Content uploaded by Anastasia H. Dalziell
Author content
All content in this area was uploaded by Anastasia H. Dalziell on May 05, 2019
Content may be subject to copyright.
Fooling the experts: accurate vocal mimicry in the song of the superb lyrebird,
Menura novaehollandiae
Anastasia H. Dalziell
*
, Robert D. Magrath
Division of Evolution, Ecology and Genetics, Australian National University, Canberra, Australia
article info
Article history:
Received 12 August 2011
Initial acceptance 19 September 2011
Final acceptance 20 February 2012
Available online 13 April 2012
MS. number: 11-00636R
Keywords:
birdsong
Menura novaehollandiae
mimicry
repertoire
rejection threshold
sexual signal
signal design
song learning
vocal accuracy
vocal mimicry
The degree of resemblance between mimics and models provides valuable insight into the evolutionary
dynamics of mimicry signalling systems, but for many systems mimetic resemblance has not been
quantified. Superb lyrebirds have a reputation for accurately imitating an astonishing variety of sounds
that they incorporate into their sexual displays. We assessed the accuracy with which males imitate the
complex song of the grey shrike-thrush, Colluricincla harmonica. We measured vocal accuracy by (1)
using playback experiments as a bioassay, to determine whether and how shrike-thrushes distinguish
between their own song and imitations of shrike-thrush songs by lyrebirds and (2) comparing acoustic
properties of mimicked and model songs. Shrike-thrushes reacted just as strongly towards mimetic song
as to their own when songs were presented alone. When mimetic song was accompanied by lyrebird
song sequences (emulating the lyrebird’s natural singing style), shrike-thrushes still usually approached
the speaker but less often than when mimetic song was presented alone or when model songs were
broadcast. Acoustic analyses showed that imitations were remarkably similar to model songs. However,
while lyrebirds maintained the structure and complexity of model songs, they sang fewer repetitions of
individual element types. This ‘abridging’of model songs is consistent with a trade-off between
demonstrating both mimetic accuracy and versatility. Overall, these results indicate strong selection on
male lyrebirds to imitate accurately the complex vocalizations of other species, and show that species
can integrate contextual information with the signal structure to distinguish between their own signals
and imitations.
Crown Copyright Ó2012. Published on behalf of The Association for the Study of Animal Behaviour by
Elsevier Ltd. All rights reserved.
Mimicry is a classic example of adaptive signal design in which
individuals benefit from imitating the signals of other species
(Bates 1862;Müller 1878;Vane-Wright 1980) or conspecifics
(Dominey 1980;Gross 1996;Norman et al. 1999;Whiting et al.
2009). Despite considerable advances over the last 150 years,
many aspects of mimetic signalling systems are only beginning to
be understood (Ruxton & Speed 2005;Chittka & Osorio 2007;
Vereecken & McNeil 2010). One area attracting recent attention is
the accuracy of mimetic signals (McGuire et al. 2006;Chittka &
Osorio 2007;Pekár et al. 2011), as it has become clear that the
degree of resemblance between mimics and models can provide
valuable insight into the selective forces driving mimicry signalling
systems (e.g. Holen & Johnstone 2004;Coleman et al. 2007;
Stoddard & Stevens 2010).
Traditional theory assumes that selection will always favour
mimics that most closely resemble the species or conspecific they
imitate, ‘the model’, but several studies have challenged the
simplicity of this view (Johnstone 2002;Sherratt 2002;Holen &
Johnstone 2004;Chittka & Osorio 2007;Harper & Pfennig 2007;
Cheney & Marshall 2009;Pekár et al. 2011). How closely the mimic
resembles the model is a function of the costs and benefits of
mimicry among all three protagonists in the mimicry complex: the
mimic, the model and the receiver of the signal. The benefits of
mimicry include gaining protection from predators, increased
access to food and reproductive advantages (Starrett 1993;Cheney
2010). The costs of mimicry include increased detection by preda-
tors, conflicting signal requirements and costs imposed by the
model (Edmunds 2000;Sherratt 2002;Holen & Johnstone 2004;
Estrada & Jiggins 2008;Stoddard & Stevens 2010). Physiological and
phylogenetic constraints on the mimic will also limit resemblance.
Furthermore, mimetic accuracy can depend on the receiver’s ability
to discriminate between models and mimics, and inaccurate
mimics may evolve as a result of weak selection on the receiver for
fine discrimination, or physiological constraints on the receiver
(Chittka & Osorio 2007;Cheney & Marshall 2009;Kikuchi & Pfennig
2010). Mimetic accuracy is the outcome of these benefits, costs and
constraints. Thus, highly accurate mimicry implies both an
*Correspondence: A. H. Dalziell, Building 116, Division of Evolution, Ecology and
Genetics, Research School of Biology, Australian National University, Canberra 0200,
Australia.
E-mail address: Anastasia.Dalziell@anu.edu.au (A. H. Dalziell).
Contents lists available at SciVerse ScienceDirect
Animal Behaviour
journal homepage: www.elsevier.com/locate/anbehav
0003-3472/$38.00 Crown Copyright Ó2012. Published on behalf of The Association for the Study of Animal Behaviour by Elsevier Ltd. All rights reserved.
doi:10.1016/j.anbehav.2012.03.009
Animal Behaviour 83 (2012) 1401e1410
advantage for accuracy in the mimic as well as strong selection on
the receiver to discriminate between models and mimics. The
nature of the differences between model and mimic are equally
informative about the selection pressures shaping the mimetic
signal.
Several species of bird from disparate families habitually mimic
the sounds produced by other species with remarkable accuracy
(Chisholm 1932), but the evolutionary drivers of vocal mimicry are
usually unclear (Baylis 1982;Kroodsma 2004;Garamszegi et al.
2007;Kelley et al. 2008). In some bird species mimicry functions
deceptively (Langmore et al. 2003,2008;Flower 2011) or may
provide increased protection from predators (Goodale & Kotagama
2006a,b). Alternatively, mimicry may not be deceptive and instead
function to attract mates or repel rivals (Dobkin 1979). This rarely
tested hypothesis is plausible in versatile mimics: oscine passerines
in which males incorporate imitations of multiple sounds into
elaborate, song-like displays (e.g. the northern mockingbird, Mimus
polyglottos, marsh warbler, Acrocephalus palustris, and European
starling, Sturnus vulgaris:Catchpole & Slater 2008). Here, the
conspecific receiver driving the evolution of mimicry gains infor-
mation about the quality of the singer by evaluating the singer’s
mimetic song. However, the structures of mimetic songs have
seldom been examined.
Assessment of mimetic accuracy could help resolve the debate
over which vocal features of versatile mimics are targeted by natural
selection. Selection may favour large repertoires rather than mimicry
per se, since repertoire size affects mate choice and territory defence
in nonmimetic birds (reviewed in Catchpole & Slater 2008). Mimicry
would thus evolve as a shortcut to gaining a large repertoire (Howard
1974) or as a by-product of repertoire acquisition itself (Hindmarsh
1986;Kelley et al. 2008). However, highly accurate mimicry
provides evidence that mimicry itself is under selection, for two
reasons. First, if repertoire size alone is important, accurate resem-
blance between model and mimetic vocalizations is unlikely to be
maintained because accumulated copying errors and differential
selection pressures between models and mimics should eventually
lead to divergent acoustic structures. Second, it is doubtful that
accurate mimicry is selectively neutral given strong evidence that
precise vocal production is challenging (reviewed in Podos et al.
2009) and that individual species appear to have physiological
adaptations optimized for species-specific vocalizations (Zollinger &
Suthers 2004). Versatile mimics, then, are probably under selection
for both mimetic accuracy and versatility, a conclusion supported by
the only study to investigate the relationship between mating
success and both mimetic accuracy and versatility (Coleman et al.
2007). Investigating mimetic accuracy is thus fundamental to
elucidating the evolution of mimicry.
Superb lyrebirds are large (880e1100 g) oscine passerines with
versatile vocal mimicry and appear to be subject to strong sexual
selection (Higgins et al. 2001). Males have extravagant tails while
females are cryptic, only females care for the young (Lill 1979), and
during the winter breeding season males are geographically clus-
tered in a lek-like pattern (Robinson & Curtis 1996). Males incor-
porate mimicry into a flamboyant vocal display with approximately
70% of vocalizations consisting of imitations of about 20 different
species of local birds (Higgins et al. 2001;Zann & Dunstan 2008).
Vocal displays are performed both as a prelude to mating and
during territorial interactions with other males, occurring most
often during the breeding season (Higgins et al. 2001). These
observations strongly suggest that mimetic songs sung by males are
a sexually selected trait and the receiver is either a rival male or
a potential mate.
Adult superb lyrebirds enjoy a reputation for highly accurate
mimicry (e.g. Attenborough 2002;Zoos South Australia 2009), but
there has been only one study of mimetic accuracy in this species
(Zann & Dunstan 2008). Acoustic analysis revealed that adult males
reproduce individual elements from two species of bird with
remarkable accuracy. Furthermore, adult males are more accurate
mimics than juveniles of indeterminate sex, suggesting that
mimetic accuracy improves with age. We built on this introductory
investigation into mimetic accuracy in lyrebirds by analysing
a complex model sound with multiple elements and by using two
methods to measure accuracy: acoustic analysis and a bioassay.
Acoustic analysis is a powerful tool to investigate the accuracy
of vocal parameters but an additional approach is to employ
a bioassay to use a bird’s assessment of mimetic accuracy (Lemaire
1975;Brenowitz 1982). The sensory systems of the model should be
attuned to its own song making it a sensitive assessor of mimetic
accuracy. Moreover, the model may be under selection to distin-
guish between conspecifics and mimics to avoid costly, unneces-
sary responses to mimetic signals. Studies of nonmimetic oscine
passerines have shown that birds respond differently to subtle
differences in species-specific songs (Stoddard 1996;Illes et al.
2006;Blumenrath et al. 2007;Botero et al. 2007;Nicholls 2008;
de Kort et al. 2009). Thus bioassays have the dual advantage of
being able to reveal differences between model and mimetic songs
that may not be detectable using acoustic analysis, and of providing
insight into signal discrimination by model species.
Our aim was to investigate the selective forces acting on
mimetic song in lyrebirds by quantifying mimetic resemblance
from the perspective of the model using a bioassay, and by directly
measuring acoustic features of model and mimetic songs. In addi-
tion, we examined how a model species distinguished between
model and mimetic songs. The grey shrike-thrush, Colluricincla
harmonica, produces a complex, multielement song and is regularly
imitated by lyrebirds (Zann & Dunstan 2008). Shrike-thrushes are
ideal subjects for a bioassay because there is no evidence from our
study or published accounts that shrike-thrushes interact directly
with lyrebirds, making it unlikely that they select for mimicry in
lyrebirds. We conducted two playback experiments. First, we tested
whether shrike-thrushes could discriminate between model and
mimetic songs. Given the likely costs of unnecessary responses, and
the sophisticated auditory systems of oscine passerines, we
assumed that shrike-thrushes would respond similarly to model
and mimic only if mimicry is highly accurate. Second, since lyre-
birds typically imitate several species in succession, we tested
whether the model was more alert to acoustic differences if songs
were presented with lyrebird mimicry of other species.
METHODS
Study Site and Species
We studied populations of superb lyrebirds and grey shrike-
thrushes in Sherbrooke Forest (37
53
0
,145
21
0
), part of the Dande-
nong Ranges National Park in Victoria, Australia. The population of
lyrebirds has been monitored for 50 years by the Sherbrooke Lyre-
bird Study Group (e.g. Kenyon 1972; N. Carter, unpublished data) and
many lyrebirds are individually colour-banded as nestlings. Further
individuals are distinguishable by plumage. Lyrebird songs were
recorded in JuneeJuly 2007 for the playback experiments, and these
were supplemented with recordings made in 2008 for the sound
analysis. Grey shrike-thrushes are socially monogamous, territorial
passerines (Higgins & Peter 2002) and both sexes sing discrete, 3 s
songs fromwhat is probably a large repertoire of song types (Higgins
& Peter 2002). The sexes can be distinguished by slight differences in
plumage (Higgins & Peter 2002) but not by song. Shrike-thrushes are
most vocal and territorially active from August to September, just
before their breeding season. Thus they sing most when lyrebirds are
relatively quiet.
A. H. Dalziell, R. D. Magrath / Animal Behaviour 83 (2012) 1401e14101402
Song types of grey shrike-thrushes can be classed into two
different groups: (1) song types beginning with repeated, low-
frequency, piping elements (Fig. 1) and (2) high-frequency song
types without introductions (Higgins & Peter 2002). Low-frequency
song types were more common in both species (unpublished data),
so only these were considered in this study. Song types were easily
identified by visual inspection of spectrograms (Fig. 1). All record-
ings of shrike-thrushes were made in AugusteOctober 2007.
General Playback Design
We carried out two experiments involving the playback of
imitations of shrike-thrushes (mimetic songs) to territorial shrike-
thrush pairs in Sherbrooke forest from September to October 2007.
Both experiments employed a matched design so that the 15
shrike-thrush pairs in each of two experiments received all treat-
ments. Territorial boundaries were determined by observing
singing shrike-thrushes before experiments. During some play-
backs both sexes responded while in others only one bird respon-
ded. Since we could not always identify each sex at all times during
playbacks we measured the pair’s response to the playback (e.g. the
latency of the first bird to approach the speaker and the combined
song output). We use the term ‘pair’throughout to refer to each
replicate. For each pair, the speaker was set up 1e2 m above the
ground, at least 30 m inside a territorial boundary and within 30 m
of the birds’current location, although birds would sometimes
move further away before an experimented started. Per pair, all
playbacks were presented from the same location and adjacent
locations were no closer than 600 m to reduce the chance that the
same pair was sampled more than once. A 15 m radius was flagged
around the speaker to facilitate measurements of approach
distance (after Vehrencamp et al. 2007). Playbacks were not started
until at least 10 min after any aggressive interactions.
Songs used in playback experiments had high signal to noise
ratios, and were recorded using a Sennheiser ME66 shotgun
microphone and a Marantz PMD670 digital recorder sampling
wave files at 44.1 kHz and 16 bits. Recordings were filtered below
400 Hz in Syrinx (J. Burt, http://www.syrinxpc.com). Playback
tracks were normalized to 90% using Audacity 1.2.6 (http://
audacutt.sourceforge.net). We played back songs at 71 dB at 9 m,
which we found to be the mean natural amplitude of shrike-
thrushes at that distance. We used a ‘Peerless’4-inch, midrange
speaker, attached via an amplifier to a Roland Edirol R-09 HR solid-
state digital player.
Each pair of shrike-thrushes was played recordings from one
individual lyrebird and one shrike-thrush, both of which were
recorded at least 3 km away from the experimental subjects. Pairs
received playbacks of a unique set of individuals. Treatments con-
sisted of a silent, 5 min, preplayback period, a 5 min playback
period and 5 min of silent, postplayback observation. The playback
period contained two sets of identical, 2 min stimuli separated by
1 min of silence. Dictated observations and vocal responses of
shrike-thrushes were recorded using the Marantz recorder and
shotgun microphone described above.
We assessed the response of shrike-thrushes to each playback
treatment by (1) counting the songs sung, (2) determining whether
shrike-thrushes type-matched broadcast songs, and (3) scoring
whether they approached within 15 m of the speaker. To determine
vocal responses we used spectrograms of the sound recordings of
the experiment and scored separately for pre-, during and post-
playback periods. Differential responses to the two experiments
resulted in additional measurements of approach behaviours
specific to each experiment (below).
Experiment 1: Mimic versus Model
We tested whether shrike-thrushes could distinguish between
model and mimicked songs. Each shrike-thrush pair received
a shrike-thrush song sung by a shrike-thrush, a lyrebird imitation of
a shrike-thrush song and a lyrebird imitation of a an eastern
whipbird, Psophodes olivaceus, duet. This last treatment controlled
for any shrike-thrush response to lyrebird mimicry of other species
or to the sounds of ecologically neutral sympatric species. Treat-
ments were played on the same day with at least 10 min between
each playback and began when the pair was within 15e50 m of the
speaker. Each 2 min stimulus contained 13 repetitions of one song,
emulating natural shrike-thrush singing (unpublished data). The
sequence of treatments was balanced between pairs to control for
6
4
2
0.5
(a)
1 1.5 2
6
4
2
0.5
(d)
1 1.5 2
6
4
2
0.5
(e)
1
Time (s)
Frequency (kHz)
1.5 2
6
4
2
0.5
(f)
1 1.5 2
6
4
2
0.5
(b)
1 1.5 2
6
4
2
0.5
(c)
Phrases
1 1.5 2
Figure 1. (aec) Three different grey shrike-thrush song types sung by grey shrike-thrushes and (def) the same song types imitated by superb lyrebirds. Songs are subdivided into
phrases containing repeats of one element type (c).
A. H. Dalziell, R. D. Magrath / Animal Behaviour 83 (2012) 1401e1410 1403
potential order effects. We also measured (4) the shrike-thrush’s
latency to approach within 15 m of the speaker, (5) the latency to
retreat from a 15 m radius circle, (6) the closest approach to the
speaker and (7) the time spent within the circle.
Experiment 2: Song Context
We tested whether the presence of lyrebird song caused shrike-
thrushes to respond differently to playbacks of either their own
species’song or imitations of their song. Each shrike-thrush pair
was played four playback treatments that differed in both the
species singing the song (shrike-thrush or lyrebird) and the
acoustic context of the song (with additional lyrebird song or
alone). Each playback consisted of a clip of one song repeated
approximately every 20 s, reflecting the song rate of shrike-
thrushes and the rate shrike-thrush songs were sung by lyrebirds.
Playbacks presented ‘alone’consisted of songs separated by periods
of silence (Fig. 2a). Playbacks with a ‘lyrebird context’consisted of
songs separated by an approximately 18 s period of lyrebird song
(Fig. 2b): a natural sequence of mimicked sounds of different
species of bird interspersed with vocalizations unique to lyrebirds.
These periods were edited to exclude vocalizations of shrike-
thrushes and extremely loud sounds. The sequence was then
amplified so that the amplitude of the shrike-thrush song was
similar to the majority of sounds in the sequence of mimicry.
The four playbacks were presented using a split-plot design and
were played over 2 days. Day A consisted of the two ‘shrike-thrush’
treatments: a model song presented (1) alone and (2) in a ‘lyrebird
context’. Similarly, day B consisted of the two ‘imitation’treat-
ments. The order of the tracks within days and the order of the
2 days themselves were balanced among pairs. This design limited
the number of days required to complete a set of treatments (and
thus avoiding long gaps between treatments caused by unsuitable
weather) while minimizing potential order effects. Within a day,
tracks were separated by at least 15 min and playbacks of groups
A and B were separated by 1e3 days. Pairs had to be between 25
and 40 m from the speaker before the playback started. In addition
to the three measurements outlined above (see General playback
design) we also measured whether pairs moved at least 5 m
towards the speaker because, unlike the previous experiment,
some shrike-thrushes flew towards the speaker but did not come
within 15 m of it (N¼9 of 60 playbacks).
Quantifying Differences in Vocal Parameters
We quantified acoustic differences between model and mimetic
songs by spectrographic analyses of 38 songs with the highest
signal to noise ratio and with one song from each individual. This
sample included 19 different song types with an example of each
song type from both the mimic and the model. We used this design
because in our sample of recordings shrike-thrushes very rarely
shared song types and there was a large repertoire of shrike-thrush
song types in both model and mimic (72 types sung by shrike-
thrushes of which 25 were imitated by lyrebirds). In two cases,
there were two songs with signal to noise ratios that were indis-
tinguishable and here we used a random number generator to
choose between examples of a particular type.
We used three different approaches to quantify differences
between model and mimetic songs. First, we compared model and
mimetic songs in seven characteristics of the entire song: the
number of different elements, the length of the song, the sound
duration (sum of the duration of all elements), the average element
duration, the number of different element types, the rate of
elements and the peak frequency (the frequency with maximum
power). Second, we compared model and mimetic songs in five
characteristics of element structure including the duration,
maximum frequency, frequency difference, peak frequency and the
RMS amplitude (the root-mean-square amplitude across the
duration of the element). Third, we tested for differences between
the two species in how consistently they produced repeated song
elements. Here we chose a subset of 16 song types in which both
lyrebirds and shrike-thrushes produced at least one repeat of the
first element in the song (e.g. Fig. 1a and d). We then compared the
last two elements in the first phrase (Fig. 1) using the spectro-
graphic cross-correlation analysis (SPCC) to obtain a peak correla-
tion score for each song. We tested whether correlation scores
8
6
4
2
2 4 6 8 10 12 14 16 18 20 22
8
6
4
2
24681012
Time (s)
(a)alone
lyrebird song
(b)
Frequency (kHz)
14 16 18 20 22
Figure 2. Sections from playback stimuli used in experiment 2 on song context. Playback songs were presented in (a) the grey shrike-thrush’s singing style (alone) and (b) the
superb lyrebird’s singing style (lyrebird song context). For the lyrebird song context, each shrike-thrush song (at 1 and 21 s) is interspersed with mimicry and lyrebird-specific
vocalizations. Depicted at the beginning and the end of each sequence is (a) a song sung by a shrike-thrush and (b) an imitation of a shrike-thrush by a lyrebird.
A. H. Dalziell, R. D. Magrath / Animal Behaviour 83 (2012) 1401e1410140 4
differed between species and within song types. SPCC was not used
in the other two analyses because of its unsuitability for analyses
involving sequences of elements and sensitivity to small trans-
positions in frequency (Charif et al. 1995).
All vocal parameters were quantified in Raven Pro 1.3 (Cornell
Laboratory of Ornithology, Ithaca, NY, U.S.A.). Analyses were con-
ducted on a spectrogram display type ‘Blackman’(filter bandwidth:
70.7 Hz; size: 1024 samples; grid resolution: 11.3 ms, 43.1 Hz).
Amplitudes of whole songs were maximized in Syrinx before
analyses. SPCC analysis was performed in Raven using a ‘biased’
normalization procedure, a band-pass filter of 400e10 000 Hz and
a linear power spectrograph of the ‘Blackman’type (filter band-
width: 120 Hz; frame length: 603 samples; grid resolution:
3.42 ms, 43.1 Hz).
Statistical Analyses
Generalized linear models (GLM) using a logit link function and
assuming a binomial distribution of error were constructed for
approach behaviours in both experiments, while a Poisson distri-
bution with a logarithmic link function was used to model the
number of birds that approached the speaker and the number of
songs shrike-thrushes sang during the playbacks. Dispersion
parameters for Poisson distribution models were estimated in
GenStat (Payne 2010a). Final models were chosen by first fitting
a saturated model. We then discarded nonsignificant interaction
terms, followed by other nonsignificant terms unless there was an
a priori reason to retain control variables such as those accounting
for the structure in the data. Since Genstat employs a sequential
sums of squares approach, we built the saturated model by first
entering control variables into the model before the variables of
interest. Control variables included the distance of the closest
shrike-thrush to the speaker before the playback started and the
presentation order of the treatments, and are reported only when
significant. Models of the number of songs sung also controlled for
the number of songs sung during the 5 min before the playback. For
all GLMs, residual plots were used to check that the assumptions of
linear models were met (Grafen & Hails 2002) and variates were
transformed as appropriate.
We employed GLMs instead of GLM mixed models because the
latter failed to converge. Instead, we controlled for repeated
sampling of pairs by incorporating the blocking term ‘pair’into
GLMs as the first term. The split-plot design of the song context
experiment (experiment 2) resulted in nesting of the context of the
playback (lyrebird or alone) in species (shrike-thrush or lyrebird),
which was in turn nested in pair. We controlled for this further level
of structure by fitting an interaction term between ‘pair’and
‘species’before fitting the context of the playback (lyrebird or
alone), and the interaction between context and ‘species’.I
nboth
experiments, we used nonparametric tests if data failed to conform
to diagnostic tests for a normal distribution of residuals.
Differences between the acoustic structures of model and
mimetic songs were analysed using univariate models of individual
parameters. Continuous measures of whole songs were first
modelled using a mixed models approach with species fitted as
afixed effect and ‘song type’modelled as a random term. In several
models, the error of this random term was large, and the variance
was small, relative to the fixed effect. Therefore, we conducted
a likelihood ratio test of the difference in deviance between a model
containing song type and one without (Payne 2010b). If the test
statistic approached zero, song type was dropped from the model
to permit more accurate estimates. The measurement ‘song
complexity’contained just three levels and was thus analysed using
a Wilcoxon signed-ranks test with the data paired according to
song type. For analyses of element structure, the random
components consisted of the phrase within the song nested in song
type to account for the structure in the data (elements, nested
within phrases, nested within song types, Fig. 1).
Performing multiple statistical tests can increase the probability
of falsely rejecting the null hypothesis (Grafen & Hails 2002) but we
were also concerned with type II errors. To balance these two
inversely related errors, we emphasize stronger effects in the
discussion (P<0.01) and report least significant differences (LSD)
and 95% confidence intervals (CI) of effect sizes where possible. In
line with recent theory (Colegrave & Ruxton 2003), our conclusions
regarding the similarity between model and mimetic song rely
most heavily on measurements with smaller confidence intervals of
the effect size (or LSD).
All statistical tests were carried out using GenStat 11.1 (VSN
International, Hemel Hempstead, U.K.) except for Wilcoxon signed-
ranks tests and associated CI, which were calculated using Stats-
Direct 2.7.8 (StatsDirect Ltd, Altrincham, U.K.).
Ethical Note
This research was conducted under permits from the Animal
Experimentation Ethics Committee of the Australian National
University and the Victorian Department of Sustainability and
Environment.
RESULTS
Experiment 1: Mimic versus Model
There was no difference between the behaviour of grey shrike-
thrushes towards mimetic and model songs in any measure of
speaker approach. Shrike-thrushes were equally likely to approach
within 15 m of a speaker broadcasting model songs as mimetic
songs (12/15 pairs in each case), but only 1/15 came within 15 m of
the lyrebird imitation of whipbirds (Fig. 3; GLM:
c
2
¼19.3,
P<0.001). Similarly, the number of birds approaching within 15 m
of the speaker was identical after playback of model and mimetic
songs, but lower following playback of an imitation whipbird duet
1
0.8
0.6 LSD
ST Imitation ST
Pla
y
back stimulus
P(approach within 15 m)
Imitation WB
0.4
0.2
0
Figure 3. Experiment 1, mimic versus model: the proportion of grey shrike-thrush
pairs that approached within 15 m of a speaker broadcasting a song sung by an
actual shrike-thrush (ST), an imitation of an ST song by a superb lyrebird (imitation ST)
or a lyrebird imitating an eastern whipbird duet (imitation WB). Columns show
predicted means generated from the model and the bar indicates the average least
significant difference (LSD) of the means at the 5% level; N¼15 pairs. Statistics are
reported in the text.
A. H. Dalziell, R. D. Magrath / Animal Behaviour 83 (2012) 1401e1410 1405
(GLM: F
2,28
¼13.1, P<0.001; imitation shrike-thrush: mean ¼1.13;
shrike-thrush: mean ¼1.13; imitation whipbird: mean ¼0.13;
average LSD ¼0.53). We did not detect anysubtle differences in the
approach responses of shrike-thrushes to playbacks of imitations or
shrike-thrush songs after excluding the control playback (Table 1),
suggesting that shrike-thrushes did not distinguish between model
and mimetic songs; however, given the broad confidence intervals
for some measurements, there was still room for smaller, unde-
tected differences in responses.
Shrike-thrushes sang a similar number of songs during playback
of mimetic songs as model songs and responded with fewer songs
to playback of an imitation whipbird duet (Fig. 4,Table 2). This
result was independent of an increase in the number of songs when
more birds were singing, and the tendency of some pairs to sing
consistently more than others (Table 2). Shrike-thrushes also
responded with more songs after playback of mimetic and model
songs than after a whipbird duet during the 5 min after the play-
back had ended (Table 2).
The only measure of shrike-thrush response that differed
between model and mimetic songs was that shrike-thrushes were
more likely to type-match model songs (two-tailed binomial test:
H
0
model ¼mimic; model: 7/15; mimic: 1/15; P¼0.013), although
the number of type matches after playback of model sounds varied
greatly (1e36).
Experiment 2: Song Context
Shrike-thrush pairs were less likely to approach a speaker
broadcasting a mimetic song when it was embedded within
a lyrebird’s natural song sequence than when it was presented
alone, but they nearly always approached model songs regardless
of context (Fig. 5,Table 3). The number of birds approaching the
speaker was not affected by the type of playback or any other of the
modelled variables (P>0.1).
Shrike-thrushes sang more songs during playback of model
songs than mimetic songs (model mean ¼22.0 songs; mimic
mean ¼14.1; LSD ¼6.72; Table 4) but there was no effect of the
context of the song (context: P¼0.2; interaction: P¼0.6). Shrike-
thrushes continued to sing more songs during the 5 min after
a playback of model songs than mimetic songs (model mean ¼20.6
songs; mimic mean ¼14.0; LSD ¼6.90; Table 4) after we controlled
for a greater song rate among pairs associated with closer proximity
to the speaker before the beginning of the playback (Table 4). Again,
there was no effect of the context of the song (context: P¼0.6;
interaction: P¼0.1). Unsurprisingly, the total number of songs sung
increased with the number of birds singing in both playback
periods (Table 4). There was no evidence that shrike-thrushes were
more likely to type-match some playback treatments more than
others (species singing the song: models ¼6/15; mimetic ¼2/15;
two-tailed binomial test: H
0
model ¼mimic: P¼0.10; song
context: alone ¼5/15; song sequence ¼4/15; two-tailed binomial
test: H
0
alone ¼song sequence: P¼0.69).
Acoustic Analyses
Of 13 measures of acoustic properties, only three differed
between model and mimetic songs. We detected no differences
between model and mimetic element structures (Tabl e 5) suggesting
that lyrebirds accurately reproduced the structure of elements
within shrike-thrush songs. In addition, spectrographic cross-
correlation analysis showed that lyrebirds produced the elements
in the first phrase of imitations just as consistently as shrike-
thrushes (lyrebird median ¼0.93; shrike-thrush median ¼0.95;
pairwise median difference ¼0.015; 95.6% CI ¼e0.050e0.11; Wil-
coxon signed-ranks test: T¼88, N¼16, P¼0.3). The only differ-
ences between mimetic and model songs were foundin measures of
length (Table 5 ); for example, mimetic songs had fewer elements
than model songs. However, the distribution of the number of
elements in the songs of both species still overlapped (Fig. 6).
Mimetic songs did not differ from model songs in the number of
different types of elements (Wilcoxon signed-ranks test: T¼4,
Table 1
Experiment 1, mimic versus model: paired comparisons of speaker approach
behaviours by grey shrike-thrushes during playback of model and mimetic songs
Dependent variable Test T/t P Mean/median (95% CI)
of difference (shrike-
thrushelyrebird)*
Closest approach (m) T50.5 0.77 0.15 (5.4 e11.6)
Latency to approach (s) T54 0.95 1.44 (149 e56.3)
Time within 15 m (s) t0.57 0.58 80.7 (224 e385)
Latency to approach
within 15 m (s)
t0.25 0.81 16.8 (128 e162)
Latency to retreat
further than 15 m (s)
t0.39 0.70 50.1 (225 e326)
Wilcoxon signed-ranks tests (T) were used instead of paired ttests (t) where data
were not normally distributed. N¼15.
*
Median differences are shown for Wilcoxon signed-ranks tests and approximate
95% CI. Means and exact 95% CI are shown for ttests.
20
LSD
ST Imitation ST
Pla
y
back stimulus
Songs sung
Imitation WB
15
10
5
0
Figure 4. Experiment 1, mimic versus model: the predicted means of the number of
songs sung by grey shrike-thrushes during the 5 min playback period in response to
playback of a song sung by an actual shrike-thrush (ST), an imitation of a ST song by
a superb lyrebird (imitation ST) or a lyrebird imitating an eastern whipbird duet
(imitation WB). The bar indicates the average least significant difference (LSD) of the
means at the 5% level; N¼15 pairs. Statistics are shown in Table 2.
Table 2
Experiment 1, mimic versus model: significant effects on the number of songs grey
shrike-thrushes sang in response to playback of their own song, an imitation of their
song by a superb lyrebird and an imitation of an eastern whipbird duet by a superb
lyrebird (‘playback type’)
Dependent variable Effects df Approximate FP
Songs during
5 min playback
Whole model 17, 27 5.43 <0.001
Pair 14, 27 4.64 <0.001
Playback type*2, 27 8.93 0.001
No. of birds singing 1, 27 9.47 0.005
Songs 5 min
after playback
Whole model 16, 28 5.20 <0.001
Pair 14, 28 5.04 <0.001
Playback type 2, 28 6.29 0.006
*
Model predictions are shown in Fig. 4.
A. H. Dalziell, R. D. Magrath / Animal Behaviour 83 (2012) 1401e1410140 6
N¼19, P¼0.75) showing that lyrebirds maintained the same song
structure as models but sang fewer repetitions of each element type
(Fig. 1).
DISCUSSION
Our playback experiments and sound analyses show that superb
lyrebirds are accurate mimics of grey shrike-thrush songs. Indeed,
their mimetic songs were accurate enough to prompt grey shrike-
thrushes to approach the playback speaker in most circum-
stances. However, there were subtle differences between the
mimetic and model songs that models responded to, particularly
once alerted to the presence of a lyrebird.
Accuracy, Signal Context and Model Response
Shrike-thrush pairs responded strongly towards both mimetic
and model songs when both songs were presented in isolation
(experiment 1: mimic versus model). We detected no differences
between the behaviour of shrike-thrushes towards mimetic lyre-
bird songs and those sung by real shrike-thrushes in any measure of
speaker approach behaviour or in singing effort. Most strikingly,
measures of approach within 15 m of broadcasts of mimic and
conspecific song were identical, and birds sang an almost identical
number of songs. The response to mimetic song was not a response
to lyrebird mimicry in general, because shrike-thrushes rarely
approached during control playbacks of lyrebird mimicry of
another species, and in that situation their song rate was half that to
lyrebird mimicry of shrike-thrushes.
Experiments using a similar bioassay to test the accuracy of
mimicry of northern mockingbirds found differences in response
according to model sex and species. Male red-winged blackbirds,
Agelaius phoeniceus (Brenowitz 1982), but not females (Searcy &
Brenowitz 1988), reacted equally strongly to model and mimetic
songs. However, these two studies measured different responses by
each sex and played back songs from onlyone mockingbird, making
it difficult to draw general conclusions. A third, well-replicated
study found that Florida scrub-jays, Aphelocoma coerulescens,
responded less strongly to mockingbird imitations of their terri-
torial ‘weep’calls than towards conspecific weep calls (Owen-
Ashley et al. 2002). These studies suggest that either mimetic
accuracy varies with the model or that discriminatory ability differs
between models.
Our second experiment showed that shrike-thrushes were more
likely to approach a speaker broadcasting mimetic songs when they
were presented alone, than together with lyrebird song containing
1
0.8
0.6
Song context
Alone
Lyrebird
LSD
0.4
0.2
P(flight towards speaker)
Imitation STShrike-thrush
(a)
0
1
0.8
0.6
Song context
Alone
Lyrebird
LSD
0.4
0.2
P(approach within 15 m)
Imitation ST
Pla
y
back stimulus
Shrike-thrush
(b)
0
Figure 5. Experiment 2, song context: the proportion of grey shrike-thrush pairs that
(a) flew at least 5 m towards the speaker or (b) approached within 15 m of a speaker,
broadcasting a song sung by an actual shrike-thrush (shrike-thrush) or an imitation of
a shrike-thrush song by a superb lyrebird (imitation ST). Each song was either
embedded in a stream of mimicry sung by a lyrebird (lyrebird song context) or
presented in isolation (alone). Columns show model predictions and the average
least-significant difference (LSD) of the means at the 5% level are shown; N¼15 pairs.
Statistics are shown in Table 3.
Table 3
Experiment 2, song context: final generalized linear models of grey shrike-thrush
approach behaviours in response to playback of shrike-thrush song sung by either
a superb lyrebird or a shrike-thrush (‘species’), embedded in either a period of
silence or a sequence of lyrebird song (‘context’)
Dependent variable Effects df Approximate
c
2
P
Flight 5m
towards the
speaker
Whole model 31 1.78 0.005
Pair 14 1.68 0.051
Species 1 7.83 0.005
Pair*Species 14 0.85 0.615
Context 1 2.66 0.103
Context*Speciesy1 9.11 0.003
Approach within
15 m of the
speaker
Whole model 31 2.54 <0.001
Pair 14 2.12 0.008
Species 1 7.99 0.005
Pair*Species 14 1.75 0.039
Context 1 w0.00 <1
Context*Speciesy1 7.76 0.005
The dependent variable was dichotomous and the experiment had a split-plot
design.
y
Model predictions are shown in Fig. 5.
Table 4
Experiment 2, song context: final generalized linear models of the number of songs
grey shrike-thrushes sang in response to playback of shrike-thrush song sung by
either a superb lyrebird or a shrike-thrush (‘species’), embedded in either a period of
silence or a sequence of lyrebird song (‘context’)
Dependent variable Effects df Approximate FP
Songs
during 5 min
playback
Whole model 16, 43 2.75 0.004
Pair 14, 43 1.97 0.045
Species*1, 43 5.59 0.023
Number of birds
singing
1, 43 11.19 0.002
Songs 5 min
after playback
Whole model 17, 42 2.54 0.007
Pair 14, 42 1.64 0.106
Initial distance
from speaker
1, 42 7.47 0.011
Species*1, 42 6.01 0.021
Number of birds
singing
1, 42 7.33 0.010
*
Model predictions are reported in the text.
A. H. Dalziell, R. D. Magrath / Animal Behaviour 83 (2012) 1401e1410 1407
imitations of other species. Discrimination was not simply based on
the presence of lyrebird song: shrike-thrushes more often
approached model than mimetic songs when both were presented
with lyrebird song. However, even when mimetic songs were
presented with lyrebird song, shrike-thrushes still approached 50%
of the time, suggesting that discriminating between mimic and
model continued to be difficult.
Two further results from our experiments suggest that while
shrike-thrushes responded strongly towards their songs sung by
both conspecifics and lyrebirds, their reaction was more intense
towards their own song. First, in the context experiment (experi-
ment 2) shrike-thrushes sang fewer songs in response to playbacks
of imitations than to model songs. This effect was not present in the
first experiment, possibly because many pairs had started incu-
bating by the time the second experiment was conducted, forcing
birds to respond strongly only to the greatest threats. Second,
shrike-thrushes were more likely to song type-match a playback of
a song sung by a conspecific than an imitation, although this effect
was only present in experiment 1. Song type matching can be
a more aggressive signal than singing a nonmatching song type
(Molles & Vehrencamp 2001;Beecher & Campbell 2005), again
suggesting that shrike-thrushes responded marginally more
strongly towards playback of models. Importantly, these effects
were subtle and the presence of lyrebird song was the only factor
influencing whether or not a shrike-thrush approached a playback
of an imitation.
That shrike-thrushes are more discriminating when mimicry
was presented with a sequence of lyrebird song shows that shrike-
thrushes have a flexible acceptance threshold and that they inte-
grate cues to distinguish model from mimetic songs. Traditional
theory argues that it is impossible to reduce the risk of responding
to a false signal, such as an imitation, without simultaneously
increasing the risk of failing to respond to a real signal, and vice
versa (Wiley 1994). However, animals can mitigate the costs of
mistakenly responding to a mimic by varying their threshold
criteria for signal identification (acceptance) depending on other
cues (‘flexible acceptance thresholds’:Reeve 1989). For example,
several host species of brood parasites are more likely to reject
unusual eggs if they have recently been exposed to a cuckoo (Davies
& Brooke 1988;Moksnes & Røskaft 1989;Davies et al. 1996;Bártol
et al. 2002), or had additional evidence that cuckoos were present
(Hauber et al. 2006;Stokke et al. 2008;Langmore et al. 2009). In
our study, shrike-thrushes appeared to use the context of the signal
to decide whether or not to approach playback of a mimetic song.
Contrary to the predictions of signal detection theory, raising their
acceptance threshold did not simultaneously reduce the chance
they would fail to respond to their own song, although further
experiments are required to explore this hypothesis. To our
knowledge, only one other study has formally tested the use of
context by model species in a vocal mimicry system (Owen-Ashley
et al. 2002). However, this study found that models responded
weakly to mimetic calls regardless of context. In contrast, our study
adds to the growing literature showing that signal acceptance
thresholds are plastic and context dependent.
Acoustic Differences Between Model and Mimetic Songs
Our comparison between the acoustic parameters of mimicked
and model songs revealed few differences. Lyrebirds accurately
reproduced the temporal patterning of elements that characterize
shrike-thrush songs. This is an unusual characteristic in birds that
10
8
4
6
2
0546
Number of elements in son
g
Lyrebird imitation
Shrike-thrush
Number of songs
78
Figure 6. The number of songs with different numbers of elements sung by both grey
shrike-thrushes and superb lyrebirds. The sample contained one song sung by a lyre-
bird and one by a shrike-thrush for 19 different song types. Statistics are shown in
Table 5.
Table 5
Differences between the acoustic structures of model and mimetic songs
Effects Estimate SED LSD df F P
Songs
Number of elements*1.2105 0.3613 0.7326 1, 36.0 11.23 0.002
Song length*0.3374 0.1072 0.2173 1, 36.0 9.91 0.003
Sound duration 0.14184 0.04958 0.1042 1, 18.0 8.19 0.010
Average element length 0.004723 0.005139 0.01080 1, 18.0 0.84 0.37
Peak frequencyy0.005280 0.008055 0.01692 1, 18.0 0.43 0.52
Element rate*0.05544 0.1660 0.3367 1, 36.0 0.11 0.74
Elements
Frequency range 119 76.4 150.7 1, 176.6 2.43 0.121
Peak frequencyy,z1.975 58.49 0.05412 1, 179.0 w0 0.973
Maximum frequency 55.7 99.06 195.5 1,177.3 0.32 0.575
RMS amplitudex501.6 444.9 877.8 1,181.9 1.27 0.261
Duration 0.00267 0.006125 0.01209 1, 178.5 0.19 0.663
Parameters were modelled in univariate linear mixed models or linear models with species as a fixed effect. Song-level analyses contained ‘Song type’modelled as a random
term unless the variance of this component approximated zero, in which case Song type was dropped from the model (Payne et al. 2010). For element-level analyses, the
phrase within the song, nested in song type, was modelled as a random term. Estimates of the differences between the means (lyrebirdeshrike-thrush), the standard error of
difference between the means (SED) and the approximate least significant differences (LSD) at the 5% level are shown.
*
These models were run without the variable ‘song type’as a random effect.
y
This variable was log transformed to normalize the residuals.
z
Frequency with maximum power.
x
Root-mean square of the amplitude across the duration of the element.
A. H. Dalziell, R. D. Magrath / Animal Behaviour 83 (2012) 1401e14101408
imitate many different sounds (versatile avian mimics), which
more commonly imitate fragments of song or short vocalizations
(Hindmarsh 1984;Ferguson et al. 2002). In addition, lyrebird
reproductions of individual elements within the model song were
indistinguishable from model elements. However, lyrebirds may
not be able to imitate all model elements as accurately as shrike-
thrush songs, as lyrebirds are less accurate mimics of the ‘whip’
element in the eastern whipbird song, perhaps because of the large
and rapid frequency change (Zann & Dunstan 2008).
In our study, mimetic and model songs differed only in the
number of elements (Table 5), causing mimetic songs to be on
average the shorter of the two. Lyrebirds abridged model songs by
reducing the number of repetitions of particular element types
while conserving the order and number of different element types
and thus the structural complexity of the model song (Fig. 1).
This suggests that differences between mimic and model were
unlikely to be caused by limitations of the lyrebird’s auditory
system. Instead, our results support previous speculation that
lyrebirds are not passive imitators of common sounds (Higgins et al.
2001;Zann & Dunstan 2008). Rather, lyrebirds appear either to
imitate selectively shorter examples of model songs or to abridge
the songs they hear. In our study, shrike-thrushes might have used
the differences in element number to distinguish between model
and mimetic songs, but confirmation requires further experiments.
Imitations of shrike-thrush songs are not the only imitations
that lyrebirds appear to shorten. Zann & Dunstan (2008) showed
that lyrebird imitations of male whipbird song had a shorter
introductory ‘tone’element than model songs, and lyrebirds appear
to truncate imitations of the vocalizations of other species such as
the ‘laugh’calls of laughing kookaburras, Dacelo novaeguineae
(Smith 1988, A. H. Dalziell, personal observation). Such shortening
of mimicked sounds may be a feature common to versatile avian
mimics. For example, imitations by northern mockingbirds of the
‘weep’calls produced by Florida scrub-jays were shorter in element
duration than the model species itself (Owen-Ashley et al. 2002).
Why are Lyrebird Imitations Shorter Than Model Songs?
Lyrebirds might sing shorter versions of model songs as a result
of selection pressures imposed by a conspecific receiver. Short
imitations may arise if male lyrebirds need to balance two
requirements: to sing highly accurate imitations, but also to
demonstrate their versatility to a conspecific receiver who may
listen for only a short time. This is a modification of the ‘Jack of all
trades’hypothesis, which predicts that mimetic accuracy is traded
off against mimetic versatility (Edmunds 2000;Zollinger & Suthers
2004). However, rather than being the result of limitations of the
vocal apparatus of the mimic (Zollinger & Suthers 2004), differ-
ences in the length of model and mimetic sounds may reflect
conflicting signal requirements. Alternatively, male lyrebirds may
sing shorter songs to facilitate discrimination by conspecifics
between lyrebirds and model species, although it seems plausible
that visual cues or other acoustic cues, such as song context, should
be sufficient for discrimination. Further research is required to
determine why lyrebirds abridge model songs and whether it is
a widespread strategy among versatile mimics.
Mimetic Accuracy in Nondeceptive Mimicry Systems
Our bioassays combined with acoustic analysis revealed
remarkably accurate mimicry of a complex auditory signal by
a nondeceptive mimic. Given the motor challenges of producing
song in birds (Gil & Gahr 2002;Catchpole & Slater 20 08;Podos et al.
2009), our results are consistent with selection for accurate
mimicry in male lyrebirds and suggest that there is selection for
discrimination in receivers. Future studies will need to conduct
playback experiments to other lyrebirds to confirm whether
conspecifics are the receivers selecting for accurate mimicry. In
nondeceptive mimicry systems the conspecific receiver conceivably
has a more complex cognitive task than a heterospecific receiver.
This is because the conspecific receiver not only has to discriminate
between model and mimetic vocalizations, but it also has to
discriminate between mimics on the basis of their mimetic ability.
Overall, our results highlight the complexity of adaptations evolved
by mimics and the value of examining the similarities and differ-
ences between model and mimetic signals.
Acknowledgments
We are grateful to Alexandra Dorland, Alex Maisey, Simon
Despoja, Sarah Stuart-Smith and Pamela Fallow for field assistance.
We thank Jan Incoll, Norm Carter and the Sherbrooke Lyrebird
Study Group for advice and permission to study lyrebirds in Sher-
brooke Forest. Parks Victoria granted us permission to work in the
Dandenong Ranges National Park and provided invaluable logistical
assistance. Robert Phillips and Bill Incoll helped with the equip-
ment. We also thank Emlyn Williams for assistance with the
statistics, the late Richard Zann for advice, Michelle Hall, Andrew
Cockburn and Naomi Langmore for equipment loans and
comments that improved the experimental design and the manu-
script. Anja Skroblin, Branislav Igic, Elsie Krebs, David Greene, Justin
Welbergen, Pamela Fallow and three anonymous referees also
provided helpful comments on the manuscript. This study was
supported by two Birds Australia Stuart Leslie Awards to A.H.D., an
Australian Geographic Society Research Award to A.H.D., an
Australian Postgraduate Award to A.H.D., an Australian National
University College of Science Scholarship to A.H.D. and an ARC
Discovery Grant to R.D.M.
References
Attenborough, D. 2002. Signals and songs. Episode 2. In: The Life of Birds. Bristol:
BBC Natural History Unit [DVD].
Bártol, I., Karcza, Z., Moskát, C., Røskaft, E. & Kisbenedek, T. 2002. Responses of
great reed warblers, Acrocephalus arundinaceus, to experimental brood para-
sitism: the effects of a cuckoo, Cuculus canorus, dummy and egg mimicry.
Journal of Avian Biology,33, 420e425.
Bates, H. W. 1862. Contributions to an insect fauna of the Amazon Valley. Lepi-
doptera: Heliconidae. Transactions of the Linnean Society of London,23,
495e566.
Baylis, J. R. 1982. Avian vocal mimicry: its function and evolution. In: Acoustic
Communication in Birds (Ed. by D. E. Kroodsma, E. H. Miller & H. Ouellet), pp.
51e83. New York: Academic Press.
Beecher, M. D. & Campbell, S. E. 2005. The role of unshared songs in singing
interactions between neighbouring song sparrows. Animal Behaviour,70,
1297e1304.
Blumenrath, S. H., Dabelsteen, T. & Pedersen, S. B. 2007. Vocal neighbour-mate
discrimination in female great tits despite high song similarity. Animal Behav-
iour,73, 789e796.
Botero, C. A., Riveros, J. M. & Vehrencamp, S. L. 2007. Relative threat and recog-
nition ability in the responses of tropical mockingbirds to song playback. Animal
Behaviour,73,661e669.
Brenowitz, E. A. 1982. Aggressive response of red-winged blackbirds to mocking-
bird song imitation. Auk,99, 584e586.
Catchpole, C. K. & Slater, P. J. B. 20 08. Bird Song: Biological Themes and Variations.
2nd edn. Cambridge: Cambridge University Press.
Charif, R. A., Mitchell, S. & Clark, C. W. 1995. Canary 1.2 User’s Manual. Ithaca, New
York: Cornell Laboratory of Ornithology.
Cheney, K. L. 2010. Multiple selective pressures apply to a coral reef fish mimic:
a case of Batesian-aggressive mimicry. Proceedings of the Royal Society B,277,
1849e1855.
Cheney, K. L. & Marshall, N. J. 2009. Mimicry in coral reef fish: how accurate is this
deception in terms of color and luminance? Behavioral Ecology,20, 459e468.
Chisholm, A. G. 1932. Vocal mimicry among Australian birds. Ibis,2, 605e624.
Chittka, L. & Osorio, D. 2007. Cognitive dimensions of predator responses to
imperfect mimicry. PLoS Biology,5, e339.
Colegrave, N. & Ruxton, G. D. 2003. Confidence intervals are a more useful
complement to nonsignificant tests than are power calculations. Behavioral
Ecology,14,446e447.
A. H. Dalziell, R. D. Magrath / Animal Behaviour 83 (2012) 1401e1410 1409
Coleman, S. W., Patricelli, G. L., Coyle, B., Siani, J. & Borgia, G. 2007. Female
preferences drive the evolution of mimetic accuracy in male sexual displays.
Biology Letters,3, 463e466.
Davies, N. B. & Brooke, M. d. L. 1988. Cuckoos versus reed warblers: adaptations
and counteradaptations. Animal Behaviour,36, 262e284.
Davies, N. B., Brooke, M. D. L. & Kacelnik, A. 1996. Recognition errors and prob-
ability of parasitism determine whether reed warblers should accept or reject
mimetic cuckoo eggs. Proceedings of the Royal Society B,263, 925e931.
Dobkin, D. S. 1979. Functional and evolutionary relationships of vocal copying
phenomena in birds. Zeitschrift für Tierpsychologie,50,348e363.
Dominey, W. J. 1980. Female mimicry in male bluegill sunfish: a genetic poly-
morphism? Nature,284, 546e548.
Edmunds, M. 2000. Why are there good and poor mimics? Biological Journal of the
Linnean Society,70, 459e466.
Estrada, C. & Jiggins, C. D. 2008. Interspecific sexual attraction because of conver-
gence in warning colouration: is there a conflict between natural and sexual
selectio n in mimet ic species? Journal of Evolutionary Biology,21,749e760.
Ferguson, J. W. H., van Zyl, A. & Delport, K. 2002. Vocal mimicry in African Cos-
sypha robin chats. Journal für Ornithologie,143 ,319e330.
Flower, T. 2011. Fork-tailed drongos use deceptive mimicked alarm calls to steal
food. Proceedings of the Royal Society B,278, 1548e1555.
Garamszegi, L. Z., Eens, M., Pavlova, D. Z., Aviles, J. & Møller, A. P. 2007.
A comparative study of the function of heterospecific vocal mimicry in Euro-
pean passerines. Behavioral Ecology,18,1001e1009.
Gil, D. & Gahr, M. 2002. The honesty of bird song: multiple constraints for multiple
traits. Trends in Ecology & Evolution,17,133e141.
Goodale, E. & Kotagama, S. W. 2006a. Context-dependent vocal mimicry in
a passerine bird. Proceedings of the Royal Society B,273, 875e880.
Goodale, E. & Kotagama, S. W. 2006b. Vocal mimicry by a passerine bird attracts
other species involved in mixed-species flocks. Animal Behaviour,72,471e477.
Grafen, A. & Hails, R. 2002. Modern Statistics for the Life Sciences. Oxford: Oxford
University Press.
Gross, M. R. 1996. Alternative reproductive strategies and tactics: diversity within
sexes. Trends in Ecology & Evolution,11,92e98.
Harper, G. R. & Pfennig, D. W. 2007. Mimicry on the edge: why do mimics vary in
resemblance to their model in different parts of their geographical range?
Proceedings of the Royal Society B,274, 1955e19 61.
Hauber, M. E., Moskát, C. & Bán, M. 2006. Experimental shift in hosts’acceptance
threshold of inaccurate-mimic brood parasite eggs. Biology Letters,2,177e180.
Higgins, P. J. & Peter, J. M. 2002. Pardalotes to Shrike-thrushes. Handbook of
Australian, New Zealand & Antarctic Birds. Melbourne: Oxford University Press.
Higgins, P. J., Peter, J. M. & Steele, W. K. 2001. Tyrant-flycatchers to Chats. Handbook
of Australian, New Zealand & Antarctic Birds. Melbourne: Oxford University
Press.
Hindmarsh, A. M. 1984. Vocal mimicry in starlings. Behaviour,90, 302e324.
Hindmarsh, A. M. 1986. The functional-significance of vocal mimicry in song.
Behaviour,99,87e10 0.
Holen, Ø. H. & Johnstone, R. A. 2004. The evolution of mimicry under constraints.
American Naturalist,164, 598e613.
Howard, R. D. 19 74. I nfluence of sexual selection and interspecific competition on
mockingbird song Mimus polyglottos.Evolution,28, 428e438.
Illes, A. E., Hall, M. L. & Vehrencamp, S. L. 2006. Vocal performance influences
male receiver response in the banded wren. Proceedings of the Royal Society B,
273, 1907e1912.
Johnstone, R. A. 2002. The evolution of inaccurate mimics. Nature,418, 524e526.
Kelley, L. A., Coe, R. L., Madden, J. R. & Healy, S. D. 2008. Vocal mimicry in
songbirds. Animal Behaviour,76,521e528.
Kenyon, R. F. 1972. Polygyny among superb lyrebirds in Sherbrooke forest park,
Kallista, Victoria. Emu,72,70e76.
Kikuchi, D. W. & Pfennig, D. W. 2010. Predator cognition permits imperfect coral
snake mimicry. American Naturalist,176, 830e834.
de Kort, S. R., Eldermire, E. R. B., Cramer, E. R. A. & Vehrencamp, S. L. 2009. The
deterrent effect of bird song in territory defense.Behavioral Ecology,20, 200e206.
Kroodsma, D. E. 2004. The diversity and plasticity of birdsong. In: Nature’s Music:
the Science of Birdsong (Ed. by P. Marler & H. Slabbekoorn), pp. 108e131.
Amsterdam: Elsevier Academic Press.
Langmore, N. E., Hunt, S. & Kilner, R. M. 2003. Escalation of a coevolutionary arms
race through host rejection of brood parasitic young. Nature,422,157e160.
Langmore, N. E., Maurer, G., Adcock, G. J. & Kilner, R. M. 2008. Socially acquired
host-specific mimicry and the evolution of host races in Horsfield’s bronze-
cuckoo Chalcites basalis.Evolution,62, 1689e1699.
Langmore, N. E., Cockburn, A., Russell, A. F. & Kilner, R. M. 2009. Flexible cuckoo
chick-rejection rules in the superb fairy-wren. Behavioral Ecology,20,978e984.
Lemaire, F. 1975. Le chant de la rousserolle verderolle (Acrocephalus palustris):
fidélité des imitations et relations avec les espèces imitées et avec les con-
génères. Gerfaut,65,3e28.
Lill, A. 1979. Assessment of male parental investment and pair bonding in the
polygamous superb lyrebird. Auk,96, 489e498.
McGuire, L., Van Gossum, H., Beirinckx, K. &Sherratt, T. N. 2006. An empirical test
of signal detection theory as it applies to Batesian mimicry. Behavioural
Processes,73, 299e307.
Moksnes, A. & Røskaft, E. 1989. Adaptations of meadow pipits to parasitism by the
common cuckoo. Behavioral Ecology and Sociobiology,24,25e30.
Molles, L. E. & Vehrencamp, S. L. 2001. Songbird cheaters pay a retaliation cost:
evidence for auditory conventional signals. Proceedings of the Royal Society B,
268,2013e2019.
Müller, F. 1878. On the advantages of mimicry in butterflies. Zoologischer Anzeiger,1,
54e55.
Nicholls, J. A. 2008. Site specificity in advertisement calls and responses to play-
backs of local and foreign call variants in satin bowerbirds, Ptilonorhynchus
violaceus.Behavioral Ecology and Sociobiology,62,831e841.
Norman, M. D., Finn, J. & Tregenza, T. 1999. Female impersonation as an alter-
native reproductive strategy in giant cuttlefish. Proceedings of the Royal Society
B,266,1347e1349.
Owen-Ashley, N. T., Schoech, S. J. & Mumme, R. L. 2002. Context-specific response
of Florida scrub-jay pairs to northern mockingbird vocal mimicry. The Condor,
104, 858e865.
Payne, R. 2010a. A Guide to Regression, Nonlinear and Generalized Linear Models in
GenStat. Hemel Hempstead: VSN International.
Payne, R. W. E. 2010b. Guide to GenStat. Part 2: Statistics. Hemel Hempstead: VSN
International.
Payne, R., Welham, S. & Harding, S. 2010. A Guide to REML in GenStat. Hemel
Hempstead: VSN International.
Pekár, S., Jarab, M., Fromhage, L. & Herberstein, M. E. 2011. Is the evolution of
inaccurate mimicry a result of selection by a suite of predators? A case study
using Myrmecomorphic spiders. American Naturalist,178 ,124e134.
Podos, J., Lahti, D. C. & Moseley, D. L. 2009. Vocal performance and sensorimotor
learning in songbirds. Advances in the Study of Behavior,40,159e195.
Reeve, H. K. 1989. The evolution of conspecific acceptance thresholds. American
Naturalist,133,407e435.
Robinson, F. N. & Curtis, H. S.1996. The vocal displays of the lyrebirds (Menuridae).
Emu,96, 258e275.
Ruxton, G. D. & Speed, M. P. 2005. Evolution: a taste for mimicry. Nature,433,
205e207.
Searcy, W. A. & Brenowitz, E. A. 1988. Sexual differences in species recognition of
avian song. Nature,332,152e154.
Sherratt, T. N. 2002. The evolution of imperfect mimicry. Behavioral Ecology,13,
821e826.
Smith, L. H. 1988. The Life of the Lyrebird. Melbourne: Heinemann.
Starrett, A. 1993. Adaptive resemblance: unifying concept for mimicry and crypsis.
Biological Journal of the Linnean Society,48, 299e317.
Stoddard, M. C. & Stevens, M. 2010. Pattern mimicry of host eggs by the common
cuckoo, as seen through a bird’seye.Proceedings of the Royal Society B,277,
1387e1393.
Stoddard, P. K. 1996. Vocal recognition in territorial passerines. In: Ecology and
Evolution of Acoustic Communication in Birds (Ed. by D. E. Kroodsma &
E. H. Miller), pp. 356e374. Ithaca, New York: Cornell University Press.
Stokke, B. G., Hafstad, I., Rudolfsen, G., Moksnes, A., Moller, A. P., Roskaft, E. &
Soler, M. 2008. Predictors of resistance to brood parasitism within and among
reed warbler populations. Behavioral Ecology,19,612e620.
Vane-Wright, R. I. 1980. On the definition of mimicry. Biological Journal of the
Linnean Society,13,1e6.
Vehrencamp, S. L., Hall, M. L., Bohman, E. R., Depeine, C. D. & Dalziell, A. H. 2007.
Song matching, overlapping, and switching in the banded wren: the sender’s
perspective. Behavioral Ecology,18, 849e859.
Vereecken, N. J. & McNeil, J. N. 2010. Cheaters and liars: chemical mimicry at its
finest. Canadian Journal of Zoology,88,725e752.
Whiting, M. J., Webb, J. K. & Keogh, J. S. 2009. Flat lizard female mimics use sexual
deception in visual but not chemical signals. Proceedings of the Royal Society B,
276, 1585e1591.
Wiley, R. H. 1994. Errors, exaggeration, and deception in animal communication. In:
Behavioral Mechanisms in Evolutionary Ecology (Ed. by L. Real), pp. 157e189.
Chicago: University of Chicago Press.
Zann, R. & Dunstan, E. 2008. Mimetic song in superb lyrebirds: species mimicked
and mimetic accuracy in different populations and age classes. Animal Behav-
iour,76, 1043e1054.
Zollinger, S. A. & Suthers, R. A. 2004. Motor mechanisms of a vocal mimic:
implications for birdsong production. Proceedings of the Royal Society B,271,
483e491.
Zoos South Australia 2009. Superb Lyrebird Imitating Construction Work. Adelaide Zoo
online video. http://www.youtube.com/watch?v¼WeQjkQpeJwY. Accessed: 11
August 2011.
A. H. Dalziell, R. D. Magrath / Animal Behaviour 83 (2012) 1401e14101410
