Doppler-Shift compensation by the mustached Bat: Quantitative Data

Mustached bats, Pteronotus parnellii, use complex biosonar signals which seem to be well suited for detecting and tracking insect prey in a cluttered acoustic environment. Each emitted pulse has a relatively long (7–25 ms) duration and contains a combination of frequency-modulated and constant-frequency (CF) components, within a series of at least four harmonics. The CF component of the second harmonic differs from individual to individual, but averages about 61 kHz. It is of special interest because the ear is sharply tuned to this frequency and the bats are able to detect Doppler shifts in the returning echoes (see Schnitzler and Henson, 1980; Henson et al. 1987).

Doppler shifts are frequency changes brought about by the relative motion between a sound source and a sound detector. In echolocating bats, the biosonar signals are Doppler-shifted in proportion to flight velocity or to movement of an echoic surface. Signals reaching the ears are shifted twice relative to the emitted signal, once when sound travels towards the target and again on its return. The flight velocity of bats varies considerably, but in most cases it probably falls in the general range 0.1–5.0 m s⁻¹. At these two extremes, the amount of Doppler shift in the echo of a 61 kHz signal would range from about 35 Hz to 1799 Hz. Some bats alter their emissions in such a way that the frequency of the returning echo is stabilized in a narrow band to which the ear is tuned. The center of this stabilized band is commonly known as the reference frequency (Schnitzler, 1968, 1970a,b). This Doppler-shift compensation process has been highly perfected in horseshoe bats (Rhinolophidae) of the Old World and in one neotropical mustached bat, Pteronotus parnellii (Schnitzler, 1970b). Doppler-shift compensation is also known for other bats that emit biosonar signals with relatively long CF components (e.g. hipposiderids), but the compensation skills of these animals appear to be much less precise than those of the horseshoe and mustached bats (Habersetzer et al. 1984).

Rhinolophids are proven models for quantifying Doppler-shift compensation skills; they will compensate under a variety of experimental conditions, including shifts due to flight velocity or produced by objects moving towards them (Schnitzler, 1968, 1973), and they will compensate in response to electronically shifted simulated echoes presented to their ears (Schuller et al. 1974, 1975). When restrained they will compensate as they are propelled towards targets (Konstantinov et al. 1978). Mustached bats, however, do not respond under most of these conditions and studies of Doppler-shift compensation have been more limited. One technique which has been used for Pteronotus parnellii involves placing the bats on a pendulum and swinging them towards a fixed target (Henson et al. 1982; Kobler et al. 1985; Gaioni et al. 1990). Under these conditions, the bats must always deal with constantly changing velocities. Velocity changes could be substantial during the intervals between successive pulses and even during the period when a pulse is being emitted and the echo is returning. A pendulum undergoing an acceleration of 3m s⁻² would produce a 44 Hz change in the frequency of a 61 kHz signal during a 20 ms pulse–echo period. This, and the much longer intervals between pulses, makes the precision of Doppler-shift compensation difficult to determine with a pendulum. Furthermore, it has been shown that Doppler-shift compensation measured during pendulum swings is different from that recorded from the same bats in free flight (Lancaster et al. 1992). This observation was the impetus for developing the constant-velocity device described in the present report.

The purpose of this report is to provide new data on the accuracy and timing of Doppler-shift compensation as bats move at controlled constant velocities. The results of this study will be compared with previous data obtained during free flight and pendulum swings.

The bats used in this study were Jamaican mustached bats, Pteronotus parnellii parnellii (Gray). Ten bats were initially tested to evaluate their willingness to emit biosonar signals while attached to the constant-velocity device (CVD); of these, four were selected for the experiments because they emitted almost continuous streams of pulses for periods of 10–20 min before the pulse emission rates became irregular and therefore unacceptable. Data for each animal were gathered during a minimum of seven runs, each at different randomly selected velocities. Experiments on each animal were repeated on different days over a period of one to several months (see Table 1).

The CVD and the restraining system (bat holder) are illustrated in Figs 1 and 2. The holder was a foam-rubber sandwich held together with tape. Only the head, neck, shoulders and wrists protruded from the holder. A small tube leading to an electret microphone (Knowles model EA1842) was positioned under the bat’s mouth to record emitted pulses. The microphone signal was amplified and bandpass-filtered by a small custom-made electronic circuit attached to the bottom of the holder. The passband of the filter was tuned for each bat to isolate the approximately 61 kHz second harmonic constant-frequency (CF) component of the echolocation pulses. This conditioned signal was passed to the recording electronics via a miniature shielded cable. The microphone signal drove a phase-locked loop circuit, which provided a graphic aid to emission frequency changes and to overall compensation performance during each trial (see Fig. 3). The microphone signal, phase-locked loop output and a tachometer output from the constant velocity device were recorded on magnetic tape (Racal Store-7D) at a speed of 15 inches s⁻¹ (38.1 cm s⁻¹). The tape was replayed at a reduced speed and the signals were displayed on an electrostatic recorder (Gould ES1000). The records could then be examined and specific trials selected; detailed analysis was carried out only for those trials during which the bats emitted pulses almost continuously (see Fig. 3).

The CVD consisted of two bicycle wheels linked by a belt made of 1.5 mm diameter steel aircraft cable. Parallel to the cable was an array of targets, which consisted of paper strips hung over a fixed wire; these were spaced about 0.5 m apart. One wheel was driven by a precision stepping motor (Compumotor model CX 57-51) in combination with a computer-controlled driver (Compumotor model 2100). Cable velocities were initially established by timing the travel of a small aluminum target attached to the moving cable. Accurate timing signals were recorded as the target passed between two precisely spaced X-band radar motion sensors (Alpha Industries, model GOS 2580). It was determined that all settings of the motor controller translated to exact cable velocities. A separate custom-built tachometer circuit, which gave a d.c. voltage proportional to the velocity of the drive wheel, was used to check the overall stability of the velocity during runs. A hook attached to the wire cable was used to snag one of a pair of plastic snap rings (Archer all-purpose cable/utility clips, catalogue no. 278-1640) attached to the top of the animal holder (Fig. 2). For each trial, the drive motor was started and the tachometer output was monitored until the motor and cable had reached the preset speed. The rings were then snapped around the moving wire and held free of contact with the cable. Once the hook had snagged the more posterior ring, the bats quickly reached the desired velocity. The anterior ring kept the bat holder in the desired orientation. During each experiment, bats were propelled at randomly selected velocities ranging from 0.1 to 5.0 m s⁻¹. The smallest velocity differences tested were increments of 0.1 m s⁻¹.

Although the bats rapidly reached the desired velocity, there was an initial period of slowing because of the sudden additional load on the motor. In addition, the long cable oscillated slightly when the hook grabbed the plastic ring; this produced noticeable oscillations in the d.c. output of the tachometer (see Fig. 3B). For these reasons, measurements of pulse frequencies were not made until oscillations had ceased and the velocity was stable.

In order to study the time course of Doppler-shift compensation, in some trials bats were shielded from Doppler-shifted echoes by a shutter which swung open at the 2 m point in each run. The shutter was attached to the bat holder and moved with it; when the shutter opened, the bat was immediately exposed to echoes from targets and began to Doppler-shift compensate (see Fig. 3A). During these shuttered runs, the pulse frequency data were collected only after the CF in the pulse emissions became stable.

The recorded signals were analyzed with a computer to measure the constant-frequency portion of each call. The microphone signal was heterodyned with a frequency about 1000 Hz below the CF and the resultant low-frequency signal was digitally sampled at 8 kHz. This was then processed with a spectrum analyzer/digital oscilloscope (Rapid System Inc., model R350) programmed to determine automatically the constant-frequency component (±15 Hz) of each pulse. Statistical analysis and computation of echo frequencies for each run were carried out with a spreadsheet program (Excel, Microsoft Corp.). The stabilized echo (reference) frequency was calculated from the Doppler formula, using: echo frequency = compensation frequency X [(344 + V_b)/(344 -V_b)], where compensation frequency is the emitted second harmonic CF (Hz), 344 is the speed of sound in air in m s⁻¹ and V_b is the velocity of the bat (m s⁻¹). Each bat’s resting frequency was determined by analysing the pulses that it emitted prior to being snagged by the hook on the CVD cable. This was a necessary step because the resting frequency was subject to change with time and activity (see Huffman and Henson, 1991,1993a).

In two bats, electrodes were chronically implanted, under anesthesia, near the cochlear aqueduct so that cochlear microphonic (CM) potentials could be recorded. CM potentials have the same frequency as the frequency of an acoustic stimulus. The cochlear resonance frequency (CRF) was measured just prior to the first run and again at the end of the experiment. This was accomplished by stimulating implanted ears with brief (2 ms) tone pulses with fast (0.1 ms) rise and decay times. The frequency of the stimulus was adjusted to be about 1.0 kHz below the CRF. Spectral analysis of heterodyned CM potentials resulted in a large peak at the cochlear resonance frequency (see Henson et al. 1990, their Fig. 4). We used this technique to establish the relationship between the resonance frequency of the ear and the stabilized frequency of echoes as a result of Doppler-shift compensation. Details concerning the electrode implantation technique, cochlear resonance measurements and spectral analysis have been published previously (Henson and Pollak, 1972; Henson et al. 1990; Huffman and Henson, 1993a,b).

When bats were attached to the CVD, they emitted pulses prior to and during each trial. The CF components of pulses emitted when the bats were not moving typically fell in a narrow 150–500 Hz band. The center frequency of this band varied among bats and in the same bat at different times (Table 1). The differences for an individual animal can be accounted for by small changes in body temperature, prior flight activity and accompanying changes in the cochlear resonance frequency, as previously described by Henson et al. (1990) and Huffman and Henson (1991, 1993a). When the bat holder was snagged by the hook on the moving cable, the animals increased their pulse repetition rate and the CF pulse component began to change almost immediately (Fig. 3B). During trials in which the ears were shielded from Doppler-shifted echoes by a shutter, Doppler-shift compensation did not begin until the shutter opened (Fig. 3A). Compensation occurred at all velocities tested, from 0.1 to 5.0 m s⁻¹ (Fig. 4). Compensation was usually accomplished by a progressive pulse-to-pulse lowering of the emitted CF from the initial resting frequency value; this occurred in small frequency steps (93±72 Hz; mean ±S.D.). For a total of 27 trials, at velocities exceeding 4 m s⁻¹, the mean time required for maximal compensation was 514±230 ms and the number of pulses required to reach this maximum averaged 10.78±5.16.

The animals compensated to hold the approximately 61 kHz reference frequency at values that averaged slightly higher than the resting frequency but, as shown in Figs 4 and 5, reference frequency values were velocity-dependent, the higher the velocity, the higher was the reference frequency (average slope 55 Hz m⁻¹ s⁻¹). For all velocities, the mean values for the reference frequency represented a compensation of less than 100 % of the Doppler shift. From the composite data in Table 1, it is interesting to note that, during unshuttered runs, the amount of undercompensation (12.8 %) was closer to perfect (0 %) than during the shuttered runs (20.5 %). In three of the four animals, undercompensation ranged from 6.6 to 9.8 % on their best trial days. For all velocities tested, the standard deviation of the echo frequencies was 89 Hz. There was a significant difference (P::s0.005) between CFs measured at velocities that differed by 0.1 m s⁻¹. A velocity difference of 0.1 m s⁻¹ corresponds to a Doppler-shift difference of about 35 Hz in the 61 kHz signals reaching the ear.

In many cases, there were noticeable interruptions in the stream of pulses emitted by the bats. In these instances, pulses emitted after an interruption were often higher in frequency than those in the preceding series of pulses; when this occurred, the bats usually brought the emitted CF back to the previous level of compensation within 10 pulses (Fig. 3A). During CVD trials, the bat’s head was free to move from side to side and it should be noted that the maximum Doppler shifts on which our calculations were based were for objects that were straight ahead of the bat.

Previous studies on many mustached bats have established that the resting frequency is generally 100–400 Hz lower than the cochlear resonance frequency (CRF) (Henson et al. 1985, 1987). It also appears that the reference frequency is intermediate between the resting and resonance frequencies. In the two animals in which cochlear microphonic electrodes were implanted, we observed these same general relationships among the different frequencies. By virtue of the progressively higher reference frequency with increases in velocity, it became obvious that the higher the velocity of the animal, the closer the proximity of the reference and resonance frequencies. Only one of our implanted animals emitted continuous streams of pulses when attached to the CVD such that the changing relationship between the reference frequency and CRF could be studied at different velocities. The graphic display of the data for this animal (Fig. 5) shows that the reference frequency and CRF values converged at velocities approaching 5 m s⁻¹. Fig. 6A shows a histogram of echo frequencies normalized to the reference frequency for our four bats during seven runs at different velocities in the range 2.4–3.6 m s⁻¹. This range of velocities corresponds to measured speeds for mustached bats during fast flight in a relatively large space.

A constant-velocity device, like the one described in this study, appears to be a good system for quantification of Doppler-shift compensation skills by selected bats. The system is well-suited for Pteronotus parnellii and, judging from the studies of Simmons (1974) and Konstantinov et al. (1978), who used somewhat similar devices, it should be useful for studies of rhinolophids and other bats that Doppler-shift compensate. The CVD has a distinct advantage over previously used pendulum systems in that the velocity and reference frequency are stabilized over long periods. In future experiments, we hope to use this device to study biosonar signal emissions and neural correlates of echolocation in different species exposed to identical movements and to varying arrays of targets and during exposure of the ear to masking sounds.

It is curious that the Doppler-shift compensation skills of three of the four bats varied so greatly between shuttered and unshuttered runs (Table 1). We have no explanation for this. The Doppler-shift compensation values determined during unshuttered CVD runs are more comparable to values obtained with free-flying mustached bats than to those measured during pendulum swings (Gaioni et al. 1990; Lancaster et al. 1992). The different level of compensation with the pendulum can be attributed to the difficulty of precise compensation in the presence of the continuous velocity changes and the relatively long (approximately 500 ms; 10 pulse) compensation times that seem to be required. When swinging on a pendulum, bats make continuous adjustments towards perfect (100 %) compensation, but because of continuous changing velocities there are continuous and variable compensation errors. Also, under these conditions, the slower the pulse repetition rate and interpulse interval, the greater is the potential for undercompensation. Furthermore, the longer the duration of an emitted signal, the greater is the total frequency shift in the echo. The somewhat inferior nature of Doppler-shift compensation during pendulum swings is also evident from the data of Gaioni et al. (1990), who calculated that the bats compensated for an average of 80 % of the echo Doppler shift (i.e. 20 % undercompensation). In their studies, they selected data from days when the performance was best. In our studies, the bats undercompensated for an average of 12.8 % of the Doppler shifts measured during regular (unshuttered) trials and, in the best runs, three of the four bats undercompensated by only 6–9 %. It should also be noted that Gaioni et al. (1990) used a secondary target consisting of thin vinyl strips attached to a rod and shaken to stimulate the bats to emit more pulses. In our experience, this technique causes a decrease in Doppler-shift compensation performance.

Some of our data for mustached bats are similar to those obtained for horseshoe bats (Rhinolophidae), which also Doppler-shift compensate with high precision (Konstantinov et al. 1978; Schnitzler, 1970a,b, 1971; Simmons, 1974; Schuller et al. 1974, 1975; Gustafson and Schnitzler, 1979). The concordant findings include (1) consistent undercompensation; (2) compensation for small (approximately 35–50 Hz) frequency shifts; and (3) changes in the reference frequency with velocity (see for example, Fig. 11 in Gaioni et al. 1990). It is interesting to note that some species which emit pulses with relatively long CF components, e.g. Hipposideros speoris and H. bicolor (now H. fulvus), also Doppler-shift compensate but they lower their emitted CF only by about 55 % of that expected if full compensation were to occur (Habersetzer et al. 1984).

The small amount of undercompensation in Pteronotus parnellii may represent an underestimation of the animal’s Doppler-shift compensation capacity. Compensation would have been closer to perfect if (1) the bats did not adjust their CF component in small-frequency steps when the emitted frequency strayed from more or less stable values; (2) the strategy were not limited to keeping the frequency at or below the resonance frequency of the ear; and (3) the animals always attended to targets that were directly in front of them. It should be noted that any echo from an off-center target must have had a smaller Doppler shift than an echo from an object directly in front of the bat. In our CVD experiments, the targets were aligned parallel to the movement of the animals and thus the targets were always slightly off center.

Probably of some importance in our experiments was the placement of a continuous array of targets parallel to the line of movement. This ensured a relatively loud continuous exposure of the ear to echoes. Konstantinov et al. (1978) found a significant difference in compensation by Rhinolophus when the targets were at different distances, i.e. during the beginning and terminal parts of a forward pendulum swing. They also noted a considerable improvement in Doppler-shift compensation on their uniform motion system after an additional target had been placed in the path of the bat.

The statistically significant change in pulse emission frequency with velocity changes of only 0.1 m s⁻¹ (Doppler-shift change of 35 Hz) implies a phenomenal frequency resolution capacity, approximately 35 Hz in 61 kHz. As unlikely as this may seem, there is ample neuroanatomical and neurophysiological evidence for exquisite receptor and neural tuning in mustached bats (Pollak et al. 1972; Bodenhamer and Pollak, 1981; Suga and Jen, 1977; Suga et al. 1975; Huffman and Henson, 1993b). The cochlea, for example, has a relatively long (approximately 800 inner hair cells) expanse, which appears to be devoted to the processing of signals in a narrow (approximately 800–1800 Hz) bandwidth (M. M. Henson, personal observation). Neural units in this band have Q-10 dB values (a measure of filter sharpness) that typically exceed 60 and may extend to 150 or more. In other regions of the mustached bat’s cochlea and in the central nervous system, the Q values are typically less than 20 (see Huffman and Henson, 1993b, their Fig. 12). The specialized high-resolution region in the mustached bat’s cochlea has been referred to as an acoustic fovea by analogy with the region of high visual resolution in the eye (fovea centralis). There is some evidence that the central part of the fovea is tuned to frequencies at or near the resonance frequency of the ear. In our laboratory, we have studied single and multiple units in the cochlear nucleus and inferior colliculus and compared the best frequencies with the cochlear resonance frequency (Huffman and Henson, 1993a,b). From data on the distribution of best frequencies it appears that sharply tuned units have best frequencies that occupy narrow bands on both sides of the resonance frequency (see Fig. 12 in Huffman and Henson, 1993b). Data from the present study support previous work indicating that mustached bats Doppler-shift compensate to hold the echo frequency at or below the resonance frequency of the ear. These and similar data have suggested that many high-Q neurons in the high-frequency (above resonance) part of the acoustic fovea are not used or are poorly stimulated by echoes from background targets. The hair cells and neurons tuned above the resonance frequency seem, then, to be reserved more for processing the Doppler shifts and acoustic glints created by the wingbeats of insects than for analyzing the frequency difference between emitted pulses and returning echoes. This has led to the suggestion that the low-frequency half of the fovea is primarily associated with the initial coding of differences between emitted CF signals and Doppler-shifted echoes. New data on the apparent ‘awareness’ of the bat’s nervous system to shifts in CRF and neural tuning complicate such interpretations (Henson et al. 1990; Huffman and Henson, 1991,1993a,b).

The change in reference frequency with velocity is potentially related to the changing properties of the ear resulting from acoustic stimulation by biosonar signals. If the resonance properties of the bat’s ear were to change as a function of exposure to the emitted pulses, and if the emission frequency were to change in accordance with shifts in cochlear tuning, then one would predict that the reference frequency would change as a function of velocity and the associated change in pulse emission frequencies. Evidence that the bats change the frequency of their pulse emissions in accordance with changes in cochlear resonance is well documented (Henson et al. 1990; Huffman and Henson, 1991, 1993a), but further studies are needed to establish a link between resonance changes and exposure of the ear to emitted signals. A potential advantage of the progressive change in reference frequency with velocity is that it could keep the approximately 61 kHz constant-frequency components of both the echo and the pulse in the acoustic fovea.

This work was supported by NIH grant DC 00114 from the National Institute on Deafness and Other Communicative Disorders.