Echolocating bats and odontocetes face the potential challenge of acoustic interference from neighbors, or sonar jamming. To counter this, many bat species have adapted jamming avoidance strategies to improve signal detection, but any such avoidance strategies in dolphins is unknown. This study provides an investigation into whether dolphins modify echolocation behavior during jamming scenarios. Recorded echolocation clicks were projected at different click repetition rates and at different aspect angles relative to two dolphins' heads while each dolphin was performing a target detection task. Changes in the timing, amplitude, and frequency of structure of the dolphin's emitted signals were compared to determine if and how dolphins modify echolocation when faced with potentially interfering conspecific echolocation signals. The results indicate that both dolphins demonstrated different responses when faced with jamming scenarios, which may reflect optimal strategies according to individual auditory perception abilities.
Odontocetes and microchiropteran bats are two groups of mammals that convergently evolved echolocation to sense their environments (Liu et al., 2010). Both groups of animals produce ultrasonic echolocation signals, but the sound generation and time-frequency characteristics of their calls are different. Odontocetes generate echolocation clicks in nasal structures (Cranford et al., 1996), whereas bats generate echolocation signals in the larynx (Neuweiler, 2000). Odontocetes produce impulsive, broadband clicks whose durations (for Tursiops truncatus) are between 40 and 70 μs (Au, 1980), and bats produce frequency modulated echolocation signals, with multiple harmonics, and whose durations can last up to several hundred milliseconds (Neuweiller, 2000).
While echolocating near conspecifics, both bats and odontocetes face the same challenge of acoustic interference from neighbors, or mutual-interference. This is especially important for species of bats that form large cave colonies up to several millions of individuals (Betke et al., 2008; Kloepper et al., 2016), or with dolphin species that form temporary groups of up to thousands of individuals (Moron et al., 2014; Butler et al., 2017). To avoid such jamming, many bats have developed a jamming avoidance response in which individuals shift their frequencies away from interfering sounds (Ulanovsky et al., 2004; Gillam et al., 2007; Bates et al., 2008). Additional strategies include altering pulse timing (Jarvis et al., 2013), directing sonar beams away from other bats (Chiu et al., 2010), or reducing or stopping pulse emission (Chiu et al., 2008; Adams et al., 2017). Other studies suggest that bats do not alter pulses to avoid jamming; rather, pulse modifications are consistent with the bats responding to conspecifics as they would any object in their flight path (Cvikel et al., 2014; Götze et al., 2016).
Thus far, all jamming investigations have been limited to bats, and no knowledge exists on how odontocetes avoid jamming, or if they are even susceptible to jamming since their echolocation signals are much shorter in duration than bat echolocation signals. Although odontocetes can modify the spectral and amplitude components of their echolocation signals, even on a pulse-by-pulse basis (Kloepper et al., 2012a,b; Kloepper et al., 2014), due to the structure of their signals they do not have the same flexibility to adjust time-frequency characteristics parameters as in bats. Additionally, bats can adjust their receiver beam pattern by actively manipulating their pinnae (Müller et al., 2008; Gao et al., 2011). Although odontocetes may control the hearing sensitivity of their sonar system (Supin and Nachtigall, 2013), no analogous mechanism to the pinnae manipulation in bats has been identified in odontocetes (Møhl et al., 1999). Therefore, any adaptations dolphins use to avoid jamming while in groups will likely occur in the transmit portion of their biosonar system.
This experiment provides the first investigation into how odontocetes modify echolocation signals when faced with different jamming scenarios. Specifically, we projected a recorded echolocation click at different click repetition rates and at different aspect angles relative to the dolphin's head. We compared changes in the timing, amplitude, and frequency structure of the dolphin's emitted signals to determine if and how dolphins modify echolocation when faced with potentially interfering conspecific echolocation signals. Additionally, we utilized two dolphins, one with normal hearing and one with hearing loss, to determine if jamming avoidance is affected by both sound production and reception.
Two male Atlantic bottlenose dolphins (Tursiops truncatus) participated in the study. Dolphin 1 (TYH) was 37 yr old with hearing loss determined by an AEP hearing test.1 Dolphin 2 (SPA) was 30 yr old with normal hearing determined by an auditory evoked potential (AEP) hearing test. Both dolphins had extensive experience with echolocation tasks. The dolphins were housed in 9 m × 9 m or 9 m × 18 m floating netted enclosures located in San Diego Bay, California, at Naval Base Point Loma. The study was approved by the Institutional Animal Care and Use Committee of the Biosciences Division, Space and Naval Warfare Systems Center Pacific, and the Navy Bureau of Medicine and Surgery. The study followed all applicable U.S. Department of Defense guidelines for the care of laboratory animals.
B. Phantom echo generator
The phantom echo generator (PEG) was similar to the system used in previous experiments (Branstetter et al., 2018; Finneran et al., 2010). The dolphin's incident signal was acquired with a Reson TC4013 hydrophone (Teledyn Reson, Slangerup, Denmark), coupled to a custom pre-amplifier (6 dB gain, 5 kHz high pass filter), digitized at 500 kS/s (National Instruments PXI-7852R data acquisition board, National Instruments, Austin, TX) containing a Virtex-5 LX50 field-programmable gate array. If the peak-to-peak amplitude of the incident signal exceeded a threshold, the Fourier transform of the signal was then multiplied by a target transfer function. The transfer function in the current study simulated a hollow sphere. The signal was then attenuated and delayed simulating the two-way acoustic propagation of a target at 10 m distance. The inverse-Fourier transform of the signal was then converted to analog, amplified (Hafler PRO2000, Tempe, AZ) and projected back to the dolphin with a separate Reson TC4013 projector. Previous research has demonstrated that dolphins echolocating on “phantom targets” behave in a similar manner as if they were echolocating on physical targets (Finneran, 2010). The advantage of the PEG system is that it allows precise experimental control over many parameters such as target range, type, and occurrence.
C. Target detection task
Each dolphin's task was to echolocate on the PEG station and report the occurrence of target echoes (i.e., target detection task) with a whistle response. If no target occurred, the dolphin would continue to echolocate without producing a whistle. During each experimental trial, the dolphin would swim and position himself on a custom “PEG station” constructed of plastic and polyvinyl chloride (PVC), and begin continuously echolocating [Fig. 1]. The PEG station contained a receiver hydrophone (h1) 1 m in front of the dolphin and a projector hydrophone (h2) 20 cm in front of the dolphin.
For each trial, there was a 70% probability of a “target trial” and a 30% probability of a “catch trial.” For target trials, the PEG system would randomly turn on after 3–6 s, producing target echoes. If the dolphin correctly produced a whistle in response to the detected echoes (hit) within 2 s of the onset of the target echoes, the dolphin received a whistle “bridge” from the trainer, followed by fish reinforcement. If the dolphin failed to whistle during the 2-s window (miss) the dolphin received a “splash” call back, returned to station, but did not receive fish reinforcement. During catch trails, the same timing interval was used as a target trial; however, the PEG system remained off and did not produce any echoes. If the dolphin continued to echolocate without producing a whistle (correct rejection) the animal received a “bridge” followed by fish reinforcement. If the dolphin whistled during a catch trial (false alarm) the animal received a “splash” call back followed by no fish reinforcement. Any whistles that occurred outside of the response window were considered false alarms.
While the dolphin was performing the task, an interfering stimulus (IS) was continuously projected into the water column [Figs. 2(a) and 2(b)]. The IS was an on-axis recording of a Tursiops truncatus click signal (1 MHz sampling rate) with a duration of 121 μs, and peak and center frequencies, respectively, of 50.8 and 59.3 kHz. For the IS, click amplitude was held constant (168 dB p-p re 1 μPa) and the inter-click interval (ICI) was controlled with custom Labview software. The digital signal was converted to analog (National instruments, USB 6251 data acquisition device, National Instruments, Austin, TX), amplified (Hafler P1000, Tempe, AZ) and projected into the water column with an ITC 1042 transducer (International Transducer Corporation, Santa Barbara, CA). The location of the IS projector [Fig. 1, h3] relative the dolphin's head was an independent variable, and systematically varied in the horizontal plane, but was maintained at a fixed radial distance of 1 m from the dolphin's head (measured from the animal's blowhole).
The experiment was conducted in two phases: Phase 1 investigated how changes in the ICI of the IS affected the emitted dolphin signal characteristics. First, the “baseline” click rate was determined for each animal by averaging the dolphin's ICIs over a session of 60 trials with no interfering stimulus. From these sessions, an average baseline click rate of 16 clicks/s and 21 clicks/sec were determined for Dolphin 1 and 2, respectively. Then, an interfering stimulus was tested, consisting of three temporal conditions of Baseline −50%, Baseline, and Baseline +50% (Table I).
|.||Dolphin 1 .||Dolphin 2 .|
|.||Click rate (clicks/s) .||ICI (ms) .||Click rate (clicks/s) .||ICI (ms) .|
|A: Baseline −50%||8||125||10.5||95.2|
|C: Baseline +50%||24||41.6||31.5||31.7|
|.||Dolphin 1 .||Dolphin 2 .|
|.||Click rate (clicks/s) .||ICI (ms) .||Click rate (clicks/s) .||ICI (ms) .|
|A: Baseline −50%||8||125||10.5||95.2|
|C: Baseline +50%||24||41.6||31.5||31.7|
Each session followed an ABBA counterbalanced format with ten trials for each temporal condition. All temporal stimulus conditions were presented with the projector hydrophone (h3) at the same azimuth of 45°. Two sessions were conducted for each condition, and we only selected trials from target present trials for final analysis. This resulted in 20 trials for each condition.
Phase 2 of the experiment investigated how changes in the azimuth of the interfering stimulus affected emitted dolphin signal characteristics. All methods were the same as described above in Phase 1 except that we maintained the baseline click presentation rate and varied the angle of the projected interfering sound. Conditions consisted of 90, 45, and 0°, presented unilaterally on the left side of both dolphins. Data for the azimuth conditions were also collected using an ABBA counterbalanced format.
Data analysis was the same for both phases of the experiment. For each click produced during a trial, the ICI, duration (based on 95% of the click energy from the waveform), peak frequency, center frequency (as defined in Au, 1993), −3 dB bandwidth, and source level (SPL p-p re: 1 μPa) were calculated. All parameters were automatically extracted with custom matlab software, which first removed the IS from the waveform before click extraction. This algorithm also removed dolphin signals that overlapped with the IS, which would result in the inflation of ICI values for subsequent clicks. To account for this, values were removed from our dataset where the ICI was greater than 100 ms. Additionally, outliers were removed from the dataset, which were defined as ICIs less than 20 ms, and click durations less than 30 μs or greater than 100 μs. These outliers represented 5.82% of Dolphin 1's data, and 4.43% of Dolphin 2's data. To further determine the effects of the IS on dolphin response behavior, the response latency (defined as the time from the first PEG echo to the last dolphin echolocation click) and number of stimulus echoes (the number of PEG echoes received before the dolphin produces a whistle response) were determined for each trial. The performance (% correct) was also determined for all treatments. Although linear regression is typically used when predictor variables are continuous, a choice was made to use analysis of variance (ANOVA) instead, and to treat levels of each predictor variables as categorical. This decision was made because, at least with respect to azimuth, acoustic patterns around the dolphin's head are non-linear. For example, the beam pattern of a dolphin's emitted click and a dolphin's perception of sounds at different azimuthal positions is largely affected by the animal's non-linear head related transfer function (Aroyan, 2001; Branstetter and Mercado III, 2006; Cranford et al., 2008; Supin and Popov, 1993). Linear regression may fail to capture significant differences related to a perceptual “sweet spot” that may affect an animal's perception of sound. All statistics were performed in SPSS v.24.
A. Overall echolocation behavior
Both dolphins maintained 100% correct performance on the target detection task for all IS treatments. For each dolphin, there was no significant difference in response latency or number of stimulus echoes before response for the different treatments of IS (p > 0.05). The dolphins did, however, demonstrate different overall acoustic response behaviors. Dolphin 1 had a significantly longer response latency time (0.437 +/− 0.127 s) and number of stimulus echoes (7.85 +/− 3.51) than Dolphin 2 [(latency: 0.234 +/− 0.030 s, echoes: 5.79 +/− 1.22); Response latency: t(150) = 13.59, p < 0.001; Echoes: t(150) = 4.88, p < 0.001]. Furthermore, Dolphin 1 produced baseline echolocation clicks with longer ICIs (Table I) and lower frequencies [Figs. 2(c) and 2(d)] than Dolphin 2 [Figs. 2(e) and 2(f)].
B. Effect of interfering stimulus inter-click interval
For Dolphin 1, there was a significant difference in all click parameters across the temporal stimulus conditions [One-way ANOVA: ICI: F(2,3997) = 8.010, p < 0.001; Duration: F(2,3997) = 14.131, p < 0.001; Bandwidth: F(2,3997) = 19.458, p < 0.001; Peak Frequency: F(2,3997) = 27.534, p < 0.001; Center Frequency: F(2,3997) = 20.514, p < 0.001; Source Level: F(2,3997) = 16.565, p < 0.001; Fig. 3].
As the click presentation rate of the interfering stimulus increased, the ICI and peak frequency emitted by the dolphin lowered. Duration also shortened between the Baseline −50% and Baseline conditions, but there was no significant difference in duration between the Baseline and Baseline +50% stimulus conditions. Center frequency and bandwidth were not different between the Baseline −50% and Baseline conditions, but the Baseline +50% conditions were significantly lower. Dolphin 1 also produced significantly louder source levels for the Baseline condition compared to the other conditions.
Dolphin 2 also demonstrated a significant difference in all click parameters across temporal stimuli [One-way ANOVA: ICI: F(2,4444) = 12.208, p < 0.001; Duration: F(2,4444) = 20.746, p < 0.001; Bandwidth: F(2,4444)= 114.804, p < 0.001; Peak Frequency: F(2,4444) = 4.085, p = 0.017; Center Frequency: F(2,4444) = 4.488, p = 0.011; Source Level: F(2,4444) = 27.156, p < 0.001; Fig. 4]. ICI increased as the click presentation rate of the interfering stimulus increased. Click durations and bandwidth were the smallest for the baseline condition. No consistent trend in peak and center frequencies were observed across conditions, and source levels were highest for the baseline condition.
The relationship between source level and the click parameters center frequency, duration, bandwidth, and ICI, separated by dolphin and stimulus condition, is illustrated in Fig. 5. Results of the multiple linear regression indicated an overall significant effect between dolphin, stimulus condition, and source level on center frequency [F(7,8625) = 8017, p < 0.001; adjusted R2 = 0.867], duration [F(7,8625) = 294, p < 0.001; adjusted R2 = 0.192], bandwidth [F(7,8625)= 1556, p < 0.001; adjusted R2 = 0.558], and ICI [F(7,8505) = 34.185, p < 0.001; adjusted R2 = 0.027]. Controlling for dolphin and stimulus condition, center frequencies (b = 1935, p > 0.001) and bandwidths (b = 3042, p > 0.001) increased with source level, durations decreased with source level (b = −1.37e−6, p > 0.001), and there was no significant effect of source level on ICI (b = 8.51e−5, p = 0.107). Controlling for stimulus condition and source level, Dolphin 1 produced lower center frequencies (b = −17497, p < 0.001), shorter durations (b = −3.933e−6, p < 0.001), narrower bandwidths (b = −3872, p < 0.001), and longer ICIs (b = 0.005, p < 0.001) than Dolphin 2. Controlling for dolphin and source level, all parameters significantly (p < 0.001) differed among individual stimulus conditions.
To examine whether the dolphins modify their ICIs within a trial, which may indicate learning or adaptation, we averaged the number of clicks within 500 ms bins throughout all the trials (Fig. 6). Dolphin 1 demonstrated a consistent reduction in ICI as time in each trial progressed for all temporal IS treatments (baseline-50%: [F(1,1204) = 166, p < 0.001, R2 = 0.121; ]; baseline: [F(1,1322) = 191, p < 0.001, R2 = 0.126; ]; baseline + 50%: [F(1,1373) = 165, p < 0.001, R2 = 0.107; ]), and Dolphin 2 also demonstrated a reduction in ICI as the time in trial progressed for all treatments (baseline-50%: [F(1,1501) = 54.8, p < 0.001, R2 = 0.035; ]; baseline: [F(1,1395) = 4.87, p < 0.028, R2 = 0.003; ]; baseline + 50%: [F(1,1409)= 89.7, p < 0.001, R2 = 0.060; ]). A Levene's test for equality of variance compared the difference in the emitted ICI variance between 0 and 500 ms (start) and 2000–2500 ms (middle/end) within the trial. Because the PEG echoes would randomly turn on after 3–6 s, the 2000–2500 ms window was chosen to represent the middle/end portion of a trial while still preserving the same sample size as in the 0–500 ms window. For Dolphin 1, the variance in emitted ICIs was the same between the start and middle/end of Baseline −50% stimulus trials [F(1,258) = 0.3275, p = 0.5676], but variance was significantly different between the start and middle/end of trials for the Baseline [F(1,268) = 4.5478, p = 0.0339] and Baseline +50% conditions [F(1,252) = 9.8830, p = 0.0019], with greater variance in ICIs for the start compared to the middle/end of trials. Dolphin 2 demonstrated significantly greater variance in ICIs for the beginning of trials for both the Baseline −50% [F(1,300) = 16.768, p = 0.0001] and Baseline conditions [F(1,292) = 7.1573, p = 0.0079], but demonstrated significantly less variance at the start of Baseline +50% condition trials compared to the middle/end of the trials [F(1,283) = 4.4163, p = 0.0365].
C. Effect of horizontal azimuth of interfering stimulus
Dolphin 1 demonstrated a significant difference in all click parameters across azimuthal stimuli [One-way ANOVA: ICI: F(2,3885) = 19.946, p < 0.001; Duration: F(2,3885) = 19.439, p < 0.001; Bandwidth: F(2,3885)= 18.763, p < 0.001; Peak Frequency: F(2,3885) = 36.470, p < 0.001; Center Frequency: F(2,3885) = 22.653, p < 0.001; Source Level: F(2,3885) = 10.706, p < 0.001; Fig. 7]. There was no consistent trend in ICI as the angle of the stimulus decreased, but duration, bandwidth, peak frequency, center frequency, and source levels were highest when the stimulus speaker angle was at 0°.
For Dolphin 2, there was a significant difference in all click parameters across azimuthal stimuli [One-way ANOVA: ICI: F(2,5761)= 20.743, p < 0.001; Duration: F(2,5761) = 18.638, p < 0.001; Bandwidth: F(2,5761) = 215.491, p < 0.001; Peak Frequency: F(2,5761) = 206.224, p < 0.001; Center Frequency: F(2,5761) = 272.784, p < 0.001; Source Level: F(2,5761) = 120.044, p < 0.001; Fig. 8]. ICIs were highest when the speaker was at 0°, and durations were shortest when the speaker was at 90°. Bandwidth, peak frequency and center frequency decreased with decreasing speaker angle. Source levels were highest when the speaker was at 90°.
Figure 9 illustrates the relationship between source level and the click parameters center frequency, duration, bandwidth, and ICI, separated by dolphin and stimulus condition. Results of the multiple linear regression indicated an overall significant effect between dolphin, stimulus condition, and source level on center frequency [F(7,9807) = 3617, p < 0.001; adjusted R2 = 0.721], duration [F(7,9807) = 415, p < 0.001; adjusted R2 = 0.228], bandwidth [F(7,9807)= 1989, p < 0.001; adjusted R2 = 0.586], and ICI [F(7,9687) = 845, p < 0.001; adjusted R2 = 0.379]. Controlling for dolphin and stimulus condition, center frequencies (b = 1777, p < 0.001), bandwidths (b = 3396, p < 0.001), and ICIs (b = 0.001, p < 0.001) increased with source level, and durations decreased with source level (b = −7.56e−7, p < 0.001). Controlling for stimulus condition and source level, Dolphin 1 produced lower center frequencies (b = −7297, p < 0.001), shorter durations (b = −4.614−6, p < 0.001), narrower bandwidths (b = −3793, p < 0.001), and longer ICIs (b = 0.023, p < 0.001) than Dolphin 2. Controlling for dolphin and source level, all parameters significantly (p < 0.001) differed among individual stimulus conditions. According to the model, larger IS angles resulted in higher emitted center frequencies for a given source level, and Dolphin 2 produced center frequency values higher than Dolphin 1, but this trend was not consistent across the range of source level values [Fig. 9(a)]. Larger speaker azimuths resulted in shorter emitted click durations and wider bandwidths than smaller speaker azimuths [Figs. 9(b) and 9(c)].
The results of comparing ICIs in 500 ms increments, as described above for Phase 1 (Fig. 6), are depicted in Fig. 10. Dolphin 1 demonstrated an increase in IPI as time in each trial progressed when the IS was positioned at 90°, and a decrease in IPI over time for azimuths of 45° and 0° (90° [F(1,1389) = 32.2, p < 0.001, R2 = 0.023; ]; 45°: [F(1,1135) = 135, p < 0.001, R2 = 0.107; ]; 0°: [F(1,1349)= 45.2, p < 0.001, R2 = 0.0324; ]). Dolphin 2 demonstrated a reduction in IPI over time for all azimuths (90° [F(1,1897) = 11.7, p < 0.001, R2 = 0.006; ]; 45°: [F(1,2035) = 39.2, p < 0.001, R2 = 0.019; ]; 0°: [F(1,1823) = 28.3, p < 0.001, R2 = 0.0153; ]). For dolphin 1, the variance in IPIs was greater at the start of the 90° [F(1,268) = 51.5665, p < 0.001] and 0° [F(1,260) = 12.7421, p = 0.0004] trials than in the middle/end of these trials, but there was no significant difference in IPI variance between the start and middle/end of trials with the speaker at 45° [F(1,252) = 0.2622, p = 0.6091]. Dolphin 2 demonstrated no significant difference in IPI variation between the beginning and middle/end of trials for all three angle conditions (p > 0.05).
The results from this experiment provide initial information on how odontocetes may adapt their echolocation signals when faced with different jamming scenarios. When the IS ICI decreased, both dolphins demonstrated a general trend of lowering the center frequencies of the echolocation signals, but different temporal strategies: Dolphin 1 decreased click ICI and Dolphin 2 increased click ICI. When the IS azimuth decreased, Dolphin 1 increased pulse duration and bandwidth and Dolphin 2 decreased bandwidth. Within trials, both dolphins lowered their ICI over trial duration and produced signals with greater ICI variance at the start of the trial compared to the middle/end of the trial.
The two dolphins demonstrated differences in overall general echolocation characteristics in both the baseline and experimental conditions. Dolphin 1 produced echolocation clicks that were lower in frequency, duration, and source level and longer in ICI than Dolphin 2 (Table I, Figs. 3, 4, 5). Dolphin 1 also demonstrated a longer response latency and number of stimulus echoes before producing a response to the behavioral task than Dolphin 2. These differences in overall behavior are likely the result of hearing loss in Dolphin 1, which has been documented to both reduce echolocation performance and result in a lowering of echolocation frequencies as the animal adapts its emitted echolocation signals to best match hearing sensitivity (Kloepper et al., 2010a,b; Li et al., 2013).
In addition to overall differences in echolocation behaviors, the dolphins demonstrated different strategies for adapting their signals when in the presence of the IS. For the temporal trials, as the IS increased in click repetition rate, or lowered in ICI, Dolphin 1 lowered the ICI and frequency of its echolocation clicks. Dolphin 2, on the other hand, increased the ICI and lowered the frequency of its echolocation clicks (Figs. 3 and 4). Assuming the task becomes more challenging for the dolphins as the IS increases in click repetition rate, each dolphin may adopt a behavioral response to overcome the challenge by producing optimal signals based on how the dolphin perceives different parts of the spectrum. Dolphin 1, for example, produced echolocation calls that were very similar in frequency to that of the IS (Figs. 1 and 3). Due to Dolphin 1's hearing loss, it would be unable to perceive echoes higher in frequency, so a response of shifting emitted call frequency away from that of the IS would not improve echo detection. Rather, Dolphin 1 instead utilized a response to increase the number of emitted clicks (and received echoes) as the task became more challenging. This behavioral response of clicking more often has been documented in bats when faced with acoustic interference (Amichai et al., 2015), which can improve the self-recognition of echoes when in an acoustically challenging environment. Dolphin 2, on the other hand, produced baseline echolocation signals that were 30–35 kHz higher in frequency than that of the IS. As the IS increased in click repetition rate and the task became more challenging, Dolphin 2 used a response of increasing peak frequency and lowering center frequency, which, due to the bimodal nature of odontocete clicks (Au et al., 1995; Houser et al., 1999) resulted in an overall increase in bandwidth (Fig. 4). Unlike Dolphin 1 that utilized a temporal response, Dolphin 2 instead utilized a spectral response, increasing ICI but increasing bandwidth, which may result in better spectral information for echo detection. This behavioral response is consistent with those used by many bats, who make spectral changes in echolocation signals to improve echo recognition when faced with jamming scenarios (Ulanovsky et al., 2004; Gillam et al., 2007; Bates et al., 2008).
With the IS changing in azimuth, the dolphins also demonstrated differences, albeit different strategies than those employed during the trials with changing the IS ICI. As the IS became more on-axis (moving towards 0°), Dolphin 1 increased the click bandwidth and duration and emitted source levels that were 7–8 dB higher than during the trials where the IS changed in ICI. Dolphin 1 also produced clicks that were approximately 20% longer in duration for the changing azimuth IS than for the changing ICI IS. These data indicate that, assuming the task becomes more difficult as the IS becomes more on-axis, Dolphin 1 switched to a behavioral response of producing longer echolocation clicks with greater bandwidth, which may help increase the spectral information available for echo detection in the presence of the IS. Furthermore, the increase in source level and signal duration can result in increasing the signal-to-noise (SNR) ratio, which further facilitates echo detection. Unlike Dolphin 1, Dolphin 2 decreased bandwidth and lowered frequency as the IS became more on-axis. Dolphin 2 produced clicks that were 20% shorter in duration, 10–15 kHz lower in center frequency, 7–8 dB higher in amplitude, and exhibited 10–15 ms longer ICIs for the changing azimuth IS compared to the trials with the changing ICI IS.
One unexplored possibility with our results is that the dolphins may be changing beamwidth or steering the beam, especially when the IS changes in azimuth. Dolphins produce highly directional echolocation signals (Au et al., 1986) and have plasticity in beam shape and direction (Moore et al., 2008). Because this study recorded echolocation signals only with a single receiving hydrophone, it cannot be ruled out that dolphins may be changing beam directionality and/or direction as a behavioral response to compensate for the IS, or to investigate the IS at off-axis azimuths. Changing the beam shape or direction would change the click parameters received by the hydrophone at 0° since the receiver would be in the off-axis portion of the transmit beam (Moore et al., 2008; Au et al., 2012). For example, the lowered frequencies produced by Dolphin 1 when the IS was at 90 and 45° could be an artifact of Dolphin 1 steering its beam to investigate the off-axis IS. If this were the case, the frequencies recorded by our hydrophone at 0° would not be reflective of the true echolocation sounds produced by the Dolphin. Therefore, further investigations into how dolphins may avoid acoustic jamming should utilize an array of receivers to investigate changes in click parameters in concert with changes in beam shape and/or direction.
Pooled analysis for both dolphins across both treatments highlight general echolocation trends (Figs. 5 and 9). Source level and center frequency were positively correlated, which is well documented among odontocetes and thought to be linked due to physiological constraints for sound production (Au et al., 1995; Au and Herzing, 2003; Kloepper et al., 2010a; Smith et al., 2016). Source level and bandwidth were also positively related, which has been documented in other odontocete species (Møhl et al., 2000), and is further supported by the link between bandwidth and center frequency (Au and Herzing, 2003). Source level and duration were inversely related, which is not well documented among odontocetes, but ICI was not influenced by source level.
For nearly all treatment conditions, clicks at the beginning of each trial exhibited greater variance in ICIs than clicks in the middle/end of the trial, which is consistent with each dolphin adjusting their ICIs to potentially time click production to avoid overlap with the IS (Figs. 6 and 10). Furthermore, both dolphins exhibited a general response of decreasing ICIs over the duration of the trial. Because the dolphins and phantom target were both stationary, this gradual reduction in ICI is unexpected and has not been documented in other stationary laboratory tasks (Turl and Penner, 1989; Finneran, 2013). This may be another behavioral response the dolphins use to increase echo recognition when faced with the IS: because the IS was produced at a constant, predictable ICI, varying the emitted ICI may help dolphins better discriminate self-echoes from the IS signals.
It is important to note that during this experiment, the dolphins maintained high levels of performance and demonstrated no significant difference in response latency or number of stimulus echoes before response for the different treatments. Therefore, the question remains: does the IS truly interfere with the dolphin's sonar. Bats, however, have still demonstrated high levels of performance when echolocating in jamming scenarios (Bates et al., 2008; Amichai et al., 2015; Jones et al., 2018), so using performance as a metric for jamming may not be sufficient. One potential approach for assessing jamming is to determine a jamming probability, or the likelihood that an echo would overlap with an interfering echolocation signal in time. This probability can be described as a function of the length of the jamming signal and the number of sources producing a jamming signal. Consider, for example, the experiment described here with a single source of jamming. Based on the IS duration of 121 μsec and the ICI ranges (Table I), the percent of total time in a trial the IS would be potentially causing acoustic jamming for a dolphin would be less than 1%. This low jamming probability highlights an important consideration for all tasks involving playback studies to investigate jamming adaptation; namely, does jamming truly occur if interfering signals do not overlap in time with signaler calls? For animals in natural group settings, jamming probabilities would increase as the number of jamming sources increase, such as with conspecifics in a megapod, so more extreme jamming scenarios may be more appropriate to investigate jamming avoidance strategies in both bats and dolphins.
We would like to thank the training staff and interns of the National Marine Mammal Foundation and the Navy Marine Mammal Program. We would also like to thank Rachel Simmons, Teri Wu, and Lara Curtis for dolphin training, and Kaitlin R. Van Alstyne for assistance with data collection. Special thanks to James Finneran for hardware and software development of the PEG system and helpful comments. We also thank Mark Xitco for logistical support. This project was supported by grants from the Office of Naval Research Young Investigator Award N000141612478 awarded to LNK and Defense Advanced Research Projects Agency 2017DARPA01 awarded to the U.S. Navy Marine Mammal Program at Space and Naval Warfare Systems Center Pacific.