Content uploaded by M. Clara P. Amorim
Author content
All content in this area was uploaded by M. Clara P. Amorim on Feb 16, 2018
Content may be subject to copyright.
Variability in the mating calls of the Lusitanian toadfish
Halobatrachus didactylus: cues for potential individual
recognition
M. C. P. AMORIM*AND R. O. VASCONCELOS
Unidade de Investigacx˜
ao em Eco-Etologia. I.S.P.A. Rua Jardim do Tabaco 34, 1149-041
Lisboa, Portugal
(Received 12 October 2007, Accepted 22 May 2008)
The mating sounds (boatwhistles) of nesting batrachoidid Halobatrachus didactylus males were
recorded in the Tagus Estuary from piers. Thirteen males with 16 boatwhistles per fish were
analysed for 20 acoustic features. All variables showed larger between-male than within-male
variation and differed significantly among individuals. Discriminant function analyses (DFA)
considering seven of these variables assigned 90–100% of boatwhistles to the correct individual,
depending on the number of males and number of sounds per male included in the model. The
acoustic features that consistently best discriminated individuals were the dominant frequency
of the middle tonal segment of the boatwhistle (P
2
) and dominant frequency modulation,
followed by P
2
pulse period, amplitude modulation and sound duration. These results suggest
the possibility of individual recognition based on acoustic cues. #2008 The Authors
Journal compilation #2008 The Fisheries Society of the British Isles
Key words: acoustic communication; Batrachoididae; individuality; signal variability; sound
production.
INTRODUCTION
The ability to discriminate between individuals or groups of individuals is
important for the establishment of social relations and implies individual dis-
tinctiveness (Bradbury & Vehrencamp, 1998). Individuality in acoustic signal-
ling (vocal signatures) arises when the within-individual variation is smaller
than the variation between individuals in one or more acoustic characteristics
or when individuals differ in the presence or absence of particular vocal fea-
tures (Beecher, 1989; Bee et al., 2001). Individual identification through vocal
signatures can mediate kin recognition (Jouventin et al., 1999), territorial nei-
ghbour recognition (Bee & Gerhardt, 2001), mate-pair recognition (Speirs &
Davis, 1991) and true individual recognition (Sayigh et al., 1999).
The existence of individual characteristics in vocal signals is well known in
various groups of animals including mammals, birds and amphibians (Bee
*Author to whom correspondence should be addressed. Tel.: þ351 218811700; fax: þ351 218860954;
email: amorim@ispa.pt
Journal of Fish Biology (2008) 73, 1267–1283
doi:10.1111/j.1095-8649.2008.01974.x, available online at http://www.blackwell-synergy.com
1267
#2008 The Authors
Journal compilation #2008 The Fisheries Society of the British Isles
et al., 2001; Christie et al., 2004) but has been poorly studied in fishes
(Amorim, 2006). The most common intraspecific variation in fish sounds is
an inverse relation of dominant frequency with fish size that may mediate
individual recognition based on size information (Myrberg & Riggio, 1985;
Myrberg et al., 1993). More elaborate (multi-featured) individual differences
in fish sounds occur in Batrachoididae (toadfishes; Barimo & Fine, 1998;
Edds-Walton et al., 2002; Thorson & Fine, 2002a,b; Fine & Thorson, 2008)
and Mormyridae (weakly electric fishes; Crawford et al., 1997; Lamml &
Kramer, 2006). In both families, territorial males rely on their advertisement
calls to attract females in turbid waters or at night (Winn, 1967; Crawford
et al., 1997). Additionally, these calls are involved in male–male competition
(Winn, 1967; Remage-Healy & Bass, 2005). It has been suggested that mating
calls may promote individual recognition in these animals. For example, differ-
ences in waveform, sound duration and distribution of energy in different har-
monic bands allow clear identification of different male toadfishes [Opsanus tau
(L., 1766) and Opsanus beta (Goode & Bean, 1880)] recorded through passive
acoustics (Edds-Walton et al., 2002; Thorson & Fine, 2002a).
Despite the clear indication of the existence of complex acoustic signals in
batrachoidids and mormyrids that may involve individual recognition, there
is to date no detailed statistical analysis of individual variation in fish sounds.
The goal of the present study is to describe in detail the boatwhistles of nesting
Lusitanian toadfish Halobatrachus didactylus (Bloch & Schneider) males and to
determine which signal properties may potentially mediate individual recogni-
tion. A comparison of the intra- with inter-male variability in 20 acoustic fea-
tures was made. Multivariate statistics were used to identify the best variables
to discriminate between individuals. A preliminary study has shown that differ-
ent nesting H. didactylus males can be recognized by ear and easily identified
through inspection of the spectrogram and oscillogram of their mating sounds
(Amorim et al., 2006). Moreover, this species has an unusual large acoustic rep-
ertoire for fishes, consisting of at least five distinct sound types (dos Santos
et al., 2000; Amorim et al., 2008), suggesting it has a complex acoustic commu-
nication system. This study provides a basis for future playback experiments in
order to test for individual recognition among nesting males and support the
use of acoustic cues in mate attraction and choice.
MATERIALS AND METHODS
STUDY SPECIES
Halobatrachus didactylus is an eastern Atlantic member of the Batrachoididae that oc-
curs in estuaries and coastal lagoons (Roux, 1986). During the reproductive season,
that lasts in Portugal from May to July (Modesto & Can ´
ario, 2003), breeding males
defend nests under rocks in shallow water. Nesting males use an advertisement call
(the boatwhistle) to attract females that results from the contraction of sonic muscles
attached to the swimbladder (dos Santos et al., 2000). Spawning females attach their
eggs to the roof of a nest and leave the area, while the resident males provide parental
care until the young are free-swimming (Roux, 1986; dos Santos et al., 2000). As in
other batrachoidids, a second type of male with different morphometric and endocrine
characteristics is thought to use a sneaking strategy for mating (Brantley & Bass, 1994).
1268 M. C. P. AMORIM AND R. O. VASCONCELOS
#2008 The Authors
Journal compilation #2008 The Fisheries Society of the British Isles, Journal of Fish Biology 2008, 73, 1267–1283
These type II males have larger testes, smaller sonic muscle mass and lower levels of
11-ketotesterone than nesting males (Modesto & Can ´
ario, 2003).
Since territorial males nest close together, boatwhistles are emitted in choruses result-
ing in a very conspicuous acoustic output (dos Santos et al., 2000). In the peak of the
breeding season, a small aggregation of males vocalizing close to the hydrophone can
reach an average of 30 boatwhistles min
1
(Amorim et al., 2006).
The boatwhistle is a tonal multi-harmonic sound lasting c. 800 ms (Amorim et al.,
2006). The fundamental frequency is c. 60 Hz [H1; Fig. 1(b)] and the dominant fre-
quency is typically either the second or the fourth harmonic (Fig. 1) (Amorim et al.,
2006).
RECORDING AND ACOUSTIC ANALYSIS OF
BOATWHISTLES
Several recording sessions lasting from 5 to 10 min were carried out during the mat-
ing season in July 2001 and July 2002 in two areas within the Tagus Estuary, Portugal:
Montijo (38°429N; 8°589W) and Barreiro (38°399N; 9°049W). These areas had been
previously identified as H. didactylus breeding areas (Amorim et al., 2006). Moreover,
nesting males were also observed to call in nests exposed at low spring tides at these
locations (pers. obs.). During recording periods, water temperature ranged between
21–22°C. A hydrophone [High Tech 94 SSQ (High Tech Inc., Gulfport, MS,
U.S.A.), with a sensitivity of 165 dB re 1 V mPa
1
, flat frequency response from 30
FIG. 1. (a), (b), (c) and (d) Oscillograms and sonograms of boatwhistles emitted by four nesting Halobatrachus
didactylus males. The middle tonal phase of a boatwhistle (P
2
) dominant frequency coincides with
the second harmonic (H2) in (a), (b) and (c), and with the fourth harmonic (H4) in male (d), which
are multiples of the fundamental frequency (H1). A power spectrum of phase 2 is given for male (b).
Dur P
1
, Dur P
2
and Dur P
3
duration of phases 1, 2 and 3 of the boatwhistle. The dotted line depicts
total boatwhistle duration. Sampling frequency 44 kHz; FFT size 8192 points; Hamming window.
ACOUSTIC VARIABILITY IN HALOBATRACHUS DIDACTYLUS 1269
#2008 The Authors
Journal compilation #2008 The Fisheries Society of the British Isles, Journal of Fish Biology 2008, 73, 1267–1283
Hz to 6 kHz 1 dB] was lowered from piers in these two locations in several sites
where acoustic activity was evident. The hydrophone was c. 150–200 mm above the
substratum. Recording sites within the same pier were at least 4 m apart and each
recording session was made from a different site. Water depth varied approximately
between 2 and 5 m depending on tide.
Sounds were recorded on tape (Sony TCD-D8, 44 kHz, 16 bit resolution; Sony,
Tokyo, Japan) and the analogue output of the recorder was digitized with a similar
sampling frequency and resolution to a computer with a sound capture device (Edirol
UA-5; Roland, Osaka, Japan). Sound files were analysed with Raven 1.2.1 for Win-
dows (Bioacoustics Research Program, Cornell Laboratory of Ornithology, Cornell,
NY, U.S.A.). A total of seven recording sessions were considered. Sound analysis
was restricted to the males calling close to the hydrophone that presented a high signal
to noise ratio (SNR) (mean ¼29 dB; the minimum SNR considered was of 18 dB)
and a maximum of three individuals was considered per recording session. Only one
recording session was considered per male. Distinction of different individuals between
years was assured by considering males from different location, i.e. Montijo in 2001
(n¼7) and from Barreiro in 2002 (n¼6). Distinction of different males in the same
recording session was based on differences in waveform envelope and relative sound
amplitude that reflected the distance of the calling male to the hydrophone (an example
is depicted in Fig. 2). As expected, relative sound amplitude of a particular male re-
mained constant throughout a recording session since nesting males are stationary
for long periods (dos Santos et al., 2000), especially in the peak of the breeding season
when territories are already established and males call to attract females while caring
for their young (Barimo & Fine, 1998; Knapp et al., 1999). To ensure that the distance
between recording sites (minimum of 4 m) sufficed to prevent considering a male twice,
boatwhistles were played back at the recording locations and recorded simultaneously
at different distances from the speaker with similar gains. The playback audio chain
consisted of a laptop computer, an amplifier (Phoenix Gold QX 4040, Portland, OR,
U.S.A.) and a speaker (Electrovoice UW-30; Lubell Labs Inc. Columbus, OH,
U.S.A.) placed 150 mm above the substratum. Played-back sounds were recorded with
a second laptop computer, a sound capture device (Edirol UA25; Roland) and three
hydrophones (High Tech 94 SSQ) placed c. 150–200 mm above the substratum and
at 05, 15 and 4 m from the speaker. Amplitude of sound playback was determined
by recording a male in a closed nest with the same recording settings as during record-
ings of sound playback. The nest was naturally occupied by the male and its entrance
was closed with a plastic mesh that allowed prey items to enter the nest but prevented
the subject male from abandoning the nest during recordings. The acoustic energy
fell off very rapidly (21 dB loss from 05 m to 4 m from the speaker) and at 4 m away
from the speaker, the boatwhistle could hardly be distinguished from the background
noise (Fig. 3). The marked sound attenuation observed in the study sites, typical of
shallow waters (Fine & Lenhardt, 1983), strongly suggests that the males considered
FIG. 2. Example of boatwhistles produced by two male Halobatrachus didactylus (A and B) that can be
distinguished by the waveform envelope and relative sound amplitude, i.e. distance from the
hydrophone. Arrows point at background boatwhistles.
1270 M. C. P. AMORIM AND R. O. VASCONCELOS
#2008 The Authors
Journal compilation #2008 The Fisheries Society of the British Isles, Journal of Fish Biology 2008, 73, 1267–1283
in the present study are distinct individuals. Other studies have used similar criteria to
the present study to identify unseen toadfishes in sound recordings (Edds-Walton et al.,
2002; Thorson & Fine, 2002a,b) and, in one occasion, male identity was confirmed
through diving (Barimo & Fine, 1998).
A total of 13 males with 16 boatwhistles per fish was analysed for 20 acoustic fea-
tures. The classification used by dos Santos et al. (2000) that considers three distinct
FIG. 3. Oscillograms and sonograms of a Halobatrachus didactylus boatwhistle played back by an
underwater speaker recorded at (a) 05, (b) 15 and (c) 4 m away from the source. The acoustic
energy of boatwhistles suffered an average attenuation of 21 dB from 05 to 4 m from the speaker
and at 4 m could hardly be distinguished from the background noise. Sampling frequency 44 kHz;
fast fourier transform (FFT) size 8192 points; Hamming window.
ACOUSTIC VARIABILITY IN HALOBATRACHUS DIDACTYLUS 1271
#2008 The Authors
Journal compilation #2008 The Fisheries Society of the British Isles, Journal of Fish Biology 2008, 73, 1267–1283
phases in the boatwhistle [beginning (P
1
), middle (P
2
) and end (P
3
)] was adopted. These
three phases differ in pulse period and dominant frequency (dos Santos et al., 2000),
with the pulse period typically decreasing and the dominant frequency increasing from
phases 1–3 (see Table I). The identification of these phases was also based on differen-
ces in sound amplitude (Fig. 1) and fine waveform structure (Fig. 4). The following
acoustic variables were measured: sound duration (ms), measured from the start of
the first pulse (when acoustic energy appears above the background noise) to the end
of the last pulse [Fig. 1 (a)]; duration of the segments P
1
,P
2
and of P
3
(ms) [Fig. 1
(a)]. Relative P
2
duration was calculated by dividing the duration of P
2
by the total
sound duration and was expressed as a percentage; pulse period in P
1
,P
2
and P
3
(ms),
calculated as the average peak-to-peak interval between six consecutive pulse units in
the middle of each segment, except in P
3
that considered the whole segment (Fig. 4);
number of pulses in P
1
,P
2
,P
3
and the total number of pulses in the whole sound; dom-
inant frequency (Hz), i.e. the frequency with maximum energy, was determined in P
1
,
P
2
,P
3
and in the entire sound. Fundamental frequency was calculated as the inverse
of the average pulse period measured in P
1
and P
2
. In batrachoidids, the fundamental
frequency of the mating signals is determined by the rate of contraction of the sonic
muscles attached to the swimbladder (Skoglund, 1961; Fine et al., 2001). These meas-
urements were confirmed with the power spectra [Fig. 1(b)] and were preferred to mea-
suring the fundamental frequency directly because in many fish this frequency band had
little energy. Dominant frequency modulation was calculated by dividing P
1
by P
2
dom-
inant frequencies and fundamental frequency modulation was calculated in a similar
way; amplitude modulation was similarly calculated by dividing the mean amplitude
(RMS) measured for the P
1
segment by the one measured for the P
2
segment; RMS
amplitude is a measurement native to Raven software. Time to maximum amplitude
was measured from the start of the first pulse to the sound peak amplitude; this is also
a measurement native to Raven software.
Temporal variables were measured from oscillograms and the dominant frequencies
from power spectra [fast fourier transform (FFT) size 8192 points; Hamming window].
STATISTICAL ANALYSIS
Mean S.D. values were calculated for the above 20 acoustic features for all males.
Overall means, S.D. and range values were subsequently calculated using each male
mean values for each variable. In order to compare between-male with within-male var-
iability for each acoustic feature the within-male coefficient of variance (C.V.
w
¼
S.D.:mean) was calculated and compared with the between-male coefficient of variation
(C.V.
b
). The C.V.
b
was obtained by dividing the overall S.D. by the respective overall
mean. The ratio C.V.
b
:C.V.
w
was calculated to obtain a measure of relative between-male
variability for each boatwhistle feature. When this ratio assumes values larger than one,
it suggests that an acoustic feature could be used as a cue for individual recognition
(Bee et al., 2001; Christie et al., 2004). Kruskal–Wallis analysis was used to test for dif-
ferences between males for each acoustic variable. Non-parametric statistics were pre-
ferred to parametric ANOVAs due to the lack of homoscedasticity of variances.
Discriminant function analysis (DFA) was carried out using SPSS 15.0 for Windows
(SPSS Inc., Chicago, IL, U.S.A.) as a multivariate tool to determine which acoustic fea-
tures best discriminate between males. DFA also gives a measure of discrimination
accuracy by revealing the percentage of sounds assigned to the correct individual. Only
seven of the 20 acoustic variable were considered for the DFA: total sound duration,
relative P
2
duration, P
2
pulse period, P
2
dominant frequency, dominant frequency mod-
ulation, fundamental frequency modulation and amplitude modulation. These variables
were chosen because they were uncorrelated, had a C.V.
b
:C.V.
w
ratio larger than one and
presented a low C.V.
w
(01; Table I). To assess the predictive accuracy of the models
obtained, a cross-validation method (‘leave-one-out’) was carried out. In this method
each sound is classified by the discriminant functions derived by the n1 remaining
sounds. Because the H. didactylus emits boatwhistles in aggregations of different sizes,
further DFA were performed to explore the variation of classification success with fish
1272 M. C. P. AMORIM AND R. O. VASCONCELOS
#2008 The Authors
Journal compilation #2008 The Fisheries Society of the British Isles, Journal of Fish Biology 2008, 73, 1267–1283
TABLE I. Means, S.D., range, within-male variability (C.V.
w
) and between-male variability (C.V.
b
) for the 20 acoustic features analysed from
13 Halobatrachus didactylus males with 16 sounds each. P
1
,P
2
and P
3
are the initial, middle and end segments in the boatwhistle (after dos
Santos et al., 2000)
Acoustic variables Overall mean S.D. (range) C.V.
w
(mean) C.V.
w
(range) C.V.
b
C.V.
b
:C.V.
w
H
a
Sound duration (ms) 76721689 (4580–10524) 010 005–020 022 223 15561
P
1
duration (ms) 2681393 (2018–3468) 009 004–016 015 171 14999
P
2
duration (ms) 43171180 (2439–6708) 017 006–033 027 162 14006
P
3
duration (ms) 675420(0
0–1471) 059 010–400 062 105 16055
Relative P
2
duration (%) 55650 (443–650) 009 003–015 009 103 9617
Number of pulses P
1
12630(8
6–184) 011 005–018 024 211 16744
Number of pulses P
2
26566 (151–376) 017 007–033 025 151 13980
Number of pulses P
3
4632(0
0–114) 067 020–400 070 104 15803
Total number of pulses 436104 (244–611) 011 006–024 024 209 15828
Pulse period P
1
(ms) 20613 (186–222) 003 001–006 006 198 16125
Pulse period P
2
(ms) 16309 (148–179) 003 001–005 005 179 14897
Pulse period P
3
(ms) 13042(4
3–220) 017 003–039 033 188 9134
b
Dominant frequency P
1
(Hz) 1693565 (700–2598) 007 000–017 033 483 18916
Dominant frequency P
2
(Hz) 1859671 (1130–2647) 008 000–037 036 467 17093
Dominant frequency P
3
(Hz) 1970669 (1238–3079) 014 004–044 034 247 13368
b
Dominant frequency
boatwhistle (Hz)
1828676 (911–2650) 010 000–040 037 355 16158
Dominant frequency modulation 096 020 (065–120) 012 002–038 021 172 14734
Fundamental frequency modulation 079 000 (074–085) 004 002–007 004 110 11883
Amplitude modulation 062 020 (037–086) 009 004–016 024 270 17296
Time to maximum
amplitude (ms)
52892056 (2087–9480) 021 011–057 039 182 15458
a
Results of Kruskal–Wallis tests (d.f. ¼12, n¼207) comparing differences between males for each acoustic feature. All comparisons are significant at P<
0001.
b
The statistical analysis excluded two males with no P
3
segment in the boatwhistle.
ACOUSTIC VARIABILITY IN HALOBATRACHUS DIDACTYLUS 1273
#2008 The Authors
Journal compilation #2008 The Fisheries Society of the British Isles, Journal of Fish Biology 2008, 73, 1267–1283
group size. Ten groups of males were considered. Each group consisted of randomly
chosen males from the initial data set in various sample sizes: three, five, eight and
11 males. Five and 10 boatwhistles randomly chosen per male for each male group were
used to further verify the change in classification accuracy with the number of sounds
considered in the analysis.
RESULTS
BOATWHISTLE STRUCTURE
The mating sounds of the H. didactylus varied considerably in duration
ranging from 317 to 1290 ms (n¼207 sounds analysed from all males), with
average values of 767 ms (Table I). The fundamental frequency (H1) and the
harmonics (multiples of H1) showed a slight frequency modulation that was
more obvious in the higher harmonics (Fig. 1). H1 was the dominant fre-
quency in only one male that exhibited eight of the 16 sounds analysed with
dominant frequencies in the H1 and the remaining in the H2 (see male 5 in
Fig. 5). H4 was the most common dominant frequency (512%) followed by
H2 (449%).
The three segments (P
1
,P
2
and P
3
) that make up the boatwhistle (dos Santos
et al., 2000) were characterized by different durations, pulse periods, relative
amplitude and dominant frequencies (Table I). The tonal phase of the boat-
whistle (P
2
) was the longest segment, lasting on average 56% of the sound,
and exhibited an intermediate pulse period and dominant frequency to P
1
and P
3
. The boatwhistle dominant frequency typically corresponded to P
2
dom-
inant frequency. Pulses in P
1
and P
3
were of a more irregular shape and had
clear starts and ends (Fig. 4), while pulses in P
2
, the tonal segment, were more
regular and fused together sometimes resembling a sinusoidal wave (Fig. 4).
The third boatwhistle segment was more variable in duration, pulse number,
pulse period and dominant frequency than the two previous segments (Fig. 4
and Table I) and was not present in all males. Two males never exhibited
FIG. 4. Oscillograms of the initial (P
1
), the middle tonal place (P
2
) and the end phase (P
3
)of
aHalobatrachus didactylus boatwhistle. Thin arrows indicate the peak amplitude of two consecutive
pulses, i.e. the pulse period. The thick arrow depicts the start of P
3
. Note the differences in the fine
waveform structure among the boatwhistle phases.
1274 M. C. P. AMORIM AND R. O. VASCONCELOS
#2008 The Authors
Journal compilation #2008 The Fisheries Society of the British Isles, Journal of Fish Biology 2008, 73, 1267–1283
the segment P
3
in their boatwhistles and in another male it was present only in
some of the calls analysed.
INDIVIDUALITY
Boatwhistles were distinct between individuals in terms of waveform (ampli-
tude modulation) and spectral characteristics (Fig. 1). Detailed waveform pat-
terns were also distinctive among calling males (Fig. 6). There was a strong
stereotypy in most acoustic variables measured, with half of these features
showing within-male C.V.s 010 (Table I). All the 20 features analysed had
C.V.
b
:C.V.
w
ratios >1, showing that they were more variable among than within
males. Consistently, the Kruskal–Wallis analyses demonstrated significant dif-
ferences among males for all features (Table I), indicating that these acoustic
variables can potentially provide recognition cues to identify calling males.
The larger relative between-male variability (larger C.V.
b
:C.V.
w
ratios) corre-
sponded to the dominant frequencies of P
1
and P
2
and of the whole signal
(Table I and Fig. 5). Most males presented dominant frequencies of P
2
and
of the whole boatwhistle either in the H2 or in the H4 and showed little
within-male variation (Fig. 5). Three males exhibited, however, higher within-
male variability in this feature (males 5, 9 and 13 in Fig. 5) because the dom-
inant frequency in different sounds corresponded to different harmonic bands.
Figure 6 also illustrates that approximately half of the males had lower dom-
inant frequencies in P
1
than in P
2
, whereas the remaining males showed an
opposite trend.
A discriminant function analysis using only seven uncorrelated acoustic fea-
tures generated a significant model (DFA, n¼207, d.f. ¼84, 1159, P<0001).
The three first discriminant functions explained almost all data variability
(91%; Table II). The sound features which weighted most heavily in explaining
variation in the first three discriminant functions were P
2
dominant frequency
followed by dominant frequency modulation for the first function, P
2
pulse
period followed by amplitude modulation for the second function and total
FIG. 5. Mean S.D. dominant frequencies of Halobatrachus didactylus boatwhistle segments initial phase
(P
1
)( ) and middle tonal phase (P
2
)( ) in the 13 males analysed.
ACOUSTIC VARIABILITY IN HALOBATRACHUS DIDACTYLUS 1275
#2008 The Authors
Journal compilation #2008 The Fisheries Society of the British Isles, Journal of Fish Biology 2008, 73, 1267–1283
duration for the third function (Table II). The highest correlations between the
discriminant variables and these discriminant functions were P
2
dominant fre-
quency for the first, amplitude modulation for the second and total sound
duration for the third discriminant functions. Classification success averaged
909% [S.D. (range) ¼104% (688–100%)] and was significantly greater
than the classification expected by chance (aprioriprobability range ¼0072–
0077; Wilcoxon test, n¼13, P<001). A clear separation between individuals
in the two-dimensional space defined by the first two discriminant functions is
depicted in Fig. 7. After cross-validation, the correct classification decreased to
ameanS.D.of85
6168% with values ranging from 563–100%.
Subsequent discriminant analyses, including the same seven acoustic features,
explored variation of classification success with fish group size (three, five, eight
and 11 males) and number of sounds (five and 10) per male. The mean percent-
age of correct classification increased in groups of fewer males from c. 90% (11
males) to 100% (three males) of boatwhistles assigned to the correct individual
(Fig. 8). There was no difference in the classification success between the anal-
yses that included 10 boatwhistles per male and those that included only five
boatwhistles, except in the sample size of eight males where mean correct clas-
sification values were 35% higher in the five boatwhistle analysis (Fig. 8; 95%
CI). Classification success was thus consistently high even when considering
few calls per individual in relatively large groups. For example, the analysis
that included 10 random groups of 11 males with five randomly assigned
FIG. 6. (a), (b) and (c) Boatwhistles from different Halobatrachus didactylus males show differences in the
waveform details in the tonal phase (P
2
).
1276 M. C. P. AMORIM AND R. O. VASCONCELOS
#2008 The Authors
Journal compilation #2008 The Fisheries Society of the British Isles, Journal of Fish Biology 2008, 73, 1267–1283
sounds, revealed a mean correct classification of 925%, which is well above
the classification expected by chance alone. As with the initial DFA, which
considered the whole data set, the acoustic features that loaded more heavily
in the first two discriminant functions of these analyses were P
2
dominant
frequency and dominant frequency modulation (typically in the first discrimi-
nant function) as well as P
2
pulse period, amplitude modulation and sound
duration.
Predictive accuracy of the above models (calculated by the cross-validation
leave-one-out procedure) considering different group sizes of randomly selected
TABLE II. Standardized canonical discriminant function analysis (DFA) coefficients,
eigenvalues and cumulative percentage of variance explained by the first three discrim-
inant functions of a DFA classifying Halobatrachus didactylus males (n¼13) by their
boatwhistle (n¼16) characteristics
Discriminant variables
Discriminant functions
First Second Third
Sound duration 034 035 122
a
Relative P
2
duration (%) 039 004 078
P
2
pulse period 044 083 004
P
2
dominant frequency 155
a
030 056
Dominant frequency modulation 126 057 058
Fundamental frequency modulation 026 060 015
Amplitude modulation 005 077
a
041
Eigenvalue 2268 1156 610
Cumulative % of variance 51177
190
8
P
2
, middle tonal segment of the boatwhistle.
a
Discriminant variable with the highest pooled within-groups correlations with the standardized
discriminant functions.
FIG. 7. Representation of the 13 Halobatrachus didactylus males (boatwhistle group centroids) in the bi-
dimensional space defined by the first two discriminant functions of a discriminant function analysis
considering seven acoustic features. Middle tonal phase (P
2
) dominant frequency correlates with the
first discriminant function and amplitude modulation with the second.
ACOUSTIC VARIABILITY IN HALOBATRACHUS DIDACTYLUS 1277
#2008 The Authors
Journal compilation #2008 The Fisheries Society of the British Isles, Journal of Fish Biology 2008, 73, 1267–1283
males also yielded high estimates of correct classifications. When 10 sounds
were considered per male, the percentage of correct classification varied from
an average of 98–85% in groups of three to 11 males, respectively (Fig. 8).
Similar results were obtained when considering five boatwhistles per male with
classification success decreasing from 95% in groups of three males to 79% in
groups of 11 males (Fig. 8). In conclusion, after cross-validation these analyses
still assigned high percentages of sounds to the correct males and considerably
more than expected by chance.
DISCUSSION
The boatwhistles emitted by the H. didactylus consisted of a relatively long
series of rapidly repeated pulses with average duration around 770 ms. These
sounds exhibited the fundamental frequencies at c. 60 Hz with typical domi-
nant frequencies represented by the second or the fourth harmonic bands.
The boatwhistle of H. didactylus was very similar to the one of O. tau, which
starts with a wide-frequency non-harmonic grunt-like phase caused by slower
and more irregular sonic muscle contractions, followed by a longer tonal
FIG. 8. (a) Variation of mean 95% CI classification success with fish group size (10 groups of three, five,
eight and 11 randomly chosen Halobatrachus didactylus males) and number of sounds per individual: 5
( ) and 10 ( ) boatwhistles. (b) Similar percentages of classification success obtained after cross-
validation ( , the classification success expected from randomly assigning calls to the different male).
1278 M. C. P. AMORIM AND R. O. VASCONCELOS
#2008 The Authors
Journal compilation #2008 The Fisheries Society of the British Isles, Journal of Fish Biology 2008, 73, 1267–1283
segment (Fine, 1978). Boatwhistles of the latter species are, however, shorter
(200–500 ms) and have higher fundamental frequencies (c. 200 Hz) than the
boatwhistles of the H. didactylus (Fine, 1978; Barimo & Fine, 1998; Edds-Wal-
ton et al., 2002). Other well-studied batrachoidids produce more divergent calls.
Opsanus beta emits a more complex courtship sound with fundamental frequen-
cies around 270–280 Hz that starts with zero to three grunts followed by a long
tonal (‘boop’) note and up to three shorter boops lasting over a second (Thor-
son & Fine, 2002a). Nesting Porichthys notatus Girard, males emit remarkably
long courtship sounds (‘hums’) that last from seconds to over an hour, with
fundamental frequencies around 100 Hz (Ibara et al., 1983; Brantley & Bass,
1994).
Clear differences were found among boatwhistles attributed to different
males that can potentially be used in individual recognition. All variables were
significantly more variable between than within males and thus could all poten-
tially provide cues to identify individuals. A DFA using a sub-set of the initial
acoustic features assigned boatwhistles to the correct male in 91% of cases, and
in 86% of cases after cross-validation, showing a high predictive accuracy.
Classification success of boatwhistles varied with sample size (number of males
and number of sounds per male) but remained high even when considering few
calls per male in large groups. In accordance with the observed C.V.
b
:C.V.
w
ratios, the most important variables to allow male identification were P
2
dom-
inant frequency followed by dominant frequency modulation (the ratio between
P
1
and P
2
dominant frequencies). P
2
pulse period, amplitude modulation and
total boatwhistle duration were also consistently important to discriminate
among individuals in the various DFAs.
In the field, males can probably only detect a maximum of eight males at
a time (maximum size of a chorus; unpubl. data) and call often in duets or
in trios, thus potentially making the task of individual recognition simpler than
the 13 males considered in the present study. Moreover, because males call
often at rates of c. 10 boatwhistles min
1
(pers. obs.) they will easily experience
more than the 16 calls from a neighbour having more opportunity to access
distinct features from stationary nesting conspecifics.
In order for the above five features to be good candidates for individual
identification, they should propagate well through the environment and should
also be recognized by the central nervous system of the receiver. Sound prop-
agation in shallow water can result in signal degradation over short distances,
including sound pressure level and frequency attenuation, and temporal pat-
terning loss (Mann, 2006; also see Fig. 3). Boatwhistles are thought to function
both to announce territorial ownership and position to other males and to
attract females as prospective mates (Winn, 1967). Because males can nest
<05 m apart (pers. obs.) environmental degradation of call properties should
not impose a major restriction between male neighbours. The effect of attenu-
ation and signal degradation, however, should be important for female attrac-
tion. This problem has probably been overcome by the increased acoustic
output resulting from H. didactylus male choruses. If there is mate choice based
on acoustic signals, it probably takes place when females are already in close
range to males with access to minimally degraded signals.
ACOUSTIC VARIABILITY IN HALOBATRACHUS DIDACTYLUS 1279
#2008 The Authors
Journal compilation #2008 The Fisheries Society of the British Isles, Journal of Fish Biology 2008, 73, 1267–1283
Differences among males in frequency attributes should be perceived by the
H. didactylus although batrachoidids are hearing generalists, i.e. they lack
morphological specializations that enhance the detection of the sound pressure
component of the acoustic signals (Fay & Simmons, 1999). According to Vas-
concelos et al. (2007), dominant and fundamental frequencies of boatwhistles
match the best hearing range of the species. P
2
dominant frequency differs
among individuals between >10 and 100% (Fig. 5), thus falling within the
range of frequency discrimination ability of hearing generalists, which is gener-
ally slightly >10% difference (Fay & Simmons, 1999). Differences in frequency
modulation should also be detected because disparities between P
1
and P
2
dom-
inant frequencies are in the majority of the studied males >10% (Fig. 5). This
variable shows high interindividual variability and dominant frequency can be
modulated upward or downward (Table I and Fig. 5). The large differences in
signal duration found in the present study (but not pulse period) should also
fall into the hearing discrimination abilities of H. didactylus since other batra-
choidids can detect small differences in signal duration (McKibben & Bass,
1998).
Acoustic recognition systems have arisen in situations where crowding, noisy
backgrounds (such as in dense colonies of birds) or darkness reduce the roles of
olfactory and visual cues or increase the risk of confusion (Beecher, 1989;
Sayigh et al., 1999). Likewise, acoustic recognition is also beneficial when vocal
animals defend long-term territories. In this context, individual recognition is
adaptive because animals can direct less aggression to familiar neighbours,
which are less likely to intrude into their territories. This phenomenon, known
as the ‘dear enemy effect’ (Fischer, 1954), has been described in several animals
(Temeles, 1994). In fishes, acoustic recognition has only been demonstrated in
a coral reef species that breed in dense colonies. Myrberg & Riggio (1985)
tested the ‘dear enemy effect’ with the bicolour damselfish Stegastes partitus
(Poey) and verified that males can recognize territorial neighbours based on
acoustic cues, probably the dominant frequency that decreased pronouncedly
with male size. Likewise, H. didactylus males establish long-term territories
forming dense breeding aggregations. In addition, they live in turbid environ-
ments where vision is impaired. Consequently, being able to discriminate
among different individuals would be beneficial in this species. A comparable
social system where individual recognition has been demonstrated is found in
anurans. Frogs and toads also form breeding choruses and establish long-term
territories during the reproductive season and may show vocal individual rec-
ognition. For example, male bullfrogs Rana catesbeiana, Shaw can learn about
individually distinct acoustic features of neighbours’ calls and a neighbour’s
position by repeatedly hearing the call from a particular location (Bee &
Gerhardt, 2001).
This study was based on short periods of recordings from unseen fish.
Although the identity of the sound producers cannot be completely ascertained,
the present results suggest that there is enough information in the mating calls
of the H. didactylus to promote individual recognition. Future work carried out
with fully identified males will need to address whether boatwhistle character-
istics are constant over longer periods of time and whether they are related to
male features.
1280 M. C. P. AMORIM AND R. O. VASCONCELOS
#2008 The Authors
Journal compilation #2008 The Fisheries Society of the British Isles, Journal of Fish Biology 2008, 73, 1267–1283
Special thanks are given to F. Almada and J. Marques for helping with recordings,
to P. Fonseca and J. M. Simo
˜es for their help with the sound playbacks and comments
on the manuscript. This study was supported by the pluriannual programme (UI&D
331/94) / FCT, and the grants POSI SFRH/BPD/14570/2003 (MCPA) and SFRH/
BD/30491/2006 (ROV) of FCT.
References
Amorim, M. C. P. (2006). Diversity of sound production in fish. In Communication in
Fishes, Vol. I (Ladich, F., Collin, S. P., Moller, P. & Kapoor, B. G., eds), pp. 71–
104. Enfield: Science Publishers.
Amorim, M. C. P., Vasconcelos, R. O., Marques, J. F. & Almada, F. (2006). Seasonal
variation of sound production in the Lusitanian toadfish, Halobatrachus didactylus.
Journal of Fish Biology 69, 1892–1899. doi: 10.1111/j.1095-8649.2006.01247.x
Amorim, M. C. P., Simo
˜es, J. M. & Fonseca, P. J. (2008). Acoustic communication in the
Lusitanian toadfish. Halobatrachus didactylus: evidence for an unusual large vocal
repertoire. Journal of Marine Biological Association of the United Kingdom 88, 1069–
1073.
Barimo, J. F. & Fine, M. L. (1998). Relationship of swim-bladder shape to the
directionality pattern of underwater sound in the oyster toadfish. Canadian Journal
of Zoology 76, 134–143.
Bee, M. A. & Gerhardt, H. C. (2001). Neighbour–stranger discrimination by territorial
male bullfrogs (Rana catesbeiana): II. Perceptual basis. Animal Behaviour 62, 1141–
1150.
Bee, M. A., Kovich, C. E., Blackwell, K. J. & Gerhardt, H. C. (2001). Individual
variation in advertisement calls of territorial male green frogs, Rana clamitans:
implications for individual discrimination. Ethology 107, 65–84. doi: 10.1046/
j.1439-0310.2001.00640.x
Beecher, M. D. (1989). Signalling systems for individual recognition: an information
theory approach. Animal Behaviour 38, 248–261.
Bradbury, J. W. & Vehrencamp, S. L. (1998). Principles of Animal Communication.
Sunderland: Sinauer Associates.
Brantley, R. K. & Bass, A. H. (1994). Alternative male spawning tactics and acoustic
signals in the plainfin midshipman fish, Porichthys notatus (Teleostei, Batrachoidi-
dae). Ethology 96, 213–232.
Christie, P. J., Mennill, D. J. & Ratcliffe, L. M. (2004). Chickadee song structure is
individually distinctive over long broadcast distances. Behaviour 141, 101–124.
Crawford, J. D., Cook, A. P. & Heberlein, A. S. (1997). Bioacoustic behavior of African
fishes (Mormyridae): potential cues for species and individual recognition in
Pollimyrus.The Journal of the Acoustical Society of America 102, 1200–1212.
Edds-Walton, P. L., Mangiamele, L. A. & Rome, L. C. (2002). Variation of pulse
repetition rate in boatwhistle sounds from oyster toadfish Opsanus tau around
Waquoit bay, Massachusetts. Bioacoustics 13, 153–173.
Fay, R. R. & Simmons, A. M. (1999). The sense of hearing in fishes and amphibians. In
Comparative Hearing: Fish and Amphibians (Fay, R. R. & Popper, A. N., eds), pp.
269–318. New York: Springer.
Fine, M. L. (1978). Seasonal and geographical variation of the mating call of the oyster
toadfish Opsanus tau L. Oecologia 36, 45–57.
Fine, M. L. & Lenhardt, M. L. (1983). Shallow-water propagation of the toadfish mating
call. Comparative Biochemistry and Physiology A 76, 225–231.
Fine, M. L. & Thorson, R. F. (2008). Effects of passive acoustics for assessing behavioral
interactions in individual toadfish. Transactions of the American Fisheries Society
137, 627–637.
Fine, M. L., Malloy, K. L., King, C., Mitchell, S. L. & Cameron, T. M. (2001).
Movement and sound generation by the toadfish swimbladder. Journal of
Comparative Physiology A 187, 371–379.
ACOUSTIC VARIABILITY IN HALOBATRACHUS DIDACTYLUS 1281
#2008 The Authors
Journal compilation #2008 The Fisheries Society of the British Isles, Journal of Fish Biology 2008, 73, 1267–1283
Fischer, J. B. (1954). Evolution and bird sociality. In Evolution as a Process (Huxley, J.,
Hardy, A. C. & Ford, E. B., eds), pp. 71–83. London: Allen and Unwin.
Ibara, R. M., Penny, L. T., Ebeling, A. W., van Dykhuizen, G. & Caillet, G. (1983). The
mating call of the plainfin midshipman, Porichthys notatus.InPredators and Prey
(Noakes, D. G. L., Lindquist, D. G., Helfman, G. S. & Ward, J. A., eds), pp. 205–
212. The Hague: Junk Press.
Jouventin, P., Aubin, T. & Lengagne, T. (1999). Finding a parent in a king penguin
colony: the acoustic system of individual recognition. Animal Behaviour 57, 1175–
1183.
Knapp, R., Wingfield, J. C. & Bass, A. H. (1999). Steroid hormones and paternal care
in the plainfin midshipman fish (Porichthys notatus). Hormones and Behavior 35,
81–89.
Lamml, M. & Kramer, B. (2006). Differentiation of courtship songs in parapatric sibling
species of dwarf stonebashers from southern Africa (Mormyridae, Teleostei).
Behaviour 143, 783–810.
Mann, D. A. (2006). Propagation of fish sounds. In Communication in Fishes, Vol. I
(Ladich, F., Collin, S. P., Moller, P. & Kapoor, B. G., eds), pp. 107–120. Enfield:
Science Publishers.
McKibben, J. R. & Bass, A. H. (1998). Behavioral assessment of acoustic parameters
relevant to signal recognition and preference in a vocal fish. The Journal of the
Acoustical Society of America 104, 3520–3533.
Modesto, T. & Can ´
ario, A. V. M. (2003). Morphometric changes and sex steroid levels
during the annual reproductive cycle of the Lusitanian toadfish, Halobatrachus
didactylus.General and Comparative Endocrinology 131, 220–231.
Myrberg, A. A. Jr & Riggio, R. J. (1985). Acoustically mediated individual recognition
by a coral reef fish (Pomacentrus partitus). Animal Behaviour 33, 411–416.
Myrberg, A. A. Jr, Ha, S. J. & Shamblott, H. (1993). The sounds of bicolor damselfish
(Pomacentrus partitus): predictors of body size and a spectral basis for individual
recognition and assessment. The Journal of the Acoustical Society of America 94,
3067–3070.
Remage-Healy, L. & Bass, A. H. (2005). Rapid elevations in both steroid hormones and
vocal signalling during playback challenge: a field experiment in gulf toadfish.
Hormones and Behavior 47, 297–305.
Roux, C. (1986). Batrachoididae. In Fishes of the North-eastern Atlantic and the
Mediterranean, Vol. III (Whitehead, P. J. P., Bauchot, M.-L., Hureau, J. C.,
Nielsen, J. & Tortonese, E., eds), pp. 1360–1361. Paris: Unesco.
dos Santos, M., Modesto, T., Matos, R. J., Grober, M. S., Oliveira, R. F. & Canario, A.
(2000). Sound production by the Lusitanian toadfish, Halobatrachus didactylus.
Bioacoustics 10, 309–321.
Sayigh, L. S., Tyack, P. L., Wells, R. S., Solows, A. R., Scott, M. D. & Irvine, A. B.
(1999). Individual recognition in wild bottlenose dolphins: a field test using
playback experiments. Animal Behaviour 57, 41–50.
Skoglund, C. R. (1961). Functional analysis of swimbladder muscles engaged in sound
production in the toadfish. Journal of Biophysical and Biochemical Cytology 10,
187–200.
Speirs, A. H. & Davis, L. S. (1991). Discrimination by the Ad
elie penguins, Pygoscelis
adeliae, between the loud mutual calls of mates, neighbours and strangers. Animal
Behaviour 41, 937–944.
Temeles, E. J. (1994). The role of neighbours in territorial systems: when are they ‘dear
enemies’? Animal Behaviour 47, 339–350.
Thorson, R. F. & Fine, M. L. (2002a). Crepuscular changes in emission rate and
parameters of the boatwhistle advertisement call of the gulf toadfish, Opsanus beta.
Environmental Biology of Fishes 63, 321–331.
Thorson, R. F. & Fine, M. L. (2002b). Acoustic competition in the gulf toadfish Opsanus
beta: acoustic tagging. The Journal of the Acoustical Society of America 111, 2302–
2307.
1282 M. C. P. AMORIM AND R. O. VASCONCELOS
#2008 The Authors
Journal compilation #2008 The Fisheries Society of the British Isles, Journal of Fish Biology 2008, 73, 1267–1283
Vasconcelos, R. O., Amorim, M. C. P. & Ladich, F. (2007). Effects of ship noise on the
auditory sensitivity and acoustic communication in the Lusitanian toadfish. The
Journal of Experimental Biology 210, 2104–2112.
Winn, H. E. (1967). Vocal facilitation and the biological significance of toadfish sounds.
In Marine Bio-Acoustics, Vol. 2 (Tavolga, W. N., ed.), pp. 283–304. Oxford:
Pergamon Press.
ACOUSTIC VARIABILITY IN HALOBATRACHUS DIDACTYLUS 1283
#2008 The Authors
Journal compilation #2008 The Fisheries Society of the British Isles, Journal of Fish Biology 2008, 73, 1267–1283
