Two adult female big brown bats (J and T, ages unknown) participated in these experiments. They were captured from local barns, as authorized by a State of RI scientific collection permit. To conserve local bat populations, the state permit limits severely the numbers of animals that can be captured in any given year and thus restricts numbers on-hand in the laboratory for research purposes. Bats were socially-housed in the laboratory in a wire frame enclosure (6′ × 8′ × 8′) within a larger colony room. They were vaccinated for rabies and individually identified by readable microchips inserted under the skin of their backs (Trovan ID-100A RFID transponder, Trovan LID-573 microchip reader). They had unlimited access to fresh water and were provided daily with sufficient vitamin-enriched live mealworms (Tenebrio larvae) to keep their body weights within a healthy range of 15–20 g. Bats were allowed to fly unfettered in a large flight room for weekly exercise; both of them echolocated and flew without difficulty. Both bats also emitted communication and echolocation sounds in their home enclosure (monitored by Wildlife Acoustics recording devices). The colony room was maintained at temperatures of 20–24 °C and 55–65% relative humidity. All experimental and husbandry procedures were approved by the Brown University Institutional Animal Care and Use Committee and adhere to US federal guidelines.

As the first step in our procedure, the two bats were trained to sit without excessive movements for periods up to 30 min in a 50-mm deep ceramic dish, with their torsos within the dish and their heads resting on the edge. The dish was placed on an elevated steel platform in a single-walled sound-attenuating, electrically-shielded recording booth (Industrial Acoustics Co, N. Aurora IL). Training took place over several weeks. At the beginning of training, a bat was placed inside the ceramic dish; this ceramic dish is familiar to the bats, being the same as the food dishes used in their social housing. In the colony room, bats often rest or sleep in their food dishes after eating their daily food allotment. If the bat began crawling out of the dish on the platform, a soft broadband “shh” sound was made by the trainer to indicate to the bat that it had made an error, then the bat was gently placed back into the dish. If the bat remained inside the bowl for 20–30 consecutive sec, it was rewarded with a mealworm, which it ate inside the dish. This was repeated multiple times, with rewards given every 30 s, until the bat stayed in the dish for a period of 2 consecutive min. At this time point, a single electrode covered in conductive paste was placed on the bat’s back. If the bat tolerated the electrode placement without excessive movement, it was given another mealworm reward immediately after. If the bats reacted by leaving the dish, a “shh” error signal was again made by the trainer, and the electrode was reattached once the bat was back in the dish and had settled down. If the bat tolerated attachment of the first electrode, a second electrode was placed on its back, and then the third electrode placed along the midline of its scalp. A mealworm reward was provided after application of each of the first two electrodes, but not the third, as we wished to avoid excessive head movement. Once all three electrodes were applied to the bat, ultrasonic FM stimuli were emitted towards the bat at full amplitude and at the same repetition rate as used in data collection. Four individual bats began training in this manner, but two were removed after showing no behavioral progress after several days of training. Training was conducted 2–3 times a week for 2–3 weeks, until the two remaining bats could reliably tolerate having three electrodes applied with conductive paste while sitting motionless in the dish, while exposed to repeating auditory stimuli at full amplitude, for at least 5 consecutive min. The two bats used in this experiment learned to sit motionless in the dish for up to 30 min. Once the bats reached the 5 min criterion, the hair on their heads and their lower back was trimmed down with safety scissors, and diluted depilatory (Nair™; Church & Dwight, Ewing NJ) was applied for 2 min to remove remaining hair. Because recording electrodes will not adhere to hair and because bats’ hair grew back within a few days after depilation, hair removal was repeated as necessary during the time frame of these experiments. Bats tolerated the hair removal procedure well, without any signs of skin irritation or permanent loss of hair.

Acoustic stimuli were generated as digital.wav files using Adobe Audition v. 12.1 (Adobe Inc, San Jose CA) at a sampling rate of 500 kHz. They consisted of FM-up and FM-downsweeps, containing one or two harmonics, at durations of 3, 2, 1, 0.7, 0.5, 0.3, 0.2, and 0.1 ms (Fig. 1). FM sweeps with a single harmonic (FM-1H) covered the frequency range of 20–100 kHz; FM sweeps with two harmonics (FM-2H) covered the frequency range of 20–50 kHz in the first harmonic and 40–100 kHz in the second harmonic. All stimuli had raised cosine envelopes with 50% rising and falling shapes, matching the envelope shape of echolocation pulses and of the signals used by Luo et al. (2019). The raised cosine shape means that signal rise-time varies along with signal duration (see Fig. 2 in Luo et al. 2019). The natural FM echolocation sounds of big brown bats are two harmonic FM-downsweeps varying in duration from about 15 ms to about 0.6 ms over a pursuit/capture sequence, depending on the surrounding environment (Surlykke and Moss 2000). Stimulus durations chosen for this experiment fall within the shorter ends of this biological range and again mimicked those used for invasive recordings (Luo et al. 2019).

Spectrograms of FM-2H logarithmic down- and up-sweeping FM sweeps recorded at the midline of the bat’s head between the two ears, after passing through the TDT tweeter. Sweeps varied in duration as shown along the upper x-axis. FM-1H stimuli (not shown) did not include the second harmonic

Example ABR waveforms recorded with pEEG electrodes (left column) and tEEG electrodes (right column). Stimuli are FM-upsweeps at 0.5 ms duration, with harmonic structure (FM-1H or FM-2H) and stimulus levels (90 dB peSPL, TDT tweeter; 98 dB peSPL, Kenwood tweeter) as indicated on the plots. pEEG data are from Bat J (top to bottom) 08-15-19, 08-15-19, and 06-24-29. tEEG data are from (top to bottom) Bat J (01-24-20), Bat J (01-28-20), and Bat T (03-14-19). The asterisk marks the second positive peak (latencies of around 4–5 ms) used for quantification of response amplitude for statistical testing (see Methods). Response amplitudes were not corrected for the different gains of pEEG and tEEG recordings

The rationale for presenting short FM sweeps varying in sweep direction and duration derives from models of basilar membrane mechanics in other mammals (Robles and Ruggero 2001). FM-downsweeps, such as those used by the bat for echolocation, contain high frequencies followed by low frequencies. The traveling wave’s direction of propagation from high frequencies at the base of the cochlea to low frequencies at the apex causes low frequencies to be delivered to their receptors slightly later than those of high frequencies. The consequence is a delayed activation of eighth nerve fibers tuned to low frequencies compared to those tuned to high frequencies. FM-upsweeps contain low frequencies followed by high frequencies. In this case, low frequencies travel to their maximal place of excitation towards the apex before high frequencies arrive at their maximal place of excitation towards the base. Here, the delayed activation of low frequency responses is counteracted. FM sweeps thus produce an asymmetric pattern of excitation, visualized as the time-place appearance of different frequencies according to the direction of the sweep (up vs down) and whether the sweep adds to the time-delay of frequencies at successive places (for downsweeps) or counteracts the time delays (for upsweeps). ABRs reflect synchronous neural activity in the ascending auditory pathway. When eighth nerve responses are dispersed in time, as in an FM-downsweep, response synchrony is weaker, leading to a lower amplitude ABR, than if all frequencies arrived at their tuned location simultaneously. At appropriate durations of an FM-upsweep, neural responses at all frequencies can be brought into alignment and summate to produce a stronger synchronous response (Dau et al. 2000). The optimal duration of an FM-upsweep that counteracts frequency dispersion along the basilar membrane and produces this stronger response is predicted to match the velocity of the travelling wave (Elberling et al. 2007). In bottlenose dolphins, the optimal FM-upsweep duration derived from surface ABR recordings lies within the range of 0.45–1.1 ms, depending on stimulus level (Finneran et al. 2017); in big brown bats, the optimal FM-upsweep duration derived from invasive recordings lies within the range of 0.5–1 ms (Luo et al. 2019).

Digitized acoustic stimuli were stored on a Dell Windows 10 laptop computer located outside the recording booth for call-up during an experiment. Stimuli were presented through an ultrasonic tweeter loudspeaker placed 45 cm away from and facing towards the midline of the bat’s head. In initial experiments, we presented sounds through a Kenwood high-frequency KFC-XT15ie tweeter loudspeaker (Kenwood Corp, Tokyo JP). The frequency response of the Kenwood tweeter varied + 2 to − 9 dB across the frequency range of 20 to 90 kHz and decreased by 25 dB at 100 kHz. In later experiments, we presented sounds through a TDT electrostatic speaker (ES1 speaker driven by an ED1 speaker driver; Tucker-Davis Technologies, Alachua FL), because of its better higher frequency response (± 9 dB over the frequency range 4–110 kHz (Table 1). As used here, the maximum output of the Kenwood tweeter was 98 dB peSPL at 20–60 kHz, and that of the TDT tweeter was 90 dB peSPL at 25–80 kHz. Stimulus levels for both tweeters were calibrated by placing a Brüel & Kjaer Model 4135 (“¼”) condenser microphone at the position occupied by the bat’s head during experiments. Stimulus amplitude is expressed as dB peSPL re 20 µPa. During experiments, the acoustic stimuli delivered to the bat were monitored using a Dodotronic Momimic ultrasonic microphone (Dodotronic, Castel Gandolfo IT) suspended over the bat’s head, whose output was connected to one channel of a Tektronix Type 2000 70 MHz 4-channel digital oscilloscope (Tektronix Inc, Beaverton OR) located outside the recording booth. Experimental parameters on each recording day are listed in Table 1.

All electronic equipment needed for sound presentation and monitoring was located outside the recording booth, connected by cables running through small openings in the booth wall. For playback during experiments, the stored acoustic stimuli were uploaded from the Dell computer to a Koolertron 15-mHz DDS Signal Generator device (500 kHz digital-to-analog sampling rate; Model GH-CJDS66-A, Shenzhen Kuleton Technology, Shenzhen China). The DDS Signal Generator device output was triggered by a Biopac MP160 data acquisition system with AcqKnowledge 5 software (Biopac Systems, Goleta CA). The analog output of the DDS device was routed into a TDT PA5 attenuator (Tucker-Davis Technologies, Alachua FL), a Harman-Kardon PM545 stereo power amplifier (Harman International Industries, Stamford CT), and finally to the tweeter inside the recording booth. The electronic trigger for producing each sound, the electrical waveform delivered to the power amplifier, and the physiological signal recorded from the bat (see below) were recorded on the other three channels of the 4-channel Textronix oscilloscope located outside the recording booth.

Two different types of electrodes were used for recording brain activity from the bat’s scalp. One type was a Natus silver-cup 6-mm diameter conventional monopolar pediatric EEG electrode (pEEG; SKU 019–772,100, MVAP Medical Supplies, Thousand Oaks CA). An electrode was placed on the posterior scalp using conductive electrode paste (Weaver Ten20™; Weaver & Co, San Diego CA). A second and third electrode were placed posterior, on the bat’s bare upper and lower back, for differential recording and grounding. The second type was a novel tripolar EEG electrode (Besio et al. 2006; CREmedical, Kingston RI), consisting of three concentric conductive contact rings—an outer ring with a diameter of 6 mm, an intermediate-sized contact ring, and a central contact point. Tripolar EEG recordings (tEEG) are based on a nested sequence of differential stages that extract voltage differences between the outer and intermediate rings, the intermediate ring and central contact, and the two rings to the central contact (Besio et al. 2006). The final output is the voltage attributable to the central contact alone, which results in suppression of common-mode artifacts not only from muscles but also from regions of the brain remote from the central contact. The tripolar electrode produces one differential signal via a custom preamplifier (CREmedical), with its differential analog tripolar output joined to the overall ground electrode for a second differential stage with a gain of 20X. This preamplifier was located on the floor inside the recording booth, and then remotely connected to the rest of the electrophysiological equipment outside the booth. The analog outputs of the pEEG and the tEEG signals were remotely connected to a Biopac MP160 System and ERS100C evoked-response hardware module (Biopac Systems, Goleta CA) for averaging, analog-to-digital conversion, and subsequent digital signal processing. The ERS100C module was set to filters from 100–20,000 Hz and included a gain of 50,000X.

Electrophysiological responses were acquired in repetitive segments of 50 ms (the acquisition window length) triggered by stimulus presentations, digitized at a sampling frequency of 10 kHz and subsequently added to the ongoing averaged electrophysiological response. A real-time display of the building-up and averaging of the response was programmed into the Biopac display on its host computer (Dell Windows 10 Laptop connected via USB for operation with the AcqKnowledge 5 program). Recorded signals were then saved as .txt files.

Bats were allowed to position themselves inside the ceramic dish until they rested comfortably. Both bats rested their heads on the rim of the dish, which was rotated if necessary to point directly towards the loudspeaker position. Either pEEG or tEEG electrodes were applied to the bat’s exposed posterior scalp with conductive paste. We recorded bats’ body temperatures before and after each recording session; temperature never varied by more than 2 °C between these two time points. If bats moved excessively during recordings, we offered them sips of water or pieces of mealworms and short breaks (during which electrodes were often reapplied). Recording sessions lasted 5–30 min, depending on the bat’s ability to remain relatively motionless and the quality of the evoked response. Differences in session length resulted in uneven sample sizes for the different stimulus types.

A FM-1H 1 ms duration upsweep at 98 dB peSPL (Kenwood tweeter) or 90 dB peSPL (TDT tweeter) was presented as a search stimulus to assess whether electrode placements yielded stable baseline and high-amplitude evoked activity. Once a good electrode site was identified, we then presented stimuli varying in harmonic structure (1H or 2H), sweep duration, and sweep direction (Table 1), with the order determined by a random number generator, at a rate of 3.2/s for 200 repetitions (on two recording days with high background noise levels, repetitions were increased to 500). Stimulus levels were set at 86 or 98 dB peSPL (Kenwood tweeter) or at 90 dB peSPL (TDT tweeter). An interval of 8–30 s (average of 10 s; long enough to save data files and upload a new sound file) separated presentations of different stimuli. In other experiments (Table 1), we presented FM sweeps at a range of levels, decreasing in steps of 10 dB. This manipulation allowed us to evaluate the presence of amplitude-latency trading, a feature of neural responses prevalent in invasive recordings (Pollak 1988; Simmons et al. 1990; Klug et al. 2000). We hypothesized that amplitude-latency trading would be observed in surface ABRs as well. At the end of each recording session, bats were given water and mealworms, and returned to their home cages.

Electrophysiological responses were averaged over the 200 or 500 stimulus presentations while being visualized in realtime, and then saved to disk unless the recording was disrupted by bat movement. These files were imported as .txt files into MATLAB 2019b (MathWorks, Natick MA) for processing using custom scripts. Each response was demeaned to remove DC offset and filtered by a 48th order linear phase bandpass FIR filter with cutoff frequencies of 300–3000 Hz. A threshold of three times the RMS value of activity in the first 1 ms of the response, consistent with the threshold used by Luo et al. (2019), was set to identify evoked responses from baseline activity. Positive and negative peaks in the averaged response were identified automatically using the findpeaks function in MATLAB, and then confirmed visually. We quantified the amplitude and latency of the largest positive peak and the subsequent negative peak within a specific, short latency range (see below) to calculate peak-to-trough amplitudes. In experiments where stimulus level was varied, we first quantified these metrics at the highest stimulus level presented, then traced any changes in amplitude or latency of this peak-to-trough response at progressively lower sound levels. Response latency was calculated from stimulus onset at the bat’s ears, taking into consideration the acoustic delay of 1.25 ms produced by the 45 cm distance between the bat and the tweeter. Responses were also visualized as heat maps in which warmer colors indicate higher positive amplitudes and bluer colors indicate lower amplitudes.