Вы находитесь на странице: 1из 13

Behav Ecol Sociobiol (2004) 55:531–543

DOI 10.1007/s00265-003-0724-y

ORIGINAL ARTICLE

Stephanie L. Watwood · Peter L. Tyack ·


Randall S. Wells

Whistle sharing in paired male bottlenose dolphins, Tursiops truncatus

Received: 30 January 2003 / Revised: 28 August 2003 / Accepted: 23 October 2003 / Published online: 22 January 2004
 Springer-Verlag 2004

Abstract The signature whistle hypothesis states that learning may enable a similar vocal convergence. Whistle
dolphins produce highly stereotyped, individually distinc- similarity was rated very high between partners and low
tive whistles when in isolation. The presence of signature between non-partners by both the quantitative technique
whistles has been called into question by recent studies and human observers. This suggests that as in songbirds
proposing that dolphins produce a shared, simple upsweep and some other mammals, adult male bottlenose dolphins
whistle when in isolation, and that whistles produced by may use vocal learning to converge on similar whistles as
socializing dolphins are shared across individuals and they develop affiliative social relationships.
social groups. This shared repertoire hypothesis suggests
that when two animals produce the same whistle type, it is Keywords Bottlenose dolphin · Vocal convergence ·
due to sharing the same common repertoire rather than Whistle matching · Tursiops truncatus
one animal learning to produce the whistle of another.
One difference between studies supporting or denying the
existence of signature whistles is the method used to Introduction
classify whistle types. We examined whistle production
by 17 free-ranging bottlenose dolphins while temporarily The signature whistle is defined as an individually
restrained. We used both a quantitative comparison distinctive, stereotyped whistle that is often the most
technique similar to that used to support the shared predominant whistle type produced by an individual
repertoire hypothesis and human judges to classify bottlenose dolphin when in isolation (Caldwell and
whistle types and quantify similarity between types. Caldwell 1965). Caldwell and Caldwell (1965) first
Contrary to recent studies that emphasize shared whistles, presented evidence of individualized whistle contours
overall whistle sharing between isolated individuals was (the rising and falling pattern of frequency modulation
low (25%) and a simple upsweep did not account for the over time) in the recorded whistles of five wild-born
most common whistle type in half of the animals. Some bottlenose dolphins held at Marineland in Florida, USA.
species of birds, bats, and primates with stable social The presence of signature whistles has been confirmed in
groups use vocal learning to converge over time to one over 120 captive bottlenose dolphins (Caldwell et al.
common group distinctive call type. We examined whistle 1990). Several studies have reported signature whistles in
similarity between adult male dolphins that are partners in free-ranging populations as well (Sayigh et al. 1990;
a close social alliance in order to test whether vocal Herzing 1996).
A few researchers have called the existence of
signature whistles into question, and have even termed
Communicated by G. Wilkinson the idea of signature whistles a “fallacy” (McCowan and
S. L. Watwood ()) · P. L. Tyack Reiss 1995; 2001). McCowan and Reiss (2001) suggested
Biology Department, that captive dolphins produce the same whistle types in all
Woods Hole Oceanographic Institution, contexts studied, and these whistle types are shared by
1–32 Redfield MS 34, Woods Hole, MA 02543, USA many individuals across different social groups. They
e-mail: swatwood@whoi.edu
point to early studies of odontocete whistles in which
Tel.: +1-508-2893463
Fax: +1-508-4572134 whistle types were reported to be shared not only across
social groups but across different species such as
R. S. Wells bottlenose dolphins, common dolphins, and pilot whales
Chicago Zoological Society, (Dreher and Evans 1964). McCowan and Reiss (1995)
c/o Mote Marine Laboratory, 1600 Ken Thompson Pkwy, Sarasota, also reported that the most common whistle produced by
FL 34236, USA
532

dolphins when in isolation was a simple upsweep whistle through learning, rather than by choosing the same call
(their type 2 whistle). They suggested that any individual type out of an established repertoire (Mammen and
variability in whistles is due to differences in acoustic Nowicki 1981; Boughman 1998). Dolphins have demon-
parameters of this type 2 whistle produced by different strated abilities of vocal learning (Richards et al. 1984),
individuals, a pattern that is similar to the isolation calls and preliminary evidence suggests that shared whistles
of many terrestrial animals. McCowan and Reiss (2001) may serve a similar function in bottlenose dolphins to
suggested that differences between their studies and shared calls in terrestrial animals. Smolker and Pepper
studies reporting signature whistles are based on differ- (1999) described whistle convergence among a trio of
ences in methods used for collecting and categorizing allied male dolphins in Shark Bay, Australia. As the
whistles, and biases by researchers who know the identity alliance bond between the three males strengthened, the
of the animal that produced each whistle prior to whistles produced by the males became more similar.
categorizing the whistles. Although earlier papers de- Alliance bonds appear to grow out of relationships
scribing signature whistles relied primarily on a single formed between non-siblings during their subadult years.
observer sorting spectrograms (who often knew back- The bonds may last for the lifetime of the adult males. In
ground information about the whistles they were sorting), Sarasota Bay dolphins, when one member of an alliance
more recent papers have used multiple judges and dies, the surviving male often pairs up with another single
examined agreement between them (Sayigh et al. 1995; male to form a new pair (Wells 1991). Life history
Janik 1999). These judges were nave to the data observations and preliminary genetic analyses indicate
collection methods and any ancillary information about that alliance partners are not closely related (Duffield and
the whistles, and therefore were unbiased in their judging Wells 2002). Several hypotheses have been suggested for
of whistle similarity. Therefore it seems unlikely that the function of the dolphin alliance, including increased
biased judges are the source of any discrepancies in predator detection and defense, cooperative foraging,
results. territory defense against other males, and increased
Both the groups that accept and those that deny mating opportunities (Wells 1991; Connor et al. 1992;
signature whistles agree that bottlenose dolphins may Owen et al. 2002). Previous studies examining signature
share some whistle types. Tyack (1986) and Janik and whistles in free ranging bottlenose dolphins have primar-
Slater (1998) reported that captive dolphins produced in ily involved adult females and their calves (Sayigh et al.
addition to their signature whistle shared whistles such as 1990, 1995), and whistle use by adult male dolphins has
an upsweep, termed a rise whistle. Rise whistles appear been largely ignored (see Smolker and Pepper 1999).
similar to the type 2 whistle of McCowan and Reiss Patterns of social relationships differ sharply between
(1995, 2001). However, Janik and Slater (1998) reported adult male and adult female bottlenose dolphins in the
that rise whistles were only common when dolphins were wild. After leaving mixed-sex subadult groups, females
swimming together. When isolated, captive dolphins often associate with other females in a similar reproduc-
tended to produce signature whistles, which were more tive state, potentially including maternally related fe-
complex and stereotyped than the simple rise whistle. males, while most males associate primarily with their
Dolphins are adept at imitating synthetic sounds alliance partners (Wells 1991; Connor et al. 1992). Due to
(Richards et al. 1984; Reiss and McCowan 1993), and the different social histories of male and female dolphins,
this ability may lead to whistle sharing through imitation. the whistles produced by dolphins may differ between the
Tyack (1986) reported that around 20% of the whistles sexes. The finding of whistle convergence among a trio of
produced by each of two captive dolphins were imitations allied males suggests a more plastic whistle repertoire
of the other’s signature whistle. Free-ranging dolphins than that of females, who have been observed consistently
have also been reported to imitate the whistles of producing the same whistle type when isolated at different
conspecifics (Janik 2000); however, it is more difficult times over a 12-year period (Sayigh et al. 1990). Sayigh et
to demonstrate true imitation of natural whistles from al. (1990) reported an increase in the whistle repertoires
wild dolphins. An alternative approach suggests that all of young male dolphins after they separated from their
the whistles produced by dolphins come from a common mothers and suggested that this increase in whistle
shared whistle repertoire and that what appears to be diversity may parallel an increase in the formation of
imitation is simply animals repeating whistles of the same social relationships. Male dolphins, therefore, are an ideal
call type. More thorough recordings of individual reper- group for examining changing patterns of whistle use and
toires are needed to distinguish these hypotheses. whistle sharing as they form strong bonds.
An important source of data in support of vocal This study examines recordings of free-ranging adult
imitation of natural calls involves the convergence of male bottlenose dolphins from Sarasota Bay, Florida,
calls that occurs in animals that maintain contact with USA, while under brief restraint for health assessment.
individuals in changing social groups. In these groups, The goal of this study is to compare the similarity of
individuals forming close associations modify particular whistle repertoires of male bottlenose dolphins between
calls that may serve as contact calls, or to coordinate alliance partners and non-partners. If the whistles of
group movements, defense of resources, or group forma- bonded males are similar, and if there is little overlap in
tion (Mammen and Nowicki 1981; Brown et al. 1988; whistle repertoires of non-partner males, then this
Elowson and Snowdon 1994). This process occurs supports the hypothesis that dolphins modify their
533

vocalizations as their social relationships change. We use Recordings from 15 male and 2 female adult bottlenose
both human judges and a quantitative technique to dolphins were collected between 1985 and 2001 (see Table 1 for
male subjects). The males composed nine distinct alliances.
categorize whistles of wild bottlenose dolphins to deter- Members of an alliance have a simple ratio coefficient of
mine if different analysis methods yield different results, association (COA) greater than 0.50 (Owen et al. 2002). Sarasota
as suggested by McCowan and Reiss (2001). We compare Bay males in long-term pair bonds have a mean annual simple ratio
whistles of all males for individual whistle stereotypy (for COA of 0.82, while the mean COA for two individuals associating
at random in a given year is 0.01 (Owen et al. 2002). In Sarasota
evidence of signature whistles) and whistle type sharing Bay, adult males generally do not associate with males other than
(for evidence of a shared whistle repertoire). In particular, their partners. The average age difference between alliance partners
we will explore (1) the degree of distinctiveness of was 1.4 years, and the average age at first pairing was 10.2 years
whistle types produced by individuals, (2) if a simple (for those whose time of alliance formation is known). As seen in
upsweep is the most common whistle type produced by Table 1, some of the males formed new pairs after their original
partner died or disappeared. Females FB07 and FB79 were
wild male bottlenose dolphins in isolation, (3) the extent temporarily restrained and recorded in the same sessions as some
of overlap in the whistle repertoires of wild male of the males. Their whistles are included in the analysis to test
bottlenose dolphins, and (4) the agreement between whether being recorded in the same session affects whistle
quantitative and visual assessments of whistle similarity similarity independent of the pair bond. FB79 was restrained at
the same time as FB66, FB76, FB94, and FB14 in 1989. FB07 and
between individuals. FB46 were restrained and recorded together in 1991.

Digitization
Methods
The recordings were played back for digitization using either a
Study subjects Samsung SV-300 W videocassette recorder or a Yamaha KX-500U
stereo cassette recorder. They were filtered with a Frequency
A temporary capture-release program for assessing the health of the Devices 9002 programmable high-pass filter set at 3 kHz. Sounds
dolphin community provided brief opportunities to record the were digitized with Dalanco Spry analog-to-digital conversion
vocalizations of restrained animals (for details see Sayigh et al. board (sampling frequency 80 kHz) or a Creative SB Live!
1990). Recordings were made using a High-Tech, Inc. hydrophone soundcard and Cool Edit Pro (Syntrillium Software; sampling
(model HTI-94SSQ or HTI-96MIN) potted into a RTV suction cup. frequency 48 kHz). Whistles were extracted from the digitized file
The soft suction cup was placed on the melon of the animal to be manually or using an automatic detector (Fripp 1999). The detector
recorded, yielding a high signal-to-noise ratio (SNR). Recordings extracted sounds whose power output was greater than a preset
used in this analysis were from periods when the recorded animal threshold background noise level. This threshold was set by first
was out of eyesight of any other animals, but often in acoustic selecting a 1-s section of the digitized file that was representative of
range. Recordings prior to 1989 were made with a Sony TC-D5 M general background noise (and contained no dolphin vocalizations).
or a Marantz PMD-430 stereo cassette recorder (nominal frequency The threshold level was then set at five standard deviations above
response of 30–15,000 Hz) onto Maxell UDXLII cassette tapes. the mean power of this noise sample. The selection of this threshold
Recordings made in 1989 and after were made with a Panasonic of five standard deviations was determined from trial and error with
AG-6400 stereo hi-fi VCR (nominal frequency response of 20– a separate data set (Fripp 1999). Since the automatic detector
20,000 Hz) onto standard VHS tapes. selected any sounds that were above a certain threshold, the

Table 1 Pairing history for the Alliancef Animal Age difference Age at alliance formation Years with COAg >0.5
15 subject males in the study.
Age estimates courtesy of Aleta 1 FB16 0 6 1987–1989a
Hohn (Hohn et al. 1989) FB10 6
2 FB10 3 10 1991–present
FB46 13
3 FB14 3 18 1991–present
FB94 21
4 FB36 2 12 1984–1999b
FB38 10
5 FB26 2 25 1982–present
FB48 23
6 FB66 1 9 1985–present
FB76 8
7 FB44 1 11 1980–1982
FB18 12 1984–1992c
8 FB44 0 19 1992–1997d
FB62 19
9 FB77 1 6 1980–1987e
FB62 7
a
FB16 disappeared in 1989
b
FB38 died in 1999
c
FB18 disappeared in 1992
d
FB62 disappeared in 1997
e
FB77 died in 1987
f
Pairs 3, 4, 5, 6,and 8 were used in the human judges comparison
g
Coefficient of association
534
detector often isolated segments of water noise, burst-pulsed calls,
echolocation, and other non-whistle sounds. Manual sorting of the
extracted segments separated the whistles from other sounds.

Similarity analysis

We employed two techniques to determine whistle similarity: a


quantitative classification technique on a large whistle sample
(modified contour similarity) and human observers judging a
smaller sample of whistles (human judges). Caldwell et al. (1990)
reported that the signature whistle of a particular dolphin can vary
in terms of frequency, duration, and number of loops while still
maintaining a distinctive frequency–time contour pattern. They
reported that the number of loops produced per signature whistle
may vary with behavioral context for each dolphin. They also found
that older animals on average produced more loops in each whistle
than younger animals, and that older age classes produced the
largest maximum number of loops in a single whistle (Caldwell et
al. 1990). Therefore, the number of loops may be a factor of the
animal’s age and behavioral context, and the same contour with
different loop numbers may represent variations in the same signal. Fig. 1A, B Whistle recordings from two animals. A Whistles with
One problem with current quantitative techniques is that they are breaks between loops within single whistles; B whistles with no
incapable of dealing with changes in the number of loops that make breaks between loops within single whistles
up multiloop whistles. Changes in loop number do not pose
problems for human judges, and Janik (1999) demonstrated that
human observers were much better at classifying stereotyped
whistles than several quantitative methods. For the human judges along the contour, and recorded the frequency at each point. This
technique, three whole whistles not divided into the component effectively normalized the whistle loops in time. The matrix of 100
loops were compared for each of ten animals (for animals see frequency measurements for each whistle loop was used to generate
Table 1). a Pearson product-moment correlation matrix in Systat 7.0 (Systat).
For the modified contour similarity technique, 15 whistles each The correlation coefficient for each whistle loop comparison is a
from 17 animals were compared. Multi-loop whistles were divided function of the linear relationship between contour lines. This
into their single loop components and the single loops were used as determines similarity of contour shape, without specific time or
the unit of comparison for the quantitative analysis. Previous frequency information. To sort the whistle loops into categories, we
studies have used breaks in the line of the fundamental frequency analyzed the correlation matrix by hierarchical clustering analysis
contour on the spectrogram as an indicator of where whistles begin using the within-groups average linkage method. We then deter-
and end. However, whistles are amplitude modulated, and factors mined the moat index (Podos et al. 1992; Smolker and Pepper
such as distance between the animal and hydrophone, recording 1999), which is calculated for each clustering level by subtracting
level, SNR, and dynamic range of the spectrogram all affect the maximum within-group linkage distance from the minimum
whether faint portions of a whistle appear in a spectrogram. between-group linkage distance and then averaging the differences
Therefore, to be robust against changes in SNR, we used the (Podos et al. 1992). The number of clusters that maximizes the
following criteria. There are often breaks in the contour visible on a moat index is the level of clustering that was accepted. This level of
spectrogram between loops within a single whistle. When there clustering was taken to indicate that there were 125 distinct whistle
were breaks, loops were distinguished from separate whistles by the loop types (WLTs) produced by the dolphins.
time between loops. The inter-loop interval was characteristically
short, while the spacing between different whistles was much
greater and more variable. Figure 1 shows spectrograms of whistles Human judges technique
from two animals. In Fig. 1A, the three whistles shown would be
broken into three loops, six loops, and four loops. However, there Whistles from ten male dolphins were used for this comparison.
were not always breaks in the contour between loops. Therefore, Only a limited number of comparisons are possible with human
within a single contour on the spectrogram, all repetitive patterns of judges because of the time it takes to make the pairwise judgments
frequency modulation were considered to be loops. In Fig. 1B, the and the rapidly increasing number of comparisons required for all
three whistles shown would be broken down into one loop, four possible pairs of N whistles. Therefore, only a small subset of
loops, and three loops. The 15 whistles from 17 animals resulted in whistles was used for this analysis. Whole whistles (not divided
a total of 631 whistle loops that were used in the quantitative into loops) from each individual were visually categorized as
comparison. described in Sayigh et al. (1990). S.L. Watwood sorted the whistles
and selected the most common whistle type for each animal. The
predominant whistle type was the fundamental frequency contour
Modified contour similarity technique shape that was most commonly produced by a given animal,
regardless of the number of loops, and was the only whistle used in
The quantitative method we used to categorize whistles followed this analysis for each individual. There was a potential bias in this
the method described by Smolker and Pepper (1999), which was a approach since the identity of the animal that produced each whistle
modified version of the contour similarity technique first described was known. However, for each animal, the determination of which
by McCowan (1995). We chose this technique because it is very whistle type occurred most often was readily apparent. Figure 2
similar to the technique used by McCowan and Reiss (2001) and shows all the whistles that were recorded from male FB48 in 1987
has been used by the only other study closely examining whistle use and female FB07 in 1991 while restrained. The outlined boxes
by wild male dolphins (Smolker and Pepper 1999). The funda- indicate categories of whistles. The whistles in box 1 for each
mental frequency of each whistle loop was first traced by hand animal were considered to represent the most common whistle type
using Matlab 5.3 (The MathWorks), which produced a contour that for each, while the whistles assigned to other numbers occurred less
preserved the time and frequency information from the spectro- often. Although the number of loops differs among the box 1
gram. A separate program then extracted 100 equally spaced points whistles, the overall shape of each contour appears similar within
535
Fig. 2 Whistles recorded from
A FB48 in 1987 and B FB07 in
1991. Boxes indicate categories
determined through visual in-
spection. In A the whistles in
box 1 are considered to be the
most common whistle type. In
B only one type was recorded,
and this is considered the most
common type. Male FB48 pro-
duced more whistle types than
FB07. FB48’s type 2 whistles
are simple upsweeps (rise
whistles). FB07’s type 1 whistle
also starts as an upsweep, but it
is more stereotyped and com-
plex than the simple, variable
upsweeps produced by FB48

each animal. The predominant whistle type for each animal although some cluster boundaries reflect close similarity
comprised from 47–100% of all whistles produced during the and others less. The fundamental frequency contours from
recording sessions. Three whistles were randomly chosen from the
predominant whistle type subset for each individual for similarity three randomly chosen whistles from each animal are
analysis. Five pairs of adult males were selected for judges’ ratings shown in Fig. 3. The numbers above each whistle loop
of whistle similarity, leading to ten individuals in the sample. correspond to the WLTs that each loop was assigned by
Pair-wise spectrogram comparisons were generated for all 30 the contour similarity analysis. Eighty-one percent of the
whistles. Whistles were not normalized in time or frequency as in
the modified contour similarity technique. Six human judges with
whistle loops were contained in the 39 most common
previous experience with animal sounds and spectrograms (but not WLTs. Each of these types had at least four whistle loops
necessarily dolphin whistles) were chosen. None of the judges had assigned to the cluster.
any prior experience with recordings from this dolphin population.
Each judge rated the similarity of whistles for all 465 pair-wise
comparisons on a scale from one (least similar) to five (most
similar). The judges were instructed only to attend to the shape of Individual distinctiveness
the fundamental frequency contour and to ignore harmonics and
patterns of concurrent echolocation in the spectrograms for Eighty-eight of the WLTs were unique to individuals
categorizing the whistles. The judges were not told before the (Table 3). Although these types accounted for 70% of the
rating began how many animals produced the whistles or the
identity of the dolphins that produced any of the whistles. Judges WLTs, they contained only 25% of the actual whistle
were divided into two groups and each group received a different loops. Ten WLTs (containing 15% of the whistle loops)
arrangement of the same spectrogram comparisons to control for were shared exclusively between male pair partners
any effects of presentation order. Ratings were averaged across (Table 3). Twenty-seven WLTs were shared either by
judges to get an average similarity rating for each individual
whistle comparison. The similarity ratings were finally averaged
non-paired animals or by both paired and non-paired
over the different whistle exemplars (three for each animal) to animals. These types contained the majority of the whistle
generate a similarity rating between the “average” whistle of each loops (60.1%). Five of the WLTs were simple upsweeps
pair of individual males. (types 34, 35, 37, 38, and 39). These types contained
32.65% of the whistle loops.

Results
Most common whistle types
Table 2 presents the most common WLTs produced by
each animal. WLT numbers that are close together reflect Ten WLTs contained the most commonly produced
types that are close together in the clustering hierarchy, whistle loops of all 17 animals: types 10, 30, 33, 35,
536

Table 2 Number of loops in 39 major whistle loop types (WLTs) produced by each animal. The remaining 86 minor types contained three or fewer whistle loops in each type
Whistle type Males Females %
Total
FB16 FB10 FB46 FB14 FB94 FB36 FB38 FB26 FB48 FB66 FB76 FB18 FB44 FB62 FB77 FB79 FB07
4 1 1 2 0.6
5 1 3 1 0.8
9 1 2 1 0.6
10 8 1 1.4
13 5 0.8
18 2 2 0.6
19 4 0.6
25 4 0.6
30 4 5 1.4
33 2 39 6.5
35 1 1 10 5 10 1 21 8 8 1 4 11 2 1 13.3
37 9 28 2 1 1 6.5
38 1 7 2 2 1 8 3.3
39 5 4 3 6 2 8 10 1 3 2 13 9.0
41 15 35 7.9
43 9 1 1.6
45 7 1.1
73 1 6 1.1
75 1 1 4 1.0
76 1 4 15 1 3.3
77 1 1 2 1 0.8
78 1 3 16 1 3.3
80 1 13 2.2
81 1 7 1.3
84 1 1 1 1 0.6
85 4 0.6
86 7 1.1
87 1 3 0.6
90 5 0.8
91 4 0.6
93 1 8 1.4
98 5 0.8
99 4 0.6
108 4 0.6
112 1 2 1 0.6
114 4 0.6
115 7 1.1
120 1 3 0.6
124 4 0.6
Major Types (39) 19 35 30 41 41 12 24 20 73 30 32 27 38 14 23 15 40 81.5
Minor Types (86) 7 10 12 1 2 7 7 5 10 1 12 13 6 19 4 0 1 18.5
Total Loops 26 45 42 42 43 19 31 25 83 31 44 40 44 33 27 15 41 100
537

Fig. 3 Fundamental frequency contours from three randomly the line. Numbers above each whistle loop indicate the whistle loop
chosen whistles from all 17 animals. Alliances are indicated by type (WLT)
dark connecting lines, alliance number is indicated in the middle of
538
Table 3 Number and percent of WLT Whistle loops
whistle loops that fall into each
WLT that are shared among No. % No. %
different groups of animals
Unique to individuals 88 70.4 159 25.2
Shared between pair partners only 10 8.0 93 14.7
Shared between non-partner animals or both non-partners and 27 21.6 379 60.1
pair partners

Table 4 Three most common whistle loop types for each animal Table 5 Number of non-partner animals (out of a total of 15) that
each male shares a percentage of whistle loops with relative to the
Most common whistle loop type amount that each male shares with his partner. Percent WLT
First Second Third Total repertoire sharing between partners and on average with all non-
partners are also given
No. % No. % No. % %
FB16 37 34.6 39 19.2 65 3.8 57.7
FB10 37 62.2 39 8.9 95 7.1 78.3
FB46 93 19.0 90 11.9 98 11.9 42.9
FB14 41 35.7 35 23.8 38 16.7 76.2
FB94 41 81.4 39 14.0 23 4.7 100.0
FB36 35 26.3 39 10.5 71 10.5 47.4
FB38 35 32.3 39 25.8 38 10.5 68.6
FB26 76 16.0 108 16.0 105 16.0 48.0
FB48 78 19.3 76 18.1 80 15.7 53.0
FB66 35 67.7 45 22.6 38 6.5 96.8
FB76 39 22.7 43 20.5 35 18.2 61.4
FB18 35 20.0 13 12.5 19 10.0 42.5
FB44 10 18.2 86 15.9 73 13.6 47.7
FB62 30 15.2 35 12.1 39 9.1 36.4
FB77 35 40.7 38 29.6 39 7.4 77.8
FB79 39 92.9 35 7.1 – – 100.0
FB07 33 95.1 35 2.4 11 2.4 100.0

37, 39, 41, 76, 78, and 93. Of these, only types 35, 37 and
39 were simple upsweeps that resemble the representative
type 2 whistle from McCowan and Reiss (2001). Nine of
the 17 animals produced one of these simple upsweep
contour types as their most common WLT.
The two females (FB79 and FB07) produced fewer
WLTs than the males (females: 2.5€0.7, range 2–3; percentage of their WLT repertoires that are more than,
males: 13.2€5.7, median 14, range 3–22; Mann Whitney equal to, or less than the percentage that they shared with
U test: U=0.5, n 1=2, n 2=15, P<0.05). Also, the females their partners. The two females are considered non-
produced more stereotyped whistles: 92.9–95.1% of their partners for all males. Males are listed twice when they
whistle loops fell under a single WLT, and 100% fell changed pair partners to consider sharing between
within three types (Table 4). This contrasts with the different partners. For example, for pair 1 FB16 was
males, where 15.2–81.4% (mean 36.2%, median 26.3%) counted as a partner to FB10 and FB46 as a non-partner.
fell in each animal’s most common WLT. Among the For pair 2, FB16 was counted as a non-partner and FB46
males, the three most common WLTs of each animal as a partner to FB10. Males shared significantly more
contained 36.4–100% (mean 62.1%, median 57.7%) of of their WLT repertoires with their partners compared
the whistle loops produced. to non-partners (Wilcoxon signed-ranks test: n=18,
P<0.001). Seven of the 18 paired males in the sample
shared as much or more of their whistle loop repertoire
Whistle type sharing with their partners than with any other animals in the
sample.
For a given animal, we calculated whistle sharing with
another animal by determining the percentage of the first
animal’s whistle loops that fell in WLTs that were Effect of recording conditions
produced by the second animal. The values are therefore
not reciprocal. A male shared on average 45% of his WLT The whistles of non-partner animals that were restrained
repertoire with his partner, and 25% with the other 16 and recorded together were compared to determine if
animals. Table 5 lists the number of animals that shared a hearing the whistles produced at that time or other factors
539
Table 6 Similarity ratings by human judges. Ratings range from 1 animal. The self-similarity rating is outlined and the partner
to 5, with 1 indicating low similarity and 5 indicating high similarity rating is shaded. The highest non-self similarity rating is
similarity. Each column contains all of the similarity ratings for one in bold type

involved in the recording sessions induced the animals to the probability of being rated most similar to a non-
produce similar WLTs. In 1991, FB46 and female FB07 partner is greater than to a partner. If one treats the most
were restrained and recorded together. FB46 and FB07 similarly rated whistle type as a choice between these two
shared no WLTs in 1991. In the other years in which outcomes, the probability that seven out of ten males have
FB46 was recorded, he never produced any of the WLTs whistles rated more similar to their partner’s whistle than
produced by FB07 in 1991. The female FB79, pair 6 to a non-partner’s whistle is highly significant (Binomial
(FB66 and FB76), and pair 3 (FB14 and FB94) were test: P<0.01).
restrained and recorded together in 1989. The percent There was strong agreement between the results from
whistle sharing between individuals had a mean of the quantitative modified contour similarity technique and
52€39% (range 0–100%). Whistle sharing between the the human observers. The matrices of similarity ratings
four males (FB66, FB76, FB14, and FB94) was compared from the two techniques were significantly correlated
for 1989 (44€37%) and the rest of the years that the two (data ranked Mantel test: 1,000 permutations, r=0.718,
pairs were not captured together. Whistle sharing for P<0.01). Pairs 3, 4, and 5, all of which had high WLT
years other than 1989 had a mean of 35€30% and was not sharing between partners (Table 5), had whistles that were
statistically different from WLT sharing in 1989 (Wil- rated highly similar to each other by human observers.
coxon signed-ranks test, n=12, P=0.3). Therefore, we Pair 8 was given a low similarity rating by the human
conclude that whistle sharing between partners is not just observers, and WLT sharing by the partners was also low.
a consequence of being recorded together, but results Interestingly, in pair 6, FB66’s whistle was rated most
from the partner relationship. similar to his partner’s whistle (FB76) by the judges, and
there were no other animals in the quantitative analysis
with which he shared more of his WLT repertoire than
Comparison between quantitative comparison with FB76. FB76’s whistle, however, was rated more
and human judges similar by judges to the whistles of several non-partners
than to FB66’s whistle, and this is reflected in the
Table 6 shows the average similarity ratings for all pair- quantitative analysis as FB76 shared more of his WLT
wise comparisons between the ten individual animals repertoire with ten non-partners than with FB66.
whose whistle similarity was judged by human observers.
The cells representing a comparison between pair partners
are shaded. Each animal’s highest similarity rating was Discussion
the self-similarity rating, indicating that individuals could
be distinguished by their whistles. For pairs 3, 4, and 5, Whistle sharing
each male’s whistle was rated more similar to his
partner’s whistle than to any non-partner’s whistle in Two separate techniques for determining whistle similar-
the sample. The whistles of FB44 and FB62 (pair 8) were ity showed that male bottlenose dolphins in strongly
rated more similar to non-partners than to each other. For bonded pairs produce similar whistles. Males shared more
pair 6, the whistle of FB66 was rated most similar to his of their WLT repertoire with their partners than with other
partner (FB76) while the whistle of FB76 was rated most animals based on the modified contour similarity tech-
similar to a non-partner (FB38). Seven out of the ten nique. In general, human judges also rated the complete
males in the sample produced whistles that were rated whistles of allied males more similar to each other than to
more similar to the whistles of their partners than to the the whistles of non-partner males. The differences in
whistles of non-partners. Since each animal’s whistles are repertoire size and variability between males and females
compared against more non-partners (8) than partners (1),
540

agrees with the results of Sayigh et al. (1990), who found social animals (Mammen and Nowicki 1981; Trainer and
a larger repertoire in male dolphin calves than in females. McDonald 1995; Boughman and Wilkinson 1998; Smolk-
Since male partners are not closely related, it is er and Pepper 1999). Detailed behavioral observations
unlikely that genetic relatedness explains the whistle and concurrent acoustic recordings are needed to deter-
similarity found here. Potentially, since partners are close mine the function of whistle use and whistle sharing in
in age, they could have experienced similar acoustic male bottlenose dolphins.
environments while young. Tyack and Sayigh (1997)
suggest that the early acoustic environment of calves
affects whistle development. Dolphin calves generally Signature whistle hypothesis
begin to produce a stereotyped whistle by the end of their
first year (Caldwell and Caldwell 1979). Miksis et al. We believe that the main result of this study, that male
(2002) showed that captive born dolphin calves develop partners produce similar whistles, is consistent with the
whistles that share more acoustic features with frequently signature whistle hypothesis. The signature whistle
heard trainers’ whistles than do free-ranging calves’ hypothesis states that each animal produces a stereotyped,
whistles (who were never exposed to trainers’ whistles). individually distinctive whistle. The overall sharing of
FB14, FB94, FB36, FB38, FB18, FB44, FB62, and FB77 whistle loops by individuals was low (25%). The whistles
are all very close in age and could have experienced produced by partners do not appear identical (Fig. 3).
similar acoustic environments. They spent at least their Sharing WLTs does not mean that two whole whistles,
subadult, if not their calf years, as residents of Sarasota made up of a series of loops, are similar. Human judges
Bay, Fla., and in at least some of the cases their mothers always rated similarity between partners to be less than
associated frequently while rearing them (Wells et al. either partner’s self-similarity rating, indicating that
1980, 1987; Wells 1991). However, FB36, FB38, and variation within individuals was lower than variation
FB94 all produce WLT repertoires that share the greatest between individuals (Table 6). We assume that the
percentage with their partners. FB44 and FB62 were in dolphins are likely capable of finer perceptual distinctions
the same subadult social group, and yet have whistles that of their own displays than the human judges. Therefore,
are not similar. Therefore, it is unlikely that a similar we believe that even though alliance partners produce
acoustic environment at an early age led to the production similar whistles, the whistles of each animal remain
of shared whistles in alliance partners. We suggest that individually distinctive. This result suggests a modifica-
the demonstrated vocal learning abilities of adult dolphins tion of the signature whistle hypothesis, that adult male
are responsible for whistle sharing. Two males may either dolphins may modify their signature whistles as a
converge on a similar whistle type through reciprocal consequence of changing social relationships.
imitation of whistle features (a process that requires time The results of this study disagree with those of
for two whistles to become similar), or one male may hold McCowan and Reiss (2001). Superficially, Table 2
his whistle constant while the other male imitates his appears similar to Table 1 in McCowan and Reiss
whistle (whistle sharing may occur at the first imitation (2001), as upsweep whistles are produced by almost all
event). Sufficient longitudinal data on whistles surround- of the animals. Their type 2 whistle may correspond to
ing the years of alliance formation were not available to types 35 and 39 in this analysis. These two WLTs were
examine which of these two processes for whistle sharing the most common; 15 out of 17 animals produced a
occurred for males as they formed an alliance. Regardless whistle that fell into these categories, and they accounted
of the mechanism, these data support the finding of for 22% of the total whistle loops in the analysis. This is
Smolker and Pepper (1999) that closely affiliated males consistent with the results of Tyack (1986), Janik et al.
produce similar whistles. (1994), and McCowan and Reiss (1995, 2001) showing
The observation of whistle sharing in pair-bonded that most dolphins produce upsweeps. However, contrary
male dolphins with close social association fits well with to the predictions of McCowan and Reiss (2001), this
other animal groups where closely affiliated animals share study does not support the assertion that all dolphins
vocal signals (Mammen and Nowicki 1981; Trainer and produce primarily the same whistle type (a simple
McDonald 1995; Boughman 1997). Dolphins represent upsweep) when in isolation. It is likely that the modified
one of the few mammalian species where adult animals contour similarity technique enhances similarity in whis-
are capable of vocal learning, and the results of this study tle loops with simple monotonic rise or fall in frequency,
demonstrate one aspect where this ability may be used in since it normalizes for differences in duration and actual
natural populations. The function of whistle sharing in frequency. For example, Fig. 4 shows the original
paired males is unclear. It may play a role in the frequency–time contour and the normalized 100-point
formation of the pair bond, as described in Smolker and contour plots of a sample of whistle loops of WLT 35 (A,
Pepper (1999). However, it could also be a mechanism for B) and WLT 39 (C, D). There are significant differences
preferred partners to stay in contact against the back- in duration and rise time between loops that are lost when
ground of rapidly changing social groups in the fission– the whistle loops are normalized in time. The Pearson-
fusion society (Wells et al. 1980). Shared whistles could product moment correlation then allows for a contour
also be used in mate attraction, competitive exclusion, or comparison that ignores absolute frequency and band-
to coordinate pair movements as has been seen in other width information. This is a characteristic of the analysis
541
Fig. 4 Extracted and normal-
ized frequency–time contours
for a sample of A, B WLT 35
and C, D WLT 39

technique that would in effect make any loops with McCowan and Reiss (2001:1160) stated that “the
monotonic changes in frequency appear more similar stereotypic whistles that signature whistle proponents
even if they differ significantly in absolute frequency and describe as signature whistles are likely to be more
time. stereotyped versions of the same whistle types used in
Animals whose most common whistles fall into the contexts other than social isolation and by multiple
same WLT as determined through the contour similarity individuals across different social groups.” This statement
analysis may still possess individually distinctive sig- implies that dolphins possess a large species-specific
nature whistles. The similarity within individuals and repertoire of whistles, and that whistle sharing among
between most pair partners is greater than the combined animals occurs by animals all choosing and producing the
similarity of the component loops. It is important to same whistle type out of a larger repertoire. They point to
remember that the animals do not produce the majority of animals in different aquariums all producing simple
the whistle loops used in this comparison singly, but upsweep whistles as evidence for this shared repertoire.
rather as parts of multiloop whistles. Therefore, we must While many of the dolphins in this study did produce
be careful in the process of moving away from human upsweeps, the results of the present analysis are incon-
observers and towards more quantitative comparison sistent with the shared repertoire hypothesis. If there were
techniques that we do not lose biologically relevant a species-specific repertoire, one would primarily expect
features. Janik (1999) showed that human observers were to find exactly the same whistle types in wild and captive
capable of classifying whistle types based on which animals. However, most studies on dolphin whistles
animal produced the whistle (without prior knowledge of emphasize their variability; comparisons between differ-
which animals made which whistles) while three separate ent studies do not show high overlap in whistle types. For
quantitative techniques (including the McCowan contour example, the whistles from the 17 free-ranging animals
similarity technique) could not. There still is no single examined here were more variable than the 12 whistle
objective, computer driven quantitative technique that has types from nine captive animals studied in McCowan and
received the same external validation that human ob- Reiss (2001). In our larger whistle sample from more
servers have (Janik 1999). Quantitative techniques such as individuals, there were no whistles resembling their types
the one used in this analysis are simply one tool used to 6, 7, 130, 131, 165, or 209. Whistles produced by animals
explore whistle production in bottlenose dolphins. in different communities would be dissimilar if animals
542

learn to produce whistles based on the whistles that they digitization and cluster analysis and Stephanie Nowacek and
hear. The whistles from the captive animals in McCowan Edward Owen calculated male pair association data. Three
anonymous reviewers, Lars Bejder, and Suzanne Yin greatly
and Reiss (2001) appear flatter (less frequency modulat- improved this manuscript. Funding for the sound data collection,
ed) than the whistles by the wild animals presented here. health assessment, and survey/observation program was provided
Since adult dolphins are capable of vocal learning by the Earthwatch Institute, National Marine Fisheries Service,
(Richards et al. 1984), perhaps the whistles produced by Chicago Zoological Society, Dolphin Quest, NSF, EPA, Ocean
Ventures Fund, ONR, and NIH. S.L. Watwood was funded by an
the adult and juvenile dolphins in the McCowan and Reiss NSF Coastal Processes Traineeship, NSF Grant No. SN-9975523,
(2001) study were influenced by the trainers’ whistles and Woods Hole Oceanographic Institution Academic Programs
similar to the dolphin calves in Miksis et al. (2002). Department. Data were collected under NMFS scientific permits
Examining more whistles from different captive and wild issued to R. Wells. This is contribution no. 10799 of the Woods
populations may illustrate the differences further. Hole Oceanographic Institution.
Additionally, animals that produce the same whistle
type at depth may not be able to rely on subtle voice cues
for individual recognition, as suggested by McCowan and References
Reiss (2001). Voice cues are used by terrestrial species
Boughman JW (1997) Greater spear-nosed bats give group-
for individual recognition; however, the effects of pres- distinctive calls. Behav Ecol Sociobiol 40:61–70
sure on diving marine mammals may make these cues less Boughman JW (1998) Vocal learning by greater spear-nosed bats.
useful. Ridgway et al. (2001) trained two adult beluga Proc R Soc Lond B 265:227–233
whales to whistle at 5, 100, 200, and 300 m below the Boughman JW, Wilkinson GS (1998) Greater spear-nosed bats
discriminate group mates by vocalizations. Anim Behav
surface. They reported that overall whistle amplitude 55:1717–1732
decreased, peak frequency increased, and amplitude Brown ED, Farabaugh SM, Veltman CJ (1988) Song sharing in a
modulation varied with depth for both subjects, and group-living songbird, the Australian magpie, Gymnorhina
whistle duration decreased with depth for one of the tibicen. Part I. Vocal sharing within and among social groups.
Behaviour 104:1–28
whales. This was suggested to be caused by increased Caldwell MC, Caldwell DK (1965) Individualized whistle contours
density of air in the nasal system, the proposed location of in bottlenosed dolphins (Tursiops truncatus). Nature 207:434–
whistle production. However, hearing threshold values 435
were not affected by depth (Ridgway et al. 2001). The Caldwell MC, Caldwell DK (1979) The whistle of the Atlantic
same animal producing the same whistle type at the bottlenosed dolphin (Tursiops truncatus)—ontogeny. In: Winn
HE, Olla BL (eds) The behavior of marine animals. Plenum,
surface and at depth will probably sound different to New York, pp 369–401
listening whales. Therefore, within-individual variation at Caldwell MC, Caldwell DK, Tyack PL (1990) Review of the
different depths may be greater than the subtle voice cue signature-whistle hypothesis for the Atlantic bottlenose dol-
differences suggested between animals. Despite differ- phin. In: Leatherwood S, Reeves RR (eds) The bottlenose
dolphin. Academic, San Diego, pp 199–234
ences in whistle amplitude, duration, and absolute Connor RC, Smolker RA, Richards AF (1992) Dolphin alliances
frequency, distinct frequency–time contours should still and coalitions. In: Harcourt AH, de Waal FBM (eds) Coalitions
be recognizable at depth. Whistle production at depth and alliances in humans and other animals. Oxford University
must be further tested in the wild, since captive animals Press, Oxford, pp 415–443
Dreher JJ, Evans WE (1964) Cetacean communication. In: Tavolga
only experience conditions near the surface. WN (ed) Marine bio-acoustics, vol 2. Pergamon, New York, pp
Both the modified contour similarity technique and 159–186
similarity rating by human judges show that adult male Duffield DA, Wells RS (2002) The molecular profile of a resident
bottlenose dolphins are more likely to share whistles with community of bottlenose dolphins, Tursiops truncatus. In:
strongly bonded partners than with other dolphins. Further Pfeiffer CJ (ed) Molecular and cell biology of marine
mammals. Krieger, Malabar, Fla., pp 3–11
study is required to determine both the mechanism and Elowson AM, Snowdon CT (1994) Pygmy marmosets, Cebuella
the function of whistle sharing in these animals. Bot- pygmaea, modify vocal structure in response to changed social
tlenose dolphins have demonstrated their vocal learning environment. Anim Behav 47:1267–1277
abilities in captive situations, and here we describe Fripp DR (1999) Techniques for studying vocal learning in
bottlenose dolphins, Tursiops truncatus. Ph.d Thesis, MIT/
evidence for this ability in free-ranging animals. Much WHOI Joint Program, Woods Hole, Mass.
like social terrestrial animals, male bottlenose dolphins Herzing DL (1996) Vocalizations and associated underwater
appear to produce signals that indicate membership in behavior of free-ranging Atlantic spotted dolphins, Stenella
particular social units. frontalis, and bottlenose dolphins, Tursiops truncatus. Aquat
Mamm 22:61–79
Hohn AA, Scott MD, Wells RS, Sweeney JC, Irvine AB (1989)
Acknowledgements We are extremely grateful to Laela Sayigh and Growth layers in teeth from known-age, free-ranging bottlenose
her students for the majority of the data collection of the whistles dolphins. Mar Mamm Sci 5:315–342
used in this study. We very much appreciate the efforts of Michael Janik VM (1999) Pitfalls in categorization of behaviour: a
Scott and Blair Irvine, as well as a dedicated group of staff and comparison of dolphin whistle classification methods. Anim
volunteers, to ensure safe and effective handling of the dolphins for Behav 57:133–143
collection of the whistle recordings over the years. Aleta Hohn Janik VM (2000) Whistle matching in wild bottlenose dolphins
provided age data for individuals for which ages were not (Tursiops truncatus). Science 289:1355–1357
determined from observations during birth years. We would like Janik VM, Slater PJB (1998) Context-specific use suggests that
to thank Nancy Dimarzio, Sarah Marsh, Jennifer Miksis, Patrick bottlenose dolphin signature whistles are cohesion calls. Anim
Miller, Susan Parks, and Rebecca Thomas for performing the Behav 56:829–838
similarity comparisons. Deborah Fripp provided assistance with
543
Janik VM, Denhardt G, Todt D (1994) Signature whistle variations Ridgway SH, Carder DA, Kamolnick T, Smith RR, Schlundt CE,
in a bottlenosed dolphin, Tursiops truncatus. Behav Ecol Elsberry WR (2001) Hearing and whistling in the deep sea:
Sociobiol 35:243–248 depth influences whistle spectra but does not attenuate hearing
Mammen DL, Nowicki S (1981) Individual differences and within- by white whales (Delphinapterus leucas) (Odontoceti, Ceta-
flock convergence in chickadee calls. Behav Ecol Sociobiol cea). J Exp Biol 204:3829–3841
9:179–186 Sayigh LS, Tyack PL, Wells RS, Scott MD (1990) Signature
McCowan B (1995) A new quantitative technique for categorizing whistles of free-ranging bottlenose dolphins Tursiops truncatus:
whistles using simulated signals and whistles from captive stability and mother–offspring comparisons. Behav Ecol So-
bottlenose dolphins (Delphinidae, Tursiops truncatus). Etholo- ciobiol 26:247–260
gy 100:177–193 Sayigh LS, Tyack PL, Wells RS, Scott MD, Irvine AB (1995) Sex
McCowan B, Reiss D (1995) Quantitative comparison of whistle differences in signature whistle production in free-ranging
repertoires from captive adult bottlenose dolphins (Del- bottlenose dolphins. Behav Ecol Sociobiol 36:171–177
phinidae, Tursiops truncatus): a re-evaluation of the signature Smolker RA, Pepper JW (1999) Whistle convergence among allied
whistle hypothesis. Ethology 100:194–209 male bottlenose dolphins (Delphinidae, Tursiops sp). Ethology
McCowan B, Reiss D (2001) The fallacy of ‘signature whistles’ in 105:595–617
bottlenose dolphins: a comparative perspective of ‘signature Trainer JM, McDonald DB (1995) Singing performance, frequency
information’ in animal vocalizations. Anim Behav 62:1151– matching, and courtship success of long-tailed manakins
1162 (Chiroxiphia linearis). Behav Ecol Sociobiol 37:249–254
Miksis JL, Tyack PL, Buck JR (2002) Captive dolphins, Tursiops Tyack PL (1986) Whistle repertoires of two bottlenose dolphins,
truncatus, develop signature whistles that match acoustic Tursiops truncatus: mimicry of signature whistles? Behav Ecol
features of human-made model sounds. J Acoust Soc Am Sociobiol 18:251–257
112:728–739 Tyack PL, Sayigh LS (1997) Vocal learning in cetaceans. In:
Owen ECG, Wells RS, Hofmann S (2002) Ranging and association Snowdon CT, Hausberger M (eds) Social influences on vocal
patterns of paired and unpaired adult male bottlenose dolphins, development. Cambridge University Press, Cambridge, Mass.,
Tursiops truncatus, in Sarasota, Florida provide no evidence for pp 208–233
alternative male strategies. Can J Zool 80:2072–2089 Wells RS (1991) The role of long-term study in understanding the
Podos J, Peters S, Rudnicky T, Marler P, Nowicki S (1992) The social structure of a bottlenose dolphin community. In: Pryor K,
organization of song repertoires in song sparrows: themes and Norris KS (eds) Dolphin societies: discoveries and puzzles.
variations. Ethology 90:89–106 University of California Press, Berkeley, Calif., pp 199–225
Reiss D, McCowan B (1993) Spontaneous vocal mimicry and Wells RS, Irvine AB, Scott MD (1980) The social ecology of
production by bottlenose dolphins (Tursiops truncatus): evi- inshore odontocetes. In: Herman LM (ed) Cetacean behavior:
dence for vocal learning. J Comp Psychol 107:301–312 mechanisms and functions. Wiley, New York, pp 263–317
Richards DG, Wolz JP, Herman LM (1984) Vocal mimicry of Wells RS, Scott MD, Irvine AB (1987) The social structure of free-
computer-generated sounds and vocal labeling of objects by a ranging bottlenose dolphins. In: Genoways H (ed) Current
bottlenosed dolphin Tursiops truncatus. J Comp Psychol mammalogy, vol 1. Plenum, New York, pp 247–305
98:10–28

Вам также может понравиться