Effects of dolphin hearing bandwidth on biosonar click emissions

Odontocetes (toothed whales) possess biological sonar (biosonar) systems that they use to interrogate their underwater environment. Dolphin biosonar sound emissions (“clicks”) are high-frequency, broadband impulses with durations of 50–80 μs, center frequencies on the order of 80–100 kHz, and centralized root-mean-square (rms) bandwidths of ∼30 kHz (Au, 1993). The presumed operating principle of biosonar involves comparing a stored replica of the emitted click to echoes returning from submerged objects and the surrounding environment. Large scale delay differences between the emitted click and returning echoes indicate the target range and fine-scale echo delay differences reveal aspects of target shape (Simmons and Gaudette, 2012; Simmons et al., 2014).

Successful operation of the biosonar system requires the animal to (1) generate the outgoing clicks and (2) to interpret the acoustic characteristics of the returning echoes. Hearing loss within the echolocation frequency range would therefore be expected to affect an animal's ability to extract information from high-frequency echoes, thus decreasing performance during biosonar tasks. This has been demonstrated in several experiments with the common bottlenose dolphin (Tursiops truncatus, Houser et al., 2005; Ibsen et al., 2009) and a false killer whale (Psuedorca crassidens, Kloepper et al., 2010b). In Houser et al. (2005), two male bottlenose dolphins participated in an open water, free swimming target search and detection task. One of the dolphins had normal hearing and the other had hearing loss above 50 kHz (Moore et al., 2004). The reduction in hearing sensitivity may have contributed to observed differences in search strategy: reduced search range, shorter interclick intervals, predominant use of wideband clicks with low-frequency peaks, and more abundant click production were seen in the animal with hearing loss compared to the normal-hearing animal (Houser et al., 2005). Ibsen et al. (2009) reported on a dolphin (“BJ”) with hearing loss above 45 kHz participating in a go/no-go task discriminating “standard” electronic (i.e., “phantom”) echoes from filtered echoes. The animal was unable to perceive differences outside of a frequency band from 29 to 42 kHz, with the low-pass frequency roughly corresponding to the animal's upper-frequency limit (UFL) of functional hearing. Finally, Kloepper et al. (2010b) documented the decline in biosonar discrimination abilities of a false killer whale (“Kina”) over a 16-year period during which high-frequency hearing loss occurred. The loss of hearing above 40 kHz resulted in a maximum performance reduction of 36% during a cylinder wall thickness discrimination task (Kloepper et al., 2010b).

In addition to decreases in auditory performance, several studies have also shown animals with high-frequency hearing loss produce echolocation clicks with relatively low center frequencies compared to conspecifics with normal hearing, presumably to improve the echo signal-to-noise ratio (SNR) within their audible frequency range. Ibsen et al. (2007) compared click emissions from a dolphin with high-frequency hearing loss (BJ, the same dolphin utilized by Ibsen et al., 2009) during target material discrimination tasks conducted in 1998 and 2003–2004 and found the click average peak frequency had shifted downward almost 3.5 octaves, from 138 kHz in 1998 to 40 kHz in 2003–2004. Using the same dolphin, Ibsen et al. (2010) found changes in the variability of click spectral content from 1998 to 2004, independent of target type or click level: in 1998, click spectra were most consistent (i.e., smallest variance in normalized spectral amplitudes; see Ibsen et al., 2010) from 90 to 100 kHz, while in 2004, click spectra were most consistent at frequencies up to 42 kHz. Together, the data from Ibsen et al. (2007, 2010) indicate consistent use of relatively low-frequency clicks by a dolphin with high-frequency hearing loss; however, since BJ's UFL was not measured in 1998, a change in click emissions directly corresponding to a change in UFL cannot be established. Kloepper et al. (2010a) reported a significant reduction in center frequency, peak frequency, and source level of clicks in the false killer whale, Kina, between cylinder wall thickness discrimination tasks conducted in 1992 and 2008. Kina's hearing ability in 1992 was not directly measured. However, in 1988, Kina could hear tones at 100 kHz (Kloepper et al., 2010a) but by 2001 her UFL had declined to ∼30–40 kHz (Yuen et al., 2005), suggesting a reduction in click frequency content as her UFL decreased with age. In the previously mentioned study by Houser et al. (2005), the two dolphins not only chose different search strategies but also demonstrated differences in the frequency content of their clicks. The animal with normal hearing produced clicks with varying peak frequencies ranging from 20 to 120 kHz, while the animal with hearing loss above 50 kHz produced most clicks with peak frequencies from 30 to 60 kHz. Results agreed with a prior study (Houser et al., 1999), where an older male dolphin with hearing loss above 50 kHz produced lower frequency clicks compared to a younger female dolphin, both performing the same biosonar target detection task. Similarly, an older Indo-Pacific humpback dolphin (Sousa chinensis) with a UFL of ∼80 kHz (∼30–40 kHz lower than measured from a younger animal) emitted clicks with peak and center frequencies ∼16 kHz lower than another younger animal (whose hearing was not tested), suggesting a reduction in click frequency with a decrease in UFL (Li et al., 2013). Finally, Finneran et al. (2016) reported that two bottlenose dolphins with full hearing bandwidth (up to ∼140 kHz) utilized clicks with center frequencies from around 80 to 90 kHz during a target change detection task, but a downward shift in center frequency from 50 to 70 kHz was observed for a dolphin with hearing loss above 80 kHz performing the same task.

Despite clear examples of changes in auditory performance and click emissions with high-frequency hearing loss, it is noteworthy that these data are limited to only a few individuals. Some studies have also reported animals with high-frequency hearing loss still producing high-frequency clicks with most of the click energy outside their functional hearing range. For example, despite having significant hearing loss above 50 kHz, a dolphin from Finneran et al. (2016) emitted clicks with center frequencies from 80 to 90 kHz, similar to the animals with normal hearing. This dolphin also tended to produce the highest amplitude clicks. Thus, even though some evidence suggests dolphins alter click frequency content to compensate for a reduced UFL, some conflicting evidence also exists. To date, there has been no large-scale, systematic investigation of the relationship between click emissions and hearing bandwidth in echolocating animals.

In the present study, we examine how specific parameters of dolphin biosonar clicks (frequency, level, and bandwidth) vary with high-frequency hearing ability. More specifically, we determine if animals with high-frequency hearing loss utilize clicks with lower center frequency and/or higher peak-peak (p-p) amplitude compared to clicks typically utilized by normal hearing animals. Compared to the previous studies of 1 to 4 individuals, this study utilized a cross-sectional design involving 16 trained bottlenose dolphins, with auditory thresholds measured as a function of frequency. Individual dolphin hearing ranges were subsequently compared to their echolocation clicks collected in the same time period.

The subjects were 16 bottlenose dolphins (5 females and 11 males, aged 4–54 years old at the time of the study, see Table I). Testing was conducted in 9 × 9 m floating, netted enclosures in San Diego Bay, CA. Data collection occurred from June to November 2019. The study followed a protocol approved by the Institutional Animal Care and Use Committee at the Naval Information Warfare Center Pacific and the Navy Bureau of Medicine and Surgery and followed all applicable U.S. Department of Defense guidelines.

Ambient noise in the testing environment was dominated by vessel traffic, other dolphins, and snapping shrimp. Noise levels during testing were quantified using a piezoelectric hydrophone (Reson TC4032, Reson Inc., Slangerup, Denmark) to measure mean octave-band sound pressure levels (SPLs) in the 20–40, 40–80, and 80–160 kHz bands. Within any single octave-band, noise SPLs across testing days were within ± 2 dB.

The UFL in each dolphin was quantified by measuring hearing thresholds as a function of sound frequency. Hearing threshold at a particular frequency was defined as the SPL at which the auditory steady-state response (ASSR) to a sinusoidal amplitude modulated (SAM) tone was just-detectable. The ASSR is a rhythmic auditory evoked potential (AEP) formed when stimuli are presented at sufficient rates so that individual, transient AEPs overlap and form a steady-state response (Regan, 1982). A plot of threshold vs sound frequency (the audiogram) was examined to identify the UFL at which hearing thresholds rapidly increased and reached a value of 120 dB re 1 μPa.

Sound stimuli were generated, and evoked responses were recorded using EVREST software (Finneran, 2008, 2009). During testing, subjects were positioned on an underwater “biteplate” (Fig. 1) at a depth of 0.9 m, located approximately 1 m in front of an ITC 5446 underwater sound projector (International Transducer Corp, Santa Barbara, CA). Stimuli consisted of SAM tones with a duration of 30 ms and modulation frequency of 1 kHz. SAM tone carrier frequencies varied from 23 to 152 kHz. Tones were digitally generated, then converted to analog at a rate of 2 MHz with 16-bit resolution (NI PXIe-6368, National Instruments, Austin, TX). Analog signals were filtered (0.2–200 kHz, 3 C module, Krohn-Hite Corp., Brockton, MA), attenuated if necessary (custom), and amplified (CC4000, Crest Audio, Meridian, MS) before being applied to the ITC 5446 projector. Stimuli were calibrated before each session using a hydrophone (TC4013, Reson Inc.) placed at the listening position, estimated as the midpoint between the dolphin's lower jaws when positioned on the biteplate. The hydrophone signal was amplified and filtered (VP1000, Reson Inc., and 3 C module, Krohn-Hite Corp.), then digitized at 1 MHz and 16-bit resolution using the PXIe-6368. After calibration, the hydrophone was moved away from the dolphin and used to monitor underwater sounds during testing. Stimuli were calibrated in terms of the rms sound pressure obtained from the averaged spectrum. For the SAM tone stimuli, this results in SPLs approximately 5 dB lower than those based on p-p equivalent SPL, which is recommended by the American National Standards Institute (ANSI) in ANSI/ASA S3/SC1.6–2018 (ANSI, 2018).

ASSRs to sound stimuli were measured using three non-invasive surface electrodes embedded in suction cups: a non-inverting electrode placed along the dorsal midline, approximately 5 cm from the posterior edge of the blowhole, an inverting electrode placed near the right external auditory meatus, and a common electrode placed in the water near the dolphin. A biopotential amplifier (ICP511, Grass Technologies, West Warwick, RI) filtered (0.3–3 kHz) and amplified (94 dB) the voltage difference between the non-inverting and inverting electrodes. The biopotential amplifier output, representing the instantaneous electroencephalogram (EEG), was digitized at 40 kHz with 16-bit resolution (PXIe-6368), segmented into 41.3 ms epochs aligned with stimulus onset, and synchronously averaged using a weighted averaging technique (Elberling and Wahlgreen, 1985). After averaging 256 epochs, a Fourier transform was performed on a 30-ms segment of the averaged EEG temporally aligned with the ASSR waveform. The resulting spectral amplitude and phase angle at 1 kHz were used to compute the magnitude-squared coherence (MSC) statistic (Dobie and Wilson, 1996; Finneran et al., 2007) based on 16 subaverages. If the MSC exceeded the critical value (α = 0.01), the ASSR was considered detected and the ASSR measurement at that frequency and SPL complete. If the MSC was less than the critical value, additional EEG epochs were averaged and the MSC calculation repeated after each integral multiple of 256 epochs. If no detection occurred after averaging 1024 epochs, the measurement was stopped and the ASSR classified as not detected at that SPL and frequency.

Hearing thresholds were obtained for all subjects using a modified up/down staircase procedure. Testing began with a SAM tone at a given frequency with a rms equivalent SPL of ∼120 dB re 1 μPa. If the ASSR in response to the stimulus was detected, the next stimulus presented was 10 dB lower than the previous. Stimulus level was decreased by 10 dB until the ASSR was no longer detected. Testing at each frequency concluded with one final measurement 5 dB above the previous undetected ASSR. The threshold for a particular frequency was defined as the lowest SPL that resulted in a detectable ASSR. The UFL of hearing was defined as the frequency with a threshold of 120 dB re 1 μPa, found by interpolation between thresholds (if necessary) using a logarithmic frequency scale.

In addition to the hearing test, all 16 subjects participated in an active biosonar task. The biosonar task was conducted in one of the netted enclosures with the dolphin positioned underwater on a biteplate oriented towards San Diego Bay (Fig. 2). The biteplate was supported by an extruded aluminum frame, similar to the hearing test apparatus, at a depth of 1 m. The task was a physical target change detection task, where the dolphin produced a conditioned vocalization when it detected a change in aspect (90° rotation) of a target consisting of two truncated cones joined at their apices [see Fig. 1(c) from Finneran et al., 2017].

Each session consisted of 40 trials. The first ten trials were a “warm-up” used to assess dolphin motivation and maintain behavioral control. During the warm-up, the target distance was incrementally increased from 2.5 to 7 m (0.5 m after each trial) away from the animal. If the dolphin's performance was ≥80% correct during the warm-up, 30 additional trials were conducted at a fixed range of 7 m. If performance during the warm-up was <80% correct, testing with that dolphin was terminated for the day. The 7-m range was determined after preliminary testing to identify the target range at which all dolphins performed the task correctly at least 75% of the time.

Each trial began with a trainer directing the dolphin to position itself on the biteplate and begin echolocating on the target. On 80% of the trials (designated as “change trials”), the target was rotated after a random interval of 2–8 s. On the remaining 20% of trials (“control trials”), the target was not rotated. The order of change/control trials was determined using a pseudorandom sequence. Dolphins were trained to respond to target rotation with a conditioned vocalization (either a whistle or burst pulse, depending on the dolphin) and to withhold the response otherwise. Responses occurring within a 1.5-s response window after target rotation were classified as “hits.” Failure to respond within the response interval was a “miss.” Responding during a control trial was a “false alarm” (FA), while withholding the response during a control trial was a “correct rejection.” Responses during change trials before the target change resulted in the trial being re-classified as a control trial and scored as a false alarm. After a correct response (hit or correct rejection), a “buzzer” was sounded signaling the dolphin to surface and receive a fish reward. After an incorrect response (false alarm or miss), a “boing” sound was played and the dolphin recalled to the surface without being rewarded. The next trial was then conducted. Testing for an individual dolphin was finished after a single 40-trial session was successfully completed.

Click emissions during each trial were recorded by a piezoelectric hydrophone (TC4013, Reson Inc.) attached to the biteplate frame, 0.75 m from the blowhole in front of the dolphin and along the main transmit axis of the biosonar beam. The hydrophone signal was amplified and filtered (5–200 kHz, VP1000, Reson Inc., and 3 C module, Krohn-Hite Corp.), then digitized at 2 MHz and 16-bit resolution using an NI PXIe-6368 multifunction data acquisition device.

Clicks from each (non-warmup) trial were analyzed by first identifying click onset using a threshold-crossing detector, then extracting a 128-μs time window encompassing the whole duration of the click. For change trials, only clicks occurring before the target rotation were extracted. If a false alarm was committed, clicks were only extracted up to the time of the false alarm. For control trials, clicks occurring during the entire trial period were extracted. Click spectra were obtained by zero-padding each click waveform to obtain 400 samples, then performing a Fourier transform and scaling the output to obtain the energy flux spectral density with a frequency resolution of 5 kHz. For each dolphin, click waveforms and spectra from all trials were pooled for analysis. Custom software was used to compute the median and inter-quartile intervals for the click waveforms and spectra, click p-p SPL, center frequency, and (centralized) rms bandwidth (Menne and Hackbarth, 1986; Au, 1993). The p-p SPL measured at the hydrophone was translated to apparent source level assuming spherical spreading with no absorption. Statistical analyses were performed using JMP statistical software (sas Institute, 2019). Relationships between dolphin age and UFL, UFL and center frequency, UFL and p-p SPL, and UFL and rms bandwidth were examined using linear regressions, with α = 0.05.

Audiograms for all dolphins (Fig. 3) were typical of those obtained using ASSR methods. Dolphin UFLs (Table I) ranged from 45 to 147 kHz. Seven dolphins demonstrated “normal” hearing bandwidth for young dolphins, with frequency cutoffs above 120 kHz. Four dolphins demonstrated some high-frequency hearing loss, with cutoffs between 80 and 101 kHz, and five demonstrated severe high-frequency hearing loss with cutoffs of 45–73 kHz. With the exclusion of TYH, all thresholds for the 15 other dolphins were within ±10 dB within the best frequency range of hearing for each animal. Age did not have a significant effect on hearing ability of subjects in this study [F_(1,14) = 1.03, p = 0.328, r² = 0.0682] with animals of varying ages occupying each of the three hearing bandwidth categories.

Performance (percent-correct) was at or above 80% for all participants, with 14 dolphins completing all 30 trials with 90% or higher correct responses. Subjects BLU and TYH completed the 30 trials with 80% and 83% correct, respectively, which could be due in part to BLU having the most severe high-frequency hearing loss and TYH having the highest thresholds. FA rates were low for 15 of the 16 subjects, with 13 subjects having no FAs, 2 subjects (APO and SPA) each with a FA rate of 1/6, and 1 subject (BLU) with a FA rate of 4/7. The number of clicks recorded for each dolphin ranged from 2507 (COM) to 7011 (IND). Median click waveforms and spectra are shown in Figs. 4 and 5. Click waveforms resembled exponentially decayed sinusoids and matched those typically reported for dolphins (Au, 1993). Spectra were broadband, with −10 dB bandwidths generally from 20 to 120–140 kHz, though spectra for some animals (e.g., BLU, COL, TYH) had relatively more energy at lower frequencies.

Dolphin click trains generally followed expected patterns, with inter-click intervals (ICIs) greater than the two-way travel time between dolphin and target. Dolphin click amplitudes measured during biosonar tasks typically fluctuate from click-to-click within a click train; however, amplitude changes from one click to the next are generally small (∼1 dB or so). In the present study, unusual click patterns were sometimes observed, where click amplitudes alternated from higher to lower amplitude over a series of successive clicks. In other words, even and odd-numbered clicks had slightly different amplitudes. This was most often seen in the dolphin OLY. On four trials with OLY, there were one or more sequences of at least 20 consecutive clicks where all the even-numbered click p-p SPLs were at least 1-dB above, or less than 1-dB below, than the odd-numbered click SPLs (Fig. 6).

Linear regressions were performed to investigate the relationships between UFL and click center frequency, p-p SPL, and rms bandwidth. Regressions were performed utilizing the mean values of each of the specific click parameters. There was a significant relationship between UFL and click center frequency [Fig. 7(a), F_(1,14) = 10.6, p = 0.00571, r² = 0.431]. Center frequency (in kHz) equaled 61.1 + 0.175 (UFL), when UFL was expressed in kHz; i.e., dolphin click center frequencies tended to decrease by ∼0.2 kHz for each 1-kHz drop in UFL. There were non-significant relationships between UFL and click p-p SPL [Fig. 7(b), F_(1,14) = 0.200, p = 0.662, r² = 0.0141] and rms bandwidth [Fig. 7(c), F_(1,14) = 1.23, p = 0.287, r² = 0.0806].

There are numerous reports of high-frequency hearing loss in odontocete cetaceans (e.g., Ridgway and Carder, 1997; Finneran et al., 2005; Houser and Finneran, 2006; Popov et al., 2007; Mann et al., 2010). Many of the individuals with hearing loss have been older (>20–30 years), suggesting age-related hearing loss (presbycusis) as the underlying cause (Houser and Finneran, 2006; Houser et al., 2008). Some dolphins, however, experience atypical sensorineural hearing loss at younger ages, i.e., due to other factors such as disease, congenital defects, noise exposure, etc. (Houser and Finneran, 2006). Of the dolphins identified with atypical hearing loss, both stranded animals and those in managed care facilities, the etiology of hearing loss is almost always unknown.

Regardless of the cause, loss of hearing bandwidth and/or elevated hearing thresholds presumably create challenges to hearing impaired dolphins performing echolocation tasks. Reduction of perceived echo frequency content decreases the information obtained in echoes and likely reduces range resolution and the ability to perceive fine target details (Simmons, 1973; Simmons and Stein, 1980; Simmons et al., 2004). Elevated hearing thresholds lower the echo SNR and reduce echo detection and range discrimination ability if click source levels are not increased to compensate (Au and Penner, 1981; Simmons, 2017). It is therefore reasonable that hearing-impaired animals would lower click frequency content and increase click amplitude, compared to normal-hearing animals, to enhance click energy at audible frequencies, thereby improving echo SNR and biosonar detection/discrimination capabilities.

UFLs in the present study ranged from 54 to 147 kHz, agreeing with previous studies that reported dolphin UFLs ranging from 50 to 169 kHz (Houser and Finneran, 2006; Popov et al., 2007; Houser et al., 2008). Though older animals typically presented with hearing loss in previous studies (Houser and Finneran, 2006; Popov et al., 2007; Houser et al., 2008; Li et al., 2013), age was not a significant indicator of hearing loss within the 16 subjects in the present study; i.e., this study contained subjects both with presbycusis and atypical hearing loss. Dolphins with high-frequency hearing loss tended to use clicks with lower center frequencies. This result is consistent with adjusting biosonar source frequency downward as a compensatory mechanism for the loss of high-frequency hearing and agrees, in part, with previous findings of odontocetes with lower UFLs utilizing lower click frequencies (e.g., Houser et al., 2005; Ibsen et al., 2007; Kloepper et al., 2010a).

The present data showed no change in click p-p SPL as a function of UFL when compared across individuals, despite some studies showing correlations between click p-p SPL and center frequency (e.g., Au et al., 1995; Finneran et al., 2014; Au et al., 2016). However, the relationship between click p-p SPL and click center frequency in hearing impaired dolphins is conflicting; Kloepper et al. (2010a) found click amplitude decreased with center frequency in a false killer whale with high-frequency hearing loss, whereas Finneran et al. (2016) found that click amplitude was higher in two hearing impaired dolphins when compared to two with normal hearing, although a lower click center frequency was only apparent in one of the animals. Prior studies showing correlations between click p-p SPL and center frequency have usually assessed their relationship within an individual animal often examining targets at a variety of distances and/or over a relatively large range of click source levels (Au et al., 1995; Finneran et al., 2014; Au et al., 2016). A subsequent regression analysis of the present data was therefore conducted relating click center frequency to p-p SPL while controlling for the individual dolphin as a random effect (i.e., linear relationships between individual click center frequencies and p-p SPL were made within each individual). The overall model demonstrated that the relationship between click center frequency and p-p SPL was strongly significant when controlled for subject [F_{(1, 72 962}) = 51 772, p < 0.0001, whole model r² = 0.76], and that ∼80% of the model variance (determined through Restricted Maximum Likelihood Estimation) could be explained by differences between individuals, but not the UFL. Thus, it is likely that comparing mean values across individuals is not the most appropriate test of whether click p-p SPL changes along with hearing-impaired shifts in click center frequency; rather, longitudinal testing in the same individual over a time period that encompasses the change from normal hearing to hearing impaired would provide a more robust assessment.

The above analysis highlights the importance of individual variation in dolphin click emissions. Despite a significant relationship between UFL and center frequency, the amount of the variability explained by the regressions was relatively low (r² = 0.431), and thus the ability to predict click properties for individual animals based on their UFL is limited. For example, contrast click center frequencies used by WHP and SPA (80 and 95 kHz, respectively) with UFLs of 143 and 101 kHz, respectively. Animal prior experiences and learning are features not captured by the present study but may play key roles in a dolphin's echolocating strategy. Dolphins may learn over time, as presbycusis reduces high-frequency hearing ability, to shift their outgoing echolocation click downward in center frequency. However, longitudinal hearing data are not available, thus the exact age that hearing loss began is unknown, leaving such questions of learning difficult to answer. Future studies involving the same individuals over time are necessary to conclusively demonstrate a progressive downward shift in center frequency with decreasing UFL from presbycusis.

Finally, we also note the role of individual variation, not just in click parameters but also in the temporal patterns of click emissions. Changes in click SPL from one click to the next are generally small—on the order of a 1 dB or so—and click SPLs typically do not oscillate from one to the next. This makes the unusual p-p SPL patterns observed in the dolphin OLY noteworthy. To our knowledge, this has not been previously reported, and we have no satisfactory explanation. The extent to which dolphins can manipulate individual click features is not known. Head movements could alter the received SPL at the hydrophone, but it seems unlikely that OLY could move his head at the rate required (without observation by the experimenter or trainer) to consistently alternate received click SPLs. Similarly, vibration of the hydrophone could affect received levels, but it is unlikely this would have occurred with OLY but no other dolphins. Future studies utilizing a contact hydrophone or digital recording tag would allow estimates of click source level independent of head movement, and thus reveal if such received SPL fluctuations are actually due to changes in emitted click source level.

Dolphins with high-frequency hearing loss tend to utilize echolocation clicks with lower center frequencies compared to animals with full hearing bandwidth, a mechanism that places more echo energy into the animal's audible range. Knowing an animal's UFL is therefore crucial to interpreting comparative studies of echolocation utilizing different individuals, or the same individual over time. Understanding how parameters of echolocation (e.g., center frequency, p-p SPL) change in response to a decreasing hearing bandwidth are likely better assessed through longitudinal studies encompassing changes in hearing bandwidth rather than through cross-sectional studies conducted at a single point in time.

The authors thank R. Dear, M. Wilson, H. Bateman, R. Breitenstein, K. Christman, L. Crafton, C. Espinoza, G. Goya, M. Graves, J. Haynesworth, D. Ram, T. Wu, and the animal care, training staff, and interns at the Navy Marine Mammal Program. Financial support was provided by the Office of Naval Research Code 32 (Mine Countermeasures, Acoustics Phenomenology & Modeling Group).