Psychophysical frequency modulation thresholds in a FM-bat, Tadarida brasiliensis

Psychophysical frequency modulation thresholds in a FM-bat, Tadarida brasiliensis

128 HEARES 01913 Psychophysical frequency modulation thresholds in a FM-bat, ~~d~~~d~b~u~~~~~n~~s Eva Bartsch and Sabine Schmidt Zoologisches Instit...

1MB Sizes 1 Downloads 60 Views

128

HEARES 01913

Psychophysical frequency modulation thresholds in a FM-bat, ~~d~~~d~b~u~~~~~n~~s Eva Bartsch and Sabine Schmidt Zoologisches Institut der Universiriir Miinchen, Miinchen, FRG

(Received 15 May 1992; Revision received 5 November 1992; Accepted 16 January 1993)

Echolocating bats hunting flying insects discriminate complex temporal patterns of acoustic stimuli. For bats using frequency modulated sonar calls (FM bats), there are no behavioral data on the perception of sinusoidally frequency modulated (SFM) stimuli. Discrimination performance for SFM stimuli of varying modulation depth was measured in 4 Tadarida brasiliemis in a two-alternative, forced choice procedure, A center frequency of 40 kHz was modulated with rates between 10 and 2000 Hz. It was found that discrimination performance improved from a mean threshold modulation depth of 3.05 kHz at a modulation rate of 2000 Hz to 1.58 kHz at a modulation rate of 10 Hz. Psychoa~ousti~al moduiation depth thresholds of i? brasilimsis are thus distindtly larger than those observed in bat species emitting eonstant frequency (CF) components followed by an FM-sweep, in active echolocation experiments. The modulation thresholds of 1 brasiliensis are discussed in connection with the ability of bats to discriminate insect wingbeats. A comparison between non-echolocating mammals and the FM bat T. brasiliensis shows that the ability to echolocate is not reflected in the modulation thresholds. Frequency modulation thresholds; Bat; Psychoph~i~;

~lutterjng target discrimination

Introduction Echolocating bats use acoustical cues as a primary source of info~ation about their environment. The emitted sonar calls return as complex echoes differing in spectral content and temporal structure, depending on the reflection properties of the objects in the acoustic path. The bats can discriminate between echoes from objects of different shape and surface structure; they determine object location in space, and motions such as wing-beats of insects or the approach or retreat of a target (cf. Schnitzler and Henson, 19801. Important information on the dynamic aspects of bat sonar may be obtained from behavioral studies in which frequency modulated signals have to be discriminated by the animals. The ability of echolocating bats to recognize oscillating targets mimicking insect wing-beats has been studied in fluttering target experiments. Echoes from fluttering targets are characterized by complicated frequency and amplitude modulations and the periodic occurrence of ‘glints’, i.e., sudden amplitude peaks and spectral broadenings (cf. Schnitzler et al., 1983). Most studies have concentrated on a small group of highly specialized bat species (CF-FM bats) that emit

Cmrespmtdence to: Sabine Schmidt, Zooiogisches Institut der UniversitIt Miinchen, Luisenstrape 14, D-8000 Miinchen 2, FRG. Fax: (49) 89 5902-450.

signals with a constant frequency component (CF) terminated by a brief downwards frequency modulated (FM) sweep. In CF-FM bats, the ability to discriminate differences in glint repetition rate is correiated to the duration of the CF component. Bats emitting long (> 10 ms) CF components are more sensitive to differences in glint repetition rate (3-11 Hz, Rhinolophus ferrumequinum, v.d. Emde and Menne, 1989; Rhinolophus rouxi, Roverud et al., 1991) than bats emitting short (< 10 ms) CF components (H~pp~s~eros lankudiva: 8-11 Hz, Roverud et al., 1991; Hiplpoideros speoris: about 15-67 Hz, v.d. Emde and Schnitzler, 1986). Similarly, bats emitting long CF components (Rh. ferrumequinum, Schnitzler and Flieger, 1983) performed better than those with short CF ~m~nents CH. l~~kudi~u, v.d. Emde and Schnitzler, 1986) in a different type of echoiocation experiment, where the perception of sinusoidally frequency modulated echoes was studied. The sensitivity to fluttering targets has thus been related to the use of the CF component (cf. Neuweiler, 1990). The auditory system of CF-FM bats shows an overrepresentation of hair cells tuned to a narrow frequency range around the individual CF (acoustic fovea; Neuweiler, 1990), and of neurons in the ascending auditory pathways (Schuller and Pollak, 1979). These neurons, in addition, are extremely sharply tuned. Doppler shifts in the CF component of the echoes caused by the flight velocity of the bat are compensated for by a correspondingly lower frequency of the

I29

emitted calls (Schnitzler, 1968), i.e., the CF component of the echoes is kept at a constant reference frequency within the acoustic fovea. The CF component of the echoes can thus be used as a carrier which is modulated by target movements. The specializations of the inner ear and the ascending auditory pathways allow a remarkable degree of temporal and spectral resolution in a narrow frequency band around the CF and may thus account for the excellent detection and classification of flying insects by CF-FM bats (Goldman and Henson, 1977; Neuweiler, 1990; v.d. Emde and Schnitzler, 1990). FM bats do not show comparable specializations. Nonetheless, the glint repetition rate discrimination performance of echolocating FM bats (Pipistrellus stenoprerus: 9 Hz, Sum and Menne, 1988; Eptesicus fuscus: 16 Hz, Roverud et al., 1991) equals that of CF-FM bats. However, it should be noted that the above experiments are not selective concerning the echo processing mechanism, as differences in modulation rate were always accompanied by differences in modulation depth (cf. Nitsche, 1987). Thus temporal as well as spectral cues may have been used for the discrimination, and it cannot be decided how the echoes were processed by the auditory system of the bats. In particular, it is not clear whether the similar performance of CF-FM and FM bats can be attributed to the same echo processing mechanisms. The present study centers on the question of how well FM bats can detect sinusoidal frequency modulations of a pure tone carrier. The bats are faced with a more artificial situation here, in that they do not normally use a pure tone signal in glint rate discrimination experiments, and in that they have to respond to acoustic stimuli without making use of their sonar. This approach, however, has been adopted since the properties of these stimuli are clearly described, i.e., the bats must base their decisions on differences in modulation depth. The threshold modulation depth at different modulation frequencies is an important measure of the dynamic frequency resolution of the auditory system, and is also relevant for a physiologically founded discussion of fluttering target perception with broadband sonar. Here, the performance of T. brusiliensis in discriminating 40 kHz pure tone from sinusoidally frequency modulated (SFM) signals about a center frequency of 40 kHz is determined as a function of modulation frequency. Materials

and Methods

Animals T. brasiliensis is a fast flying bat with highly variable sonar calls of durations between 1 ms and 10 ms adapted to the pursuit of airborne prey in open spaces;

calls of constant frequency of about 50 kHz are reported from field observations of animals in clutter-free environments (Simmons et al., 1978). In cluttered surroundings or when tracking a target, frequency modulated (FM) signals with 2-3 harmonics are used. Their fundamental typically sweeps downwards from about 50 to 28 kHz. When leaving their roosts and in the search phase, the duration of the sweeps may increase, and their modulation may be more shallow. In addition, the FM component can be preceeded or followed by components of nearly constant frequency (authors’ observations and of Simmons et al., 1978). This wide repertoire of sonar calls is well matched to a broadly tuned audiogram (Henson, 1970; Schmidt et al., 1990); both are typical for many bat species using FM signals for orientation (Neuweiler, 1984). Four naive T. brasiliensis (Microchiroptera, Molossidae; 2 females, 2 males) from Austin, TX were used. The bats were kept in a weakly illuminated (15 W bulb) flight room at constant temperature (28-30°C) with free access to water. The animals were tested on six days a week and fed exclusively on a diet of mealworms during the experimental sessions. Since two individuals died during the experiments, complete threshold data are presented from only two bats. Stimuli The bats were trained to discriminate pure tone pulses from sinusoidally frequency modulated @FM) pulses in a two-alternative, forced choice procedure. The rewarded stimulus was a sequence of 40 kHz pure tone pulses. The unrewarded stimulus was a sequence of SFM pulses of the same mid frequency. Both modulation depth and modulation frequency were varied. Modulation frequency, i.e., the rate of occurrence of the modulations, was set to 10, 20, 50, 100, 200, 500, 1000 and 2000 Hz. All pulses had a trapezoidally shaped envelope; their duration was 500 ms, with rise/fall times of 50 ms. Each pulse was followed by a pause of 400 ms. The sound pressure level of the stimuli was set to 60 dB SPL, corresponding to a sensation level of about 40 dB (Schmidt et al., 1990). The pure tone carrier frequency was produced by a Wavetek 136 function generator. Controlled by an electronic switch, its VCG (voltage control gate) input was either grounded (to emit the carrier frequency) or connected to another function generator (Wavetek 186) set to the respective modulation frequency, thus creating a SFM signal. Carrier and modulation frequencies were monitored on Black Star Meteor 100 frequency counters. An attenuator (Elementa; minimum stepwidth 0.1 dB) at the output of the modulation frequency generator was used to adjust the amplitude at the VCG input in order to control the modulation depth of the SFM signal.

130

In this paper, modulation depth (A fT) is given as the maximum frequency deviation of the SFM stimulus from its mid frequency. The presented modulation depths varied between t_ 0.45 kHz and ) 7.6 kHz. Pulse sequences were generated by feeding the pure tone or the SFM signal, respectively, to an envelope shaper (on-off ratio exceeding 60 dB) triggered by a pulse generator (devices custom-made). Then, pure tone and SFM stimuli were automatically distributed by an electronic switch to two output channels. In each channel, an attenuator (Elemental, used to adjust the stimulus level (stepwidth 1 dB), was followed by a power amplifier (Toellner 7606) driving a custom-made ultrasonic speaker (Polarization voltage 200 V DC). For a control experiment, lOO%, 75% and 50% sinusoidally amplitude modulated (SAM) pulses (carrier frequency 40 kHz) of a modulation frequency of 200 Hz were generated using the VCA (voltage controlled amplitude) input of the Wavetek 136 function generator. Calibrations Modulation depth was calibrated at the input of the power amplifiers using a custom-made frequency to voltage converter. Calculated modulation depths (cf. Long and Clark, 1983) corresponded to the measured values at all relevant attenuator settings. The sound pressure level of the stimuli was adjusted before each experimental session using a B&K 4135 l/4 inch microphone mounted at the starting position of the bats (see below) and a B&K 2610 measuring amplifier. The frequency response of the ultrasonic speakers was regularly checked (signal analyzer HP 3561A); at 40 kHz, both speakers differed by less than 0.5 dB; within * 10 kHz around 40 kHz, variations of less than rt 3 dB were measured. An analysis of the frequency content of the stimuli showed that the spectral splatter in 1 kHz bands around the signals was more than 40 dB lower than signal level. Harmonic distortions were also more than 40 dB down. Apparatus The experiments were performed in dim light (15 W bulb) within a cage (310 cm X 220 cm X 190 cm) made of mosquito net (mesh 2 mm) in a room with a wall coating of foam rubber wedges. Room temperature was about 28°C. The Y-shaped experimental platform, mounted on a rod 90 cm above the floor, consisted of a central starting platform (SP) connected to a slightly lower (0.5 cm) right and left response platform (see Fig. 1). The two ultrasonic speakers were mounted at a distance of 140 cm, 110 cm above the floor, on a table covered with sound absorbent material, each 22” left and right to the axis of the starting platform. The observer was positioned behind the starting platform from where she

Fig. 1. Experimental setup (drawing not to scale): The cross (+) gives the head position of the bat on the starting platform (SP) when deciding between stimuli alternately presented at the right and left loudspeakers (LS). All distances are given in cm (for further details, see text).

could run the equipment using a remote control. Stimuli were visually monitored by the observer using an Hitachi VC6025 oscilloscope. Target presentation To start an experimental trial, the animal moved from the observer’s hand to the starting platform. There, the bat had to compare the stimuli presented alternately at both loudspeakers until it made its decision. In a trial, the rewarded stimulus was presented at one of the speakers. Successfully trained bats followed the presentation of the stimuli by turning their head towards the active speaker. These animals did not echolocate in this situation, as was routinely checked with a bat detector (Petterson D 90). The bat indicated its decision by completely crawling down onto the corresponding response platform. In case of a correct choice, a piece of mealworm was given with forceps. At the end of each trial, the bat climbed back into the observer’s hand. In one session 20-65 trials could be run. The channel for the presentation of the rewarded stimulus was selected according to a pseudorandom scheme in which a bat using a position or alternation strategy would score 50% correct. The rewarded stimulus was not presented more than four times in consecu-

131

tive trials on the same channel in order to prevent the formation of side preferences by the bats. Within one experimental session, up to seven different modulation depths were presented. Modulation frequency was kept constant within a session.

At least 30 decisions of a bat collected in consecutive experimental sessions were combined in a data point (percentage of correct choices) for each modulation depth at each modulation frequency. For some modulation frequencies, two independent experimental series were obtained. Data points at a given modufation frequency were considered as a function of the logarithm of modulation depth A f, (cf. Fig. 2a-d). The data points were approximated by a Gaussian error function (cf. Finney, 1971) between chance level (50%. lower asymptote) and the best measured pcrformance for each individual (bats 1 and 2: 90%; bat 3:

a 100

96%; bat 4: 94%; upper asymptote), i.e., the upper asymptote was set to the expectation vatues for simple FM versus pure tone discrimination tasks. The quality of the approximation was determined by a chi2-test (Sachs, 1978). The threshold was defined at the point of maximal steepness of the fit curves (bats 1 and 2: 70%; bat 3: 73%; bat 4: 72%). This point represents the most probabfe threshold value. The standard error of the threshold values was calculated from the quadratic deviation of the data points from the fit. Then, threshold modulation depth A f was considered as a function of modulation frequency. The mean performance of the four bats was calculated from all threshold vahres at a given modulation frequency. Threshold values were weighted by the number of trials run for threshold determination. In addition, the standard error of these mean thresholds was calculated based on the standard error of the individual thresholds. Finally, the mean thresholds were approximated by linear regression.

b

Bat 1

7

-3

0

3

6

9

It

15

16

8at

401 -3

rel.att. idBl At,

IkHzl

2

.

I

I

0

3

6.8

d9

I



6

9

12

15

3.0

2.3

1.4

11

12

15

i

18

rel.att. [de1 AfT

[kHzl

d

; 70 I $ 60 aQ 50

40-l -3

,

/

0

3

73

41

40’

, 6 2.9

9 20

12

15

18

rel.att.

1.4

1.0

0 7

*f T

[dBl

[ktfzl

-3

0

7.6

3

53

6 3.6

9 2.5

$6

rel.att.

1.6 *f

T

idBj

[kHzl

Fig. 2. Typical examples of Gaussian fit curves from the four bats (a) - (d). The upper abscissa gives the relative attenuation (rel.att. [dB]) at the VCG input of the function generator generating the frequency modulated signal. Relative attenuation was varied in steps of 3 dB. On the lower abscissa, the corresponding moduiation depth fd f Ik&]) is shown. The data points (triangles) are plotted as a function of the relative attenuation and approximated by a Gaussian error function (continuous line). The dotted line marks the threshold criterion. The error bar represents the standard error of the threshold value calculated from the residual deviations of the data points from the fit curve. Please note that the modulation depth A f, decreases logarithmically with increasing attenuation; therefore, when measured in kMz, the standard error of the threshold modulation depth is not symmetrical about the threshold value.

132

the data points, while 4 fit curves differ i’rom their respective data points with an error probability I’ c 5c.~ (chi’-test, Sachs, 1978). These thresholds arc marked by open symbols in Fig. 3a and b). At all testcci modulation frequencies, the performance of the bats reached chance level at the lowest modulation depths presented (error probability P < 5%; Keller, 1967). The standard errors of the threshold values which vary between 0.05 kHz and 3.18 kHz are given in Table I.

Results Modulation thresholds

Threshold modulation depths A f for the discrimination of SFM from pure tone stimuli were obtained at different modulation frequencies from four bats (see Fig. 3a-d and Table I>. Altogether, 30 threshold values were determined; 26 of the Gaussian approximations sufficiently represent a

. A

b

7 Bat

4-

1

.

Bat 2

4-

a/ .

3-z

3-

.

2 2z

.

.

.

z. E 2a'

l-

l-

ti 0-l

0-I

I

10

5

ta

50

modulation

100

200

frequency

500

10002000

5

, 10

[Hz1

, 20

r 50

modulation

100

200

frequency

500

10002000

[Hz]

d

C

Bat 3

4-

4

Bat 4

37

v.

z

‘I

26

l-

-,

, 5

10

20

50

modulation

e

/,

100

200

frequency

500

I/

10002000

[Hz1

“i

10

20

50

modulation

100

200

frequency

500

10002000

[Hz1

4-

05 5

10

20

50

modulation

100

200

frequency

500

10002000

[Hz]

Fig. 3. Performance of four bats in discriminating 40-kHz pure tone pulses from sinusoidally frequency modulated (SFM) stimuli around the same mid frequency: In (a) - (d) threshold modulation depths (A f [kHz]) are plotted as a function of modulation frequency [Hz]. Threshold values obtained in the first experimental series are represented by up triangles, values obtained in the second series by down triangles. Open symbols represent thresholds for which the datapoints were not sufficiently approximated by the fit curve (chi’-test; P < 5%). The solid line connects the weighted mean threshold values for each animal. in (e) weighted mean threshold modulation depths from all animals (circles) are given with the standard errors of the means (error bars). In addition, the linear regression line on the mean thresholds is given with its 95% confidence interval (cf. ziifel, 1988).

133

An in~uence of the modulation frequency on these standard errors was not found. Bats 1 and 2 for which complete threshold data were obtained tended to be more sensitive to SFM stimuli at the low modulation frequencies (Fig. 3a and b). For bats 3 and 4, which died in the course of the experiments, thresholds could be determined at only three modulation frequencies, rendering it impossible to establish a comparable trend (Fig. 3c and d). Threshold modulation depths at a given modulation frequency vary widely between different experimental series in one animal (threshold differences between 0.11 kHz and 2.01 kHz), and among the bats (threshold differences between 0.01 kHz and 2.12 kHz). The large scattering of the thresholds in independent experimental series cannot be explained by learning, because threshold values are not systematically lower in later series (see distribution of up and down triangles in Fig. 3a-d). Since threshold differences from two measurements within an animal are in the same range as those

TABLE 1 Threshold modulation depths (A f [kHz]f at all tested modulation frequencies (fmod [Hz]) and their standard errors to higher (SE+) and lower (SE - ) values for bats 1-4 * Bat I 1 I 12 1

12

I I 12 1 12 1 I, 2 2 2

2 2 2, 2 2 2 3 3 -L 3 3 3 -z 4 4 4

fmod IHal

to 20

so 50 100 100

200 500 500

1000 1000 2oou 2000 IO 20 SO 100 200 200 500 1000 2 000 50 SO 100 200 200 SO 100 200

N

df [kHzl

SE+ [kHzl

SE&HZ1

211 313 212 129 261 95 443 183 207 183 259 52 232 212 233 237 199 394 266 209 178 175 s2 180 236 330 228 90 197 263

1.78 0.60

1.05 0.58 0.82 1.38 1.20 1.06 0.37 0.85 1.85 3.15 1.07 1.06 0.80 0.56 0.75 3.18 0.40 0.85 0.18 0.85 0.60 0.87 0.08 0.65 0.78 0.60 1.12 2.18 1.07 0.50

0.68 0.18 0.50 0.74 0.60 0.74 0.23 0.70 1.15 1.85 0.73 0.16 0.40 0.44 0.45 0.75 0.33 0.70 0.17 0.20 0.57 0.13 0.05 0.45 0.20 0.45 0.70 1.27 0.78 0.40

1.53 1.64 2.10 2.64 1.68 2.95 1.60 3.65 1.78 2.04 4.05

1.39 1.35 2.3s 2.03 2.70 2.45 2.86 3.90 2.03 1.58

1.75 1.05 1.55

2.28 2.97 2.83 2.90

* N gives the number of trials on which threshold determinations were based: 2 characterizes the second experimental series.

TABLE Ii Mean modulation thresholds (A f [kHz]) and Weber fractions (WR) * determined from pooled data of four T. hrusiliensis as a function of modulation frequency (fmod [Hz]) fmod [Hz]

N

J f [kHz]

WR

10 20 50 100 200 500 1000 2000

432 546 930 988 1864 599 620 45Y

1.58

0.040

0.92 1.94 2.03 2.20 2.45 2.94 3.05

0.023 0.049 0.05 1 0.055 0.061 0.074 0.076

* The Weber fraction is defined as the ratio of the threshold modulation depth to the carrier frequency; N gives the total number of trials on which threshold determinations were based.

among the different bats. and since these differences can be accounted for by the standard errors of the individual threshold determinations (Table I), all obtained thresholds at a modulation frequency were pooled for further analysis. The resultant mean threshold modulation depths increase from 1.58 kHz at a modulation frequency of 10 Hz to 3.05 kHz at a modulation frequency of 2000 Hz (see Fig. 3e and Table II). An approximation by linear regression for threshold modulation depth on the logarithm of the modulation frequency (correlation coefficient r = 0.926) reveals a conspicuous increase in threshold vaiues. In particular, the standard error intervals of mean threshold modulation depths at 10 Hz and 20 Hz modulation frequency do not overlap with those at 1000 Hz and 2000 Hz modulation frequency. In addition, the 95% confidence interval around the regression line has been calculated (see Fig. 3e). Since this interval contains no straight line parallel to the modulation frequency axis, the increase of threshold modulation depth with modulation frequency is significant with an error probability of 5%. The mean threshold modulation depth at the modulation frequency of 20 Hz differs considerably from the regression line. This deviation, however, lies within the standard error and is therefore not significant. Corresponding to the threshold modulation depths, the Weber fractions (WR) increase with modulation frequency from 0.04 at a modulation frequency of 10 Hz to 0.076 at a modulation frequency of 2000 Hz. The lowest WR is 0.023 at the modulation frequency of 20 Hz. Controt experiments After the experiments with SFM stimuli had been completed, SAM stimuli of a modulation frequency of 200 Hz were presented to bat 1 in four consecutive experimental sessions. Discrimination performance was

134

62% correct (number of trials N = 871 for 100% SAM, 61% correct (N = 36) for 75% SAM, and 52% correct (N = 21) for 50% SAM. This performance is not significantly different from random choice (error probability P= I%, cf. Keller, 1967). This shows that bat 1 was unable to discriminate between SAM and pure tone stimuli without additional training. It can be concluded from these data that the minor amplitude modulations caused by nonlinearities of the speakers did not provide additional cues for the discrimination of SFM from pure tone stimuli.

Discussion In the following, we compare the threshold modulation depths for SFM stimuli obtained from T. brusiliensis to modulation and frequency discrimination thresholds of non-echolocating vertebrates and to the frequency discrimination thresholds of the CF-FM bat Rfa. ~er~~equ~~u~. Then, the present data are discussed in the context of fluttering target discrimination in CF-FM and FM bats, and possible mechanisms for glint rate discrimination in FM bats are considered in view of our results. Similar to our results in T. brasi~ie~sis, the threshold modulation depths in man increase as a function of modulation frequency for modulation frequencies above about 5 Hz. In man, thresholds above a critical modulation frequency decrease again, which is explained by the perception of side bands of the SFM stimuli outside the critical band around the center frequency (Zwicker and Fastl, 1990). Our results do not exclude the existence of an analogous critical modulation frequency in T. bras&en&s. Indeed, the critical modulation frequency in man increases with the width of the critical band (70 Hz for a critica bandwidth of 160 Hz at a mid frequency of 1 kHz, and 300 Hz for that of about 1800 Hz at a mid frequency 8 kHz; Zwicker and Fastl, 1990). In T. brusiliensis, the critical bandwidth around 40 kHz amounts to about 16 kHz (Schmidt et al., 19901, which may be sufficiently large for the critica moduiation frequency to lie well above 2000 Hz *. In experiments in which pure tone stimuli have to be discriminated from SFM stimuli of the same center frequency (symmetrical SFM), threshold modulation

* The prominent decrease in threshold modulation depth of nearly 2 kHz from 1000 Hz to 2000 Hz modulation frequency in bat 2 does not indicate that the critical modulation frequency has been passed, because the threshold value at the modulation frequency of 2000 Hz results from a poor fit curve (cf. Fig. 3b). Besides, threshold differences of up to 2 kHz at the same modulation frequency were found in independent experimental series (cf. bat 1, Fig. 3a).

depths are usually given as 2 d f values. i.e.. as the difference between the maximal and minimal I’& quency of the modulated stimuli. In man and starling, the 2 d f threshold modulation depths at low modulntion frequencies are 2-3 times larger than the just noticeable frequency differences between pure tones Gchorer, 1989; Zwicker and Fast], 1990: Langemann, 1991). On the other hand, in an experiment in which the mid frequency of the SFM stinlulus was set to a higher frequency than the pure tone reference, so that the lowest frequency of the SFM stimulus was identical to the frequency of the pure tone (asymmetrical SFM), the 2 A f threshold values are by a factor of 2 smaller than those found for symmetrical SFM stimuli in both man and starling (Langemann, 1991). These differences in threshold modulation depth may stem from the fact that the auditory system compares the actual deviations (k-d f) from the frequency of the pure tone reference rather than determining the maximum frequency differences (2 3 fl in the modulated stimulus when symmetrical SFM stimuli are used. For the present discussion, we will use the best A f threshold values at low modulation frequencies as a measure of the just noticeable frequency differences. Just noticeable frequency differences at different reference frequencies near 40 kHz are available from a number of species. The just noticeable frequency differences in other insectivorous mammals are similar to the minimal threshold modulation depth obtained for T. brasiliensis in this study; at a reference frequency of 42 kHz, thresholds of 1260 Hz (Tupaia glis, Heffner et al., 19691, 1000 Hz (~e~iech~nus aur~~u.s,Ravizza et all 1969b) and 1512 Hz (Didelphis cirginianus, Ravizza et al., 1969a) were found. In rodents, comparable thresholds of 1680 Hz (Cauia porcelfus, Heffner et al., 19711 and 2520 Hz (Rattus mmegicus, Kelly, 1970) were reported for guinea pigs and rats at a reference frequency of 42 kHz, while laboratory mice with thresholds of about 350 Hz proved more sensitive to differences in frequency (Mus musculus, reference frequency 40 kHz, Ehret, 1975). In the cat, (Felis catus, Elliott et al., 19601, a mean frequency discrimination threshold of 575 Hz was found at a reference frequency of 32 kHz. In non-echolocating mammals, the Weber fractions (WR), which permit a comparison of discrimination thresholds at different reference frequencies, generally vary between about 0.01 and 0.10 (calculated from Fay, 1988). The Weber fractions for low modulation frequencies in the present study (see Table IT) fit well to these values. The ability to echolocate is thus not reflected in the frequency modulation thresholds of 7: brasiliensis.

In contrast, just noticeable frequency differences of only 40-60 Hz (corresponding to a WR of less than

0.~7) were reported for frequencies close to the CF component in the CF-FM bat Rh. ~~r~~~~~~~~~~ (Heilmann-Rudolf, 1984). This discrimination ability may be attributed to the specializations found in CFFM bats. Similar discrimination thresholds of only 60100 Hz (corresponding to WRs smaller than 0.0031, however, were also found for test tone frequencies between 20 kHz and 80 kHz, i.e., for frequeneies outside the physiologically specialized region. The thresholds in this frequency discrimination experiment may have been influenced by the temporal pattern of stimulus presentation, as short stimuli (lo-20 ms> were presented with a repetition rate of 25 and 50 Hz. The additional spectral information provided by this stimulus presentation pattern is perceived as roughness by humans (Zwicker and Fastl, 1990). Thus the results of Heilmann-Rudolf may be due to the animals’ use of differences in roughness between a series of tone pulses of the same frequency and a series of tone pulses with two alternating frequencies, rather than to a particularly good frequency discrimination ability over the complete hearing range in Rh. ferrumequinum. It is not surprising that the threshold modulation depths J f of 7: brasiliensis are more similar to the frequency discrimination performance of non-echoloeating mammals in behavioral experiments than to that of CF-FM bats near the frequency of the CF component, in the view of the following physiological findings: The Qro dB values, which characterize the frequency resolution of single units, determined in the anterior ventral cochlear nucleus (AVCN) and the inferior colliculus (ICI of T. brasiliensis range between 3 and 10 Giefer, 1990; Pollak et al., 1978; Pollak and Bodenhamer, 198 1). Comparable values were also found in other FM bats (Suga, 1973; Vatcr et al., 1979) and are commonly reported from non-echolocating mammals (Aitkin et al., 1975; Ehret and Moffat, 1985). In this respect, the auditory system of T. brasiliensis resembles that of non-echolocating mammals more than that of the highly specialized CF-FM bats, which show Q,,, dB values of up to 600 for the neurons tuned to the CF component (Neuweiler, 1990). The threshold modulation depths determined in this study are also compatible with other physiological data. In the IC of 7: hrasiliensis, a population of cells with the typical broadly tuned response to pure tones was described (Bodenhamer and Pollak, 1981). However, these cells were excited only by a narrow frequency band of a width of 0.5-2.5 kHz, when FM broadband sweeps were used as stimuli. The sweep rate and the intensity of the stimuli did not influence the midfrequency and the width of the excitatory frequency band. Similarly, neurons in the IC of the cat were described (Ehret and Merzenich, 1985) that responded to a frequency band of constant width independent of

stimulus intensity. Ehret and Merzenich (1985) suggested that these neurons constitute a neuronal correlate for the bandwidth of frequency processing in mammals. If we assume that the neurons described in both studies are homologous, the intensity independent bandwidth of these neurons may determine the ability to discriminate between a pure tone and a SFM stimulus around this pure tone. Indeed, the width of 0.5-2.5 kHz of the excitatory frequency bands in i? brasiliensis fits well to the threshold modulation depths determined in the present study. In the next section, we investigate the relationship between threshold modulation depth and the performance of actively echolocating bats in fluttering target experiments. The present results are comparable to those of fluttering target experiments on the assumption that the bats make use of the prominent frequency modulations occurring in such experiments. The performance of CF-FM and FM bats is now discussed separately. CF-FM bats may obtain information on fluttering targets from the frequency modulation of the echoes in two ways: In situations where sudden spectral broadenings, i.e., glints, occur, the repetition rate of the glints may be analyzed in the time domain by the auditory system. In addition, differences in modulation depth, i.e., frequency structure, can be used for discrimination. This second mechanism may apply exclusively when glints are missing, e.g. in sinusoidally FM echoes or when the glint rate cannot be used for target discrimination, e.g. when the bats discriminate between different insects with the same wing-beat frequency (see v.d. Emde and Schnitzler, 19’301. In two studies, the ability of CF-FM bats to perceive sinusoidally frequency modulated echoes from a slowly (< 200 Hz1 oscillating loudspeaker membrane was tested as a function of modulation frequency. tl. lankadil:a (v.d. Emde and Schnitzler, 1986) showed a similar increase in threshold modulation depth with increasing modulation frequency as does T. bru.~i~i~rzsisin the present study. Threshold modulation depths at comparable modulation frequencies (IO-100 Hz), however, are about 10 times larger in T. brusiiiensis (7’. brasiliensis: d f = 920-2030 Hz, corresponding to a WR of 0.023-0.051; H. lankadira: 3 f = 90-300 Hz, corresponding to a WR of 0.001-0.0041. In contrast, threshold modulation depths in Rh. ~errumequjnum (Schnitzler and Flieger. 19831, decreased with increasing modulation frequency. Extremely small threshold modulation depths of 30-60 Hz (corresponding to a WR of 0.0004-0.0007) were measured at modulation rates below 30 Hz, decreasing to only 5-10 Hz (corresponding to a WR of O.O(~OO~0.0~01) at a modulation rate of 200 Hz. Thus, modulation rate had an opposite effect on threshold frequency modulation depth in different species of CF-FM bats.

136

On the other hand, a similar increase in threshold modulation depth with modulation frequency is found in man @wicker and Fastl, 1990) and in both the CF-FM bat H. lunkadiva and the FM bat ‘r. brasiliensis.

In the glint rate discrimination experiments, propellers rotating with different speed had to be discriminated. Here, both differences in glint rate and the concomitant differences in modulation depth may have been used as cues by the bats. The maximal modulation depths occurring in these experiments depend on the carrier frequency of the reflected signal, the geometric dimensions of the rotating propeller and its rotation frequency (cf. Nitsche, 1987). Thus, for a given propeller and a given carrier frequency, each glint rate results in a characteristic maximal modulation depth. In order to test the assumption that CF-FM bats used the differences in modulation depth to discriminate between different glint rates, the threshold modulation depths were calculated taking into account the dimensions of the different propellers and the frequencies of the CF components in the different bat species. For Rh. rouxi, (Roverud et al., 1991), threshold modulation depths of 80-220 Hz (WR = ~,~1-~.~3) were obtained (Nitsche, 1987). For H. lankudiua, (Roverud et al., 1991), we calculated values of 200-300 Hz (WR = 0.003-0.004). For H. speoris, (v-d. Emde and Schnitzler, 1986) the estimated threshold modulation depths amounted to about 2000 Hz (WR = 0.016). These modulation depths are compatible with the results obtained in the experiments with an oscillating loudspeaker membrane as puttering target at a corresponding modulation frequency of 50 Hz. In these experiments, ti. ferrumequinum, a species closely related to Rh. roui, discriminated modulation depths of 20-45 Hz (WR = 0.0~2-O.~, Schnitzler and Flieger, 1983) and H. lankudiva of 300 Hz (WR = 0.004; v.d. Emde and Schnitzler, 1986). H. speoris failed to discriminate the presented modulation depth of about 930 Hz (v.d. Emde and Schnitzler, 19861, which is below the threshold modulation depth of about 2000 Hz in this species assessed from the glint rate discrimination experiment. The assumption that modulation depth is used by CF-FM bats to discriminate between different glint rates is thus supported. A comparison of the WRs presented above shows that CF-FM bats differ considerably in their sensitivity to perceive modulation depth. While the WR of T. brasiliensis (0.049, see Table II) is more than 10 times larger than those determined for H. lankadiva and the rhinolophid species, no remarkable differences in WRs between H. speoris and T. brasiliemis are found. For bats using broadband FM sonar calls, the abilities to discriminate between fluttering targets and to perceive m~ulation depth are not as obviously related as in the CF-FM bats. It is, however, unlikely that FM

bats can reconstruct the wing-beat rates of prey insects. because the durations of the sonar calls are too short and their repetition rates too low to convey information about successive glints or of complete wing-beat cycles. On the other hand, threshold modulation depth may again constitute an appropriate measure to understand the performance in the glint rate discrimination experiments, as FM sweeps reflected from flying insects are Doppler shifted depending on the dimensions and velocity of the wings (Kober and Schnitzler, 1990). It is therefore assumed in the following that FM bats analyze a set of stationary frequency patterns or *still images’ to recognize insects (see v.d. Emde and Schnitzler, 1990). The echoes may be analyzed in two ways: the bats may (1) detect the shifts in frequency at the low frequency end of consecutive echoes or (2) discriminate changes in echo spectra arising from the interference between echoes from stationary and oscillating parts of the puttering targets. The first hypothesis (1) which emphasizes the importance of the low frequency end of the FM sweep is consistent with physiological studies. Indeed, the highest Q,, dB values and the most sensitive neuronal thresholds in the ascending auditory pathways of T. brasiliensis and other species of FM bats are found at frequencies between 20 and 30 kHz (Siefer, 1990; Pollak et al., 1978; Suga, 1973; Vater et al., 1979). Thus, in these bats, the highest sensitivity and the best frequency resolution of the auditory system are found in a frequency range that mainly covers frequencies at the low frequency end and below the bats’ echolocation calls. For frequencies above 30 kHz, the Qn, dB values in FM bats decrease again, in contrast to non-echolocating mammals (e.g. cat and house mouse) where Q,, dB values increase with the best frequencies of the neurons (Aitkin et al., 1975; Ehret and Moffat, 1985). At the glint rate discrimination threshold, the differences in modulation depth at the low frequency end of the FM sweeps amount to about 160 Hz, corresponding to a WR of 0.006, in Eptesicus fuscus (Roverud et aI., 1991; Nitsche, 1987) and to about 100 Hz, corresponding to a WR of 0.003, in Pipistrellus stenapterus (Sum and Menne, 19881. These small frequency differences differ considerably from the threshold modulation depths we determined in T. brasiiiensis around a center frequency of 40 kHz. Although one may expect improved threshold modulation depths for frequencies about 20 kHz, i.e., in the frequency range for which higher QIo tJNvaiues were found for i? brasiliensis, the large discrepancy in the results suggests that the performance of the FM bats in glint discrimination experiments can not be accounted for by a detection of frequency shifts at the low frequency end of consecutive echoes.

137

The second hypothesis (2) supposes that the broadband structure of the sonar calls is used by FM bats to discriminate between glint rates. The bats may have discriminated changes in echo spectra arising from the interference between echoes from stationary and oscillating parts of the fluttering targets (see Sum and Menne, 1988). The auditory system may extract a specific timbre, depending on glint rate, by integrating informations from the different frequency channels. An excellent performance in glint rate discrimination experiments can be achieved by using this mechanism, even if the tuning of the individual channels is relatively broad. In view of the similar tuning characteristics of auditory neurons in different species of FM bats, it can be assumed that the threshold modulation depths determined for T. hrasiliensis are also typical for other FM bats. Then, our results are compatible with the second hypothesis, i.e., that the tested FM bats used broadband changes in echo spectra for glint discrimination. It remains an open question, however, whether FM bats use broadband changes in echo spectra for glint rate discrimination when hunting flying insects in open spaces, where echoes from an interfering stationary background do not exist. Thus the animals may still rely on frequency shifts at the low frequency end of consecutive echoes for the pursuit of prey in their natural habitat., Finally, if FM bats identify a set of ‘still images’ rather than dynamical echo patterns, they are under no evolutionary pressure to develop especially high sensitivity to modulation depth at prey wing-beat frequencies, an adaptation found in CF-FM bats. Thus, it is not surprising to find in T. brasifiensis the regular increase of threshold modulation depth as a function of modulation frequency also observed in non-echoloeating mammals.

Acknowledgments

This study was funded by SFB 204, ‘Gehiir’, Miinchen. We thank Professor G. Pollak for kindly providing the bats and M. Potke for technical assistance. We also like to thank Professor G. Neuweiler and Dr. K.M. Schmidt for critically reading the manuscript.

References Aitkin. L.M.. Webster. W.R., Veale. J.L. and Crosby, D.C. (1975) Inferior colliculus I: Comparison of response properties of neurons in central, pericentral and external nuclei of adult cats. J. Neurophysiol. 3X. 1196- 1206.

Bodenhamer,

R.D.

and

Pollak,

G.D.

(1981)

Time

and

frequency

domain processing in the inferior colliculus of echolocating bats. Hear. Res. 5, 317-335. Ehret, G. (1975) Frequency and intensity difference limens and nonlinearities in the ear of the housemouse (Mus muscu[us). J. Comp. Physiol. 102, 321-336. Ehret, G. and Merzenich. M.M. (10%) Auditory midbrain responses parallel spectral integration phenomena. Science 227. 1245-1247. Ehret, G. and Moffat. A.J.M. (198.5) Inferior colliculus of the house mouse. II. Single unit responses to tones. noise and tone-noise combinations as a function of sound intensity. J. Camp. Physiol. A 156, 619-635. Elliot. D.N., Stein. L. and Harrison, M.J. (lY60) Determination of absolute intensity thresholds and frequency difference thresholds in cats. J. Acoust. Sot. Am. 32, 380-384. Emde, G. von der and Schnitzler. H.U. (IYXh) Fluttering target detection in Hipposiderid bats. J. Camp. Physiol. A 15Y. 7655772. Emde, G. von der and Menne, D. (1989) Discrimination of insect wingbeat-frequencies by the bat Rhinolophus frmrmequinum. J. Camp. Physiol. A 164, 663-671 Emde, G. von der and Schnitzler, H.U. (1990) Classification of insects by echolocating greater horseshoe bats. J. Camp. Physiol. A 167. 423-430. Fay, R.R. (1988) Hearing in vertebrates. A psychophysics databook. Hill and Fay Associates, Winnetka, IL. pp. 451-462 Finney. D.J. (1971) Probit analysis. 3rd edition. Cambridge University Press, Cambridge, UK. Goldman, L.J. and Henson. O.W.Jr. (lY77) Prey recognition and selection by the constant frequency bat. Prr,ronotus pundii. Behav. Ecol. Sociobiol. 2, 41 I-419. Heffner, H., Ravizza, R. and Masterton. B. (1969) Hearing in primitive mammals III: Tree shrew (‘licpuia ,qIis). J. Aud. Res. 9, 12-1X. Heffner. R., Heffner. H. and Masterton. B. (1971) Behavioral measurement of absolute and frequency difference thresholds in guinea pig. J. Acoust. Sot. Am. 4Y. IXXX- 1X9.5. Heilmann-Rudolf. U. (11)X4) Das Frequenrunterscheidungs-vermiigen bei der @open Hufeisennase. Rhinolophs ~&x~ntequincmi. Dissertation, Eberhard-Karls-Universitat Tiibingen. Henson. O.W. Jr. (1970) The ear and audition. In: W.A. Wimsatt (Ed.). Biology of bats. Academic Press. New, York and London. pp. 1X1-263. Kelly. J.B. (1970) The effects of lateral lemniscal and neocortical lesions on auditory absolute thresholds and frequency difference thresholds in the rat. Ph.D. thesis. Vanderbilt University. university microfilms 70-16, 429. Kober. R. and Schnitzler, H.U. (1090) Information in sonar echoes of fluttering insects available for echolocating hats. J. Acouat. Sot. Am. X7, 8X2-X96. Keller, S. (1967) Tafeln zur Beurteilung von Haufigkeiten in: Graphische Tafeln zur Beurteilung statistischer Zahlen. Steinkopf Verlag, Darmstadt. Langemann. U. (1991) FrequenLunterscheidung beim Staren (Sturnrrs ~,itlguri\ ): Eine Schwellenbestimmung mit verschiedenen Reizmustern. Diplomarheit, Institut fir Zoologie der Technixhen Universitat. Munchen, FRG. Long. G.R. and Clark. W.W. (1983) Detection of frequency and rate modulation by the chinchilla. J. Acoust. Sot. Am. 75, 1184-l 100. Neuweiler. G. (1984) Foraging. echolocation and audition in bats. Natutwissenschaften 71, 446-455 Neuweiler. G. (1990) Auditory adaptations for prey capture in echolocating bats. Physiological Reviews 70, 615-641. Nitache. V. (1987) Das Unterscheidungsvermigen fir Schlagfrequenzen mechanisch simulierter Fliigel in storungsfreier und akustisch gestiirter Umgehung bei der FM-CF-FM Fledermaus Rhinolophus rouxi (Microchiroptera). Diplomarbeit. Technische Universitat, Miinchen. FGR. Pollak, G. and Bodenhamer. R. (1981) Specialized characteristics of

138

single units in the inferior colliculus of the mustache bat: frequency representation, tuning and discharge patterns. J. Neurophysiol. 46, 605-620. Pollak, G., Marsh, D.S., Bodenhamer, R. and Souther, A. (1978) A single-unit analysis of inferior colliculus in unanesthetized bats: response patterns and spike-count functions generated by constant-frequency and frequency-modulated sounds. J. Neurophysiol. 41, 677-691. Ravizza, R., Heffner, H. and Masterton, B. (1969a) Hearing in primitive mammals I: Opossum (Dide&zti r+niunusf. J. Aud. Res. 9, 1-7. Ravizza, R., Heffner, H. and Masterton, B. (1969b) Hearing in primitive mammals II: Hedgehog (Hetniechinus aurifus). J Aud. Res. 9, B-11. Roverud, R.C., Nitsche, V. and Neuweiier, G. (1991) Discrimination of wingbeat motion by bats, correlated with echolocation sound pattern. J. Camp. Physioi. A 168, 259-263. Sachs, L. (1978) Statistische Auswertungsmethoden. Springer-Verlag, Berlin, Heidelberg, New York. Schmidt, S., Thaller, J. and Piitke, M. (1990) Behavioral audiogramm and masked thresholds in the free-tailed bat, Tadaridu hrasiiiensti. In: N. Elsner and G. Roth (Eds.), Brain - Perception Cognition. Proc. 18th Gottingen Neurobiot. Conf, Thieme, Stuttgart, p. 146. Schnitzler, H.U. (1968) Die Ultraschall-Ortungslaute der HufeisenFledermiuse (Chiroptera-Rhinolophidae) in verschiedenen Orientierungssituationen. Z. Vergl. Physiol. 57, 376-408. Schnitzier, H.U. and Henson W.Jr. (1980) Performance of airborne animal sonar systems. In: R.G. Busnel and J.F. Fish fEds.1. Animal Sonar Systems, Plenum Press, New York and London. pp. 109-181. Schnitzler, H.U. and Flieger, E. (1983) Detection of oscillating target movements by echolocation in the greater horseshoe bat. J. Comp. Physiol. A, 385-391.

Schnitzier, KU.. Menne, D., Sober, R. and flebiich. k. it?83) f‘iw acoustical image of fluttering insects in ec~lo~o~~ting hats. In: F Huber and H. Mark1 (Edsf. Neuroethology and behavictra1 physiology. Roots and growing points, Springer. Berlin, Ifcidelbcrg. New York, pp. 2X-250. Schorer, E. (lY8Y) Vergleich eben erkennbarer Unterschiede und Variationen der Frequenz und Amplitude von Schallen. Acustica 6X, 183-199. Schuller, G. and Pollak, G. (1979) D~sproportionatc frequency representation in the inferior colliculus of Doppler-compens;lting greater horseshoe bats: evidence for an acoustic fovea. .T,romp. Physiol. 132, 47-54. Siefer, W. (1990) Frequenzreprasentation im peripheren Hiirsystem der Fiedermaus Tudurida brasiliensis. Diplomarbeit, Ludwig-Max imilians-Universit~t~ Miinchen. Simmons, J.A., Lavender, W.A., Lavender, B.A., Childs, J.E., Huiebak, K., Ridgen. M.R., Shermann, J. and Wooiman, B. (1978) Echolocation by free-tailed bats (Tadurida). J. Comp. Physiol. 125, 291-299. Suga. N. (1973) Feature extraction in the auditory system ot bats. in: A.R. Moller (Ed.), Basic mechanisms in hearing. Academic Press, New York. pp. 675-744. Sum, Y.W. and Menne, D. (1988) Discriminati(~n of fluttering targets hy the FM-bat Pipistrelius stenopterus? J. Camp. Physiol. A ih3, 349-354. Vater, M., Schlegel, P. and Zoller, H. (1979) Comparative auditory neurophysiology of the inferior coiliculus of two bats, Colossus ater and A4&ssus mofossus (I). J. Comp. Physiot. 131, 137-160. zljfei, P. (1988) Statistik in der Praxis. UTB Gustav Fischer Verfag. Stuttgart. Zwicker, E. and Fastl, H. (1990) Psychoacoustics: Facts and models. Springer-Verlag, Berlin, Heidelberg. New York.