182
P.L. Tyack and c.w. Clark
for tonal and pulsed calls. I will use Recchia's work as an example of an
attempt at a comprehensive analysis of the repertoire of an odontocete. Her
recordings came from three different aquaria: the Aquarium for Wildlife
Conservation, Brooklyn, New York (five adult belugas); Point Defiance Zoo
and Aquarium, Tacoma, Washington (three adult belugas); and the Vancouver Public Aquarium, Vancouver, British Columbia (five adult belugas).
Recchia first developed a subjective classification for the beluga repertoire,
based upon aural impressions and examination of spectrograms. Figure 4.5
illustrates nine examples of the 14 mutually exclusive call types she identified. However, the calls of belugas were so variable and intergraded that she
was very uncomfortable with the standard practice of sorting calls into discrete categories by visual inspection of spectrograms. In order to test this
categorization using quantitative techniques, Recchia extracted time-frequency parameters from each of 603 of these signals. Two of the call types,
buzzsaws and jaw claps, could be isolated by multivariate discriminate analysis, but few of the others were well separated by this analysis. A tree-based
classifier was only able to classify 72 % of the calls to nodes that matched
Recchia's subjective categories, and all classes but one were split into more
than one node. Recchia concluded that it was by no means clear whether the
signal types identified by visual inspection of spectrograms represent welldefined acoustic categories. It is possible and indeed likely that the auditory
processing of belugas involves feature detectors that enable categorization
of many of these call types, but it is impossible to achieve a valid categorization without testing how belugas themselves perceive or use these
sounds. In addition, it is very difficult to compare subjective call types analyzed from the same or different species by different researchers. Recchia
concluded that the jaw clap was the only call type that could confidently be
called equivalent across studies. Many of the call types identified by Recchia
appear to have parallels in the vocal repertoire of bottlenose dolphin. These
include pulsed sounds such as clicks,jaw claps, and tonal sounds such as whistles and chirps. However, again, without better definition of the acoustic
structure of the calls, how they are produced, and how they are perceived, it
is difficult to make valid comparisons.
Beluga whales have good high-frequency hearing (White et al. 1978;
Johnson et al. 1989) and they have been demonstrated to be as skilled at
echolocation as bottlenose dolphins (Au et al. 1985). However, there is also
an important difference in sonar processing in belugas compared to bottlenose dolphins. Beluga whales also are capable of emitting "packets" of
clicks with interpulse intervals less than the round trip travel time to the
sonar target (Turl and Penner 1989). Beluga whales do not have to process
each pulse independently, waiting for the echo return before emitting the
next pulse, but rather they may be able to process whole series of pulses in
ways that improve their sonar performance. While it is not consistent with
usage among marine mammalogists, these signals appear to fit the original
sonar definition of burst-pulsed signals.
P.L. Tyack and c.w. Clark
for tonal and pulsed calls. I will use Recchia's work as an example of an
attempt at a comprehensive analysis of the repertoire of an odontocete. Her
recordings came from three different aquaria: the Aquarium for Wildlife
Conservation, Brooklyn, New York (five adult belugas); Point Defiance Zoo
and Aquarium, Tacoma, Washington (three adult belugas); and the Vancouver Public Aquarium, Vancouver, British Columbia (five adult belugas).
Recchia first developed a subjective classification for the beluga repertoire,
based upon aural impressions and examination of spectrograms. Figure 4.5
illustrates nine examples of the 14 mutually exclusive call types she identified. However, the calls of belugas were so variable and intergraded that she
was very uncomfortable with the standard practice of sorting calls into discrete categories by visual inspection of spectrograms. In order to test this
categorization using quantitative techniques, Recchia extracted time-frequency parameters from each of 603 of these signals. Two of the call types,
buzzsaws and jaw claps, could be isolated by multivariate discriminate analysis, but few of the others were well separated by this analysis. A tree-based
classifier was only able to classify 72 % of the calls to nodes that matched
Recchia's subjective categories, and all classes but one were split into more
than one node. Recchia concluded that it was by no means clear whether the
signal types identified by visual inspection of spectrograms represent welldefined acoustic categories. It is possible and indeed likely that the auditory
processing of belugas involves feature detectors that enable categorization
of many of these call types, but it is impossible to achieve a valid categorization without testing how belugas themselves perceive or use these
sounds. In addition, it is very difficult to compare subjective call types analyzed from the same or different species by different researchers. Recchia
concluded that the jaw clap was the only call type that could confidently be
called equivalent across studies. Many of the call types identified by Recchia
appear to have parallels in the vocal repertoire of bottlenose dolphin. These
include pulsed sounds such as clicks,jaw claps, and tonal sounds such as whistles and chirps. However, again, without better definition of the acoustic
structure of the calls, how they are produced, and how they are perceived, it
is difficult to make valid comparisons.
Beluga whales have good high-frequency hearing (White et al. 1978;
Johnson et al. 1989) and they have been demonstrated to be as skilled at
echolocation as bottlenose dolphins (Au et al. 1985). However, there is also
an important difference in sonar processing in belugas compared to bottlenose dolphins. Beluga whales also are capable of emitting "packets" of
clicks with interpulse intervals less than the round trip travel time to the
sonar target (Turl and Penner 1989). Beluga whales do not have to process
each pulse independently, waiting for the echo return before emitting the
next pulse, but rather they may be able to process whole series of pulses in
ways that improve their sonar performance. While it is not consistent with
usage among marine mammalogists, these signals appear to fit the original
sonar definition of burst-pulsed signals.