Directional Sensitivity of Echolocation System in Bats Producing Frequency-Modulated Signals

For echolocation, bats emit sound and listen to echoes (Griffin, 1958). To investigate the directional sensitivity of their echolocation system it is necessary to measure the directional properties of both the orientation sounds and the ears. The field of the orientation sound around the head has been measured in Eptesicus and Chilonycteris^† (Simmons, 1969), Rhinolophus (Schnitzler, 1968) and Megaderma (Möhres, 1966). In these measurements the animal with its head unrestrained emitted sounds toward a target. Thus these data might include the effect of head movement. Sokorov & Makarov (1971) fixed the head of Rhinolophus and measured the sound field produced by it. Recently, it has been found that sounds very similar to species-specific orientation sounds are elicited by electrical stimulation of the dorsal part of the reticular formation in the midbrain and/or the central grey matter of an unanaesthetized bat. In this case the head was immobilized by cementing it to a nail (Suga & Schlegel, 1972). This technique makes possible accurate measurements of the sound fields produced by bats.

For sound localization at least two types of cues are conceivable: interaural pressure and time (or phase) differences. In man the interaural pressure difference (IPD) is used mainly for location of high-frequency sound, while the interaural time difference (ITD) is used mainly for low-frequency sound (Mills, 1958, 1960 ; Stevens & Newman, 1936). Since the bats of the genus Myotis are small and their orientation sounds are higher than 10 kHz, the IPD may be a primary cue for echolocation. The IPD as a function of azimuth is obtained from the directional sensitivity curve of the auditory system, which can easily be measured by recording summated or single neural activity (Grinnell, 1963; Grinnell & Grinnell, 1965; Neuweiler, 1970; Suga, 1964).

To investigate the directional sensitivity of the echolocation system and the IPD as a function of azimuth, we therefore measured the sound fields produced by bats, and also the directional sensitivity of the auditory system in terms of summated neural activity.

Orientation sounds of Myotis are frequency-modulated (FM) so that an ITD can be coded by a group of neurones with different best frequencies on both the left and right sides. Possible roles of the ITD as well as the IPD are also discussed in relation to echolocation.

Three grey bats (Myotis grisescens) and five little brown bats (M. lucifugus) were used. These two species are closely related, and there were no noticeable differences in orientation sounds and audiograms. Under ether anaesthesia a nail 1·5 cm long was mounted on the skull with acrylic adhesive and dental cement. Then the pinnae were carefully relocated to their original position by tightening the skin with sutures. A few hours to a day later the animal (now recovered) was placed on a 7·5 cm diameter vinyl ball floating on water. The head of the animal was fixed by fastening the nail to a metal rod with a set screw. The midpoint of the ears was placed at the centre of the acoustic perimeter and the eye-nostril line was made horizontal. The nail, the rod supporting it, and the floating ball were at least 20· posterior to the animal ‘s mouth.

Without anaesthesia, a small hole was made in the skull through which a double steel electrode with a few megohms resistance was inserted for the electrical stimulation of the midbrain to elicit FM sounds similar to species-specific orientation signals (Suga & Schlegel, 1972). After fixing the electrode in place with dental cement, the micromanipulator used for positioning it was removed. The electrical stimulus applied to the brain was a short train of pulses delivered twice per second. Each train consisted of 10 electric pulses, each having a duration of 0·1 msec and an amplitude of 1–15 V. The inter-pulse interval was 1·7 msec.

The sounds elicited by the bat were recorded with two quarter-inch condenser microphones (Brüel & Kjaer 4135). One,’reference microphone’ pointed towaid the bat ‘s mouth and was positioned 5° right, 5° down from the eye-nostril line, and 70 cm anterior to the mouth. The other,’the scanning’ microphone, mounted on a movable aluminium arm in order to scan the bat ‘s perimeter at a 50 cm radius, was also aimed at the mouth. Since the sounds emitted by the bat were frequency-modulated, the interference between the signal directly incident upon the microphones and the echoes was easily detected, because the overlap of the delayed echoes with the signal produced ripples in the envelope of the recorded signal. To minimize these echoes, objects in the sound-proofed room were covered with cotton. An Ampex FR-100 tape recorder (frequency response: 300–300000 Hz at 60 in/sec) simultaneously recorded the following signals: electrical stimuli applied to the brain, bat ‘s sound monitored with the reference and scanning microphones, and electric signals indicating the position of the scanning microphone. At each position of the scanning microphone about 15 FM sounds emitted by the bat were recorded.

The recorded signals were played back at of the original tape speed. A 2 sec square pulse triggered by the electrical stimulus controlled the shutter of a camera, while the pulse synchronized with the onset of the bat ‘s sound triggered the sweep of a cathode-ray oscilloscope. The sound monitored with the reference microphone was fed into the period meter in order to examine the frequency sweep. The output of the period meter and the tape-recorded sounds were simultaneously displayed on the oscilloscope screen and were photographed. A 3125 Hz signal, which corresponds to 100 kHz at the original tape speed was also fed into the period meter prior to the tape-recorded sound and its output was also displayed for purposes of calibration (Fig. 1). In addition, the signal indicating the location of the scanning microphone was. displayed on an oscilloscope screen. In Fig. 1 the frequency of a bat ‘s sound sweeps from 100–43 kHz. Since there were no noticeable harmonics for the components higher than 50 kHz, the radiation patterns of sounds higher than 50 kHz could be studied by measuring the peak-to-peak amplitudes. In our experiments the amplitude of the 55, 75, and 95 kHz components were measured in FM sounds monitored with the reference and scanning microphones, and the difference in amplitude at the two microphones was expressed in either percentage of, or decibels below, maximum. Assuming sagittal symmetry, each point in our data represents the average of six measurements.

To study the relationship between mouth movement and sound emission a photoelectric displacement transducer was connected to the lower jaw of one animal by a rubber band, and a quarter-inch microphone was placed 10 cm anterior to the bat. The outputs of the transducer and microphone were simultaneously displayed on an oscilloscope screen (Fig. 2).

For the measurement of the directional sensitivity of the auditory system a tungsten electrode was inserted into the nucleus of the lateral lemniscus to record its activity evoked by acoustic stimuli. This is hereafter called the LL response (equivalent to N₄). The LL response was selected, because it is more directional than the summated neural activity of primary auditory neurones (N₁) as a result of binaural interaction occurring in medullary auditory nuclei (Grinnell, 1963). Instruments for acoustic stimulation and recording were the same as those described in the previous paper (Suga, 1971). A condenser loudspeaker was moved around the head in the horizontal plane at 66 cm radius, and thresholds of LL responses to 35, 55,75, and 95 kHz sounds were measured as a function of azimuth. These acoustic stimuli had a rise-decay time of 0·2 msec and a duration of 4 msec.

When an electrical stimulus was applied to the dorsal part of the reticular formation in the midbrain, the bat opened its mouth 50–60° and emitted an FM sound with a latency of 30–50 msec. This was very similar to the species-specific orientation sound used in the searching phase of echolocation. The amplitude of the emitted sound ranged from 100–110 dB SPL (sound pressure level referred to 0·0002 dyne/cm² r.m.s.) at 10 cm in front of the mouth. Its duration was between 3 and 5 msec. The frequency always swept downward from 100—ca. 45 kHz. If such an FM sound was emitted at different phases of mouth movement, the sound field measured would show large variations; the relationship between the sound emission and mouth movement was thus first studied. As shown in Fig. 2, the sound emission was not strictly locked to a certain phase of the mouth movement, but it occurred anywhere between slightly earlier and slightly later than the peak opening of the mouth, so that a given component of the FM sound might be emitted at any aperture of the mouth between 30–60°. Thus, the sound field varied for each sound. This variation was within ±12% of the mean amplitude.

Fig. 3 shows an example of sound fields produced by M. grisescens (Mg 1). The contour map shows that the sound is strongly radiated in a direction of 5–10° downward from the eye-nostril line and the main lobe is sharper at higher frequencies. Figs. 4,–6 show radiation patterns of the 55, 75, and 95 kHz components, respectively, in the horizontal and vertical planes produced by Mg 1. All directional patterns were normalized with respect to the maximum amplitude for each component. The radiation angle at half-amplitude of maximum was 38° lateral, 18° up and 50° down at 55 kHz, 34° lateral, 8° up and 32° down at 75 kHz, and 30° lateral, 5° up and 2 down at 95 kHz (Figs. 4·6). Thus, the width of the main lobe becomes narrower wi an increase in frequency. At 95 kHz two prominent side lobes were present: one oriented 25° upward and the other 50° downward (Fig. 6). Two other bats showed different values, which were broader in angle at half-amplitude than Mg 1.

Since the measurement of sound pressure was performed in 5° or 10° steps, there was a possibility that sharp lobes which might exist for high frequencies were missed. The fields of 55, 75 and 95 kHz sounds were therefore measured around a loudspeaker with an aperture comparable to that of the bat ‘s mouth. This measurement indicated that the presence of such lobes in addition to those in Fig. 4 was unlikely. The main lobe calculated by Strother & Mogus (1970) is also compatible with our data. Furthermore, there was a possibility that the ball under the bat modified the sound field in the hemisphere in front of the bat. The sound field around the loudspeaker was also measured with and without the ball. The result demonstrated that the sound field in front of the bat was not noticeably different.

When a tungsten-wire electrode was inserted into the lateral lemniscus, the LL response (lateral lemniscal evoked potential) to a tonal stimulus was 1–2 mV in amplitude. The loudspeaker was then moved in the horizontal plane including the eye-nostril line, and the threshold of the LL response to the sound was measured as a function of azimuth angle. The DSA was expressed by the reciprocal of the threshold. Figs. 7, 8 show DSA curves measured with either 35 and 55 kHz or 75 and 95 kHz tones. Each curve is the average of measurements with 5 different bats. The LL response was most sensitive to sounds delivered from the contralateral side, so that impulses for the lateral lemniscus mainly originated from the contralateral ear. The maximum sensitivity appeared at 38 ± 7·5° azimuth to the midline for 35 kHz, 27 ±9·0° for 55 kHz, 18 ±7·5° for 75 kHz and 5·0 ±8·1° for 95 kHz. The slope of the DSA curve toward and beyond the midline was 0·28 ± 0·03, 0·43 ± 0·03, 0·42 ± 0·03 and 0·58 ± 0·03 dB/degree for 35, 55, 75, and 95 kHz, respectively. Since higher-frequency sounds were more directional, it was expected that the slope of the DSA curve would increase with frequency, but it was not anticipated that the maximally sensitive direction would move medially with increasing frequency.

The DSE was obtained by adding the two curves for the directionality of the orientation sound (DO) and the DSA. As shown in Figs. 7, 8, the slopes of the DSE curve on both sides of its peak are about 0·6 dB/degree at 55 kHz, 0·8 dB/degree at 75 kHz, and 1·0 dB/degree at 95 kHz. These results confirm that the DSE increases in sharpness with frequency. The peak of the DSE curve is 15° lateral for 55 kHz, 15° lateral for 75 kHz, and 2·5° lateral for 95 kHz. It was noticed that the peak for a 95 kHz sound was very close to the midline.

Since the IPD appeared to be the essential cue for echolocation in bats of the genus Myotis, it was calculated as a function of azimuth angle for 35, 55, 75, and 95 kHz. We assumed sagittal symmetry of the auditory system. The DSA or DSE curves in Figs. 7, 8 were inverted at zero degrees. The IPD curves in Fig. 9 were then obtained by subtracting the inverted curves from the non-inverted ones. The IPD for a 35 kHz sound changed linearly with azimuth between 0–50° at a rate of 0·4 dB/degree. Beyond 6o° it changed at a rate of 0·2 dB/degree. For a 55 kHz sound the slope of the IPD curve was 0·7 dB/degree between 0-30° lateral, and 0-2 dB/degree beyond 30°. For 75 and 95 kHz sounds the slope of the IPD curve was 0·3–0·4dB/degree between 0° and 30° –40° lateral, but it becomes nearly zero beyond 40°. These data indicate that a 55 kHz sound gives a larger IPD cue than do 35, 75, and 95 kHz sounds and that the IPD cue for sounds higher than 75 kHz is very poor at positions more lateral than 30°.

Sound fields produced by different species of bats have been measured by Möhres (1966), Schnitzler (1968), Simmons (1969), and Sokolov & Makarov (1971). Except for Sokolov & Makarov ‘s extensive study, the animals did not have a head-holding device when they emitted orientation sounds toward a target. The azimuth angle at half-amplitude of maximum is 23° at 28–30 kHz for Eptesicus fuscus and Pteronotus pamellii rubiginosus (previously called Chilonycteris rubiginosa) which emit sounds through the mouth (Simmons, 1969), and 21–23° at 83 kHz for Rhino lophus ferrumequinum which emits signals through the nostrils (Möhres, 1966; Schnitzler, 1968), and 35° at 60–80 kHz for Megaderma lyra (Mohres, 1966). According to Sokolov & Makarov (1971) who fixed the head of Rhinolophus, the azimuth angle at a half-amplitude is 28°, but it is 30° after amputating the nose-leaf. Thus, the effect of the nose-leaf on horizontal radiation is very small. This is unexpected data for us, because we had the impression that Rhinolophus emitted an orientation sound in a narrower beam than that of Myotis and Eptesicus, since Rhinolophus emits a high-frequency sound through nostrils which are surrounded by the nose-leaf.

In our data on Mg 1, the horizontal angle at a half of maximum amplitude is 34° at 75 kHz and 30° at 95 kHz, which are comparable to the data obtained by Sokolov & Makarov (1971). Conversely, the azimuth angle at half-amplitude at 55 kHz in Mg 1 (38°) is much wider than that at 28–30 kHz in Eptesicus and Pteronotus (Simmons, 1969).

The vertical distribution of an orientation sound has been measured only with Rhinolophus. The vertical angular width at a half-amplitude of maximum at 83 kHz is 43° according to Schnitzler (1968) and 73° according to Sokolov & Makarov (1971). The large angular width obtained by Sokolov and Makarov is due to the fusion of the main and side lobes. Interestingly, the dissection of the nose-leaf greatly reduces the vertical width down to 46° because of the reduction in the side lobes. Rhinolophus moves the nose-leaf during echolocation, probably to control the vertical distribution of orientation sounds. In our data with Mg 1, the vertical angular width of the main lobe at half-amplitude was 68° at 55 kHz, 40° at 75 kHz and 30° at 95 kHz.

Since the DSE curves showed sharp slopes (0·6 dB/degree at 55 kHz and 1·0 dB/degree at 95 kHz), the amplitudes of echoes at the ears greatly decrease as echo sources move away from the maximally sensitive direction. Accordingly, clatter from irrelevant sources at the ears is greatly reduced. If the peaks of the DSA curves were more lateral than those in Figs. 7, 8, the IPD and the dynamic ranges of the IPD curves would increase, but the peaks of the DSE curves would decrease and move laterally. In other words, the IPD and the DSE are mutually dependent. For echolocation the peak of the DSA curve should be thus neither too lateral nor too medial. The dynamic ranges of the IPD curves greatly depend on the difference between the slopes of the DSA curves toward the ipsi-and contralateral sides. To increase the dynamic range, the slope lateral to the peak of the DSA (or DSE) curve should be smaller than that on the medial side. This was found true for lower frequencies but not for higher ones.

For comparison with the present data, a DSA curve (Neuweiler, 1970) and a directionality curve of orientation sound (Schnitzler, 1968) of Rhinolophus ferrumequinum are replotted in Fig. 10A. The frequency of sound used for these measurements was 83·3 kHz. The slope of the DSE curve is 0·5–0·6 dB/degree. The IPD varies linearly with azimuth angle from 0–30° lateral at a rate of 0·5 dB/degree. It does not, however, change as the azimuth angle increases beyond 30° (Fig. 10B). The IPD curve of Rhinolophus is thus similar to that of Myotis (whether the difference between the two curves in Fig. 10 B is significant or not is subject to future experimentation).

Rhinolophus moves the pinnae synchronously with vocalization and does not change the frequency of CF-FM (constant-frequency plus frequency-modulated) sound more than a few kHz during echolocation (Schnitzler, 1968), while Myotis does not move the pinnae during vocalization and greatly drops the frequency of FM sounds when hunting insects and when landing (Griffin, 1958). Thus, the dynamic range of the IPD curve may alternately shift left or right in Rhinolophus, while it broadens in Myotis.

Both the narrow dynamic range of the IPD curve for higher frequencies and the decrease in frequency of FM orientation sounds during the approach and terminal phases appear to be disadvantageous for echolocation and target characteristics. However, Myotis may have not only this disadvantage, but also some advantage. During a searching phase Myotis emits FM sounds sweeping from 100–45 kHz. At these frequencies, the slope of the DSE cune toward a lateral side is very sharp, so that any clatter which may be produced by irrelevant objects and/or other bats more lateral than 30° is greatly attenuated. Since the distance to a target is long in the search phase, a wide dynamic range for echolocation may not be needed by the bats, so the narrow dynamic range may not necessarily be disadvantageous for echolocation. The bats can efficiently find small objects in front of them with a sound of such a high-frequency, although higher-frequency sounds are attenuated by the air more than lower-frequency sounds. The target characteristics may be performed during and/or just before the approach phase. During the terminal phase and the latter part of the approach phase Myotis emits FM sounds sweeping down about one octave between 50–15 kHz (e.g. from 40–20 kHz) at a high repetition rate. In these phases tracking of the target may be most important. Since the distance to the target is short, the direction of the target may quickly change, so that a wide dynamic range for echolocation and the frequent emission of orientation sounds may be more important to the bat.

In man the just-detectable IPD is 0·5 dB at 2–3 kHz and 50 dB above the sensation level which corresponds to a 1·2–2·5° lateral shift of a sound source from the midline (Mills, 1960). In our data on Myotis the slope of the IPD curve around the midline was 0·4, 0·7, 0·3, and 0·4 dB/degiee for 35, 55, 75, and 95 kHz sounds, respectively. If one assumes that the just-detectable IPD of Myotis is 0·5 dB, the bat may be able to detect a 0·7–1·7° azimuth difference around the median plane. Harrison & Downey (1970) measured the IPD with Phyllostomus hastatus and obtained the minimum detectable angular difference of 2·0° at 10 kHz and 2·3° at 20 kHz. They assumed the just-detectable IPD to be 0·5 dB. Behavioural experiments with bats indicate that Eptesicus fuscus with an interaural distance of 14 mm can detect a 6·8° azimuth difference and P. hastatus with an interaural distance of 22 mm can detect a 4·6° difference with their orientation sounds (Peff & Simmons, 1971). These behavioural data can be explained by the IPD cue without assuming that the bat can detect a much smaller IPD than can man. There is no doubt that the IPD is the essential cue for echolocation. Since Myotis can catch two separate Drosophila within a half second (Griffin, Webster & Micheal, 1960), the minimum detectable azimuth difference in Myotis may be smaller than the behavioural values cited above.

Pumphrey (1948) stated that sound localization is accomplished by the binaural comparison of intensity ratios at least at three different frequencies to which the ear shows different polar diagrams of sensitivity. As a matter of fact, the polar diagrams of the ears of owls (Payne, 1961) and bats (Neuweiler, 1970; Grinnell & Grinnell, 1965) change extensively with the frequency of sound and also with the position of the external ear. This is also true in man (Sivian & White, 1933), so that the spectrum density of a complex sound differs at the two ears (Nordland & Fritzell, 1963). The source of a complex sound is thus more easily located than that of a pure tone. Thus, FM sounds are apparently better signals for echolocation than CF sounds.

Responses of auditory neurones vary not only in discharge pattern, but also in latency with stimulus amplitude. In Myotis the difference in latency between the left and right LL responses to a tonal stimulus due to the IPD appears to be much larger than the ITD and may greatly contribute to the information processing for sound localization. Since this interaural latency difference is due to the IPD, it is not included in the following discussion about the ITD.

In man, the just-detectable ITD is at best 5 μsec (9 μsec on the average) at a 75 % correct point for a noise with a band width of 0·15–17 kHz and 60–80 dB above the sensation level (Klumpp & Eady, 1956). This ITD Corresponds to placing a sound source 1·1° lateral from the median plane (Fedderson et al. 1957). Fedderson et al. (1957) calculated the ITD as a function of azimuth for a sphere with a diameter of 17·5 cm, which is equal to the interaural distance in man, and found that the calculated values (9 μsec/degree) are very similar to the values measured with microphones placed in the ears of human subjects. In Myotis grisescens and M. lucifugus, the interaural distance is ca. 9 mm. The ITD for Myotis as a function of azimuth angle can be obtained by multiplying the ITD for man by the ratio between the interaural distances, 0·05. The calculated ITD curve showed a slope of about 0·45 μsec/degree from the median plane to 40° lateral. If one assumes that the just-detectable ITD in Myotis is 5 μsec as in some human subjects, the theoretical limit for the just-detectable azimuth difference would be 11°. In order to detect a 0·7–1·7° azimuth difference with the ITD cue, the just-detectable ITD should be at least 0·3–0·8 μsec or less. The just-detectable ITD of Myotis should therefore be assumed to be 8–17 times smaller than that of man, if the ITD cue is to be considered equally important to the IPD cue for echolocation. No data have been obtained yet to indicate whether the above assumption is reasonable or not. According to the above assumption the minimum time-pressure trading ratio is 0·6–1·0 μsec/dB, which is much smaller than the value for man, 9–11 μsec/dB (Hershkowitz & Durlach, 1969 ; Gilliom & Sorkin, 1972).

In Eptesicus and Phyllostomus the just-detectable azimuth difference is 6–8° and 4–6°, respectively (Peff & Simmons, 1971). Since the interaural distance is about 14 mm for Eptesicus and 22 mm for Phyllostomus, the ITD at the just-detectable azimuth difference is about 5 μsec. Thus, the ITD cannot be ruled out as a cue for sound localization, although it is apparently inferior as a cue to the IPD.

Primary auditory neurones of mammals commonly show phase-locked responses to sounds lower than 5 kHz (Rose et al. 1967). In bats, orientation sounds are higher than 10 kHz, so that phase-locked responses may not play any role in echolocation. Since the orientation sound of Myotis is always frequency-modulated by about one octave, the ITD is coded by many neurones tuned at different frequencies within the sound. If the ITD plays an important role in echolocation, FM sounds are much better signals than CF sounds or noise bursts.

These experiments were supported by the National Science Foundation (Research Grant GB-13904-A1). We wish to thank Dr J. Simmons for his valuable discussion and Mr T. Gass for his assistance. We also thank Drs G. Neuweiler and H.-U. Schnitzler for permitting us to reproduce their data.