The offset auditory brainstem response in bottlenose dolphins (Tursiops truncatus): Evidence for multiple underlying processes

Recordings of the auditory brainstem response (ABR) have been used to examine the function of dolphin audition both during active biosonar tasks [i.e., ABRs to dolphins' outgoing biosonar “clicks” and returning echoes, e.g., Supin et al. (2003), Li et al. (2012), Finneran et al. (2013), and Supin and Nachtigall (2013)] and passive listening [i.e., using experimentally controlled acoustic stimuli, Ridgway et al. (1981), Supin et al. (2001), and Mooney et al. (2012)]. ABRs examined in these studies are typically associated with the onset of sound stimuli, a pattern that parallels data from studies with humans and other terrestrial mammals [see Burkard and Don (2007) and Jones et al. (2019)]. Only a few dolphin studies have either noted (Popov and Supin, 1997; Finneran et al., 2020) or specifically examined (Burkard et al., 2020) ABRs to the offset of sound stimuli. Based on the limited existing data, offset ABRs in bottlenose dolphins (Tursiops truncatus) behave similarly to those in terrestrial mammals; offset ABR amplitudes are typically smaller than those for onset ABRs and increase with decreasing rise-fall time (RFT) and increasing stimulus level (Burkard et al., 2020). Further examination of the offset ABR in dolphins is of interest from the perspective of comparative sensory physiology, as some of the properties of the onset ABR that have resulted from the evolution of biosonar [e.g., large amplitude, rapid temporal resolution; see Supin et al. (2001)] may also be present in offset ABRs.

A few of the observations by Burkard et al. (2020) warrant more detailed investigation in dolphins. First, results with stimuli of varying frequency content and for dolphins with varying degrees of high-frequency hearing impairment suggested some difference in the specific cochlear places for onset and offset responses. This is similar to results with mice, where offset ABRs to tones appear to be generated by cochlear frequency regions near those responsible for the onset ABR, although with more limited contributions from the stimulus center frequency (Henry, 1985a,b). Second, dolphin offset ABRs for toneburst stimuli were sufficiently different from those for noiseburst stimuli to suggest potentially different mechanisms in underlying response generation [see also Brinkmann and Scherg (1979) and Laukli and Mair (1985)]. The experimental manipulations presented here delve further into the topics of dolphin offset ABR generation (as assessed through response morphology) and place specificity. Offset ABRs are reported for seven bottlenose dolphins in response to toneburst and noiseburst stimuli, in some cases using highpass noise (HPN) masking techniques (Teas et al., 1962; Don and Eggermont, 1978; Laukli and Mair, 1985; Finneran et al., 2016). Simultaneous presentation of HPN during stimulus presentation masks possible responses from cochlear regions basal (i.e., higher in frequency) to the masker cutoff, and thus the ABR arises from cochlear regions more apical (i.e., lower in frequency) to the masker cutoff than the highpass frequency, under the assumption that apical spread of the noise is negligible. The results provide insight into the mechanisms responsible for generating offset responses in dolphins, and the findings are contrasted with the more widely studied onset ABR.

Measurements were conducted with seven bottlenose dolphins at the U.S. Navy Marine Mammal Program (MMP) in San Diego Bay, CA (subject codes COL, IND, SHA, SPA, SPO, TRO, WHP). The dolphins ranged in age from 4 to 39 years and had upper-frequency limits (UFLs) of hearing from 64 to 147 kHz (Table I), defined as the frequency at which each dolphin had an electrophysiological hearing threshold of 120 dB re 1 μPa [Strahan et al. (2020); see ANSI (2018)]. Subjects with UFLs≥120 kHz were considered to have normal hearing (NH) (subjects SHA, SPO, TRO, and WHP) while the other dolphins were designated as hearing-impaired (HI) (subjects COL, IND, and SPA).

Testing took place in 9 m × 9 m floating netted enclosures in San Diego Bay. Dolphins were trained to position themselves on a “biteplate” located 1 m under water and attached to an extruded aluminum frame. The biteplate was located directly in line with the main transmitting axis of a piezoelectric sound projector (ITC 5446, International Transducer Corporation, Santa Barbara, CA) that was located 1 m in front of the dolphin. Underwater ambient noise during testing was dominated by snapping shrimp, other dolphins, and vessel traffic. Median noise pressure spectral densities were approximately 70 dB re 1 μPa²/Hz at 20 kHz and decreased linearly with the logarithm of frequency to approximately 55 dB re 1 μPa²/Hz at 150 kHz.

Stimuli were either tonebursts with center frequencies of 40 kHz, or spectrally “pink” noisebursts (i.e., power spectral density level that decreased 3 dB/octave) with frequency content varying across experiments. Due to the many stimulus manipulations used in this study, the general details are provided in these “general methods” sections, while more specific details are given in the sections for each experiment. Example calibrations for 40-kHz toneburst and noiseburst stimuli are shown in Fig. 1, but see Burkard et al. (2020) for more detailed examples of calibrations from the same environment and with comparable equipment and stimuli (although note that a cosine rise/fall function was used in that study, while the current study used a linear rise/fall function).

Tonebursts and noisebursts were generated [0.5–2 MHz digital-to-analog (D/A) rate, 16 bit] by either a continuous tone or band limited noise [National Instruments (NI) USB-6251 or PXIe-6368 data acquisition (DAQ) card, National Instruments Corporation, Austin, TX] sent to an NI PXI-7852R DAQ card with a Vertex-5 LX50 Field Programmable Gate Array (FPGA). Stimuli were digitized by the FPGA (500 kHz, 16 bit), multiplied by an envelope function consisting of a linear rise and fall and a constant amplitude plateau, and then converted to analog (500 kHz, 16 bit). All stimuli were attenuated (PA5, Tucker-Davis Technologies, Alachua, FL), filtered (0.2–200 kHz, 8-pole Butterworth, 3C module, Krohn-Hite Corporation, Brockton, MA), and amplified (31 dB, CC4000, Crest Audio, Meridian, MS), before being transmitted to the dolphin using the ITC 5446 transducer at a rate of 20/s. Where HPN was used simultaneously with stimulus presentation, the Gaussian noise masker was generated at 500 kHz (16 bit) with an NI PXIe-6361 DAQ card, attenuated using a PA5 attenuator, bandpass filtered 0.2–200 kHz with a 3C module, and added to the stimulus chain (SM5 Signal Mixer, Tucker-Davis Technologies) immediately prior to amplification with the CC4000 amplifier.

Calibration was conducted without the dolphin present, using a TC4013 hydrophone (Teledyne-Reson, Slangerup, Denmark) placed at the estimated midpoint of the ears when stationed on the biteplate. A Reson VP1000 voltage preamplifier (32 dB, 2-kHz highpass) was used to amplify the signal prior to digitization (16 bit) using either the USB-6251 or PXIe-6368 DAQ card. For noiseburst stimuli and masking noise, an equalization procedure was used to account for the frequency response of the ITC 5446 and underwater multipath effects [as described by Finneran et al. (2018)]. This procedure resulted in “pink” noise with equal SPL in proportional octave bands. Masking noise was calibrated in terms of mean 1/3-octave band SPL over the passband. Tonal and noise stimuli were calibrated in terms of rms SPL (dB re 1 μPa) in the stimulus plateau.

ABRs were recorded using surface electrodes embedded in silicone suction cups placed on the skin surface of the dolphin. A non-inverting electrode was placed approximately 5 cm behind the blowhole along the midline, and an inverting electrode was placed immediately behind the right external auditory meatus. A common electrode was placed into the seawater near the experimental apparatus. The incoming signal from the electrodes was filtered (0.3–3 kHz) and differentially amplified (94 dB) using an ICP511 biopotential amplifier (Grass Technologies, West Warwick, RI). The signal was digitized at 100 kHz (16 bit) using the PXIe-6368 DAQ card and stored to hard disk. The EEG was digitally bandpass filtered (0.3–3 kHz) using a zero-phase implementation of a 6th-order Butterworth filter prior to averaging to visualize ABRs. The ABRs for each condition were typically based on two recordings, each comprised of approximately 512 individual stimulus presentations (epochs). These time-domain data were coherently averaged to produce an averaged ABR. Eight ABRs per condition were collected in cases where an especially high signal-to-noise level (SNR, in dB) was desired (see below).

For ABR amplitude analyses, the multiple replicate recordings were averaged together to create a final average for a condition. Amplitudes for onset and offset ABRs were calculated as the rms amplitude in the time interval from 1 to 7 ms relative to the rise and fall of the acoustic stimulus arriving at the dolphin. This window was chosen to begin the calculation immediately before the first peaks of the ABR and end following the last peaks, to provide a measurement that was relatively independent of waveform morphology, and was used effectively by Burkard et al. (2020). Residual background noise in the averaged ABRs was estimated by calculating the rms amplitudes from the corresponding time periods of the “± average” (Schimmel, 1967; Wong and Bickford, 1980; Finneran et al., 2019). The SNR level was calculated for each dolphin/stimulus condition according to

where A is the rms amplitude of the averaged onset or offset ABR, N is the rms amplitude of the background noise in the corresponding time window, and the overbar represents the mean value (averaged across animals). For this study, ABR data were only analyzed in cases where the mean SNR level for a condition was greater than 3 dB.

Responses to 40-kHz tonebursts were measured in HPN conditions (120 dB re 1 μPa, 1/3-octave band SPL) with highpass cutoff frequencies between 28 and 134 kHz. The tonebursts had 32-μs rise/fall times, 8-ms plateaus, and SPLs of 115 dB re 1 μPa. Examples of two 512-epoch records obtained for a NH (WHP, UFL of 143 kHz) and a HI (COL, UFL of 83 kHz) dolphin are shown in Fig. 2, and a summary of the rms amplitudes of the onset and offset ABRs for all of the dolphins is shown in Fig. 3. In Fig. 2—and for all other figures including ABR waveforms—positivity at the non-inverting electrode is plotted upwards. Both the onset and offset ABRs increased in amplitude with increasing noise highpass frequency across the tested range for the NH dolphins, or increased to a plateau near the 67-kHz highpass for the HI dolphins. The amplitudes of the offset ABRs were lower than those for the onset responses at all conditions (Fig. 3, top row). The variability was larger for the HI dolphins than for the NH dolphins, presumably due to the differences in UFL among those subjects. However, this variability was reduced when the responses were calculated as a percentage of the unmasked response for each dolphin (Fig. 3, middle row). When the offset response amplitudes were calculated as a percentage of the onset response for a particular condition (Fig. 3, bottom left), the relative amplitude of offset ABRs increased with increasing highpass frequency, especially for the HI dolphins.

Measurements were made with noiseburst stimuli and HPN in two dolphins, one NH (TRO) and one HI (COL). Noisebursts were spectrally pink from 20 to 40 kHz, with the same envelope and SPL as the 40-kHz tonebursts described above (32-μs rise/fall, 8 ms plateau, 115 dB re 1 μPa). Stimuli were simultaneously presented with a pink HPN masker (120 dB re 1 μPa, 1/3-octave band SPL) with a varying highpass cutoff frequency. A high-pass subtractive masking method was used to obtain the derived-band ABRs within half-octave (or octave, for the 20–40 kHz bandwidth of the stimulus) frequency regions of the hearing range (Teas et al., 1962; Don and Eggermont, 1978; Laukli and Mair, 1985; Finneran et al., 2016). There is a 3-dB decrease in the SNR of the ABR that results from the subtraction of waveforms using this procedure, therefore eight 512-epoch replicates were obtained to improve the SNR. Figure 4 shows the derived-band ABRs obtained for TRO and COL for the 20–40 kHz noiseburst. The amplitudes of the onset and offset responses as a percentage of the corresponding ummasked response further suggested that the offset ABRs were generated at slightly higher-frequency, more basal cochlear regions than the onset ABR. Interestingly, the summed percentages for the derived-band ABRs for TRO totaled more than 100% of their unmasked counterpart. This appeared to result from the fact that waveforms for the 40–57 kHz and 57–80 kHz derived-band ABRs showed differences that led to destructive interference in the unmasked response.

A close inspection of the unmasked offset ABRs in Fig. 4 reveals a difference in the morphologies of the offset and onset responses. As most clearly observed for COL, the first main voltage deflection in the offset ABR (wave P1) appears to be in the vertex negative direction, as opposed to the normal vertex positive P1—visible in the onset ABR—that arises from the compound action potential of the auditory nerve (Popov and Supin, 1991). This could be interpreted as an absence of that particular wave in the offset response, or, alternatively, this morphology difference could arise from waveform inversion. The latter interpretation of waveform inversion was previously suggested by Brinkmann and Scherg (1979); the authors noted that human offset ABRs for noisebursts (but not tonebursts) appeared to be inverted relative to onset ABRs, and they speculated that this could arise from synchronous reduction in neural firing rate at the offset of a stimulus. To further examine if this is the case for dolphins, we conducted a set of experiments aimed at examining ABRs to noisebursts. It was hypothesized that, with increasing noiseburst bandwidth, the offset ABR would become more distinct from the toneburst evoked ABR and potentially show the polarity inversion suggested by Brinkmann and Scherg (1979).

Measurements were conducted with all seven dolphins in ambient bay noise only using noisebursts that were identical in envelope and plateau SPL to those for noisebursts in the measurements described above (32-μs rise/fall, 8-ms plateau, 115 dB re 1 μPa). The only differences were that the noiseburst highpass cutoff was set to 40 kHz, and the lowpass was progressively increased in 1/4-octave steps up to 160 kHz. Thus, the stimuli progressively increased in bandwidth as the lowpass frequency was increased. Two 512-epoch replicate measurements were obtained per condition. Figure 5 shows examples of records for a NH (SPO, 147-kHz UFL) and a HI (SPA, 101-kHz UFL) dolphin, and Fig. 6 plots the mean amplitudes of the ABR peaks for all dolphins as a function of noiseburst bandwidth. The most obvious pattern was that onset ABR amplitudes increased with increasing bandwidth. However, the offset ABR amplitudes remained essentially constant with changes in bandwidth. The frequency at which onset amplitudes reached a plateau differed between the two groups: amplitudes plateaued above 134-kHz lowpass for the NH dolphins, and above approximately 95-kHz lowpass for the HI dolphins.

To better visualize the polarity of the ABR morphologies in noise, eight 512-epoch replicates were collected for 40-kHz toneburst and noiseburst stimuli with TRO (NH, 144-kHz UFL) and COL (HI, 83-kHz UFL). The stimuli were identical in temporal envelope to those from the previous measurements (32-μs rise/fall, 8-ms plateau) but the plateau SPLs were increased to 124 dB re 1 μPa to increase ABR amplitudes (and therefore SNR). Figure 7 shows the averaged ABRs for TRO and COL, with responses to the toneburst as the top traces and those to noisebursts of increasing bandwidth below. The rms amplitudes for the responses are shown in Fig. 8. As observed in the previous measurements, the onset ABR increased with increasing stimulus bandwidth (although not past an 80-kHz lowpass cutoff frequency for COL, due to his hearing impairment). A different pattern was observed with the offset responses: the offset ABR was of higher amplitude for the toneburst and narrowest-band noise, decreased in amplitude with increasing noise bandwidth near the 40- to 57-kHz lowpass cutoff transition point, and remained at essentially a constant amplitude with increasing bandwidth. Also, shown on the right in Fig. 7 is a comparison of the onset and offset ABRs for each condition, with the offset either with its normal polarity (0°, positivity at vertex plotted upwards) or inverted (180°, positivity at vertex plotted downwards). The onset and offset ABRs for the lowest frequency, lowest bandwidth stimuli were relatively similar in amplitude and morphology. However, as the bandwidth of the stimuli increased, the offset response changed in morphology. Notably, there was a progressive reduction and eventual disappearance of P1 in the offset ABR with increasing bandwidth, even as P1 in the onset response increased in amplitude. Additionally, regardless of the offset polarity used for comparison, the latencies of the peaks and troughs in the offset ABRs appeared to shift relative to the onset ABR (shifted to shorter latencies for 0° and longer latencies for 180°).

The results collectively suggest that the “offset ABR” in bottlenose dolphins may not arise from a single underlying mechanism but may be the result of at least two distinct mechanisms. Voltages from multiple processes (not to mention those from multiple neural structures) summing at the level of the recording electrodes may be responsible for the varied ABR patterns observed with different stimulus manipulations in the current study. The following discussion will focus on two proposed mechanisms of offset response generation based on interpretations of the current results and on prior data from bottlenose dolphins and terrestrial mammals. Much of the following discussion is based off of the more detailed measurements made with only two dolphins (TRO and COL). This sample size is obviously small (but is a common limitation of studies with marine mammals), and the conclusions must therefore be taken with some caution in the absence of replication with more individuals.

One pattern of offset ABR was observed primarily when using toneburst (i.e., narrowband) stimuli. This response was clearly visible in the data using 40-kHz stimuli, and based on experimental manipulations using HPN, it was more prominent at slightly higher frequencies than the onset ABR. Similar to the onset ABR, the amplitude of this offset ABR increased as masking noise high-pass frequency was increased, which may point to the larger contributions of more basal cochlear regions to the dolphin ABR (Popov and Supin, 2001; Finneran et al., 2016). The offset response here is most likely evoked by stimulus offset at cochlear regions that are more basal than those underlying the onset ABR [i.e., it is an “onset” response, as previously described by Brinkmann and Scherg (1979), Laukli and Mair (1985), and Burkard et al. (2020)], and in some cases, appeared to have contributions from cochlear regions up to the UFL (i.e., Fig. 4). This basal spread of activation may result from non-linearities in the response of the basilar membrane (Rhode and Robles, 1974). It is also possible that the few decibels of “overshoot” relative to plateau SPL and a widening of stimulus bandwidth occurring at the short stimulus fall time could contribute to these offset ABRs (Burkard et al., 2020).

This type of response (termed here as an across place offset response) was clearly visible for 20–40 kHz noisebursts only when the amplitude was increased from 115 to 124 dB re 1 μPa for the final experiment (i.e., data in Figs. 7 and 8). The rms SPL method used for calibrating stimuli could have led to this result. Frequency-specific toneburst stimuli have proportionally more energy in a restricted frequency region than the octave band (20–40 kHz) of noise with the same rms SPL. “Upward,” basal activation due to high stimulus levels [i.e., a loss of place specificity, Finneran et al. (2016)] might therefore require relatively higher rms SPL levels for a broadband noiseburst relative to a toneburst with a single carrier frequency. In other words, had the spectrum level been held constant across noise bandwidth (rather than SPL), the pattern of offset response amplitude may have differed from that observed in Figs. 7 and 8.

The contributions of the proposed second type of offset ABR (termed within place) had two critical features. First, the amplitude of the offset response initially decreased and then remained essentially constant, while the onset response amplitudes increased with bandwidth of the noiseburst stimuli (i.e., the lowpass cutoff). The direct relationship between onset ABR amplitude and stimulus bandwidth is explained by growth in the contribution (i.e., in-phase voltage summation) of neural activity from increasingly more basal cochlear regions as the noiseburst lowpass frequency increases (Don et al., 1994; Popov and Supin, 2001; Burkard and Don, 2007). The initial decrease followed by constant amplitude in the offset ABR is counterintuitive and suggests that the response in these dolphins is not primarily generated by more basal cochlear activation—as previously suggested by data from cats (Laukli and Mair, 1985). This is also suggested by the similar amplitudes of the offset ABRs at all stimulus conditions for TRO (NH) and COL (HI) in the final experiment, despite the obvious differences in basal contributions to the onset ABR as stimulus bandwidth increased.

Second, this bandwidth-independent offset ABR appears to have a morphology that is different than that of the onset ABR, as well as to the offset ABRs for the 40-kHz toneburst and the narrowest-band noiseburst in Fig. 7. The most notable of these differences are the apparent absence of a P1 and smaller vertex-negative deflection near 4 ms after stimulus offset [an absence of P1 in an offset response to noiseburst stimuli was also suggested in a previous study with dolphins; Finneran et al. (2020)]. Of course, if the offset ABR were inverted relative to that of the onset response (Brinkmann and Scherg, 1979) and there were even a small shift in peak latencies, it would be difficult to definitively conclude that P1 is absent as opposed to appearing as an initial negative deflection. This can be seen in the comparisons of the offset waveforms (0° and 180°) with the onset waveforms in Fig. 7, where for the broadest-bandwidth stimuli, the peaks for neither offset ABR polarity line up precisely with peaks in the onset ABR. Unfortunately, cross correlation analyses of the onset and offset responses (data not shown) were not able to definitively answer if the latter is shifted in latency, inverted, lacking P1, or a combination of these features, as they are to some degree intertwined in time-domain analyses. However, the last prominent peak near ∼4–5 ms in the onset ABR and the 0° offset ABR both being positive suggests that the “missing P1” explanation may be more parsimonious than an “inversion” explanation that would require a synchronous reversal of the underlying generator dipoles.

It is hypothesized that the within place offset response is a “true” offset response that is synchronous with the cessation of sound and is not inverted relative to the onset ABR [as proposed by Brinkmann and Scherg (1979)]. Based on the current observations (i.e., Figs. 7 and 8), this response can be obscured in the time domain by the across place response when more basal regions of the cochlea are available for response generation. When noisebursts with an increasing lowpass frequency are used, these more basal regions appear to be “occupied” (or “masked”) by these higher noiseburst frequencies, eliminating across place responses and revealing the within place offset responses. Whether this represents a simultaneous masking effect through residual stimulation from the stimulus plateau and the continued effect of noise in the fall from stimulus plateau to ambient pressure—as opposed to a forward masking from only the higher-amplitude stimulus plateau—is not known. It is also not currently known how the reporting of pink noise stimuli in terms of rms energy as opposed to, e.g., 1/3-octave band SPL might have affected the amplitudes of ABRs across bandwidth. The latter form of equalization might increase ABR amplitudes at the higher bandwidths used in the study as energy per 1/3-octave band would not decrease with increasing bandwidth.

The apparent independence of the within place offset responses on noiseburst bandwidth, combined with the potential lack of a defined P1 suggests that it is not initiated in the eighth nerve, but at a higher level in the ascending auditory nervous system [see Kopp-Scheinpflug et al. (2018) for a review of offset responses from the cochlea and auditory nerve versus those from the central auditory nervous system]. Here, again, the similarity of the offset ABR amplitudes for TRO (NH) and COL (HI), despite marked differences in high-frequency (basal) cochlear function and consequent bandwidth-related differences in the onset ABR amplitudes [see Popov and Supin (2001)], suggests a generation site further up the ascending auditory pathway than the cochlea and auditory nerve. A synchronous response at sound offset could be preserved from, e.g., the cochlear nuclei through the more rostral brainstem structures that generate the later ABR peaks [e.g., superior olivary complex, lateral lemniscus, inferior colliculus; see Supin et al. (2001)].

As the offset response to tonebursts is very sensitive to stimulus fall times [and offset responses to tonebursts are substantial only for short fall times (32 or 64 μs), Burkard et al. (2020)], how either of these offset responses would provide information useful to the dolphin, particularly during echolocation, is an open question. The click signals and returning echoes used in echolocation are short in duration [<100 μs, ignoring temporal distortions (Au, 1993; Wahlberg et al., 2011)], and completely within the temporal window responsible for the onset ABR (Moore et al., 1984; Finneran et al., 2018; Jones et al., 2019; Finneran et al., 2020). Offset responses may therefore be of little importance to high-frequency echolocation in this species. As the present data suggest that the within place offset response may not require the highest frequencies present in dolphin biosonar (i.e., its amplitude was not highly dependent on noise lowpass frequency or the UFL of the dolphins), perhaps the within place response primarily encodes the offset of longer-duration communication signals at lower frequencies (see Tyack, 1998).

ABRs to toneburst and noiseburst offsets in bottlenose dolphins suggested that offset ABRs may represent voltage contributions from at least two different processes. The first hypothesized type, across place, is akin to the onset ABR. It is mostly observed for tonebursts (although it can be generated by narrowband noisebursts of sufficient level), is generated from activation of neural units at the level of the auditory nerve and projected to auditory centers higher in the ascending auditory pathway, and features notable contributions from frequencies above the stimulus (i.e., more basal cochlear locations). The second hypothesized type, within place, potentially results from excitatory offset responses at auditory centers higher than the cochlea (the cochlear nucleus and above). Within place responses are most commonly observed for broadband noiseburst stimuli. For toneburst stimuli, the within place response might be obscured by the larger amplitude across place response. For broadband noisebursts, broadband masking from the stimulus could eliminate the across place response, allowing the within place response to be observed. Within place offset responses may be of more interest for future research than across place offset responses, given the similarity of the latter to extensively studied onset ABRs. Experiments of interest could manipulate noiseburst parameters to examine the contribution of lower frequencies to the within place offset responses, for example, presenting noiseburst stimuli in which the highpass frequency is systematically decreased.

The authors thank R. Dear, M. Wilson, H. Bateman, R. Breitenstein, K. Christman, L. Crafton, L. Curtis, C. Espinoza, G. Goya, M. Graves, J. Haynesworth, D. Ram, T. Wu, and many interns of the animal care and training staff at the Navy MMP. The comments of Dr. Alexander Supin and Dr. Joaquín Valderrama substantially improved the quality of this manuscript during the review process. The experimental protocol for this study was 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. The study was funded by the U.S. Navy Living Marine Resources Program. This is contribution 301 of the National Marine Mammal Foundation.