and taste

can twist the relationship of Mullerian co-mimics. Behav. Ecol.

Sociobiol., 62, 1267–1276.

Joron, M. & Mallet, J. (1998). Diversity in mimicry: paradox or

paradigm? Trends Ecol. Evol., 13, 461–466.

Kapan, D.D. (2001). Three-butterfly system provides a field test of

mullerian mimicry. Nature, 409, 338–340.

Kokko, H., Mappes, J. & Lindstrom, L. (2003). Alternative prey

can change model-mimic dynamics between parasitism and

mutualism. Ecol. Lett., 6, 1068–1076.

Lindstrom, L., Alatalo, R.V., Lyytinen, A. & Mappes, J. (2004). The

effect of alternative prey on the dynamics of imperfect Batesian

and Mullerian mimicries. Evolution, 58, 1294–1302.

Lindstrom, L., Lyytinen, A., Mappes, J. & Ojala, K. (2006). Relative

importance of taste and visual appearance for predator educa-

tion in Mullerian mimicry. Anim. Behav., 72, 323–333.

Mallet, J. (1999). Causes and consequences of a lack of coevolution

in Mullerian mimicry. Evol. Ecol., 13, 777–806.

Mallet, J. (2010). Shift happens! Shifting balance and the evolution

of diversity in warning colour and mimicry. Ecol. Entomol., 35,

90–104.

Marshall, A.K. (1908). On Diaposematsim with reference to some

limitations of the Mullerian hypothesis of mimicry. Trans.

Entomol. Soc. Lond., 1908, 93–142.

Mikaberidze, A. & Haque, M. (2009). Survival benefits in mimicry:

a quantitative framework. J. Theor. Biol., 259, 462–468.

Muller, F. (1878). Uber die vortheile der mimicry bei schmetter-

lingen. Zool. Anz., 1, 54–55.

Muller, F. (1879). Ituna and Thyridia: a remarkable case of mimicry

in butterflies. Trans. Entomol. Soc. Lond., 1879.

Nicholson, A. (1927). A new theory of mimicry in insects. Aust.

Zool., 5, 10–104.

Nur, U. (1970). Evolutionary rates of models and mimics in

Batesian mimicry. Am. Nat., 104, 477–486.

Owen, R.E. & Owen, A.R.G. (1984). Mathematical paradigms for

mimicry: recurrent sampling. J. Theor. Biol., 109, 217–247.

Rowland, H.M., Ihalainen, E., Lindstrom, L., Mappes, J. & Speed,

M.P. (2007). Co-mimics have a mutualistic relationship despite

unequal defences. Nature-London, 448, 64–67.

Rowland, H.M., Hoogesteger, T., Ruxton, G.D., Speed, M.P. &

Mappes, J.M. (2010). A tale of two signals: mimicry between

aposematic species enhances predator avoidance learning. Behav.

Ecol., 21, 851–860.

Sanders, K.L., Malhotra, A. & Thorpes, R.S. (2006). Evidence for a

Mullerian mimetic radiation in Asian pitvipers. Proc. R. Soc. Lond.

B: Biol. Sci., 273, 1135–1141.

Sargent, T.D. (1995). On the relative acceptabilities of local but-

terflies and moths to local birds. J. Lepid. Soc., 49, 148–162.

Sbordoni, V., Bullini, L., Scarpelli, G., Forestiero, S. & Rampini, M.

(1979). Mimicry in the burnet moth Zygaena ephialtes – population

studies and evidence of a Batesian-Mullerian situation. Ecol.

Entomol., 4, 83–93.

Sherratt, T.N. (2003). State-dependent risk-taking by predators in

systems with defended prey. Oikos, 103, 93–199.

Sherratt, T.N. (2008). The evolution of Mullerian mimicry. Natur-

wissenschaften, 95, 681–695.

Sherratt, T.N., Speed, M.P. & Ruxton, G.D. (2004). Natural

selection on unpalatable species imposed by state-dependent

foraging behaviour. J. Theor. Biol., 228, 217–226.

Skelhorn, J. & Rowe, C. (2007). Predators� toxin burdens influence

their strategic decisions to eat toxic prey. Curr. Biol., 17, 1479–1483.

Speed, M. (1993a). When is mimicry good for predators? Anim.

Behav., 46, 1246–1248.

Speed, M.P. (1993b). Muellerian mimicry and the psychology of

predation. Anim. Behav., 45, 571–580.

Speed, M.P. (1999a). Batesian, quasi-Batesian or Mullerian mimicry?

Theory and data in mimicry research. Evol. Ecol., 13, 755–776.

Speed, M.P. (1999b). Robot predators in virtual ecologies: the

importance of memory in mimicry studies. Anim. Behav., 57,

203–213.

Speed, M.P., Alderson, N.J., Hardman, C. & Ruxton, G.D. (2000).

Testing Mullerian mimicry: an experiment with wild birds. Proc.

R. Soc. Lond. B: Biol. Sci., 267, 725.

Symula, R., Schukte, R. & Summers, K. (2001). Molecular phylo-

genetic evidence for a mimetic radiation in Preuvian poison

frogs supports a Mullerian mimicry hypothesis. Proc. R. Soc. Lond.

B: Biol. Sci., 268, 2415–2421.

Turner, J.R.G. (1977). Butterfly mimicry: the genetical evolution of

an adaptation. Evol. Biol., 10, 163–206.

Turner, J.R.G., Kearney, E.P. & Exton, L.S. (1984). Mimicry and

the Monte Carlo predator: the palatability spectrum and the

origins of mimicry. Bot. J. Linn. Soc., 23, 247–268.

Waldbauer, G.P. & Cowan, D.P. (1985). Defensive stinging and

Mullerian mimicry among eumenid wasps (Hymenoptera, Ves-

poidea, Eumenidae). Am. Midl. Nat., 113, 198–199.

Wallace, A. (1882). Dr. Fritz Muller on some difficult cases of

mimicry. Nature, 26, 86–87.

Zar, J.H. (1999). Biostatistical Analysis, 4th edn. Prentice Hall, New

Jersey, USA.

S U P P O R T I N G I N F O R M A T I O N

Additional Supporting Information may be found in the

online version of this article:

Data S1 Mimicry between unequally defended prey can be

parasitic: Evidence for quasi-Batesian mimicry.

As a service to our authors and readers, this journal provides

supporting information supplied by the authors. Such

materials are peer-reviewed and may be re-organized for

online delivery, but are not copy-edited or typeset. Technical

support issues arising from supporting information (other

than missing files) should be addressed to the authors.

Editor, Gregory Grether

Manuscript received 16 June 2010

First decision made 16 July 2010

Second decision made 2 September 2010

Manuscript accepted 16 September 2010



ecology letters, (2010) 13: 1494–1502 doi: 10.1111/j.1461...

Documents