Forward masking enhances the auditory brainstem response in the free-tailed bat, Tadarida brasiliensis, during a critical time window for sonar reception

Forward masking is an auditory phenomenon in which the response to a sound transiently alters the physiological response to subsequent sounds (Zwislocki et al., 1959; Abbas and Gorga, 1981; Carlyon, 1988; Relkin and Turner, 1988). The main effect is a brief reduction in auditory sensitivity lasting ten to hundreds of milliseconds, but under certain conditions forward masking can also produce a response enhancement (Henry, 1991a,b). Forward masking effects are problematic for biosonar because outgoing pulse emissions may interfere with the ability to detect and discriminate faint echoes returning soon afterwards. Echolocating bats have adaptations for attenuating the responses to loud outgoing pulse emissions (Suga and Schlegel, 1972; Jen and Suga, 1976), however, it has also been hypothesized that the suppressive effects of temporal masking could benefit biosonar by providing an automatic gain control mechanism that stabilizes response amplitudes within an optimal dynamic range (Supin et al., 2008, 2009). Despite the potential significance of forward masking effects on biosonar performance, surprisingly little is known about how forward masking effects influence auditory response properties or sonar performance in echolocating bats.

Auditory brainstem responses (ABRs) are acoustically evoked electrical responses that can be recorded with intracranial, subdermal, or superficial electrodes. The response consists of a stereotyped waveform generated by synchronous neural activity in successive early stages of auditory processing, and has proven to be a useful tool for comparing hearing sensitivities across animals. ABRs have been measured in several bat species (Wenstrup, 1984; Burkard and Moss, 1994; Obrist and Wenstrup, 1998; Boku et al., 2015; Simmons et al., 2015), but only one investigated forward masking effects in ABRs, finding that a masker could produce either suppression or enhancement depending on timing and level of the masker relative to the probe (Grinnell, 1963b). More recently, ABRs have been used to measure the time course of masking effects in echolocating cetaceans. Beluga and false killer whales displayed only suppression effects (Supin et al., 2007; Supin and Popov, 2015), but in dolphins it was discovered that the suppression could be quickly replaced by a pronounced enhancement window under certain circumstances (Finneran et al., 2016). These apparent differences among cetaceans may be due to experimental methods or potentially reflect species-specific physiological adaptations for sonar use in different contexts.

Here, we used ABRs to measure temporal masking effects in the free-tailed bat (Tadarida brasiliensis). To discriminate the overlapping second response from the continuing first response which may last 10–15 ms, we used a double-pulse stimulation protocol similar to the methods used in whales (Supin et al., 2007; Supin and Popov, 2015) in which averaged ABR responses to a leading masker pulse were digitally subtracted from responses to masker-probe stimulus pairs varying in inter-stimulus interval, thereby revealing in greater detail the response to the trailing probe pulse in each pair. We document an initial suppression phase that recovered within 4 ms and was replaced by an enhancement phase extending 4–15 ms beyond the first pulse, which corresponds to a critical time period for echolocating bats during which increased sensitivity could be very beneficial. The results indicate that bats have an adaptation for enhancing sonar receiver gain during a critical timeframe when faint echoes return from nearby small objects.

These experiments used seven wild-caught adult Mexican free-tailed bats (Tadarida brasiliensis, four females, three males) housed socially in an artificial bat vivarium in which they could fly and had access to food and water ad libitum. Husbandry and experimental procedures were approved by the Texas A&M institutional animal care and use committee and complied with all National Institute of Health guidelines for the care and use of animals.

ABRs were recorded intracranially from the surface of the midbrain inferior colliculus. Animals received a prophylactic dose of buprenorphine analgesia before being anesthetized with isoflurane (5% induction, 2.5% maintenance) and positioned in a stereotaxic apparatus for surgery and throughout the recording session. A small (4 mm²) region of scalp over the midbrain was reflected laterally and a 0.5 mm diameter craniotomy was drilled directly above the inferior colliculus. A low impedance (<100 kΩ) silver wire was lowered by micromanipulator to just touching the surface of the brain. A silver reference electrode was inserted beneath the skin behind one ear and a stainless-steel subdermal needle electrode was inserted between the scapulae as ground. At the end of the experiment the hole was sealed with dental cement and the tissue folded back in place and closed with veterinary adhesive (Vetbond), and animals recovered in isolation before eventually being returned to social housing. Bioelectrical signals were recorded using the TDT (Tucker Davis Technologies, Alachua, FL) Auditory Workstation, which included an RZ6 Multichannel I/O processer, RA4PA Medusa preamplifier, and RA4LI low-impedance headstage. Incoming signals were digitized at 25 kHz and bandpass filtered between 300 and 3000 Hz. Each ABR was the averaged response from 256 stimuli delivered at 17 per second.

Acoustic stimuli were generated digitally using the TDT BioSigRZ software suite at 200 kHz sample rate. For audiograms we employed 5 ms tone pips (cosine window, 1 ms rise/fall time) ranging in frequency from 10 to 80 kHz. Stimulus levels were controlled by a combination of digital and analog programmable attenuators on the TDT RZ6. The stimuli were delivered through an analog amplifier (STR-DE598, Sony, Tokyo, Japan) to a planar ribbon tweeter (PTMini-6, Dayton Audio, Springboro OH) positioned 10 cm directly ahead of the bat and equidistant from each ear. Stimulus amplitudes were calibrated with a Brüel & Kjær type 4939 1/4 in. microphone before each experiment. Sound pressure levels (SPLs) were compensated for the frequency response of the tweeter to be flat ±3 dB from 10 to 80 kHz at levels up to 85 dB and are presented relative to the peak equivalent root mean square of pure tones at 20 μPa (dB SPL re 20 μPa). To establish thresholds at each frequency the sound level was reduced in 5 dB steps until a response was no longer observable and the prior stimulus level was defined as threshold. P2/3 amplitude, usually the largest peak in response to 5 ms tone pips, had to be at least four standard deviations beyond noise within the predicted latency time window and be repeatable to be accepted as an evoked response.

Forward masking was investigated using 2 ms downward frequency-modulated (FM) sweeps (50–20 kHz) mimicking the principle harmonic of the free-tailed bat sonar pulse (Schwartz et al., 2007). To measure forward masking effects, we replicated the “double-click” protocol of Supin et al. (2007). ABR waveforms typically extend 7–15 ms beyond the first stimulus, obscuring measurements of the response to a second stimuli arriving within this time frame: the double-click protocol utilizes digital subtraction of the averaged response to the first stimulus from the averaged response to pairs of stimuli to isolate the response to the second stimulus across a range of inter-stimulus intervals. Figure 1 provides a representative example of how the response to the second stimulus was isolated when the responses overlapped in time. To account for progressive changes in response amplitudes over the duration of an experiment, the averaged responses to a single stimulus were re-recorded in between each double-pulse stimulus presentation and served as the template subtracted from the immediately succeeding double-pulse stimulus. In these experiments the inter-stimulus interval varied from 1.0 to 15 ms, and each interval pair was tested across a range of amplitudes spanning from 45 dB SPL (limited by ABR threshold at start of FM sweep) up to 85 dB SPL, corresponding to an ecologically meaningful range of echo amplitudes based on the typical pulse emission amplitudes (95–115 db SPL when recorded in the lab) for this species (Tressler and Smotherman, 2009).

The variables measured were the latencies and peak-to-following trough amplitudes of waves labeled P1, P2/3, and P4 [Fig. 2(a)] following the nomenclature of previous studies in bats (Burkard and Moss, 1994; Boku et al., 2015; Simmons et al., 2015). Peaks 2 and 3 were only discriminable at higher intensities, which is why they were combined as P2/3. A fifth, more variable peak (P5) was also common at high stimulus levels but was not analyzed.

ABR waveforms [Fig. 1(a)] varied with the acoustic parameters of the stimulus, but generally conformed to previous descriptions of ABRs in bats and other mammals. The audiogram derived from ABR thresholds using 5 ms tone pips [Fig. 2(b)] reached its lowest value of 15 dB SPL at 20–25 kHz. Below 20 kHz the threshold increased by approximately 3 dB/kHz down to the lower limit of the tweeter, and above 25 kHz the threshold increased by approximately 1.3 dB/kHz to the upper limit of the tweeter. The audiogram closely matched a previous estimate compiled from single neuron thresholds in the free-tailed bat inferior colliculus (Pollak et al., 1978), but ABR audiograms overestimate behavioral thresholds by about 20 dB (Obrist and Wenstrup, 1998).

Forward masking results derived from pairs of FM sweeps are summarized in Fig. 2(c). Initially, a similar degree of suppression was evident in the amplitudes of all peaks of the second response (RII) relative to the first response (RI), but relative amplitudes recovered to near one within approximately 4 ms. At 6 or 8 ms all three peaks of RII exceeded their mean corresponding levels in RI, but P4 was the most elevated. P1 reached its highest relative amplitude (1.3 ± 0.1) at 8 ms, and P2/3 reaches its highest relative amplitude (1.5 ± 0.3) at 6 ms IPIs, but both returned to near RI amplitudes within 10–15 ms. P4 amplitudes increased significantly more than P1 and P2/3, doubling relative to RI (2.0 ± 0.2) at 8 ms, and remained elevated relative to RI at 15 ms. A one-way repeated measures analysis of variance comparing the three data sets in Fig. 1(c) confirmed that the three peaks changed differently across intervals (F_(2,7) = 4.6, P = 0.03); however, while P1 was not significantly different from P2/3 (P = 0.55), P4 was significantly different from both P1 and P2/3 (P = 0.01). Changing stimulus amplitude did not significantly alter the results within the range tested; RII was enhanced to a similar degree at all stimulus amplitudes.

Figure 2(d) illustrates representative data on the effect of stimulus amplitude on the latency of the successive ABR peaks. Amplitude-latency relationships are known to vary between species and stimulus parameters, and may therefore reflect subtle yet important differences in auditory processing characteristics. In responding to tone pips over a 60-dB range, all three peaks displayed statistically similar slopes of approximately −25 μs/dB (P1, −23.9 ± 4.0 μs/dB; P2/3, −24.9 ± 5.6 μs/dB; P4, −25.2 ± 8.1 μs/dB: F_(2,6) = 0.68, P = 0.93). However, over the first 20–30 dB above threshold slopes were higher, ranging from −40 to −100 μs/dB (mean −57.2 ± 7.5 μs/dB) and became much shallower at higher intensities, approaching ∼−10 μs/dB at 55–85 dB SPL. FM sweeps were tested over a range of 30–70 dB above threshold and their amplitude-latency functions displayed slopes in the range of −6 to 9 μs/dB.

The primary aim of this study was to quantify forward-masking effects in free-tailed bats. The issue is important for understanding how the vertebrate auditory system may be adapted to improve sonar processing of faint echoes returning rapidly from nearby objects. First, it was confirmed that the ABR recovers from forward masking's suppressive effects in a remarkably short time (<4 ms). Suppression is primarily ascribed to neural refractory periods and the recovery typically takes tens to hundreds of milliseconds in other mammals (Henry, 1991b,a). This fast recovery rate in bats supports the idea that echolocating mammals possess physiological adaptations to minimize suppression duration, although the precise mechanism is unknown. Bats possess active mechanisms to dampen perception of the outgoing pulse (Suga and Schlegel, 1972; Jen and Suga, 1976), but this alone does not appear sufficient to explain the rapid recovery from suppression.

The recovery transitioned into an amplification of the entire ABR response amplitude from 4 to 15 ms after the first stimulus, with the greatest effect seen in P4. The time-course and amplitude of this enhancement was consistent with Grinnell (1963b), who likewise found that P4 recovered more quickly and sometimes became “supranormal” by as much as 100%. However, Grinnell did not observe enhancement in the other peaks, which may be a species difference but also possibly attributable to the differences in equipment and methodology. The observation that P1 and P2/3 were modestly enhanced is important because P1 presumably reflects the compound action potential of the VIIIth nerve, and thus any enhancement would be endogenous to the cochlea. However, P4 arises from within the brainstem auditory pathway, so the extra enhancement seen in P4 may indicate a central origin for this amplification. Thus, since we observed a small coincident increase in P1 and P2/3 and an additional distinct increase in P4, bats may possess both central and peripheral mechanisms for enhancing sonar receiver gain shortly after pulse emissions. Any peripheral mechanism of response enhancement seems likely to be at least indirectly affiliated with the active biomechanical amplification processes that exist in the mammalian cochlea (Henry, 1991b; Ashmore et al., 2010), although these mechanisms may be modified to support echolocation in bats. Accounting for the extra response enhancement seen in P4 is more difficult. One possibility is that pairs of pulses selectively recruit an additional pool of neurons to the ABR waveform not otherwise excited by a single acoustic stimulus. Echolocating bats possess a specialized population of neurons in their ascending auditory system known as delay-tuned neurons that are selectively responsive to acoustic stimuli mimicking naturalistic pulse-echo sequences, and such neurons exist in the auditory midbrain areas that contribute to the P4 waveform (Dear and Suga, 1995; Park et al., 1996; Yan and Suga, 1996; Portfors and Wenstrup, 1999). Hypothesizing that the recruitment of delay-tuned neurons in the midbrain contributes to the enhanced ABR amplitudes also provides an explanation for why the enhancement can be observed during passive listening.

Last, we examined pulse-amplitude trading relationships in the free-tailed bat because these have been used previously to infer mechanical tuning properties of the bat cochlea (Boku et al., 2015) which as described above may contribute to species-specific differences in forward masking effects (Henry, 1991b). Our measurements of the amplitude-latency trading effects in free-tailed bats were in accordance with similar measurements in other FM bat species, which range from −7 to −9 μs/dB in the Japanese house bat (Boku et al., 2015) to −14 to −18 μs/dB in big and little brown bats (Grinnell, 1963a; Burkard and Moss, 1994). The values reported here are slightly higher than reported for other species, and notably single neuronal recordings from the free-tailed bat IC also reported a broad range of values (−10 to −180 μs/dB, mean of −47 μs/dB) (Pollak, 1988). It has been hypothesized that amplitude-latency trade-offs may be frequency dependent, by which the steeper slopes reported here are consistent with Tadarida brasiliensis using lower frequency search phase sonar pulses (≈20 kHz) than the other FM bats studied. Amplitude-latency relationships derived from sonar clicks in dolphins also fall in a similarly short range of −5 to −12 μs/dB (Finneran et al., 2017). In this case, using FM sweeps within a behaviorally relevant range of amplitudes produced results similar to other studies in FM bats.

This research was supported by the National Science Foundation (Grant No. IOS 1354381) and U.S. Office of Naval Research (Grant No. ONR N00014-17-1-2736). K.B. received support from the Texas A&M Institute for Neuroscience.