Effects of echo phase on bottlenose dolphin jittered-echo detection

Toothed whales and microchiropteran bats actively sense their environments utilizing sophisticated biosonars (Madsen and Surlykke, 2013). The biosonar operating principle presumably involves comparing an emitted sound pulse to returning echoes, with large-scale echo delays indicating target range and fine scale delay differences revealing spatial properties of the target. An echolocating animal must therefore be able to resolve small differences in echo delay to perceive and discriminate fine-scale spatial features—i.e., to have high-resolution in the biosonar “image.”

Two different types of experiments have been used to measure echo-delay resolution in bats and dolphins. In the first, animals are trained to discriminate physical objects or electronic (“phantom”) echoes that differ slightly in range or echo delay. These studies have generally revealed echo-delay resolution on the order of tens to hundreds of microseconds (see Murchison, 1980; Simmons and Grinnell, 1988; Masters and Jacobs, 1989). In the second type of experiment, animals discriminate between electronic echoes with fixed delay and those with delay that alternates, or “jitters,” from one echo to the next (Simmons, 1979; Menne , 1989; Moss and Schnitzler, 1989; Simmons , 1990; Moss and Simmons, 1993; Masters , 1997; Simmons , 1998; Simmons , 2003; Simmons , 2004; Finneran , 2019; Finneran , 2020). The jittered-echo paradigm results in much smaller estimates of echo-delay resolution; jittered-echo delay resolutions have been reported as ∼1.3 μs in bottlenose dolphins (Tursiops truncatus, Finneran , 2019; Finneran , 2020) and ≤0.4 μs in big brown bats (Eptesicus fuscus, Simmons, 1979; Menne , 1989; Moss and Schnitzler, 1989; Simmons , 1990; Moss and Simmons, 1993; Simmons , 1998; Simmons , 2003; Simmons , 2004). Smaller echo-delay resolutions are found in the jittered paradigm, at least in part, because the jittered-echo paradigm is less influenced by animal head movements (Simmons , 1990). The jitter paradigm has also been controversial because of the very low echo delay resolutions reported for the big brown bat [10 ns (Simmons , 1990; Simmons , 2004)] and the apparent sensitivity of both bats and dolphins to changes in echo phase (Simmons, 1979; Menne , 1989; Simmons , 1990; Moss and Simmons, 1993; Finneran , 2019; Finneran , 2020), despite the likely absence of neural phase locking at the high frequencies present in echolocation signals [see Köppl (1997)].

The present paper is focused on the bottlenose dolphin's sensitivity to echo fine structure by examining the effects of echo phase on dolphins' ability to discriminate jittering from non-jittering echoes. The experimental approach utilized a phantom echo change detection paradigm similar to previous studies examining jittered-echo delay resolution in dolphins (Finneran , 2019; Finneran , 2020). Hardware/software enhancements in the present study significantly reduced the inherent jitter in phantom echo generation (compared to the previous system) and allowed unconstrained manipulations of echo phase shifts. Test conditions were similar to those utilized by Menne (1989), where echo delay and phase were independently manipulated. The goal was to determine the effects of manipulating echo phase on jittered-echo delay resolution and test the trading relationship between echo delay and echo phase; i.e., would certain phase and delay combinations that result in similar echo fine structure lead to reduced discrimination performance?

Tests were conducted within a 9 m × 9 m floating, netted enclosure at the US Navy Marine Mammal Program facility in San Diego Bay, California. Three male bottlenose dolphins participated in the task: COM (age 5–7 years during the period of testing), IND (18–19 years), and SPO (4–6 years). Upper-frequency hearing limits were approximately 130 kHz for COM, 64 kHz for IND, and 140 kHz for SPO, defined as the frequency at which electrophysiological auditory steady-state response thresholds reached a sound pressure level (SPL) of 120 dB re 1 μPa [measured September 2021, as in Strahan (2020)]. These hearing limits indicate full hearing bandwidth for COM and SPO and substantial high-frequency loss for IND. Both COM and SPO had extensive experience with the jitter paradigm (Finneran , 2020). IND had no prior experience with the jitter paradigm, but had previously participated in a phantom echo change detection task where he was required to detect a change in the time separation of a two-highlight echo, analogous to a physical cylinder wall thickness discrimination task (Branstetter , 2020).

During each trial, the dolphin positioned itself on an underwater “biteplate” supported by an extruded aluminum frame, constructed so the areas around the dolphin's melon and lower jaw were unobstructed (Fig. 1). The biteplate was oriented so the dolphin faced San Diego Bay through an enclosure gate opening containing a clean, netted panel. Two piezoelectric transducers (TC4013, Reson Inc., Slangerup, Denmark) were positioned ∼0.75 m in front of the biteplate: one functioned as the click receiver and the other as the echo projector. The nearest land mass in the direction of the dolphin's biosonar transmissions was ∼1.1 km distant. Mean water depth was ∼7 m. Background ambient noise at the test site resulted from snapping shrimp, other dolphins, and occasional contributions from passing vessels and aircraft. From 20–140 kHz, median ambient noise pressure spectral density levels decreased from 63 to 50 dB re 1 μPa²/Hz; the mean pressure spectral density level over this frequency range was 55 dB re 1 μPa²/Hz.

The biosonar task was an echo change detection task, identical to Finneran (2019) and Finneran (2020). The dolphins' task was to produce echolocation clicks, listen to returning phantom echoes, and produce a conditioned acoustic response (IND, SPO: whistle, COM: burst pulse) if the echoes changed from “non-jittering” (echo A) to “jittering” (alternating between echoes B₁ and B₂). For non-jittering echoes, echo delay was fixed at 13.2 ms (∼10 m simulated target range). For jittering echoes, echo delay and/or echo phase varied on alternate echoes. Jitter delay was defined as the difference between the echo delays for B₁ and B₂.

Table I shows the combinations of echo delay and phase for echoes B₁ and B₂ in each of the four experiments (see Fig. 2). Experiment 0 was the baseline condition and featured jittered echo delay only (no phase shift). In experiment 1, B₁ and B₂ were 180° out-of-phase (i.e., a polarity jitter). Experiment 2 featured various combinations of jittered delay and phase. During experiment 2a, jittering echoes had +45°/−45° phase shifts, with B₁ having longer echo delay and +45° phase shift. Phase shifts were reversed in experiment 2b (i.e., B₁ had –45° phase shift). In experiment 2c, B₁ had phase change only and B₂ had delay change only. Finally, in experiment 3, the phase shift of each echo (A, B₁, and B₂) was randomized on an echo-to-echo basis. For each experiment, at least 38 trials were conducted for each value of jitter delay for each dolphin (see Table I), resulting in a total of 450 to 1300 trials for each dolphin in each experiment. Echo amplitudes were ∼15 dB above echo-detection thresholds. For SPO and COM echo sound exposure levels were approximately –70 dB relative to the received click. This resulted in an echo SNR of roughly 30–40 dB for click p-p SPLs from 215 to 225 dB re 1 μPa. This value should be interpreted cautiously since the ambient noise was neither spectrally flat nor Gaussian. For IND, relative echo levels had to be higher ( –35 dB) because of his high-frequency hearing loss and the transmitting response of the TC4013 echo projector (most echo energy was above his audible range).

Experimental sessions consisted of 60 to 110 trials, divided into 10-trial blocks with constant jitter delay within each block. Jitter delay magnitude was reduced after each block until the final block, which served as a “cooldown” and had jitter delay matching the first block (cooldown data were not analyzed). Approximately 70% of the trials were echo change trials, where the echoes changed from echo A to B₁/B₂ after 3 to 8 s. The echo change was followed by a 1.5-s response interval. The remaining trials were control trials, where echo A was presented for the entire 4.5- to 9.5-s trial. Responses within the response interval following an echo change were classified as hits. Responses during a control trial were false alarms. Withholding a response resulted in a correct rejection during a control trial or a miss during an echo change trial. Responses during an echo change trial, but before the echo-change, resulted in the trial being re-classified as a control trial and the response scored as a false alarm. For COM and SPO, 90% of correct responses during experiments 0–2 were rewarded with one fish. This was done to familiarize the dolphins with not receiving a reward on a small fraction of trials to allow later testing with probe trials (to be reported separately). For IND, and COM and SPO during experiment 3, all correct responses were rewarded. The dolphins were recalled with no fish reward after incorrect responses.

Biosonar echoes were created using a phantom echo generator (PEG) implemented in software using a multifunction data acquisition board featuring a field-programmable gate array (FPGA). Clicks emitted by the dolphin were captured by the click receiver, amplified and filtered (5–200 kHz bandwidth: VP-1000, Reson Inc., and 3C module, Krohn-Hite Corporation, Brockton, MA), then input to either an NI PXIe-7852R device (National Instruments, Austin, TX) containing a Virtex-5 LX50 FPGA (experiments 0–2) or an NI PXIe-7856R containing a Kintex-7 160 T FPGA (experiment 3). Analog hydrophone signals were digitized with 16-bit resolution at either ∼741 kHz (1.35-μs interval, PXIe-7852R) or 1 MHz (PXIe-7856R). If the digitized hydrophone signal exceeded an amplitude threshold, the click waveform was extracted and convolved with a target impulse response to create the echo waveform. The echo waveform was then scaled in amplitude, delayed, and converted to analog (PXIe-7852R, 741 kHz, 16 bit or PXIe-7856R, 1 MHz, 16 bit). The analog echo waveform was then filtered (5–200 kHz, 3C module), amplified (M7600, Krohn-Hite Corp.), and used to drive the echo projector. The click and echo waveforms were also digitized at 2 MHz and 16-bit resolution by an NI PXIe-6368 multifunction data acquisition device and stored for later analysis. Operation of the PEG was verified by replacing the dolphin click signal input to the PEG with an on-axis dolphin click recording and broadcasting the analog echo waveforms using the echo projector. The resulting acoustic echoes were measured each day using the click receiver hydrophone and periodically measured using a separate TC4013 hydrophone positioned at the “listening position”—the midpoint between the dolphin's lower jaw when on the biteplate (without the dolphin present).

Dolphin performance was quantified using the hit rate (number of hits divided by the number of echo-change trials) and false alarm (FA) rate (number of false alarms divided by the number of control trials). To obtain a measure of sensitivity independent of response bias, d′ was calculated (Stanislaw and Todorov, 1999). During calculation of d′, response rates of 0 were replaced with 0.5/N and rates of 1 were replaced with 1–0.5/N, where N is the number of echo-change or control trials (Stanislaw and Todorov, 1999).

Dolphin clicks were quantified by extracting individual clicks from the digitized click receiver hydrophone signal and computing the inter-click interval (ICI), peak-to-peak (p-p) sound pressure level (SPL), energy spectrum, center frequency, and (centralized) rms bandwidth (Menne and Hackbarth, 1986; Au, 1993; Simmons , 2004). Clicks within the first 3 s of each trial (less than the minimum time for the echo change) were not analyzed. Mean values of the click waveform and energy spectrum, and distributions for the remaining quantities, were computed for each dolphin/experiment. Representative echoes for each dolphin were obtained by replacing the dolphin click input to the PEG with the mean click waveform for that dolphin, then broadcasting the resulting analog echo and coherently averaging the acoustic echo measured at the listening position. To assess the degree of similarity between the jittering echo waveforms, the cross correlation function (XCF) between echoes B₁ and B₂ was computed. For IND, representative echoes were digitally low-pass filtered at 64 kHz (zero-phase, 8th order Butterworth) before correlation to simulate the effects of his high-frequency hearing loss.

Figures 4 and 5 show the hit rate, FA rate, sensitivity (d′), and jittering-echo XCF for each dolphin during experiments 0 and 1, respectively. For experiment 0 (jitter delay only), performance of all three dolphins tended to decrease as jitter delay decreased towards zero. Thresholds for COM and SPO based on d′ = 1 were 0.7 μs, lower than previously reported [1.2 and 1.5 μs for COM and SPO, respectively, based on 50% correct (Finneran , 2020)]. In addition to the performance drop near zero jitter delay, performance decreased for both COM and SPO at 8 μs, where the fine structure of the two jittering echoes (B₁ and B₂) had significant overlap, but envelopes were shifted (see inset graph in Fig. 4). Performance also declined slightly for SPO at 16 μs. Performance for IND was much worse: he had higher false alarm rates and a jitter delay threshold of 4.1 μs. For all dolphins, performance roughly followed the XCF between B₁ and B₂: where correlation was high, performance was poor.

Results for experiment 1 matched previous data for COM and SPO (Finneran , 2020). Both dolphins were nearly 100% correct with no jitter delay but polarity jitter only (identical envelopes but different fine structure for B₁ and B₂), and performance was poor at 5 μs jitter delay (different envelopes but similar fine structure); i.e., high correlation between echoes B₁ and B₂ resulted in low sensitivity.

Figure 6 shows dolphin performance with echo phase shifts of (a) B₁ = +45° and B₂ = –45° and (b) B₁ = –45°, B₂ = +45°. As with the polarity flip condition (Fig. 5), with no jitter delay, echo waveforms were visually distinct (i.e., low correlation) and performance was nearly perfect. However, performance dropped at 2.5-μs jitter delay when B₁ = +45° and B₂ = –45° [Fig. 6(a)], and at 7.5 μs when B₁ = –45° and B₂ = +45° [Fig. 6(b)]. For both conditions, jittering echo waveforms were highly correlated, and the degree of correlation matched the drop in performance; i.e., higher correlation and lower d′ at 2.5 μs in Fig. 6(a) compared to that at 7.5 μs in Fig. 6(b). Figure 6(c) shows effects of a phase shift only in B₁ (phase = +90°) and a delay change only in B₂ (delay = –20 to 20 μs). Both dolphins showed high performance with B₂ delay of 0 but drops in performance when B₂ delay = −7.5 and 2.5 μs. Both conditions corresponded with high correlation between echoes B₁ and B₂; these conditions represent combinations of delay and phase change where their effects cancelled.

When all echoes were given random echo-to-echo phase shifts, performance of all dolphins dramatically decreased (Fig. 7): jitter delay thresholds increased to 50 and 48 μs for COM and SPO, respectively, and 140 μs for IND.

Click waveforms were typical of those for bottlenose dolphins (Fig. 8, insets). Click spectra (Figs. 8 and 9) were similar across experiments except for IND during experiment 0, where click energy was shifted towards lower frequencies. Maximum estimated click source levels were high (p-p SPLs >220 dB re 1 μPa at 1 m) and there was a tendency for click SPL and center frequency to increase with experiment, though it is not clear if this was related to task difficulty or an order effect. Electronic echoes from the PEG mirrored the click waveform; however, acoustic echoes were colored by the echo projector transmitting frequency response. As a result, acoustic echoes contained more cycles and relatively more high-frequency content compared to the clicks. Click ICI distributions for COM and SPO were much broader for experiments 1–3 compared to experiment 0. All three dolphins exhibited variable click patterns, similar to those seen previously in COM and SPO (Finneran , 2020). Temporal gaps in click emissions, defined as ICI > 150 ms, were relatively common for COM, but less so for the other dolphins: mean ± SD temporal gaps per trial were 0.91 ± 1.2, 0.25 ± 0.52, and 0.14 ± 0.37 for COM, IND, and SPO, respectively. It was also common to see click rates slow down, then speed up during trials, as evidenced by local maxima in the ICI during a trial. The mean ± SD number of ICI local maxima per trial was 4.9 ± 3.1, 1.2 ± 2.1, and 5.1 ± 3.3 for COM, IND, and SPO, respectively.

Results of the present study show clear sensitivity to echo fine structure in the jitter detection task. This sensitivity is revealed in experiments 0–2 as a decrease in dolphin performance when the fine structure of the jittering echoes was similar, even when echo envelopes were different [e.g., Figs. 4 and 6(c)], high performance when envelopes were identical but fine structure differed (Figs. 5 and 6) and drops in performance when the effects of echo delay and phase shift cancelled each other [Fig. 6(c)]. Disruption of echo fine structure through phase randomization (Fig. 7) dramatically increased jitter detection thresholds, from 0.7 to 4 μs in experiment 0 up to ∼50–140 μs in experiment 3. The similarity in fine structure between the jittering echoes can be quantified by their cross correlation (after low-pass filtering the echoes for IND). When the correlation between echoes was high (≳0.8), the dolphins had lower values for sensitivity (d′) and committed more errors. Sensitivity to echo fine structure in the present study is consistent with the performance of a hypothetical coherent receiver; however, a coherent receiver is not necessary to obtain the present results, only that the auditory system is sensitive to echo fine structure.

For experiments 0 and 1, results for COM and SPO were consistent with previous studies in bats showing sub-microsecond jitter delay thresholds [≤0.4 μs (Menne , 1989; Moss and Schnitzler, 1989; Simmons , 1990)] and detection of 180° phase jitter without accompanying time jitter (Menne , 1989; Simmons , 1990; Moss and Simmons, 1993). Detection thresholds for COM and SPO in experiment 0 were, however, lower than previously measured (0.7 vs 1.2–1.5 μs), and the decrease in sensitivity at 8 μs was not previously observed (Finneran , 2020). The reasons for these differences are not clear but may be related to the greater experience of the dolphins with the jitter paradigm, lower inherent jitter in the PEG during the current experiment, and/or improvements in the hardware apparatus (no physical obstructions near the lower jaw). Performance for IND was significantly worse than for COM and SPO, likely as a result of his high-frequency hearing loss. Low-pass filtering the received echoes at his upper-frequency cutoff resulted in broader peaks in the cross correlation between jittering echoes, meaning greater jitter delays were required to significantly reduce the correlation between the jittering echoes. It is also noteworthy that the XCF for the filtered jittering echoes had a secondary peak near 20 μs, which may explain the drop in sensitivity for IND as jitter delay approached 20 μs. The decrease in jitter-delay acuity with removal of high-frequency echo content found in IND matches previous results in bats, where low-pass filtering jittering echoes reduced performance (Simmons , 2004).

One of the goals of the present study was to examine the trading relationship between echo phase and delay; specifically, whether the effects resulting from changes to echo delay and echo phase could cancel. Although Menne (1989) reported that bats could discriminate non-jittering echoes from those that jittered by ±45° or 0°–180°, they did not find combinations of echo delay/phase jitter where their effects canceled. In contrast, the present study did reveal combinations of delay/phase jitter where the effects canceled; e.g., in Fig. 2(c) a 90° phase shift in B₁ and a –7.5 or 2.5 μs delay of B₂ resulted in significant overlap in the fine structure of B₁ and B₂ and drops in dolphin performance. These data remain consistent with a coherent receiver and support the hypothesis (Simmons , 1990) that the lack of a comparable performance drop in the data from Menne (1989) was a result of the specific jitter delays that were tested; i.e., the tested jitter delays were not close enough to the specific delay required to cancel the phase shift.

Applying random phase shifts to all echoes dramatically affected performance for all dolphins (Fig. 7). The resulting thresholds for COM and SPO (50 and 48 μs) matched dolphin physical object range discrimination thresholds measured by Murchison (1980) at 1, 3, and 7 m and then extrapolated to 10 m. This fits an interpretation that echo fine structure is not available to dolphins performing a range comparison task in the way that it is during a jittered-echo discrimination task. Worse performance for IND was, as observed in experiments 0–2, likely a result of his high-frequency hearing loss and potentially a longer effective echo duration.

Reports of low jitter detection thresholds and sensitivity to echo phase have resulted in controversy over the interpretation of data arising from the jitter paradigm. Several alternate explanations for the jitter results have been proposed, including unintended cues arising from differences across hardware echo generation channels and spectral cues arising from temporal overlap between jittering echoes and non-jittering environmental reflections of the transmitted biosonar pulse [e.g., Pollak (1993) and Beedholm and Mohl (1998)].

In the present study, the use of a single channel PEG with digital implementation of echo delay removed the potential for cues related to the differences across hardware channels or impedance mismatches as echo delay changed: both jittering and non-jittering echoes were produced from the exact same hardware, with differences in echo delay and phase arising in software. However, the jitter detection threshold of 0.7 μs for the normal-hearing dolphins was similar to thresholds previously measured in a simulated cylinder wall-thickness discrimination experiment [∼0.4–1 μs when echo highlights were within the temporal window (Branstetter , 2020)]. This raises the question of whether the dolphins in the present study were utilizing changing spectral cues arising between the overlap of the jittering echoes from the PEG with (non-jittering) environmental echoes from the dolphin's click.

To reduce the potential for overlap of jittering and environmental echoes, mean echo delay was set at ∼13 ms (∼10 m range) in the present study, well beyond the distances associated with the apparatus and transducers. At the time of PEG echo reception, reverberation levels from the emitted click were <−15 dB relative to received echo levels for COM and SPO (<−50 dB for IND) and well below ambient noise levels. It is unknown whether overlap of environmental reverberation at these levels with the jittering echoes could provide a sufficiently large and stable spectral cue to explain the dolphins' high performance at certain combinations of echo phase/delay (but not all combinations). The pronounced errors at non-zero delay values (i.e., Figs. 5 and 6) are, likewise, not wholly consistent with the dolphins responding to a spectral change due to overlap of jittering echoes with static environmental echoes. Even with the strong correlation of echo fine structure at these delay values, a shift in delay would result in differences in the interference patterns of the spectra for B₁ vs B₂ and a static environmental echo [i.e., peaks and notches (Au and Pawloski, 1992; Dubrovsky , 1992; Accomando , 2020; Branstetter , 2020; Accomando , 2022)]. Dolphins have been shown to be very sensitive to such spectral changes in complex echoes (Au and Pawloski, 1992; Branstetter , 2020), and it seems less likely that the dolphins would ignore potential spectral cues at localized non-zero delay values but not at nearby delay values where performance is near perfect. It is more likely that the drop in performance is due to high correlation between B₁ and B₂.

As an additional test as to whether environmental echoes provided cues to the presence of jittering echoes, we conducted a limited number of trials with COM where Gaussian, spectrally white broadband masking noise was electronically added to the PEG echoes before transmission. The resulting acoustic masking noise had spectral content similar to the echoes (both dictated by the transducer transmitting voltage response). The masker level was set by first increasing the noise until COM's thresholds for jitter delay detection (with no phase jitter) began to be affected. The noise was then lowered by 6 dB, and the PEG output and noise level were both increased by 30 dB; the intent was that the noise would mask environmental echoes but not significantly mask echoes from the PEG. With these settings, data were collected using the conditions from experiment 2a. Results (Fig. 10) showed a similar pattern to the original measurements (albeit with higher FA rate likely caused by masking noise): high sensitivity with ±45° phase jitter only and a drop in performance when jitter delay was 2.5 μs. The measurements with masking noise therefore suggest that the dolphins were not utilizing the overlap of environmental and jittering echoes as a cue to the presence of jittering echoes.

Therefore, although it is difficult to prove the dolphins in this study were not utilizing the overlap of environmental and jittering echoes, given the relatively low reverberation levels at the time of echo reception, the reduction of performance at non-zero delay values, performance in the presence of masking noise, and similarity to results of multiple bat studies, the most parsimonious explanation is that the dolphin auditory system is sensitive to the echo fine structure when detecting echo jitter. This does not prove sensitivity to the phase of steady-state or longer duration signals, the operation of a coherent receiver, or the use of fine structure in other tasks (e.g., range comparison), but only that the auditory system can encode aspects of echo fine structure in a way that is perceptually available in some echolocation tasks.

Dolphins' echo delay resolution is <1 μs when assessed with the jittered-echo paradigm and dolphins are capable of detecting changes in echo phase. Performance in the jitter task can be predicted by the cross correlation function between jittering echoes: when jittering echoes are highly correlated, animals commit more errors. Performance in the jitter task is thus consistent with that of a coherent receiver; however, a coherent receiver is not necessary to obtain the present results, only that the auditory system is sensitive to echo fine structure.

The data that support the findings of this study are available from the corresponding author upon reasonable request.

The authors thank H. Bateman, J. Bennett, R. Breitenstein, L. Crafton, R. Dear, C. Espinoza, M. Graves, D. Ram, K. Stuhldreher, M. Wilson, T. Wu, and the animal care staff, training staff, and interns at the Navy Marine Mammal Program. K. Christman, K. Donohoe, E. McGarvey, and N. Packard assisted with data collection. Financial support was provided by the Office of Naval Research Code 32 (Mine Countermeasures, Acoustics Phenomenology & Modeling). The authors have no conflicts of interest to disclose. 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.