ANIMAL BEHAVIOUR, 2003, 66, 893–902

doi:10.1006/anbe.2003.2250

The energy cost of song in the canary, Serinus canaria

SALLY WARD*, JOHN R. SPEAKMAN† & PETER J. B. SLATER*

*School of Biology, University of St Andrews
†School of Biological Sciences, University of Aberdeen and Rowett Research Institute

(Received 20 November 2002; initial acceptance 6 January 2003;

final acceptance 26 February 2003; MS. number: 7535)

Although sound production requires energy, it has been unclear how much singing increases metabolic
rate in passerine birds. We measured the rate of oxygen consumption of two breeds of canary that sang
inside a respirometry chamber. Metabolic rate increased with the proportion of time that birds spent
singing. Average metabolic rate during singing at 15–20C was 1.05–1.07 times that of standing quietly in
the same temperature range or 2.2–2.6 times basal metabolic rate (BMR). Whether an increase in
metabolic rate during song of this order would represent a fitness cost to free-living passerine birds would
depend upon the circumstances. Singing rather than perching during the day would raise metabolic rate
only slightly. Singing at night or at dawn, instead of sleeping with a metabolic rate closer to BMR, would
cause a greater increase in metabolism. Birdsong could act as a condition-dependent signal, since birds
that are easily able to achieve energy balance could afford the cost of singing, but those close to their
energy limits might not.
 2003 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.

Birdsong is a conspicuous example of a display that
can function both to deter rivals and to attract mates
(Catchpole & Slater 1995). The possible costs of birdsong
are not well understood. Most models of animal signal-
ling assume that displays should be costly if they are to
carry honest information about the quality of the signal-
ler (Zahavi 1975; Clutton-Brock & Albon 1979; Grafen
1990a, b; Godfray 1991). Hypotheses on how animal
displays may have evolved assume explicitly or implicitly
that signals should be costly (Fisher 1930; Zahavi 1975;
Grafen 1990a, b). The costliness of signals is also an
important assumption of stochastic dynamic program-
ming models of animal signalling (e.g. Hutchinson et al.
1993).
Song production obviously requires energy. Muscles
associated with control of respiration, the syrinx, the
upper vocal tract and movements of the bill are all used
during singing (Goller & Larsen 1997; Suthers et al. 1999;
Larsen & Goller 2002). Singing could also be costly in
other ways: time spent singing is not available for other
activities, singing might attract predators, and song learn-
ing and production involves specialized areas of the brain
that could be costly to develop or maintain (Gil & Gahr
Correspondence: S. Ward, School of Biology, Bute Medical Buildings,
University of St Andrews, St Andrews, Fife KY16 9TS, U.K. (email:
sw29@st-andrews.ac.uk). J. R. Speakman is at the Aberdeen Centre for
Energy Regulation and Obesity, School of Biological Sciences, Zoology
Building, Tillydrone Avenue, Aberdeen B24 2TZ, U.K. and Division of
Appetite and Energy Balance, Rowett Research Institute, Greenburn
Road, Bucksburn, Aberdeen, AB21 9SB, U.K.
893
0003–3472/03/$30.00/0  2003 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.
2001). Field observations of birds imply that singing has
an energy cost. Birds sing more when in good body
condition or following supplementary feeding and less
after cold nights or following adverse experimental treat-
ments (Lambrechts 1996). However, these data do not
show whether singing itself is energetically costly or
whether time that could have been spent singing, at
however low a rate of energy expenditure, was better
spent foraging during periods of energy shortage.
Measurements of the rate of oxygen consumption of
captive animals show that sound production can cause
large factorial increases in metabolism for insects,
anurans and bats (Ryan 1988; Speakman et al. 1989;
Prestwich 1994; Kotiaho et al. 1998). However, one can-
not infer from these data how energetically costly singing
should be for passerine birds, because resting metabolic
rates and sound production mechanisms differ between
taxa. Singing inside respirometry chambers appears to be
energetically costly in Carolina wrens, Thryothorus ludovi-
cianus (Eberhardt 1994), but not in zebra finches, Taen-
iopygia guttata, Waterslager canaries, Serinus canaria, or
European starlings, Sturnus vulgaris (Oberweger & Goller
2001; Franz & Goller 2003). Singing by free-living com-
mon nightingales, Luscinia megarhynchos, increases their
overnight rates of mass loss, implying that song has an
energy cost in this species (Thomas 2002). In contrast,
begging by nestling birds is thought to be energetically
cheap (Chappell & Bachman 2002) as is crowing by
cockerels, Gallus gallus domesticus, and junglefowl, G.
gallus spadiceus (Chappell et al. 1995; Horn et al. 1995).
894 ANIMAL BEHAVIOUR, 66, 5
Conclusions that bird vocalization is energetically
costly (Eberhardt 1994) and that it is energetically cheap
(McCarty 1996) have both been questioned (Gaunt et al.
1996; Verhulst & Wiersma 1997; Weathers et al. 1997).
Whether singing is energetically costly or not therefore
remains controversial, both because some of the empiri-
cal data are contradictory and because it is unclear how
much energy expenditure must increase during an
activity before this will impose a fitness cost. Verhulst &
Wiersma (1997) proposed that whether a particular
activity is energetically costly should best be considered
by calculating its marginal cost, that is, how much of the
available energy that activity requires. A problem with
this approach, however, is that it is unclear how to
determine how much energy is available to an animal.
Previous studies of the energy cost of singing inside
respirometry chambers have not included measurement
of a possible anaerobic component to metabolism during
vocalization (Weathers et al. 1997), for which the oxygen
debt would be repaid by increased aerobic metabolism
between songs. Nor has it been assessed whether birds
sang as loudly inside respirometry chambers as when they
were not enclosed. It is important that song volume is not
reduced inside respirometry chambers, since louder
sounds contain more energy. The energy cost of singing
was found to increase with song amplitude in a single
European starling that sang both quiet and loud songs
(Oberweger & Goller 2001).
We measured the energy cost of singing by two breeds
of canary (roller and Fife Fancy; body masses given in
Table 1) from the rate of oxygen consumption of birds
singing inside a respirometry chamber. Each song bout
lasted several minutes and included both singing and
pauses between songs, so our measurements would
include any increase in gas exchange between songs that
compensated for an anaerobic component to metabolism
during song. The structure of the song differs substan-
tially between the two breeds of canary that we studied.
The song of the Fife Fancy is louder, higher-pitched and
contains more phrases than that of the roller (Güttinger
1985; Mundinger 1995). We compared the cost of singing
between these two breeds to assess whether song com-
plexity influenced the energy cost of singing. To ensure
that the measurements were representative of song in
more normal situations, we compared the characteristics
of song recorded inside the respirometry chamber and in
the birds’ home cages. We recorded the rate at which
birds moved so that we could separate the energy costs of
singing from those of moving. We also measured basal
metabolic rate (BMR) so that we could express metab-
olism during singing as a multiple of BMR. Our results
allow us to assess whether singing is energetically costly
or cheap in passerine birds.
METHODS
Birds and Husbandry
We purchased nine male roller canaries from local
breeders and purchased or bred 49 male Fife Fancy
canaries (hereafter called Fife canaries). We kept the birds
singly (to prevent aggression between birds) in cages
measuring 6030 cm and 35 cm high except during
breeding, when a pair of birds shared a double cage. The
male birds could see and hear each other. The Fife
canaries could also see females. The diet was mixed seed
(Super Canary for Fife canaries and Roller Mix for roller
canaries, Haiths, Cleethorpes, U.K.) fed ad libitum and
supplemented with broccoli, egg food (Sluis, Woodlea
Birds, Maldon, U.K.) mixed with hardboiled egg, soaked
seed (Easysoak, Haiths, Cleethorpes, U.K.) and vitamins
(Daily Essentials, The Birdcare Company, Nailsworth,
U.K.). The birds had constant access to water and crushed
oyster shell. Water baths were provided once a week. The
photoperiod (daylight fluorescent strip lights) was
10:14 h light:dark in October, was increased evenly to
16:8 h in January, remained constant until June and was
decreased evenly to 10:14 h by October. Room air tem-
perature was 15–20C. The two breeds of canary were kept
in separate rooms. Data were collected between December
and May. Bird husbandry and our experimental pro-
cedures were approved by the Home Office, as was
returning the birds to local canary breeders at the end of
our work.
Respirometry
We used as subjects the 12 male Fife canaries that sang
most frequently in their home cages, and all nine male
roller canaries. Each bird spent 30–90 min inside the
respirometry chamber during an experiment. The
respirometry chamber was placed in the room in which
the birds were normally kept. Food (broccoli and mixed
seed) was provided inside the chamber. Eight Fife canaries
and six roller canaries sang inside the chamber.
To measure the rate of oxygen consumption of the
canaries, we used an open-flow respirometry system con-
nected to an oxygen analyser (model 1100A or Xentra,
Servomex, Crowborough, U.K.). Air was pumped (Charles
Austen, Byfleet, U.K., diaphragm pump) through an
ABS plastic chamber (1712 cm and 15 cm high, Ensto,
Briticent, Christchurch, U.K.) with a 1-cm-diameter
wooden perch 3 cm from the floor. The upper 9 cm of the
chamber was transparent. We used a wet type gas flow
meter (model DM3A, Zeal, London, U.K.) to measure the
flow rate of ambient air into the chamber. The meanSD
flow rate of dry air at standard temperature and pressure
(STPD) across experiments was 1480450 ml/min
(N=206). Flow rates varied by less than 0.5% within
experiments, and these changes were taken into account
in the calculations of the rate of oxygen consumption.
Gases were dried (silica gel) before and after passing
through the flow meter and after passing though the
chamber. The zero point of the oxygen analyser was set
each week using oxygen-free dry nitrogen gas (BOC,
Guildford, U.K.) and the span was set before each exper-
iment using ambient air. The output from the oxygen
analyser was sampled at 30 Hz using a microcomputer
with an analogue-to-digital converter (PC-ADH24, Bede
Technology, Sunderland, U.K.) and averaged every 2 s
by customized software written in BASIC. The ambient
oxygen content of the air that was pumped through the
Table 1. Metabolic rates (W) of Fife canaries and roller canaries at night in the thermoneutral zone (BMR), and during the day while sitting (on a perch with the tarsi covered by the body
feathers and the body in a relatively horizontal posture), standing (in the singing posture with the tarsi exposed and the body held relatively vertically), eating and singing
Metabolic rate (W) Factorial increase t test, sing versus stand

Mass Proportion Sing/ Sing/ Sing/ Effect

Bird (g) BMR Sit Stand Eat Sing song BMR Sit Stand t df P size, r

Fife canaries
Dor31 20.1 0.25 — 0.72±0.06 0.74±0.07 0.84±0.10 0.39±0.09 3.41 — 1.18 3.91 22 0.0007 0.64
(13) (4) (14) (14)
DP37 17.1 0.28 — 0.55±0.08 0.57±0.04 0.50±0.03 0.21±0.09 1.82 — 0.92 −1.37 6 0.22 −0.49
(6) (4) (6) (6)
DW2 17.7 0.33 — 0.60±0.07 0.70±0.09 0.62±0.09 0.40±0.09 1.89 — 1.03 0.41 5 0.70 0.18
(3) (11) (7) (7)
J6 22.4 0.31 0.54±0.05 0.60±0.05 0.62±0.07 0.61±0.10 0.24±0.09 1.98 1.14 1.02 0.22 2 0.85 0.15
(2) (8) (11) (3) (3)
J9 21.5 0.23 — — 0.59±0.11 0.73±0.07 0.35±0.07 3.13 — — — — — —
(17) (7) (7)
J22 23.7 0.35 0.52 — 0.73±0.09 0.80±0.06 0.18±0.06 2.28 1.54 — — — — —
(1) (21) (2) (2)
LY4 21.5 0.37 — 0.62±0.05 0.77±0.07 0.71±0.09 0.33±0.11 1.95 — 1.14 2.91 15 0.01 0.60
(6) (4) (19) (19)
TW47 28.3 0.25 0.46±0.05 0.66±0.02 0.78±0.09 0.73±0.04 0.56±0.17 2.93 1.59 1.11 4.19 14 0.0002 0.80
(2) (7) (15) (10) (10)
Mean 21.5±3.5 0.30±0.05 0.51±0.04 0.62±0.06 0.69±0.08 0.69±0.11 0.33±0.12 2.46±0.67 1.42±0.25 1.07±0.10 (6) — — — 0.31±0.47
(8) (8) (3) (6) (8) (8) (8) (8) (3) (6) (6)
Roller canaries
Var 19.8 0.29 — 0.73±0.16 0.75±0.17 0.68±0.11 0.39±0.09 2.32 — 0.93 −0.69 7 0.51 −0.25
(10) (19) (21) (21)
208 21.9 0.35 — 0.69±0.08 0.72±0.09 0.71±0.08 0.54±0.15 2.00 — 1.02 0.3 7 0.8 0.11
(5) (7) (17) (17)
1725 19.9 0.32 0.43±0.05 0.51±0.09 0.58±0.06 0.62±0.01 0.36±0.10 1.92 1.44 1.21 1.23 4 0.3 0.52
(5) (4) (11) (2) (5)
109 18.7 0.25 0.40±0.03 0.52±0.06 0.53±0.06 0.56±0.06 0.32±0.07 2.20 1.40 1.06 1.08 6 0.32 0.40
(5) (17) (11) (5) (5)
23 21.7 0.22 — 0.56±0.06 0.56 0.62±0.11 0.54±0.17 — — 1.11 0.63 5 0.56 0.27
(16) (1) (5) (5)
209 17.9 0.38 — 0.64±0.09 0.75 0.71±0.08 0.44±0.18 — — 1.11 2.43 23 0.02 0.45
(14) (1) (24) (24)
Mean 20.0±1.6 0.30±0.06 0.41±0.02 0.61±0.11 0.65±0.10 0.64±0.07 0.43±0.09 2.11±0.15 1.42±0.02 1.05±0.07 — — — 0.25±0.29
(6) (6) (2) (4) (6) (4) (6) (4) (2) (6) (6)

Only data from bouts of sitting, standing and singing during which the birds did not move are included, except for the data given in italics for two roller canaries (birds 23 and 209) for which
movement rates did not differ between singing and standing birds. Proportion of song refers to the proportion of time for which a bird sang during a song bout. The t tests compare metabolic
rate between perching and singing within individual birds. Sample sizes are given in parentheses. Means are given±SD. The mean values given for each breed of canary are the means of the
mean values for each bird. The durations of measurements of metabolic rate across Fife canaries were 3.0±2.4, 6.6±1.5, 6.8±1.6 and 5.9±2.1 min for sitting, standing, eating and singing,
respectively. The durations of measurements of metabolic rate across roller canaries were 7.5±2.6, 5.2±3.0, 6.4±5.2 and 6.5±1.0 min for sitting, standing, eating and singing, respectively.
WARD ET AL. THE ENERGY COST OF CANARY SONG
895
896 ANIMAL BEHAVIOUR, 66, 5
chamber was measured before and after each bird was
placed in the chamber. We used these measurements to
compensate for any drift in the output of the oxygen
analyser during an experiment.
We calculated the rate of oxygen consumption by
multiplying the difference between the fractional oxygen
content of ambient and excurrent air by the incurrent
flow rate (corrected to STPD). Carbon dioxide was not
absorbed; this maximizes the accuracy of calculated
energy expenditure when the respiratory quotient is not
measured (Koteja 1996; Speakman 2000). We converted
rates of oxygen consumption to energy equivalents using
an oxycalorific value of 20.92 J/ml (derived from Brody
1945 for a respiratory quotient of 1). We measured basal
metabolic rate (BMR) by placing each bird in the
respirometry chamber overnight at 30C (within the
thermoneutral zone for passerine birds of similar mass to
canaries; Calder & King 1974) inside a constant tempera-
ture incubator (model INL-401N-010, Gallenkamp,
Loughborough, U.K.). BMR was calculated from the
lowest metabolic rate over 5 min.
To determine the washout characteristics of the
respirometry chamber, we reduced the oxygen content of
the gases inside the chamber to 20.8% (a level similar to
that experienced by the birds during experiments) using
nitrogen gas (BOC). We then pumped ambient air
through the chamber at rates within the range used
during experiments until the oxygen content of the
excurrent gases returned to that of ambient air. A Fife
canary that had died of natural causes was placed on the
perch inside the chamber to make the internal volume
and air flow characteristics of the chamber as similar as
possible to those during experiments with living birds. A
meanSD of 5200200 ml (N=3) of ambient air was
passed through the chamber before the oxygen content
of the gases leaving the chamber recovered by 95% from
the level to which it had been perturbed. We therefore
assumed that the equilibrium rate of oxygen consump-
tion of a bird for a given behaviour was reached after that
behaviour had been performed consistently for long
enough for 5200 ml of gas to have passed through the
chamber (hereafter called the chamber equilibration
time). The meanSD chamber equilibration time was
3.81.1 min (N=206 experiments). The combined time
taken for gases to pass from the respirometry chamber to
the oxygen analyser and for the analyser to respond to
any change in concentration of the gases was 3–5 s,
depending on the flow rate. This response time was
taken into account when we matched the data on the
oxygen content gases leaving the chamber with bird
behaviour. We did not use an instantaneous correction
(Bartholemew et al. 1981) to calculate the rate of oxygen
consumption, because this was unnecessary when bouts
of behaviour were long enough to allow us to measure
equilibrium rates of oxygen consumption.
Bird Behaviour
We recorded bird behaviour on VHS video (JVC GR-A
X280 video camera connected to a Panasonic NV-FS90
video recorder) during experiments. Bird behaviour was
categorized as singing, standing (in the singing posture
with the tarsi exposed and the body held relatively
vertical), sitting (with the legs and feet covered by the
body feathers and the body in a relatively horizontal
posture) or eating. We defined song duration as the
period during which a bird produced a series of phrases in
rapid succession. Song duration was always shorter than
the equilibration time of the chamber. Our measure-
ments of the energy cost of singing therefore refer to the
average metabolic rate during bouts of songs. Song bouts
began or ended when the bird perched, sat, ate, moved (if
the bird had not otherwise moved during the singing
bout) or sang at a rate that differed by more than 50%
from that during the rest of the bout. Movement was
assessed from the rate at which birds moved their feet (i.e.
to hop between the perch and the chamber floor or to
rotate by 180 upon the perch).
Song Duration and Amplitude
We recorded Fife canary song using a Tandberg TM6
microphone connected to a computer (Dell Dimension
4100, Creative Sound Blaster AudioPCI 64V soundcard)
running Goldwave (version 4.23, Goldwave Inc., St
John’s, Canada) software with a sampling rate of 22 kHz
and 16-bit resolution. The centre of the microphone was
5 cm in front of and 2 cm below the bird’s bill when the
bird sang facing the microphone while standing on
the perch inside the respirometry chamber. We used the
video equipment described above to assess when the birds
sang in the same position relative to the microphone
inside their home cages. We recorded song from eight
Fife canaries standing in this position relative to the
microphone inside the respirometry chamber and from
seven of these birds in their home cages.
We calculated the proportion of time spent singing
during each bout of song from the summed durations
of all songs divided by the duration of the bout. The
duration of each Fife canary song was measured from the
digitized recordings to the nearest 0.1 s using Avisoft-
SASLab Pro (Berlin, Germany, version 3.94c). The dur-
ation of each roller canary song was measured from the
videos, with a stopwatch. Comparison of the meanSD
duration of Fife canary songs that were measured both
with the stopwatch and from the digitized recordings
showed that song duration was measured to the nearest
0.40.1 s (N=230 songs) using the stopwatch. The pro-
portion of singing during song bouts measured using the
stopwatch differed by 23% (N=8 bouts) from that
measured from the digitized recordings.
All eight Fife canaries included three types of phrase
during the first 5 min of song that we recorded from each
bird in his home cage. These were a series of elements
with continuously descending pitch, an ‘A’ phrase (Vallet
& Kreutzer 1995; Vallet et al. 1998), and a series of
elements in which descending and low pitch notes alter-
nated (Fig. 1). To measure the peak amplitude and the
frequency of the peak amplitude of these elements, we
used the spectral characteristics function from the log-
arithmic power spectrum in Avisoft-SASLab Pro. We
sampled 3–15 repetitions of each element of the three

Figure 1. The three phrases sung by Fife canaries for which
characteristics were analysed.
phrases on six randomly selected occasions from record-
ings of Fife canaries singing inside the respirometry
chamber and in their home cages. We calculated the
mean peak amplitude, the mean frequency at the peak
amplitude and the mean repetition frequency for each
bird singing in both situations.
We compared the volume of recordings of Fife canary
song played with and without the lid on the respirometry
chamber to determine the effect of enclosure inside the
chamber upon the apparent volume of canary song. We
used three to five repetitions of a randomly selected
example of each of the three Fife canary phrases shown in
Fig. 1 sung by each of seven birds. We played the songs
through a general-purpose driver speaker from a Sony
SRS-P3 speaker (60 cm diameter, 0.4–20 kHz output)
placed in a box (88 cm and 5.5 cm high) from which
sound was directed upwards into a plastic funnel (7 cm
diameter initially, narrowing gradually to 1 cm diameter
over 4.5 cm, 1 mm wall thickness). The sound from the
speaker was directed horizontally for 2.5 cm before being
broadcast through an opening of 0.5 cm internal diam-
eter, 11.5 cm from the floor of the chamber (a similar
height to the head of a canary). The volume of the songs
played from the speaker was similar to that of singing
canaries and was recorded with the same equipment as
was used to record the birds.
Data Analysis
We used t tests to compare metabolic rate during two
behaviours in individual birds, song characteristics under
two circumstances or characteristics of the two breeds of
canary. All but two (of 40) data sets on the metabolic rate
of individual birds during a particular activity (Table 1)
were normally distributed (Kolmogorov–Smirnov nor-
mality tests). The data sets that were not shown to be
normally distributed both had N<5. We assumed that the
inferences drawn from t tests of metabolism during dif-
ferent activities would be robust to these deviations from
normal distributions of data (Zar 1996, page 97). The
overall significance of tests across birds was assessed with
Z tests following Rosenthal (1991). We calculated Z, the
standard normal deviate, (Rosenthal 1991, equation 5.14)
and r, the effect size, (Rosenthal 1991, equation 2.16) for
each bird. We determined the overall value of Z across
birds from the unweighted values of Z from each bird
(Rosenthal 1991, equation 5.4) and calculated the overall
effect size from the mean value of r across birds. We used
WARD ET AL. THE ENERGY COST OF CANARY SONG 897
a paired t test (with song element as the unit of indepen-
dent data) to test whether differences in volume between
song inside the respirometry chamber and in the birds’
home cages were the same as the differences in volume
between song played from a speaker with and without the
lid on the respirometry chamber. We used a Bonferroni
correction to adjust
when more than one test was
performed on related data. All statistical tests used were
two tailed. All means are presentedSD unless otherwise
stated.
We used least-square regression to assess whether meta-
bolic rate was related to the proportion of time spent
singing for individual birds (Zar 1996). All computations
were carried out using arcsine-transformed proportion of
time spent singing. This transformation is appropriate
before analysis of data that are expressed as a proportion
(Sokal & Rohlf 1995, page 421). The regression coef-
ficients were back-transformed before presentation. We
used GLM ANCOVA in which bird was entered as a
random factor, and the proportion of time spent singing
was a covariate to assess whether the relation between
metabolic rate and the proportion of time spent singing
differed between birds (Zar 1996). The gradients of the
relations between metabolism and the proportion of time
spent singing did not vary between birds, but the inter-
cepts differed. We therefore present the relations for
individual birds. We used these relations to calculate
metabolic rate during continuous singing, with an associ-
ated 95% prediction interval (Zar 1996, equation 16.29).
Analyses were carried out with Minitab 12.22 (Minitab
Inc, State College, Pennsylvania, U.S.A.).
RESULTS
Song Characteristics
Fife canaries sang slightly shorter songs inside the
respirometry chamber than in their home cages
(respirometry chamber: 5.92.2 s, range 0.4–56.0 s, N=8
birds; home cage: 7.32.5 s, range 0.2–49.9 s, N=7 birds;
Z test: Z=5.9, N=7 birds, P<0.001). Mean roller canary
song duration inside the respirometry chamber was
8.72.6 s (range 0.7–68.5 s, N=6 birds). The phrase rep-
etition frequency and the frequency at the peak ampli-
tude of the elements of these phrases did not differ
between Fife canary song in the respirometry chamber
and in the birds’ home cages (repetition frequency,
Z=0.9, N=3 phrases, P=0.36; frequency at peak ampli-
tude, Z=1.7, N=5 elements, P=0.08). The peak amplitude
of song recorded inside the respirometry chamber was
5.53.6 dB greater than that recorded inside the birds’
home cages (Z test: Z=2.8, N=5 elements, P=0.007). The
volume of Fife canary song played from the speaker was
3.72.9 dB greater when the lid was on the chamber
than when it was removed (Z=5.1, N=5 elements,
P<0.001). The difference in recording volume between
Fife canaries singing inside the respirometry chamber and
in their home cages did not differ from the difference in
volume between recordings of song played through the
speaker with and without the lid on the respirometry
chamber (paired t test: t4 =0.9, N=5 elements, P=0.4).
898 ANIMAL BEHAVIOUR, 66, 5

1 t12 =1.96, P=0.07; bird 23: t6 =0.1, P=0.9). When the data
(a) from these two birds were included in the analysis,
0.8 metabolism during singing was 1.050.07 times greater
than that during standing (Z=1.8, N=6 birds, P=0.03;
0.6 Table 1, Fig. 2b). Metabolism during singing without
moving did not differ from that during eating (Z=0.2,
0.4 N=4 birds, P=0.4). Metabolic rate during singing without
moving was 2.110.15 (N=4 birds) times BMR and
Metabolic rate (W)

0.2 1.420.02 (N=2 birds) times that during sitting.

8 3 6 8
0
1 Singing
(b)
0.8
0.6
0.4
0.2
6 2 5 5
0 shown in Table 2 (GLM ANCOVA: bird: F2,68 =13.5,
BMR Sit Stand Sing
Figure 2. Mean+SD metabolic rate (W) of (a) male Fife canaries and
(b) male roller canaries at night in the thermoneutral zone (BMR),
and during the day when sitting, standing without moving, singing
without moving and eating. Numbers inside columns show the
number of birds that performed each behaviour.
Metabolism During Singing and Other Activities
Some of the Fife canaries moved at different rates
during bouts of singing and standing (t tests: two birds
that moved at different rates: t33 =2.1, P=0.04, t24 =2.8,
P=0.01; other birds: t3–20 =0.2–2.8, P=0.07–0.9), so in our
analyses we used only data from periods during which the
birds did not move (Table 1). Fife canary metabolism
during singing was 1.070.10 times that during standing
(Z test: Z=3.6, N=6 birds, P<0.0001) but did not differ
from that during eating (Z=0.3, N=8 birds, P=0.7; Table
1, Fig. 2a). Metabolic rate during singing was
2.460.67BMR (N=8 birds) or 1.420.25sitting
metabolic rate (N=3 birds; Table 1).
The metabolic rate of four roller canaries that sang and
stood without moving did not differ between singing and
standing (Z test: Z=0.8, N=4 birds, P=0.2; Table 1). The
rates of movement of two additional roller canaries did
not differ between singing and standing (t tests: bird 209:
8
Metabolism and Proportion of Time Spent
Metabolic rate increased with the proportion of time
spent singing for the three Fife canaries from which we
collected most data (Table 2). Metabolic rate was not
related to the proportion of time spent singing in the
other birds (regression: P>0.08 in all cases, sample sizes in
Table 1). The intercepts, but not the gradients, of the
relationships between metabolic rate and the proportion
of time spent singing varied between the three birds
6
Eat P<0.001; proportion of time spent singing: F1,68 =23.1,
P<0.001; birdproportion of time spent singing:
F2,68 =1.0, P=0.4). We extrapolated the relationships
between metabolic rate and the proportion of time
spent singing (measured range 0–0.75) to predict
metabolic rate during continuous singing (Table 2, Fig. 3).
The mean metabolic rate during continuous singing
would be 0.880.10 W (equivalent to 1.300.12 times
metabolic rate during standing or 3.140.78BMR,
N=3 birds).
Comparison Between Breeds
Fife canaries and roller canaries did not differ in mass,
BMR or the proportion of time spent singing (t tests of
mean values for each bird of each breed: t12 =0.2–1.6,
P>0.1 in all cases), or metabolic rate during singing
without moving or factorial increase in metabolism
between BMR and singing without moving (t tests:
t11 =0.5–0.7, P>0.4 in all cases) or metabolic rate during
standing without moving (t test: t9 =0.6, P=0.6). Our
measurements of metabolism during standing without
moving, singing without moving and the factorial
increase in metabolism during singing over that during
standing by roller canaries did not differ from those of

Table 2. Relations between the proportion of time spent singing and the metabolic rate of three male Fife canaries
Metabolic rate for continuous song
Mean (95% prediction range)

Intercept Gradient Metabolism Metabolism

Bird ±SE ±SE R2 P N (W) (×stand)

Dor31 0.73±0.02 0.25±0.09 0.26 0.007 27 0.98 (0.75–1.22) 1.35 (1.02–1.67)

LY4 0.63±0.03 0.25±0.10 0.20 0.02 25 0.88 (0.64–1.12) 1.39 (1.01–1.77)
TW47 0.67±0.01 0.11±0.03 0.48 0.002 17 0.78 (0.69–0.87) 1.16 (1.03–1.30)

1 roller canaries, was consistent with the greater proportion
0.8
Metabolic rate (W)
0.6
0.4
0.2
0 demanding display.
0 0.2 0.4 0.6 0.8
Proportion of time spent singing
Figure 3. Metabolic rate (W) in relation to the proportion of time
spent singing in eight Fife canaries (shown by different symbols).
Solid lines and points show the relations for the three birds shown in
Table 2. Dotted lines show extrapolation of these relations to
continuous singing. Open points show data from the other five
birds.
Waterslager canaries (a different strain of roller canary)
measured by Oberweger & Goller (2001; t tests: t6 <1.2,
P>0.3).
DISCUSSION
Canary Metabolism During Singing
Recordings of Fife canary song from inside the
respirometry chamber were louder than those from the
birds’ home cages. This was due to acoustic effects of
enclosure inside the chamber, since the volume of canary
song played from a speaker increased to similar extent
when the lid was placed on the chamber. The volume,
frequency and repetition frequency of a representative
sample of Fife canary song phrases did not alter inside the
respirometry chamber, although songs were slightly
shorter than in birds’ home cages. This difference was
unlikely to influence the metabolic cost of singing. Our
measurements of the cost of singing are therefore
representative of song in more normal situations.
Metabolic rate increased with the proportion of time
spent singing in Fife canaries from which we had most
data, as would be expected if standing between songs
required slightly less energy than singing. Variation
between birds in the gradients of the relations between
metabolic rate and the proportion of time spent singing
would have suggested that some birds produced song
more efficiently than others. Since the gradients of these
relations did not vary between birds, there was no evi-
dence for individual variation in the efficiency of song
production. The variation between birds in the intercepts
of these relations reflects individual differences in meta-
bolic rate while standing. Such differences may be related
to a combination of factors such as mass (Calder &
King 1974), body composition (Daan et al. 1990) and
dominance (Buchanan et al. 2001).
The slightly greater mean factorial increase in metab-
olism during singing by Fife canaries, compared with
WARD ET AL. THE ENERGY COST OF CANARY SONG 899
of the time for which Fife canaries sang. The substantial
differences between the volume, frequency and structure
of the songs of Fife canaries and roller canaries (Güttinger
1985; Mundinger 1995) did not lead to measurable differ-
ences in the metabolic cost of singing. Our data thus
confirm the conclusions of Oberweger & Goller (2001)
and Franz & Goller (2003), that changes in song structure
and complexity do not alter the energetic cost of singing.
Female preferences for greater song complexity (Searcy
& Yasukawa 1996) are not therefore because singers of
more complex songs are producing a more energetically
1 Weathers et al. (1997) suggested that bird vocalization
may have an anaerobic component so that measurements
of oxygen consumption made during singing may under-
estimate the energy cost of song production. We think
that it is unlikely that canaries would use anaerobic
metabolism during song, because they sing for such a
high proportion of the time that there would be little
opportunity to pay back any oxygen debt incurred during
singing. Oberweger & Goller (2001) calculated the cost of
singing by comparing the metabolic rate of Waterslager
canaries immediately before they sang with that during
song. Any anaerobic component to metabolism during
singing might not have been included in their measure-
ments of the cost of song, since the measurement period
ended before any oxygen debt would have been repaid
during the subsequent intersong interval. The metabolic
rates of our roller canaries did not differ from those
measured by Oberweger & Goller (2001). The way in
which we calculated the energy cost of singing would
have included repayment of any oxygen debt incurred
from anaerobic metabolism during song, since intervals
between songs were also included in our bouts of singing
behaviour. Canaries therefore do not have a measurable
anaerobic component to metabolism during song. Our
measurements of the energy cost of singing also represent
the overall effect upon metabolism of short-term fluctu-
ations in gas exchange caused by changes in the
respiratory patterns of singing birds (Franz & Goller
2003).
The metabolic rate during song bouts did not differ
from that during eating in our canaries, despite what
intuitively appeared to be much greater effort while
singing (the whole of the bird shook vigorously) than
while eating (the bird leaned forwards occasionally to
pick up food which was manipulated in the bill while
otherwise remaining stationary). Singing by both breeds
of canary raised metabolism only slightly (mean factorial
increase 1.05–1.07) over that of birds that stood in the
same posture as was used during song. Metabolic rate
during singing was a greater multiple of BMR (a factorial
increase of 2.1–2.5). Singing was more costly in relation
to BMR than in relation to standing, partly because
metabolic rate increases as temperature falls below the
thermoneutral zone (Calder & King 1974), and we
measured the cost of singing at lower temperatures (15–
20C) than we measured BMR (30C). The factorial
increase in metabolism during song relative to BMR is
likely to vary with temperature, because the energy cost
900 ANIMAL BEHAVIOUR, 66, 5
of thermoregulation changes with temperature. Singing
caused a factorial increase in metabolism over sitting that
was intermediate between those relative to standing and
BMR partly because the thermoregulatory costs of sitting
are likely to be lower than those of standing birds,
because sitting birds covered their legs. The legs can make
a substantial contribution to heat loss (Ward et al. 1999).
Metabolic rate during sitting was also likely to be lower
than that of singing or standing because the birds
appeared to be less alert during sitting. The energy cost of
song production itself should therefore be assessed by
comparing metabolism during song with that during
standing in the same posture (mean factorial increase
1.05–1.07). Our data thus support the hypothesis that
song production by canaries is energetically cheap.
Does Singing Generally Increase Metabolic Rate in
Birds?
Singing increases the rate of oxygen consumption of
Carolina wrens, European starlings, zebra finches and two
breeds of canary singing inside respirometry chambers
(Eberhardt 1994; Oberweger & Goller 2001; Franz &
Goller 2003; this study). Singing also increases the rate
of overnight mass loss (which is presumably correlated
with the rate of consumption of body reserves and there-
fore with energy expenditure) of free-living nightingales
(Thomas 2002). In contrast, crowing by junglefowl or
cockerels inside a respirometry chamber does not cause a
measurable increase in the rate of oxygen consumption
(Chappell et al. 1995; Horn et al. 1995). Although cock-
erel crows are loud and inefficient (Brackenbury 1977),
they were performed for a much lower proportion of the
time than that for which our canaries sang (0.05–0.06 for
crowing: Chappell et al. 1995; Horn et al. 1995; 0.33–0.43
for singing: this study). Cockerel crows also contain
substantially less acoustic power in relation to the mass of
the bird (60 mW/kg) than the songs of typical passerine
birds such as the song thrush, Turdus philomelos, winter
wren, Troglodytes troglodytes, and European robin, Erith-
acus rubecula (870, 600 and 300 mW/kg, respectively;
Brackenbury 1979). The substantially higher proportion
of time spent singing and greater volume of passerine
song in relation to their body mass probably explain why
the cost of passerine song has been detected over other
fluctuations in metabolism, but the cost of crowing by
junglefowl and cockerels has not.
The measurements of metabolic rate of singing
Carolina wrens (Eberhardt 1994) have generally been
regarded as anomalously high (Gaunt et al. 1996;
Oberweger & Goller 2001; Thomas 2002). The reported
factorial increase in the metabolism of Carolina wrens
during song (3.61.1BMR, with a maximum value of
9BMR: Eberhardt 1994) was greater than that of Fife
canaries (2.40.6BMR, this study), roller canaries
(2.40.4BMR: this study; Oberweger & Goller 2001),
zebra finches (2.00.2BMR: Oberweger & Goller 2001)
or European starlings (2.10.06BMR: Oberweger &
Goller 2001; ANOVA with species, or breed of canary, as a
factor and the mean factorial increase in metabolism
during singing of individual birds as independent data:
F4,27 =4.3, P=0.01 with Tukey post hoc multiple compari-
sons). The cost of singing measured for Carolina wrens
may be greater than that of the other species because of
difficulties in measuring transient changes in oxygen
consumption in a relatively large respirometry chamber,
because the birds moved as well as sang during measure-
ments, or because of insufficiently detailed mathematical
treatment of the data (Frappell et al. 1989; Gaunt et al.
1996; Ortigues et al. 1997). Movements during song bouts
were not monitored during Eberhardt’s (1994) study of
Carolina wrens, so the costs of any movements would
have been included in the measured cost of singing (Horn
et al. 1995; Gaunt et al. 1996). If free-living Carolina
wrens move during song bouts, the cost of movements
should be included in the cost of the display, but in this
case the cost of the display would clearly be greater than
that of song alone. The increased energy expenditure of
sage grouse, Centrocercus urophasianus, that displayed at
higher rates was attributed partially to higher movement
rates (Vehrencamp et al. 1989). Song production is likely
to make a minor contribution to the cost of display by
birds such as skylarks, Alauda arvensis, that sing while
flying (Hedenström & Alerstam 1996).
Because consistent results have been obtained from
more than one breed of canary and two other passerine
species (Oberweger & Goller 2001; Franz & Goller 2003;
this study), it is likely that the modest cost of singing,
relative to that of standing during the day, in these
species is representative of most passerine birds. Winter
wren song is louder in relation to the mass of the bird
than that of many other passerines (Brackenbury 1979),
and singing by wrens of all genera is considered to be
particularly loud (Brewer 2001). Wren song may therefore
be energetically more costly than that of most other
passerine birds, but further empirical data are required to
determine whether this is the case.
The Energy Cost of Song and the Dawn Chorus
Singing appears to increase metabolism only slightly
over that of a bird standing during the day, but the
factorial increase over BMR is much greater. The
additional energy cost of singing at night (rather than
roosting with a metabolic rate closer to BMR) would be
substantially greater than that of singing during the day
(rather than standing). This is because the energy cost of
the activity that singing would replace would be much
lower at night than during the day. Similarly, singing
during the dawn chorus could cause a greater factorial
increase in energy costs than singing during the day if
birds sing rather than sleep at this time. The energy cost
of singing at night or at dawn would presumably be a
factorial increase of between 1.4 (the difference between
sitting and singing during the day in canaries) and 2.2–
2.6 (the difference between BMR and singing in canaries).
The extent of the increase in metabolism during singing
at night would depend on the difference in temperature
(and hence thermoregulatory costs) between the micro-
habitats in which birds would roost or sing, changes in
thermoregulatory costs due to altering posture and
altered alertness. Sound production itself would make a

minor contribution to metabolism. Our results are there-
fore consistent with the energetic cost hypothesis for the
evolution of the dawn chorus (McNamara et al. 1987)
although they do not exclude other explanations for the
dawn chorus such as enhanced sound transmission
(Brown & Handford 2003) or reduced foraging success
(Kacelnik & Krebs 1983) at this time.
Is Singing Energetically Costly for Passerine Birds?
Whether singing is energetically costly for a passerine
bird will depend on two factors: how much song produc-
tion increases metabolic rate and the proportion of the
energy available to a bird that is used by singing. It is this
marginal energy cost (Verhulst & Wiersma 1997) that will
influence whether song production imposes a fitness cost
upon the singer. We have shown that singing causes only
a small increase in the metabolic rate of canaries and that
this result is probably typical of most passerine birds.
Despite this, whether the energy cost of singing
imposes a fitness cost upon the singer will depend upon
the circumstances. If a bird is close to starvation, foraging
during the day or energy conservation by roosting at
night is clearly more important than singing. The mar-
ginal cost of singing by a bird under such circumstances
would be very great. At the other extreme, a bird with an
ample food supply and time for foraging, as well as
negligible predation risk either during singing or forag-
ing, would not experience a reduction in fitness by
singing. Singing would have a low marginal cost for such
an individual. The time and energy constraints for most
birds will be intermediate between these two extremes. It
is presently unclear how to assess how much energy is
potentially available to an individual bird for singing. The
marginal costs of singing for different individuals will
remain unknown without this information, but we can
still conclude that singing is a condition-dependent sig-
nal. Singing will impose a greater fitness cost upon birds
that have difficulty in achieving energy balance, or in
foraging without increasing their predation risk, even
though song production itself has a modest energy cost.
Birds potentially compensate for an increase in metab-
olism caused by song production by reducing other
aspects of their metabolism such as self-repair (Sheldon &
Verhulst 1996). Birds are able to perform metabolic com-
pensation in other circumstances such as during flight
(Deerenberg et al. 1998). Since singing leads to only a
small overall increase in metabolism compared with that
required during flight, there would appear to be less need
for compensatory reductions in metabolism during sing-
ing than while flying. However, compensatory reductions
in other components of metabolism during song remain a
possibility.
Although singing has an energy cost, it is an energeti-
cally cheap signal if compared with alternative behaviour
that might replace the functions of song. Flying through-
out the area within which song could have been heard to
locate potential mates or rivals would be much more
energetically costly than singing. Flight costs 10BMR
for a bird of the mass of a canary (Butler & Bishop 2000)
compared with 2.2–2.6BMR for singing. Flying could
WARD ET AL. THE ENERGY COST OF CANARY SONG 901
also be less effective than singing, since a bird can only be
in one place at a time, but woodland birds can often
be heard simultaneously in all directions for greater
distances than they can be seen.
Acknowledgments
We are very grateful to R. Donaldson, H. Hodge, D.
Lumsden, D. Pointon and W. Turnbull for their advice on
canary husbandry, to I. Maynard, V. Murray, A. Oliver, G.
Thomson and R. Webster who helped to care for the
birds, to M. Coutts and G. Taylor for electronic support,
to Lucy Gilbert, Vincent Janik, Natasha LeBas and Joe
Tomkins for discussions, to two anonymous referees for
suggesting improvements to the manuscript and to the
BBSRC for funding our work.
References
Bartholemew, G. A., Vleck, D. & Vleck, C. M. 1981. Instantaneous
measurements of oxygen consumption during pre-flight warm-up
and post-flight cooling in sphingid and saturnid moths. Journal of
Experimental Biology, 90, 17–32.
Brackenbury, J. H. 1977. Physiological energetics of cock-crow.
Nature, 270, 433–435.
Brackenbury, J. H. 1979. Power capabilities of the avian sound
producing system. Journal of Experimental Biology, 78, 163–166.
Brewer, D. 2001. Wrens, Dippers and Thrashers. London:
Christopher Helm.
Brody, S. 1945. Bioenergetics and Growth. New York: Reinhold.
Brown, T. J. & Hanford, P. 2003. Why birds sing at dawn: the role
of consistent song transmission. Ibis, 145, 120–129.
Buchanan, K. L., Evans, M. R., Goldsmith, A. R., Bryant, D. M. &
Rowe, L. V. 2001. Testosterone influences basal metabolic rate in
male house sparrows: a new cost of dominance signalling? Pro-
ceedings of the Royal Society of London, Series B, 268, 1337–1344.
Butler, P. J. & Bishop, C. 2000. Flight. In: Sturkie’s Avian Physiology
(Ed. by G. C. Whittow), pp. 391–436. London: Academic Press.
Calder, W. A. & King, J. R. 1974. Thermal and caloric relations of
birds. In: Avian Biology. Vol. 4 (Ed. by D. S. Farner & J. R. King),
pp. 259–413. New York: Academic Press.
Catchpole, C. K. & Slater, P. J. B. 1995. Bird Song: Biological Themes
and Variations. Cambridge: Cambridge University Press.
Chappell, M. A. & Bachman, G. C. 2002. Energetic costs of
begging behaviour. In: The Evolution of Begging: Competition,
Co-operation and Communication (Ed. by J. Wright & M. L.
Leonard), pp. 143–162. Netherlands: Kluwer.
Chappell, M. A., Zuk, M., Kwan, T. H. & Johnsen, T. S. 1995.
Energy cost of an avian display: crowing in red junglefowl. Animal
Behaviour, 49, 255–257.
Clutton-Brock, T. L. & Albon, S. D. 1979. The roaring of red deer
and the evolution of honest advertisement. Behaviour, 69, 145–
170.
Daan, S., Masman, D. & Groenewold, A. 1990. Avian basal
metabolic rates: their association with body composition and
energy expenditure in nature. American Journal of Physiology, 259,
R333–R340.
Deerenberg, C., Overkamp, G. J. F., Visser, G. H. & Daan, S. 1998.
Compensation in resting metabolism for experimentally increased
activity. Journal of Comparative Physiology, B, 168, 507–512.
Eberhardt, L. S. 1994. Oxygen consumption during singing by
male Carolina wrens Thryothorus ludovicianus. Auk, 111, 124–130.
Fisher, R. A. 1930. The Genetical Theory of Natural Selection. Oxford:
Clarendon.
902 ANIMAL BEHAVIOUR, 66, 5
Franz, M. & Goller, F. 2003. Respiratory patterns and oxygen
consumption in singing zebra finches. Journal of Experimental
Biology, 206, 967–978.
Frappell, P. B. R., Levin, H. A. & Baudinette, R. V. 1989. Under-
standing respirometry chambers: what goes in must come out.
Journal of Theoretical Biology, 138, 479–494.
Gaunt, A. S., Bucher, T. L., Gaunt, S. L. L. & Baptista, L. F. 1996.
Is singing costly? Auk, 113, 718–721.
Gil, D. & Gahr, M. 2002. The honesty of bird song: multiple
constraints for multiple traits. Trends in Ecology and Evolution, 17,
133–141.
Godfray, H. C. J. 1991. Signalling of need by offspring to their
parents. Nature, 352, 328–330.
Goller, F. & Larsen, O. N. 1997. A new mechanism of sound
generation in songbirds. Proceedings of the National Academy of
Sciences, U.S.A., 94, 14787–14791.
Grafen, A. 1990a. Sexual selection unhandicapped by the Fisher
process. Journal of Theoretical Biology, 144, 473–516.
Grafen, A. 1990b. Biological signals as handicaps. Journal of Theor-
etical Biology, 144, 517–546.
Güttinger, H. R. 1985. Consequences of domestication on the song
structures of the canary. Behaviour, 94, 254–278.
Hedenström, A. & Alerstam, T. 1996. Skylark optimal flight speeds
for flying nowhere and somewhere. Behavioral Ecology, 7, 121–
126.
Horn, A. G., Leonard, M. L. & Weary, D. M. 1995. Oxygen
consumption during crowing by roosters: talk is cheap. Animal
Behaviour, 50, 1171–1175.
Hutchinson, J. M. C., McNamara, J. M. & Cuthill, I. C. 1993. Song,
sexual selection, starvation and strategic handicaps. Animal Behav-
iour, 45, 1153–1177.
Kacelnik, A. & Krebs, J. R. 1983. The dawn chorus in the great tit
(Parus major): proximate and ultimate causes. Behaviour, 83,
287–309.
Koteja, P. 1996. Measuring energy metabolism with open-flow
respirometric systems: which design to choose? Functional Ecology,
10, 675–677.
Kotiaho, J. S., Alatalo, R. V., Mappes, J., Nielsen, M. G., Parri, S. &
Rivero, A. 1998. Energetic costs of size and sexual signalling in a
wolf spider. Proceedings of the Royal Society of London, Series B,
265, 2203–2209.
Lambrechts, M. M. 1996. Organization of bird song and constraints
on performance. In: Ecology and Evolution of Acoustic Communi-
cation in Birds (Ed. by D. E. Kroodsma & E. H. Miller), pp. 305–320.
Ithaca, New York: Cornell University Press.
Larsen, O. N. & Goller, F. 2002. Direct observation of syringeal
muscle function in songbirds and a parrot. Journal of Experimental
Biology, 205, 25–35.
McCarty, J. P. 1996. The energetic cost of begging in nestling
passerines. Auk, 113, 178–188.
McNamara, J. M., Mace, R. H. & Houston, A. I. 1987. Optimal daily
routines of singing and foraging in birds singing to attract a mate.
Behavioral Ecology and Sociobiology, 20, 399–405.
Mundinger, P. C. 1995. Behaviour-genetic analysis of canary song:
inter-strain differences in sensory learning, and epigenetic rules.
Animal Behaviour, 50, 1491–1511.
Oberweger, K. & Goller, F. 2001. The metabolic cost of bird-
song production. Journal of Experimental Biology, 204, 3379–
3388.
Ortigues, I., Dussap, C-G. & Anglaret, Y. 1997. Comparison
of various methods of calculating the instantaneous res-
piratory gaseous exchanges from discrete measurements in
respirometry chambers. Journal of Theoretical Biology, 185, 489–
501.
Prestwich, K. N. 1994. The energetics of acoustic signalling in
anurans and insects. American Zoologist, 34, 625–643.
Rosenthal, R. 1991. Meta-analytic Procedures for Social Research.
Thousand Oaks, California: Sage.
Ryan, M. J. 1988. Energy, calling, and selection. Am. Zool., 28,
885–898.
Searcy, W. A. & Yasukawa, K. 1996. Organization of bird song and
constraints on performance. In: Ecology and Evolution of Acoustic
Communication in Birds (Ed. by D. E. Kroodsma & E. H. Miller),
pp. 454–473. Ithaca, New York: Cornell University Press.
Sheldon, B. C. & Verhulst, S. 1996. Ecological immunology: costly
parasite defences and trade-offs in evolutionary ecology. Trends in
Ecology and Evolution, 11, 317–321.
Sokal, R. R. & Rohlf, F. J. 1995. Biometry. New York: W. H. Freeman.
Speakman, J. R. 2000. The cost of living: field metabolic rates
of small mammals. Advances in Ecological Research, 30, 177–
297.
Speakman, J. R., Anderson, M. E. & Racey, P. A. 1989. The energy
cost of echolocation in pipistrelle bats Pipistrellus pipistrellus.
Journal of Comparative Physiology, A, 165, 679–685.
Suthers, R. A., Goller, F. & Pytte, C. 1999. The neuromuscular
control of birdsong. Philosophical Transactions of the Royal Society
of London, B, 354, 927–939.
Thomas, R. J. 2002. The costs of singing in nightingales. Animal
Behaviour, 63, 959–966.
Vallet, E. & Kreutzer, M. 1995. Female canaries are sexually
responsive to special song phrases. Animal Behaviour, 49, 1603–
1610.
Vallet, E., Beme, I. & Kreutzer, M. 1998. Two-note syllables in
canary songs elicit high levels of sexual display. Animal Behaviour,
55, 291–297.
Vehrencamp, S. L., Bradbury, J. W. & Gibson, R. M. 1989. The
energetic cost of display in male sage grouse. Animal Behaviour,
38, 885–896.
Verhulst, S. & Wiersma, P. 1997. Is begging cheap? Auk, 114,
134–134.
Ward, S., Rayner, J. M. V., Möller, U., Jackson, D. M., Nachtigall,
W. & Speakman, J. R. 1999. Heat transfer from starlings Sturnus
vulgaris during flight. Journal of Experimental Biology, 202, 1589–
1602.
Weathers, W. W., Hodum, P. J. & Anderson, D. J. 1997. Is the
energy cost of begging by nestling passerines surprisingly low?
Auk, 114, 133–133.
Zahavi, A. 1975. Mate choice: a selection for a handicap. Journal of
Theoretical Biology, 53, 205–214.
Zar, J. H. 1996. Biostatistical Analysis. New York: Prentice Hall.