Role of the temporal window in dolphin auditory brainstem response onset

Concerns regarding the effects of anthropogenic noise on marine mammals have led to increasing efforts to characterize marine mammal hearing and the effects of noise on their auditory systems. Because limited numbers of marine mammals are available for behavioral conditioning and psychophysical testing, marine mammal hearing is often characterized using measurements of auditory evoked potentials (AEPs). Most AEP measurements in marine mammals have focused on measuring the auditory brainstem response (ABR), a series of voltage deflections in the averaged electroencephalogram (EEG) within the first 6–8 ms after sound onset (Burkard and Don, 2007). Although the ABR is known to be an “onset” response—i.e., a sustained stimulus produces an ABR only at the onset (and potentially offset) of the stimulus (Hecox et al., 1976; Brinkmann and Scherg, 1979; Suzuki and Horiuchi, 1981; Burkard and Don, 2007)—few studies have investigated the manner in which specific features of the stimulus onset affect ABR morphology, peak amplitudes, and latencies.

Supin and Popov (2007) measured auditory steady-state responses (ASSRs, formed by repetitive, overlapping AEPs; Regan, 1982; Lins et al., 1995) in bottlenose dolphins (Tursiops truncatus) while manipulating the stimulus sound pressure level (SPL) and duration. Stimuli consisted of repetitive, 64-kHz tonebursts that were enveloped by a single cycle of a raised cosine function; therefore, changes in duration also affected the envelope risetime. The results showed that ASSR amplitude increased with decreasing toneburst risetime and duration (Supin and Popov, 2007); however, since changes in toneburst envelope also affect frequency bandwidth, the results were confounded by the stimulus bandwidth changing with toneburst envelope manipulations.

Two recent studies examined the effects of noiseburst onset properties on bottlenose dolphin ABRs (Finneran et al., 2018; Jones et al., 2019). Noisebursts were used, rather than tonebursts, to allow envelope properties to be manipulated without affecting the stimulus bandwidth. Both studies used spectrally “pink” noisebursts with 10- to 160-kHz bandwidths and 2-ms plateau durations. The main difference between the studies was in the stimulus onset envelopes: Finneran et al. (2018) gated the noisebursts with linear rise/fall envelopes, while Jones et al. (2019) utilized raised-cosine rise/fall envelopes. The results of both studies mirrored earlier measurements from cat auditory nerve and cortex neurons (Heil, 1997b,a; Heil and Irvine, 1997) and chinchilla nearfield responses from the auditory nerve and inferior colliculus (Phillips et al., 2001). For linear-onset stimulus envelopes, ABR peak latencies were functions of the rate-of-change of the envelope sound pressure rather than the risetime or plateau sound pressure alone (Finneran et al., 2018). For cosine-onset envelopes, latencies were best described by the maximum value of the second derivative of the pressure envelope function (Jones et al., 2019). Heil (1997b) reported similar behavior in first spike latencies in cortical neurons and suggested that the shape of the envelope at stimulus onset determines the peak latency, i.e., latencies were best-described using the parameter that resulted in the stimulus envelopes from the various conditions being most similar at stimulus onset.

Both prior dolphin studies found that ABR peak amplitudes were better described by the envelope sound pressure at the end of a fixed, 260-μs time interval rather than the envelope plateau pressure, risetime, maximum rate-of-change of the envelope pressure, or (for cosine envelopes only) the maximum value of the second derivative of the envelope pressure (Finneran et al., 2018; Jones et al., 2019). The 260-μs value was selected for three reasons: (1) Different patterns in ABR peak amplitudes were observed by Finneran et al. (2018) depending on whether the envelope risetime was greater/less than ∼250 μs, (2) bottlenose dolphin performance during biosonar echo detection tasks featuring multi-pulse echoes can be modeled by an energy detector with an integration time of 264 μs (Au et al., 1988), and (3) a duration of 260 μs is close to estimates of the temporal resolution and the time period within which interaction of acoustic signals occurs, referred to as the “critical interval” in dolphin hearing (Vel'min and Dubrovskii, 1975, 1976; Moore et al., 1984). The critical interval was first broadly established from backward masking and pulse-pair discrimination experiments as having a duration of 200–300 μs (Vel'min and Dubrovskii, 1975). Specific numeric estimates for the critical interval from a pulse-pair discrimination experiment ranged from 230 ± 40 μs to 260 ± 25 μs, depending on whether the interval was defined by 75% correct discrimination or chance performance, respectively (Vel'min and Dubrovskii, 1976; Dubrovskiy, 1990). The critical interval was also independently estimated by Moore et al. (1984). This study utilized a backward masking paradigm where a bottlenose dolphin used its biosonar to detect the presence of a physical object while noiseburst maskers were presented at various time delays relative to echo reception. The results showed monotonically decreasing performance as the echo-masker delay decreased from 500 μs down to 100 μs with a 70% correct threshold of 265 μs (Moore et al., 1984). Together, these studies show that temporal resolution of the bottlenose dolphin auditory system can be modeled using a “temporal window” with an effective duration of ∼260 μs (Moore et al., 1988).

The findings of Finneran et al. (2018) and Jones et al. (2019) indicate that stimulus onset properties within a short temporal window are critical for determining farfield ABR peak amplitudes and latencies. However, neither dolphin study systematically varied the stimulus duration within the temporal window, and certain stimulus parameters may have potentially confounded data interpretation (Finneran et al., 2018; Jones et al., 2019). For example, both studies utilized a stimulus plateau duration of 2 ms so the effects of shorter-duration stimuli (i.e., with durations less than the temporal window) could not be investigated. Since the 2-ms plateau duration was shorter than the duration of the bottlenose dolphin ABR, ABRs to stimulus offset, if present, would have overlapped with the later waves of the onset ABR and possibly affected the measured peak amplitudes and latencies (e.g., see Burkard et al., 2020).

The present study further examines the relationships between stimulus onset properties and ABR peak amplitudes and latencies in dolphins. Methods were designed to eliminate potential confounds and test predictions from the earlier studies (Finneran et al., 2018; Jones et al., 2019) and explicitly study the effects of short-duration stimuli (durations< 260 μs). Two experiments were conducted using noiseburst stimuli with linear-onset envelopes. In the first [see Fig. 1(a)], the rate-of-change of the stimulus envelope pressure was held constant and plateau duration was fixed at 8 ms (longer than the ABR duration), while risetime and plateau pressure were systematically varied. Based on earlier studies, ABR peak latencies were hypothesized to remain constant and ABR peak amplitudes to increase until the risetime exceeded the temporal window duration, after which the ABR peak amplitude was expected to remain constant. In the second experiment [see Fig. 1(b)], stimulus plateau pressure and risetime were fixed while plateau duration (and, thus, total duration) varied. ABR peak amplitudes were hypothesized to increase as stimulus duration increased within the temporal window (i.e., to increase with energy within the temporal window, based on Au et al., 1988), and remain constant afterward.

Six bottlenose dolphins participated in the experiments (Table I). The upper-frequency limit (UFL) of hearing for each dolphin was defined as the frequency at which underwater ASSR thresholds (see ANSI, 2018) reached 120 dB re 1 μPa (Strahan et al., 2020). Based on these data, three dolphins were classified as having “normal hearing” (NH) with UFL ≥ 120 kHz, and three were considered “hearing impaired” (HI) with UFL ≤ 100 kHz.

Tests were conducted within a 9 m × 9 m floating, netted enclosure at the U.S. Navy Marine Mammal Program facility in San Diego Bay, CA. During each trial, the dolphin positioned itself on an underwater “biteplate” attached to an extruded aluminum frame at a depth of 1 m. A piezoelectric, underwater sound projector (ITC 5446, International Transducer Corp., Santa Barbara, CA) was suspended 1 m in front of the biteplate. Background ambient noise at the test site was dominated by contributions from snapping shrimp, other dolphins, and passing vessels and aircraft. The median ambient noise pressure spectral density levels were ∼67 dB re 1 μPa²/Hz at 20 kHz and decreased linearly with the logarithm of the frequency to ∼55 dB re 1 μPa²/Hz at 150 kHz.

Noiseburst stimuli were created by multiplying continuous Gaussian pink noise by an envelope function, P(t), consisting of a linear rise, constant amplitude plateau, and linear fall. Continuous noise was digitally synthesized using a reverse–fast Fourier transform (FFT) technique and then converted to analog at a rate of 500 kHz and 16-bit resolution (PXIe-6368, National Instruments, Austin, TX). Analog noise was then digitized (500 kHz, 16 bit) by a PXI-7852R (National Instruments) containing a Virtex-5 LX50 FPGA (Xilinx, San Jose, CA), which multiplied the noise by the envelope function and converted the noisebursts to analog (500 kHz, 16 bit). Analog noisebursts were filtered (0.2–200 kHz, eight-pole Butterworth, 3 C module, Krohn-Hite Corporation, Brockton, MA), attenuated (PA5, Tucker-Davis Technologies, Alachua, FL), amplified (CC4000, Crest Audio, Meridian, MS), and applied to the underwater sound projector. Noise was digitally compensated for the sound projector frequency response and multipath effects to achieve spectrally pink conditions from 20 to 160 kHz (see Au and Floyd, 1979; Finneran et al., 2018). Noisebursts were generated and presented at a rate of ∼20 s⁻¹.

During experiment 1 (see Table II), noiseburst risetime varied from 32 to 528 μs. Plateau SPL varied with noiseburst risetime from 102 to 126 dB re 1 μPa in order to maintain a constant pressure envelope rate-of-change (dP/dt) of 4 kPa/s. Plateau duration was constant at 8 ms. For experiment 2 (see Table III), risetime was fixed at 32 μs and plateau SPL was fixed at 115 dB re 1 μPa, resulting in a pressure envelope rate-of-change of ∼18 kPa/s. Plateau duration varied from 0 to 724 μs. Figure 2 shows representative examples of acoustic noiseburst envelopes and 1/3-octave pressure spectrum levels measured during calibrations. From 20 to 160 kHz, noiseburst 1/3-octave SPLs were within ±5 dB from 32 to 160 kHz, and SPLs were within ±2 dB (i.e., most deviations were in the lower frequency bands, where frequency compensation of the noise was confounded by ambient noise). For a 32-μs risetime, some overshoot of the plateau pressure occurred after the initial envelope rise. The overshoot was on the order of 20%, which amounts to ∼1.5 dB.

It is, at times, more meaningful to characterize acoustic transients in terms of energy rather than sound pressure or intensity (Urick, 1983; Au, 1993). Sound exposure (E) is defined as

where T is the integration time and p(t) is the instantaneous sound pressure (ANSI, 1994). For a plane wave (a reasonable assumption for the present study), E is proportional to sound energy per unit area (Urick, 1983). For the noiseburst signal, the average value of E is obtained by replacing the instantaneous pressure p(t) in Eq. (2) with the pressure envelope function P(t), i.e., $E=∫P2(t) dt$⁠. For the linear-onset envelopes used here, regardless of the integration time, E is proportional to the product of the envelope plateau mean square pressure, $PP2$⁠, and a time quantity related to the integration time, risetime, and/or plateau time. The acoustic level metric for E is sound exposure level (SEL) with units of dB re 1 μPa²s. For a linear-onset envelope, regardless of the integration time, SEL is equal to the sum of the plateau SPL and the logarithm of a time quantity related to integration time, risetime, and/or plateau time.

ABRs were measured using two surface electrodes embedded in suction cups and placed on the dolphin's head and dorsal surface: a non-inverting electrode located on the midline, approximately 5 cm posterior to the blowhole, and an inverting electrode near the right external auditory meatus. A third, common electrode was located in the seawater near the dolphin. A biopotential amplifier (ICP511, Grass Technologies, West Warwick, RI) filtered (0.3–3 kHz) and amplified (94 dB) the electrode signals. The differential voltage between the inverting and non-inverting electrodes, representing the instantaneous EEG, was digitized at 100 kHz with 16-bit resolution using the PXIe-6368 (National Instruments). During each measurement, noisebursts were presented and EEG data streamed to computer disk for 26 s (512 stimulus presentations). ABRs were obtained by first digitally bandpass filtering the EEG data from each measurement from 0.3 to 3 kHz using a zero-phase implementation of a sixth-order Butterworth filter and then synchronously averaging 10-ms epochs of EEG data temporally aligned with each noiseburst. Four measurements were conducted for each combination of subject and noiseburst envelope (see Tables II and III).

The four ABRs for each condition were coherently averaged to produce a single averaged ABR (based on a total of 2048 epochs) for each condition, as well as two “subaverages,” each comprising half the available epochs (with the epochs arranged sequentially into the two subaverages). The two subaverages were subtracted to create the “±average,” which provides an assessment of the residual background noise in the averaged EEG (Schimmel, 1967). The signal-to-noise ratio (SNR) for each averaged ABR was estimated as the ratio of the root-mean-square (rms) values of the ABR and the residual background noise within a time window from 1 to 7 ms relative to the stimulus onset.

Amplitudes and latencies for the ABR peaks P1, N2, P4, and N5 (see Popov and Supin, 1990) were based on the local minimum or maximum within time intervals near visually identified peaks in the averaged ABRs. P1 was defined as the positive peak closest to the P1-N2 zero crossing. N2 was defined as the minimum amplitude between P1 and P3. To account for acoustic/electronic delays during stimulus generation, measured latencies were corrected by subtracting the time of arrival measured at the listening position for an acoustic click projected from the same apparatus (680 μs).

To control for the variance in ABR amplitudes and latencies between individual dolphins, values for each subject were normalized before averaging. Peak-to-peak (p-p) amplitudes were normalized by first fitting each dolphin's ABR amplitude versus stimulus risetime (experiment 1) or total duration (experiment 2) data with the exponential growth function,

where t is the risetime (experiment 1) or duration (experiment 2), and the fitting parameters A and τ represent the final value and time constant, respectively (Originlab, 2019). Amplitudes for each dolphin were normalized by dividing by the best-fit value of A for that dolphin and then averaging. Peak latencies for each dolphin were normalized by subtracting the mean latency for that dolphin. Normalized latencies were then averaged and fit using linear or nonlinear regression (Originlab, 2019).

Figure 3 shows example ABR waveforms measured during experiment 1 for one HI dolphin (OLY) and one NH dolphin (WHP). The ABR morphology matched the typical pattern for transient ABRs seen in dolphins with six prominent positive and negative waves between ∼1 and 6 ms after stimulus onset (e.g., Ridgway et al., 1981). Data quality was good with all signal-to-noise-ratio-levels (SNRLs) ≥ 6 dB. ABRs for the HI animals were generally smaller than those for the NH animals with the exception of COL who, despite poor high-frequency hearing, had good sensitivity below ∼60 kHz.

Figure 4(a) shows p-p amplitudes for P1-N2 (upper panel) and P4-N5 (lower panel) as functions of noiseburst risetime. ABR amplitudes for all dolphins followed the same trend of increasing with risetime up to ∼100–250 μs and then remaining constant; however, there was high variability across subjects. Fits of Eq. (2) to the individual data were generally good [P1-N2, mean adjusted R² = 0.813, standard deviation (SD) = 0.151; P4-N5, mean adjusted R² = 0.926, SD = 0.0455]. Individual time constants (τ) ranged from 35 to 128 μs for P1-N2 and from 36 to 83 μs for P4-N5. There were no obvious trends in values of τ between NH and HI animals, and differences between mean values of τ for NH and HI animals were non-significant for both P1-N2 [t(4) = 1.41, p = 0.233] and P4-N5 [t(4) = −0.127, p = 0.905]. Normalized data for both the NH and HI groups were, therefore, averaged together. Normalized p-p amplitudes are shown in Fig. 4(b). Fits of Eq. (2) to the mean, normalized amplitudes were good (adjusted R² = 0.991 and 0.989 for P1-N2 and P4-N5, respectively) with best-fit values (±half the 95% confidence interval) for τ equal to 64 ± 6.0 μs and 55 ± 5.5 μs for P1-N2 and P4-N5, respectively.

Figure 5(a) shows peak latencies for P1 (upper panel) and N5 (lower panel) as functions of noiseburst risetime. Individual latencies showed similar patterns with small increases (<100 μs) over the ∼500 μs range of risetimes. For P1, HI animals tended to have longer latencies; however, this pattern did not hold for N5, where OLY often had the shortest latencies. Individual latencies were normalized by subtracting the mean latency for that individual and ABR peak. Normalized latencies are shown in Fig. 5(b). Linear-log fits to the mean, normalized latencies were fair with R² = 0.632 and 0.395 for P1 and N5, respectively. Best-fit slopes were small with 12 μs/doubling of risetime (∼2 μs/dB change in the plateau SPL, p = 0.00596) for P1 and 9.4 μs/doubling of risetime (∼1.6 μs/dB change in the plateau SPL, p = 0.0517) for N5.

Figure 6 shows example ABR waveforms measured during experiment 2 for one HI dolphin (SPA) and one NH dolphin (SHA). ABRs were similar to those in experiment 1 but with larger amplitude resulting from the consistently short risetime and relatively high plateau pressure. The minimum SNRL was 12 dB. As in experiment 1, for the same stimulus conditions, ABRs for the HI animals were generally smaller than for the NH animals.

Figure 7(a) shows ABR p-p amplitudes for each individual dolphin for experiment 2 as a function of the noiseburst total duration. For durations < ∼200–250 μs, ABR amplitudes increased with noiseburst duration. Unlike experiment 1, where a clear plateau tended to exist for longer risetimes, ABR amplitudes decreased as duration increased above a few hundred microseconds. Fits of Eq. (2) to the individual data were fair (P1-N2, mean adjusted R² = 0.699, SD = 0.209; P4-N5, mean adjusted R² = 0.725, SD = 0.0427), likely because of the lack of a stable plateau at long durations. As in experiment 1, there were no obvious trends in the time constant τ between NH and HI animals, and differences between mean values were non-significant for both P1-N2 [t(4) = 0.853, p = 0.442] and P4-N5 [t(4) = 0.0669, p = 0.950] and, thus, normalized data for both the NH and HI groups were averaged together.

Figure 7(a) shows that for constant plateau pressure and at the shorter total durations, a linear-log relationship exists between ABR amplitude and stimulus duration. This suggests that the ABR amplitude may be linearly related to the stimulus SEL. The normalized ABR p-p amplitudes are, therefore, plotted versus the SEL in Fig. 7(b). Mean, normalized ABR amplitudes increased linearly with SEL up to noiseburst durations of 246 μs: linear fits to the data over this region were good with r² = 0.985 and 0.995 for P1-N2 and P4-N5, respectively. For durations > 246 μs, normalized ABR p-p amplitudes decreased with increasing SEL. For P4-N5, a local minima existed for a duration of 576 μs, beyond which amplitudes increased again. To evaluate if the decreases in p-p amplitudes shown in Fig. 7 reflect both underlying peak amplitudes, ABR zero-peak amplitudes for P1, N2, P4, and N5 are shown in Fig. 8. This shows decreases in ABR peak amplitude with durations >246 μs for N2, P4, and N5 but not P1, i.e., the decrease in P1-N2 for durations > 264 μs in Fig. 7 resulted from a decrease in the N2 peak amplitude not the P1 peak amplitude.

Figure 9(a) shows peak latencies for P1 (upper panel) and N5 (lower panel) measured for the individual dolphins during experiment 2. Individual latencies changed by relatively small amounts (<60 μs) over the ∼750 μs range of stimulus durations. As in experiment 1, individual latencies were normalized by subtracting the mean latency for that individual and waveform. Normalized latencies are shown in Fig. 9(b) as functions of the SEL. Linear fits to the mean normalized latencies were good with r² = 0.872 and 0.721 for P1 and N5. Best-fit values for the slopes were significantly different from zero (p < 0.001 and 0.01 for P1 and N5, respectively), although the slopes were small: 3.8 μs/dB (for both P1 and N5). It is noteworthy that the trend of increasing latency with an increasing SEL is opposite the typical trend of a decreasing ABR peak latency with an increasing stimulus SPL in numerous mammalian species (including dolphins).

Temporal resolution of the auditory system can be modeled by smoothing the time-domain acoustic signal by the temporal window function before subsequent processing (Moore et al., 1988). In this sense, the temporal window acts as a time-domain analog to the critical band (i.e., peripheral auditory filter bandwidth; Penner et al., 1972) within which auditory events “merge.” Detection of multiple biosonar echoes falling within the temporal window can be modeled by an energy detector (Au et al., 1988), and echoes falling within the temporal window interact to create spectral cues that may be utilized for discrimination; these cues appear to be absent if echoes fall outside the temporal window (Dubrovskiy, 1990; Dubrovsky et al., 1992; Branstetter et al., 2020), and it has been suggested that dolphins can only form an integrated biosonar “image” of an object if the echo highlights fall within the temporal window (Vel'min and Dubrovskii, 1975).

Estimates of the dolphin temporal window duration vary depending on the particular study and the method used to define the window, but values typically range from 200 to 300 μs (Vel'min and Dubrovskii, 1975, 1976; Moore et al., 1984; Au et al., 1988). For example, Vel'min and Dubrovskii (1976) reported chance performance during a passive hearing, pulse-pair discrimination task when the pulse separation was 260 ± 25 μs, whereas Moore et al. (1984) reported that a 265-μs time separation between signal and backward masker resulted in a 70% correct performance during a biosonar go/no-go detection task.

The results of experiment 1 show, for a fixed envelope pressure rate-of-change, ABR amplitude increases exponentially with risetime with a time constant (τ) of 55–64 μs. In this context, the time constant represents the risetime at which the ABR amplitude reaches 63% of its final, maximal value; at a risetime of 5τ (275–320 μs), the ABR amplitude reaches 99% of the final value. The ABR data from the present study thus match dolphin temporal window estimates (Vel'min and Dubrovskii, 1976; Moore et al., 1984; Au et al., 1988): a temporal window duration of 260–265 μs would be equivalent to 4.1τ–4.8τ from the present study, at which point mean ABR amplitudes reach 96%–98% of their final values.

The results of the present study are consistent with previous data showing that noiseburst envelope properties within the first ∼260 μs dictate ABR amplitude and latency (Finneran et al., 2018; Jones et al., 2019). For linear envelopes, ABR peak latencies can be described by the envelope pressure rate-of-change dP/dt (Finneran et al., 2018). Data from the present study generally fit this pattern: dP/dt was constant for each experiment, and the resulting latencies showed little change with increasing risetime (experiment 1, Fig. 5) or noiseburst duration (experiment 2, Fig. 9).

It is not clear why small increases in mean latency occurred with increasing noiseburst risetime (experiment 1) or duration (experiment 2). Stimulus off-time decreased slightly (by ∼1 ms) as stimulus risetime increased from 32 to 528 μs (experiment 1) or plateau duration increased from 0 to 724 μs (experiment 2). Decreased off-times can lead to increased adaptation and larger peak latencies; however, adaptation normally causes ABR peak amplitudes to decrease, not increase, as in the present study (e.g., Burkard et al., 2017). Although the broadband noise stimulus used here stimulated nearly all of the cochlear partition responsible for ABR generation (Finneran et al., 2016), the plateau SPL increases in experiment 1 could have resulted in an apical spread of basilar membrane activation into the 10–20-kHz region (below the noiseburst high-pass cutoff). This could explain both the longer-latency and larger-amplitude ABRs as contributions from lower-frequency (more apical) cochlear regions would be activated slightly later that those at higher frequency cochlear regions, and could lead to an increase in ABR p-p amplitude. However, the manipulations in experiment 2 did not feature a change in stimulus level that would result in an increased spread of activation into apical regions, suggesting another process instead of or in addition to the apical spread of excitation. This could happen due to single units firing at longer latencies (in a probabilistic fashion) as stimuli are on for a longer time period or the firing of additional units with higher thresholds of activation, i.e., SEL thresholds that are not met until stimuli achieve a longer duration (Heil and Neubauer, 2001).

The previous study with linear rise envelopes (Finneran et al., 2018) showed that ABR amplitude was better predicted by the envelope SPL at the end of a 260-μs window compared to the plateau SPL or risetime alone. For a constant dP/dt, ABR amplitude would, therefore, be expected to increase with risetime up to 260 μs and then remain constant. This was, in fact, observed during experiment 1 (Fig. 4). For a constant plateau SPL and risetime < 260 μs, ABR amplitudes would be expected to remain constant for durations > 260 μs. The present data (experiment 2, Fig. 7) differ somewhat: when the duration, exceeded 260 μs, ABR amplitudes decreased with increasing duration, with P4-N5 reaching a local minimum near 500 μs. The local minimum for P4-N5 near 500 μs—approximately half the dominant period (i.e., the interpeak interval) in the ABR—suggests that the decrease in p-p amplitude may have resulted from interference between the (onset) ABR and a second ABR originating at the beginning of the stimulus offset. This interpretation is supported by recent observations of offset ABRs in dolphins in response to broadband noisebursts with 32-μs cosine fall times (Burkard et al., 2020). Furthermore, offset ABRs to broadband noise appear to either have a reduced or absent P1 or be inverted in polarity (e.g., see Brinkmann and Scherg, 1979; Burkard et al., 2020). Either of these conditions would explain (1) the local minimum observed in the amplitude of P4-N5—but not P1-N2—at a stimulus duration of 576 μs (Fig. 7) and (2) the decline in amplitudes of the ABR peaks N2, P4, P5—but not P1—as noiseburst total duration exceeds ∼250 μs (Fig. 8).

Unlike previous studies, the present study (experiment 2) examined ABRs in response to stimuli with durations within the ∼260-μs temporal window. Here, ABR amplitude linearly increased with the logarithm of duration, suggesting a relationship between ABR amplitude and a metric such as stimulus SEL. When plotted versus SEL, mean ABR amplitudes increased linearly within the temporal window. The SEL computed over the temporal window duration was also found to be a good predictor of the ABR amplitude measured in experiment 1, as well as previous dolphin ABR data obtained with linear and cosine-enveloped noisebursts (Fig. 10; Finneran et al., 2018; Jones et al., 2019). However, fits to the SEL [i.e., linear-log fits to $∫P2(t) dt$] were not necessarily better than fits to $∫P(t) dt$ or $∫P3(t) dt$ computed over the temporal window (Fig. 10). The latter expressions arise from the work of Heil and colleagues, who examined the ability of the integral of P(t) to predict first spike timing and the integral of P³(t) to predict neural firing rate and psychophysical auditory thresholds (Heil et al., 2011; Heil et al., 2013; Heil and Matysiak, 2020). The ability of all of these expressions to more or less equally predict ABR amplitudes suggests that the critical element is the integration of some feature of the stimulus envelope over the temporal window. Regardless of the exponent to which the pressure envelope function is raised, the integral over time can explain the exponential growth observed in the ABR data (Figs. 4 and 7).

The transient ABRs have been used to study auditory system processing and measure hearing thresholds in both marine and terrestrial mammals (including humans). These studies have typically used toneburst stimuli with the toneburst duration often specified as a fixed number of cycles regardless of center frequency. In this fashion, the ratio of bandwidth to center frequency remains constant when the toneburst frequency changes. In these situations, toneburst duration will decrease with increasing frequency, thus, stimulus duration may be less than the temporal window duration in some situations, e.g., at 40 kHz, with a period of 25 μs, a toneburst with <10 cycles would have a duration less than the ∼250-μs temporal window duration in bottlenose dolphins.

In the present study, the ABR amplitudes in bottlenose dolphins decreased as the noiseburst duration decreased below ∼250 μs. Although not measured, it seems likely that thresholds to noisebursts with durations below ∼250 μs would be elevated in dolphins compared to those measured with longer duration noisebursts. Measurements similar to those of the present study but obtained using toneburst stimuli would be necessary to determine the manner in which ABR amplitudes and thresholds change with toneburst duration. These data would be necessary to understand if decreases in ABR amplitudes with increasing frequency actually reflect a change in subject hearing sensitivity as opposed to effects of stimulus duration.

Dolphin ABR amplitudes to broadband noisebursts are dictated by the stimulus envelope within the auditory temporal window, estimated to have a duration of ∼250 μs. ABR p-p amplitudes can be predicted by the SEL calculated over the temporal window aligned with the stimulus onset; however, there is no clear evidence that acoustic energy actually governs the ABR amplitude. In fact, the time integral (over the temporal window) of a quantity related to the stimulus pressure envelope appears to be the most critical parameter for predicting the ABR amplitude.

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 staff, training staff, and interns at the Navy Marine Mammal Program. 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 all applicable U.S. Department of Defense guidelines. Financial support was provided by the U.S. Navy Living Marine Resources (LMR) Program.