Neurophysiological Studies On Echolocation Systems in Awake Bats Producing CF-FM Orientation Sounds

Microchiropterans emit species-specific orientation sounds for echolocation. The bats which produce frequency-modulated (FM) signals are commonly called ‘FM bats’, and those which use signals consisting of constant-frequency (CF) and FM components are called ‘CF-FM bats’. Among CF-FM bats, there are a few subtypes in terms of signal variations during the searching phase of insect hunting by sonar. For instance, Pteronotus parnellii always emits a long CF sound ending with a short FM sound (Schnitzler, 1970). Pteronotus suapurensis produces a sound consisting of a very short CF component followed by a short FM component. Noctilio leporinus, on the other hand, independently changes the durations of the CF and FM components within a range of 0–15 msec, so that some orientation sounds appear to consist of only CF or FM components (Suthers, 1965). Such differences in CF components among species may indicate that the vocalization systems vary anatomically, and that the information obtained by the CF sound varies.

A CF sound is an ideal signal for Doppler measurement (e.g. the measurement of the relative velocity between a bat and a target) but it is poor for target ranging and for determining target characteristics. The CF sound is a good signal for echo-detection when the size of the target is similar to or larger than the wavelength of the signal and the integration period for sensation is the same as or longer than the duration of the CF component. The CF-FM bats presumably use the CF component for its inherent advantages. Rhinolophus ferrumequinum (Schnitzler, 1968; Simmons, 1974) and P. parnellii (Schnitzler, 1970) adjust the frequency of the transmitted CF signal to receive a Doppler-shifted echo at a certain preferred frequency. In these animals, the threshold curves of the auditory system show specialization for the reception of a sound at the preferred frequency (Ajrapetianz & Vasilyev, 1971; Grinnell, 1970; Neuweiler, 1970; Pollak, Henson & Novick, 1972). The CF sound is undoubtedly not merely a by-product of laryngeal activity. It is used for the measurement of target-velocity, for target-detection, and perhaps for compensation of errors in echo-analysis caused by the Doppler effect. The CF-FM bats which are known to alter vocalizations to receive Doppler-shifted echoes at a certain preferred frequency are well suited for behavioural and neurophysiological studies on the coupling between the auditory and vocalization systems.

In Myotis, P. parnellii, and P. suapurensis, it has been found that electrical stimulation of the dorsal part of the midbrain reticular formation elicits the emission of species-specific orientation sounds (Suga et al. 1973). The neural circuit for the emission of orientation sounds is apparently species-dependent. In the present paper, we describe (1) how electrically elicited sounds from different species of CF-FM bats are modified by different types of acoustic stimuli and (2) what kind of specialization exists in the auditory system and its relation to vocal behaviour.

Experimental subjects were ten Pteronotus parnellii (previously called Chilonycteris rubiginosa) and ten P. suapurensis from Panama and two Noctilio leporinus from Trinidad. The animals were etherized only during surgery. On P. suapurensis, a nail 1·8 cm long was mounted on the exposed skull with dental cement and was fixed on to a metal rod with a set screw, to immobilize the head. P. parnellii and N. leporinus are larger animals, so their heads were immobilized stereotaxically with three blunt needles mounted on micromanipulators and pressed against their skulls.

In order to stimulate the brain with electric pulses, a small hole was made in the skull covering the inferior colliculus. A pair of tungsten wire electrodes was inserted into the midbrain reticular formation through this hole. The tips of the electrodes Were less than 50 μm in diameter and were 0·2–0·5 nun apart. The electric stimuli used to evoke vocalization were short trains of monopolar electric pulses delivered O’5/sec. Each train consisted of 10 electric pulses which had 0·1 msec duration and 1·15 V amplitude (the estimated amount of current was 0·1–1·5 μA). The inter-pulse interval was 1·7 msec. Sounds emitted by the bat were picked up with a quarter-inch microphone (Briiel & Kjaer 4135) placed 8–10 cm anterior to the bat ‘s mouth and were recorded with a tape recorder having a frequency response of 50–150000 Hz at a tape speed of 30 in/sec (Ampex FR-100). These sounds were roughly analysed with a zero-crossing period meter for observation during the experiment. Later these were analysed at or of the original tape speed with a Kay sonagraph.

The cochlear microphonic (CM) and the summated activity of primary auditory neurones at the stimulus onset (Ni) were recorded with a tungsten wire electrode placed at the rim of the round window through the dorsoposterior part of the auditory bulla. The summated activity of the lateral lemniscal neurones (LL) at the stimulus onset was recorded with a tungsten-wire electrode which was placed in the nucleus of the lateral lemniscus through the inferior colliculus. Single unit activity of primary auditory neurons was recorded with a micropipette electrode inserted into the cochlear modiolus through the cochlear nucleus after aspirating the lateral part of the cerebellum. The recording of all the above electrical activities was performed without anaesthetic in a sound-proofed room, the inner wall of which was covered with fibreglass to reduce echoes. The electronic instruments used to generate acoustic stimuli were the same as those in previous experiments (Suga, 1968). For the measurement of threshold curves, pure tone bursts with an 0·5 msec rise-decay time and a 4 msec duration were repeatedly delivered at a rate of 0·5/sec, unless otherwise described. The amplitude of the tone bursts delivered from a condenser loudspeaker was measured with a quarter inch microphone (Brüel & Kjaer, 4135) placed at the bat ‘s ears, and it was expressed in dB SPL (sound pressure level referred to 0·0002 dyne/cm² r.m.s.)

When electric stimuli were applied to the dorsal part of the midbrain reticular formation or the lateral part of the central grey matter, near the boundary between the superior and inferior colliculi, each bat emitted sounds very similar to its speciesspecific orientation sounds as recorded in our laboratory. The latency of this vocal response ranged between 25 and 60 msec. Each sound emitted by P. parnellii for each electric stimulus consisted of a 10–30 msec CF component followed by a 2–3 msec FM component (Fig. 1A). The frequencies of the first, second and third harmonics in the CF component were about 31, 62, and 93 kHz, respectively. The second harmonic always predominated, while the first harmonic was sometimes faint. The third harmonic was always present. The second harmonic in the FM component swept downward from 62 to 50 kHz. Each sound emitted by P. suapurensis in response to electrical stimulation was 1–3 msec in duration and consisted of CF and FM components (Fig. 1B). The duration of the CF component emitted by P. suapurensis was very short compared to that of P. parnellii. The first, second and third harmonics were of approximately equal strength in the sounds of P. suapurensis. N. leporinus emitted either an FM sound or a compound CF-FM sound for each electric stimulus (Fig. 1, C₁ and C₂). The frequency of the CF component was about 53 kHz. It was not clear whether there are separate anatomical areas for the emission of purely FM sounds and for the emission of CF or CF-FM sounds. During the searching phase of hunting N. leporinus emits both CF and CF-FM sounds. It emits only FM sounds in the approach and terminal phases of the echolocation of prey (Suthers, 1965). The electric stimuli did not evoke purely CF sounds, although they often evoked a short CF sound followed by an FM sound, with an intervening silent period of about 4 msec (Fig. 2B).

When the electric stimuli (delivered at a rate of 0·5/sec) were attenuated, the amplitudes of electrically evoked sounds became small, and the latencies of these vocal responses became long and showed large fluctuations. Under these conditions, 2 msec tone pulses (at a repetition rate of 250/sec) or continuous tones were delivered in addition to electric stimuli. Then, the electrically evoked sounds became larger in amplitude and more regular in occurrence. The number of individual sounds emitted for each electric stimulus often increased. Hereafter, such a behavioural response is called ‘a vocal response to acoustic stimuli’. The vocal response to acoustic stimuli varied with the parameters of the acoustic stimuli. For instance, an increase in amplitude of the tone pulses caused an increase in the amplitude of the vocal response of P. parnellii (Fig. 2 A). The frequency and amplitude of the pure tone pulses were varied so that the tuning curve of the vocal response could be measured. The vocal response to acoustic stimulation was sharply tuned to 62–63 kHz (Fig. 3). As the frequency of the stimulus increased toward the best frequency, the threshold of the vocal response decreased at a rate of 1200 dB/octave. The threshold increased at a rate of 1300 dB/octave above the best frequency. The Q-10 dB value (the best frequency divided by the band-width of the tuning curve at 10 dB above the minimum threshold) was about 50. The minimum threshold was about 23 dB SPL. The threshold was higher than 65 dB SPL at frequencies below 50 and above 75 kHz. When a continuous pure tone was delivered instead of the tone pulses, it was found that the threshold of the vocal response to a continuous pure tone was 10–25 dB higher than that for the tone pulses, although the continuous pure tone had energy of 3 dB larger than the tone pulses had.

A response representing the summated activity of primary auditory neurones (N₁) appeared at both the onset and cessation of stimulation. N₁ at the onset is hereafter called ‘N₁-on’ and N₁ at the cessation is ‘N₁-off’. The threshold curve for N₁-on was quite different from that of N₁-off (Fig. 3). N₁-on was tuned to about 30 –45, 63–64 and 95–100 kHz, while N₁-off was sharply tuned at 24–27, 59–61 and 90–95 kHz. The threshold curve of the vocal response to acoustic stimulation was not similar to these neural tuning curves, but its single sharp notch did appear at 62 kHz, between the sharp tuning curves of N₁-on and N₁-off.

Instead of pure tone pulses, either upward or downward sweeping FM tone pulses were repetitively delivered as acoustic stimuli. The vocal response to acoustic stimulation in P. parnellii were significantly stronger to the FM sounds than to pure tone pulses except for those around 62 kHz (Fig. 3 C). In particular, downward sweeping FM sounds were most effective in evoking the vocal responses. The thresholds for downward sweeping FM sounds were 10–20 dB lower than those for upward sweeping FM sounds at the same frequencies and about 30 dB lower than those for pure tone pulses, except for frequencies immediately around 62 kHz. These data suggest that the vocalization system is coupled with the auditory system through particular types of auditory neurones specialized for CF or FM detection.

As already described, the N₁ threshold curve of P. suapurensis showed a broad notch tuned at about 50 kHz. There was no sharp notch as with P. parnellii (Fig. 4). The Q-10 dB value was roughly 1·7. The threshold of the acoustically elicted vocal response in this species of bat was high, being 70–80 dB SPL for pure tone pulses, 65–70 dB SPL for upward-sweeping FM sounds and 50–60 dB SPL for downwardsweeping FM sounds. These data indicate that the vocalization system of P. suapurensis is not strongly coupled through single neurones sensitive to a particular pure tone, but there is some coupling through neurones particularly sensitive to downward sweeping FM sounds.

N. leporinus often emits CF orientation sounds or sounds consisting of CF and FM components (Suthers, 1965). The threshold curves of N₁-on and of the vocal responses did not show any sharp notch, different from P. parnellii. In the vocal response of N. leporinus, not only the amplitude of the electrically elicited sound, but also its duration increased prominently. The increase in duration was mainly due to the lengthening of the CF component (Fig. 2B). The threshold of the vocal response was lower for an FM sound sweeping from 60 to 30 kHz than for pure tones or upward sweeping FM sounds over the same band.

The middle-ear muscles of mammals contract synchronously with the emission of sounds (Carmel & Starr, 1963; Henson, 1965; Salomon & Starr, 1963). In P. parnellii and P. suapurensis, cochlear microphonic potentials were recorded with an electrode implanted on the round window or in the cochlear modiolus. When a steady pure tone, as a testing stimulus, was delivered from a loudspeaker while keeping the position of the bat ‘s head and the loudspeaker constant, the cochlear microphonic (CM) was constant in amplitude, provided there were no gross body movements. However, the CM varied greatly when the animal moved. The middle-ear muscles apparently contracted together with gross body motion. When the bat vocalized, the CM evoked by an emitted orientation sound was consistently recorded in addition to the CM evoked by the steady pure tone. The CM evoked by the emitted sound could be eliminated with an electronic filter to a satisfactory extent. It was then found that the CM of the steady pure tone became smaller in synchrony with the vocalization, as first described by Henson (1965) in Tadarida.

In Fig. 5 A, vocalization was initiated by electrical stimulation to the dorsal part of the reticular formation in the midbrain of P. parnellii. When electric stimuli failed to evoke vocalization, the middle-ear muscles did not contract (A1). When they evoked vocalization, however, the middle-ear muscles synchronously contracted with the vocalization (A2–A4). The muscles started to contract 1 msec earlier or nearly at the same time as the onset of vocalization. When the vocalization was not strong, the muscle contraction reached a plateau after about 4 msec from its beginning (A2, A3). When the emitted sound was strong, however, the amount of contraction increased and reached a plateau only after 6–10 msec (A4). Since the CM evoked by the FM component of the emitted sound was not completely removed from the pure tone CM by the filter, it was not clear whether the muscle contraction further increased at the time of emission of the FM component. The middle-ear muscles more often appeared to begin to relax prior to the emission of the FM component, rather than to further contract. The middle-ear muscle contraction terminated within 1 msec after the termination of the first vocalization. When the bat produced more than one sound for each electric stimulus, the middle-ear muscle contraction lasted for more than 10 msec after the second and after the third sounds (A2–A4). The time-course of the contraction of the middle-ear muscles in P. parnellii was thus quite different from that observed in Tadarida brasiliensis, which emits FM signals for echolocation (Henson, 1965)-

In P. suapurensis, the middle-ear muscles started to contract about 4 msec prior to vocalization and relaxed in 3 msec after it. The maximum contraction always occurred in the initial part of the vocalization and the relaxation started to occur during its later part (Fig. 5 B). Thus, the time-course of the contraction of the middleear muscles of P. suapurensis was more similar to that of the FM bat, T. brasiliensis, than to its relative, P. parnellii. In the above experiments, the middle-ear muscles ionically contracted during the delivery of the continuous pure-tone, test signal and attenuated its transmission through the middle ear by 5–10 dB. Consequently, the amount of attenuation of self-stimulation in the natural condition could not be properly studied. The above experiments, however, indicated that the amount of attenuation of vocalizations was more than 10 dB in P. parnellii and more than 15 dB in P. suapurensis.

Since the loudest sounds which normally stimulate the ears are self-vocalized sounds, the animals have a neural circuit which reduces the amount of self-stimulation by contracting the middle-ear muscles. These muscles also contract as an acoustic reflex when intense sound is delivered. The shortest latency of this reflex in the bats was 6–7 msec measured in terms of the attenuation of the cochlear microphonic evoked by a tone burst (Fig. 5 C). These muscles would play only a minor role in attenuating the amount of stimulation provided by orientation sounds emitted by other bats flying nearby, because the duration of orientation sounds is short relative to the reflex time. In P. parnellii the threshold of the middle-ear muscle reflex in terms of the attenuation of CM was higher than 60 dB SPL, when measured 8 h after cessation of ether administration. The reflex was not observed for sounds higher than 90 kHz, even when the amplitude was increased up to 100 dB SPL. As shown in Fig. 6, the threshold of the reflex was high at 60 kHz but low near 56 and 62 kHz. The threshold curve for the middle-ear muscle reflex between 50 and 70 kHz was more like that of Npon than N₁-off.

In P. suapurensis, the threshold curve of the middle-ear-muscle reflex was broad with no abrupt peak and notch, different from P. parnellii (Fig. 7 A). The threshold was about 63 dB SPL for sounds between 23 and 60 kHz. The reflex was observed for sounds up to 120 kHz. This high threshold of the reflex may be due partly to an after-effect of anaesthetic and also to the ratio of the signal (CM) to noise. When the amplitude of a 100 msec tone burst was kept at 90 dB SPL and the amount of the CM attentuation by the middle-ear muscles was measured at its maximum point, the curve in Fig. 7B was obtained. The amount of attenuation was large at 55–60 kHz, the frequency of the strongest CF harmonic in the bat ‘s sounds. Above these frequencie attenuation quickly diminished; the efficiency of the reflex declined. This is expected, because generally the lower the frequency of sound, the greater the attenuation by the middle-ear muscles (Møller, 1965; Starr, 1969).

Since echolocating bats emit sounds and then listen to immediately occurring echoes, the responses of auditory neurones to one sound following another have been studied in anaesthetized non-vocalizing bats (Friend, Suga & Suthers, 1966; Grinnell, 1963; Grinnell & Hagiwara, 1972; Suga & Schlegel, 1973). In normal conditions, contraction of the middle-ear muscles (Henson, 1965) and modification of the activity of neurones at higher levels in the auditory system both occur synchronously with vocalization (Suga & Schlegel, 1972, 1973). The responses of auditory neurones to synthetic echoes following sound emission were studied in unanaesthetized, vocalizing bats of the species, P. parnellii.

The frequency of the CF component in the second harmonic of the orientation sound of P. parnellii is about 62 kHz, and the terminal FM component sweeps down from 62 to 45–50 kHz. When a pair of either pure tones at 62 kHz or FM sounds sweeping down from 63 to 45 kHz (0·5 msec duration) were used for measurement, the recovery of the lateral lemniscal evoked potential (LL) was very fast (Fig. 8 A). A 50% recovery occurred at about 2 msec interstimulus delay, regardless of whether the stimulus was a CF or FM sound. When a CF sound was repetitively delivered along with intermittent spontaneous vocalization by the bat, probably for echolocation, the recovery cycle was much longer than the above (Fig. 8 B). The bat ‘s emitted sounds were obviously different from the artificial stimulus in spectrum and duration. A 50% recovery of the LL response for a tone of 32 kHz occurred at about 25 msec delay from the beginning of vocalization, or at about 9 msec delay from the termination of the vocalization. The N₁ response for a 62 kHz sound, however, was not reduced more than 76% during vocalization. When the vocalization was evoked with electrical stimulation rather than spontaneously occurring, the recovery of the LL response was much slower than the recovery after the bat ‘s autonomous vocalization (Fig. 8C). This might be due to the emission of a stronger sound and/or some effect of the electric stimulus. As shown by the curves in Fig. 8 B and C, the LL response to a tone pulse delivered immediately before electrically initiated vocalization was smaller than the control value. Such a reduction of the LL response is very prominent before the vocalization initiated by electric stimuli, indicating that the electrical stimulation evoked an unnatural phenomenon.

As previously described, the threshold curve of N₁-on in P. parnellii showed a sharp notch at 63–64 kHz, and the curve of N₁-off showed a sharp notch at 59–61 kHz, a frequency region where N₁-on had a relatively high threshold (Fig. 6C and D). Both the on- and off-responses were most prominent for a tone burst with an abrupt (0·01 msec) rise-decay time. Apparently, the scatter of sound energy to adjacent frequencies at the sharp onset and cessation of the stimulus simultaneously excited many primary auditory neurons. The interval between N₁-on and N₁-off was identical to the duration of the stimulus, within an error range of 0·1 msec. In lengthening the rise-decay times from 0·01 to 5 msec, the off-response usually disappeared completely or weakened, while the on-response still remained clear. The off-response was more sensitive to a change in stimulus amplitude than was the on-response. The best frequency to evoke the off-response was 59·61 kHz regardless of the decay time.

Off-responses to pure tone bursts were also found in P. suapurensis. As opposed to p. parnellii, the threshold curve of the off-response for P. suapurensis did not show any sharp peak or notch. Lengthening the decay time shifted the off-response curve toward lower frequencies (Fig. 9). Off-responses in P. suapurensis appeared mainly to arise from physical properties of the acoustic stimulus (i.e. acoustic transients) and the ear. The threshold curve for on-responses did not shift to lower frequencies with the lengthening of rise time from 0·2 to 1·0 msec.

The properties of primary auditory neurons in P. parnellii should thus be studied in order to explore the neural basis for the sharp tuning curves of the N₁-on and N₁-off responses. Such a study was not performed here because this species was in short supply. For this reason only, P. suapurensis was used in order to check whether the N₁-off response was due to the off-discharges of single neurones rebounding from neural inhibition. In the cochlear nerve, the responses of 35 single neurones to 50 msec pure tone bursts with an 0·2 msec rise-decay time were studied. Neurones showed tonic on-responses to pure-tone stimuli within their excitatory area and post-excitatory inhibition after the stimulus, but off-discharges were not evident. For pure-tone stimuli at frequencies just outside the excitatory area of a given single neurone, about half of the neurones showed a clear off-response in addition to a phasic on-response. Background discharges, if any, were not affected at times between the on- and off-discharges. As far as our small samples are concerned in P. suapurensis, the off-response was not a rebound from neural inhibition but rather the response to mechanical transients of the stimulus and/or the ear.

When a 10 msec tone burst (0·2 msec rise-decay time) with frequency sweeping from 80 down to 40 kHz was delivered to P. parnellii, the on-response was prominent but the off-response was indistinct. The off-response to the 59–61 kHz component was not observed in addition to the N₁-on. In the orientation sound of P. parnellii, a CF component at about 62 kHz is always followed by an FM component sweeping from 62 down to 45–50 kHz. When the animal emitted such a sound or when a similar artificial sound was delivered from a loudspeaker, the N₁ and LL responses appeared at the onset and also at the termination of the stimulus. To what extent did the response to the terminal portion of the sound consist of the off-response to the CF component and to what extent was it the response to the FM component ?

P. parnellii, in flight, adjusts the frequency of the emitted CF component in an orientation sound in order to receive Doppler-shifted echoes from moving targets at a preferred frequency, about 62 kHz; the bat compensates for echo Doppler shifts by changing its transmitted frequency. For example, the frequency of a CF component could be 58 kHz in an emitted sound and 62 kHz in an echo (Schnitzler, 1970). Echoes almost always return soon enough to overlap the emitted sound, due to the length of the signal (8–30 msec). Accordingly, when an 15-msec, 58 kHz tone burst was delivered along with an 15-msec, 62 kHz tone burst, timed so that the onset of the 62 kHz burst was delayed by 2–5 msec from the onset of the 58 kHz burst, the LL-off response to the 58 kHz sound was greatly attenuated (Fig. 10A, a). A 3 msec FM sound was attached to the end of the 58 kHz sound without an amplitude change occurring at the boundary. The response to this CF-FM tone burst is shown in Fig. 10A, b₁. The response to the terminal portion of the CF-FM burst was smaller than the off-response to the 48 kHz tone burst. This response at the terminal portion was not appreciably reduced by simultaneous delivery of a 62 kHz sound (Fig. 10A, B2). These masking experiments indicate that the response to the terminal portion of the CF-FM burst primarily consisted of the response to the FM component and not the off-response to the CF burst. In Fig. 10 A, one should also notice that the LL-off response to the 62 kHz sound was greatly reduced by the 58 kHz tone burst and by the CF-FM sound.

The same masking experiments were repeated with recording of both N₁ and LL responses in order to clarify whether the masking of off-responses by the 62 kHz sound occurred at the periphery and/or in the midbrain. One recording electrode was placed near the auditory nerve in order to record N₁, and the other was inserted into the nucleus of the lateral lemniscus in order to record the LL response. These two electrodes were connected to a single differential amplifier. As shown in Fig. 10B, a, both the N₁-off and the LL-off responses to a 58 kHz tone burst were greatly attenuated by the 62 kHz tone. Both N₁ and LL responses to the terminal portion of the CF-FM sound were not appreciably reduced by the 62 kHz tone (Fig. 10B, b). Thus, the observed masking phenomenon was due to peripheral events. The N₁ and LL responses to the terminal portion of the CF-FM signal were mainly the response to the FM component.

Vocalization occurred when the electrical stimuli were applied to the central grey matter and/or the midbrain reticular formation. Vocalization also occurs, however, in association with emotional distress. Consequently, there is a problem as to whether the electrical stimuli directly excite a part of the vocalization system or instead evoke an emotional change, which in turn initiates vocalization. In our experiments, (1) the shortest latency of the vocal response was 25 msec, (2) the vocal response occurred synchronously with each electrical stimulus and immediately stopped when the stimulus stopped, (3) no gross body movements were associated with this vocalization when the electrodes were placed at a low threshold area, and (4) the properties of electrically evoked sounds were similar to those of the species-specific orientation sounds. The vocalization was appropriately different from species to species; the neural circuitry in the vocalization system in the midbrain or the larynx apparently differs systematically from species to species.

The change in the electrically evoked vocalization in the presence of acoustic stimuli indicates that the bat may have attempted to echolocate with the electrically evoked sounds. When the acoustic stimuli were delivered, echoes from objects in front of the bat (such as a loudspeaker and a steel support rod) might be masked. The animal could increase the amplitude and perhaps the number of outgoing sounds in order to detect echoes, overcoming the masking tone. In this situation, the vocal response to a masking signal should theoretically be related to the spectrum of the masking sound, to the information-bearing elements in the emitted CF-FM signals, and also to properties of the auditory neurons used for echo analysis. The observed thresholds for vocal responses to acoustic stimuli might be interpreted in the following way: in the CF component of the orientation sound of P. parnellii, the 62 kHz second harmonic is the principal information-bearing element, and the first and third harmonics in the CF component are not so important, because the threshold of the vocal response was low at 62 kHz only. Thresholds for FM sounds sweeping down from high to low frequencies were lower than thresholds for upward sweeping FM sounds and for pure tones, except for pure tones at about 62 kHz. The FM components in the first, second, and third harmonics of the orientation sound all appear to be significant to the bat.

The very sharp tuning curve for the vocal response at about 62 kHz and the low threshold for the downward sweeping FM sounds probably indicate that acoustic control of the vocal response was mediated by particular types of auditory neurones; one type of auditory neurone with a very narrow tuning curve at about 62 kHz and the other type being more sensitive to FM sounds than to pure tones. In bats of the genus Myotis, which emits only FM sounds for echolocation, the narrowest tuning curve obtained in the inferior colliculus showed a Q-10 dB value of 120, a low frequency slope of about 2000 dB/octave and a high frequency slope of about 1300 dB/octave (Suga, 1964). P. parnellii apparently can utilize the activity of such a neurone sharply tuned at a certain frequency and can ignore the activity of all other neurones tuned at other frequencies. In anaesthetized bats of the genus Myotis, FM-specialized neurons were found to be about 3% of the total in the inferior colliculus (Suga, 1965 a, 1968) and not more than 14% of the total in the auditory cortex (Suga, 1965 b). In P. parnellii, the responses of single neurones to acoustic stimuli have not yet been studied, but there is little doubt that the bat has FM-specialized neurones, because the animal emits FM signals for echolocation and because such neurones have been found not only in bats, but also in cats (Whitfield & Evans, 1965; Watanabe & Ohgushi, 1968). The low threshold vocal responses to FM sounds are probably mediated by FM-specialized neurons.

In P. suapurensis, the threshold of the vocal response was significantly higher for pure tones than for FM sounds, regardless of frequency. The relative significance or function to P. suapurensis of the CF components of the sounds for echolocation may be different from that to P. parnellii. On the other hand, the FM components of the orientation sounds are probably equally essential, considering the low thresholds for vocal responses to FM signals in both species. N. leporinus often emits an orientation sound consisting of CF and FM components. Its vocal responses were similar to those of P. suapurensis. Vocal responses to acoustic stimuli in these species of bats also indicate that the vocalization system is coupled with the auditory system through particular types of auditory neurons. The coupling may depend upon the information-bearing elements in the signals with which the animals are most concerned at a given moment.

In some of the bats producing CF-FM signals for echolocation, it has been found that their auditory system is sharply tuned at a frequency of a predominant CF component in the signal (Ajrapetianz & Vasilyev, 1971; Grinnell, 1970; Grinnell & Hagiwara, 1972; Neuweiler, 1970; Pollak et al. 1972). Pollak et al. (1972) found that the CM threshold curve of unanaesthetized P. parnellii was so sharply tuned at 62 kHz that a Q-10 dB value was 310. In our data obtained from unanaesthetized P. parnellii, 3–15 h after stopping ether administration, the CM threshold curve was similar to theirs. The presence of a sharply tuned notch at 61 kHz was repeatedly confirmed. The Q-10 dB value was, however, about 50. For a 61 kHz, 4 msec tone burst with a 0·2 msec rise-decay time, the envelope of the CM was quite different from the stimulus envelope, showing a slow increase in amplitude at the onset and a slow decrease after the cessation. Such a peculiar CM response was not observed for 50 and 70 kHz tone bursts. A similar CM response to the 61 kHz tone burst which was Viso observed by Pollak and his co-workers (personal communication) indicates that the ear of P. parnellii has some mechanical resonating element(s) sharply tuned at 61 kHz.

In P. parnellii, the threshold curves of N₁-on and N₁-off both were different from the CM threshold curve. N₁-on was sharply tuned at 63–64 kHz and broadly tuned at 30–40 kHz, but it was insensitive to 59–60 kHz sound. On the other hand, N₁-off was sharply tuned at 59–61 kHz. Some of the differences in the threshold curves may be explained by the position of the recording electrode, but the differences between N₁-on and N₁-off threshold curves are not so explained, because they consistently existed regardless of the position of the recording electrode. In Rhinolophus, N₁-on is sharply tuned at 83·3 kHz, but N₁-off does so at 81·5 kHz, to which N₁-on is insensitive (Neuweiler, 1970; Neuweiler, Schuller & Schnitzler, 1971). Comparative studies on several species of neotropical bats indicated that on-and off-responses are tuned at the same frequency in Saccopteryx (Grinnell, 1970), but off-responses are tuned for frequencies higher than those for which on-responses are tuned in Hipposideros and Aselliscus (Grinnell & Hagiwara, 1972). The off-responses are not related to the insensitivity of on-responses in these bats.

Off-responses to tone bursts have been recorded from the peripheral auditory system of certain species of bats (Grinnell, 1970; Grinnell & Hagiwara, 1972; Neuweiler, 1970). In order to discuss the functional role of the off-responses, we first wanted to study the properties of primary auditory neurones, but we could get only preliminary data because of the number of bats available. It may be none the less important to discuss the implications of our data on single unit activity and summated evoked potentials and to speculate on the origin of off-response.

Off-discharges as rebounds from inhibition caused by tonal stimuli have been found in the cochlear nucleus of Myotis lucifugus (Suga, 1964). Off-discharges have often been found at higher auditory nuclei in various types of animals (e.g. in the bat, Chilonycteris rubignosus, Grinnell, 1970; in monkeys, Katsuki, Suga & Kanno, 1962; in cats, Whitfield & Evans, 1965). Off-discharges observed in the brain are excluded in the following discussion, because these are easily explained by a rebound from neural inhibition.

In bats (Frishkoff, 1964), cats (Sachs & Kiang, 1968), and monkeys (Nomoto, Suga & Katsuki, 1964), primary auditory neurones commonly show ‘two-tone suppression (or inhibition)’ which is still observed after severing the efferent nerve fibres. Two-tone suppression appears to be due to non-linearity in the cochlea (Engebretzen & Eldredge, 1968; Pfeiffer, 1970). At the termination of two-tone suppression, primary auditory neurons show off-discharges as a rebound from the two-tone suppression (Sachs & Kiang, 1968; Arthur, Pfeiffer & Suga, 1971). The off-discharges observed in the primary auditory neurones of P. suapurensis were not associated with such suppression. According to the present knowledge, it is unlikely that hair cells located in a particular portion of the basilar membrane release inhibitory substances that cause hyperpolarization of afferent fibres during acoustic stimulation and then depolarization as a rebound from hyperpolarization at the cessation of the stimulus. The off-responses observed in our experiments are clearly not related to the activity of the olivo-cochlear bundle, because the latency of the N₁-off response was the same (within 0·1 msec) as the latency of the N₁-on response. Furthermore, the shape of the threshold curves of N₁-on and N₁-off cannot be explained by this. The second possibility is that the off-response is the response to an acoustic transient at the cessation of the stimulus. The number of primary auditory neurons studied here was small, but these data suggest that the N₁-off response was neither a rebound from neural inhibition nor from two-tone suppression. In P. suapurensis, N₁-off responses appeared to be due to a mechanical transient in a receptor and/or a stimulus.

These experiments were supported by the National Science Foundation (Research Grant GB-13904-A1 and GB-40018).