Echolocation in bats
Echolocation in bats is...
How bats avoid obstacles
History of the discovery
Robert Galambos in 2006: There are at least two histories of how it came to be known that flying bats use high-frequency auditory echolocation to avoid obstacles as they fly in the dark. The one you are reading is a short one; the long one began as a chapter in my 1941 PhD thesis and was later published in ISIS, the history of science journal (in 1942, vol 34).
The long one starts in 1794 with experiments by Lazzaro Spallenzani in Italy, who decided several years later with Louis Jurine, a Swiss biologist, that bats use their ears, but neither of them had a clue as to how they did it. They certainly did not suggest bats use high frequency sounds people cannot hear, a strange fact that must mean the best thinkers of the time didn’t realize such sounds must exist. Throughout the next 150 years, at least 20 authorities addressed the problem experimentally or philosophically, and proposed theories ranging from wings sensitive to air motion, through the idea all five senses are involved to some extent, on to postulating a mysterious sixth sense we don’t have and so will never understand.
Some of the late comers, when discussing their own experiments, also arrived at the hearing hypothesis, and H. Hartridge, a physiologist who did not report an experiment of his own, actually suggested in 1920 that “bats during flight emit a short wave-length note and this sound is reflected from objects in the vicinity”—an example of having the right answer at the wrong time, because in biology a claim without an experiment to support it is just a more or less interesting collection of words.
In 1937 Donald R. Griffin, a Harvard graduate student, learned that a physics professor, G. W. Pierce, had just invented a device that converts ultrasonics into frequencies people can hear. For years the professor had listened to insects singing their songs outside his summer cottage up in New England somewhere, and, curious whether those cheeps and chirps included what were then called “supersonic” sounds, he put together two boxes filled with vacuum tubes, one to convert ultrasonics into audible sounds, the other to generate a wide band of ultrasonic sounds for use in his laboratory.
Griffin, who was already an expert on bats, having published several papers on their migration habits, approached Professor Pierce in 1937 with the idea they set a bat free to fly around in his lab with the ultrasonic converter turned on, and just listen. They did this, heard nothing from flying bats, and published the finding in 1938 (J. Mammology, vol 19). Don later blamed the negative result on certain properties of the microphone, but when we made the measurement together a year or so later we heard a symphony of bat cries whenever the same microphone pointed at a flying animal. Shall we blame the previous failure on some unknown instrumental error in the 1937 setup?
I was also a graduate student at the time, working with my professor, Hallowell Davis, using a then-new physiological method to test animal hearing. When Don learned this in 1939 he asked me if I would put a bat through the procedure I was using on guinea pigs and cats. When Professor Davis said “Go ahead” I did, and a few weeks later I had learned the bat ear is really special compared to that of people, cats, and guinea pigs. The normal adult human upper frequency limit is about 20,000 Hz, and the cat, dog, and guinea pig upper limits were then in the 4-5 KHz range. The bat, I learned after moving to the Pierce lab to use his high frequency detector, responded at 98 KHz, the limit of the instrument.
That unique upper limit of the bat hearing spectrum made the answer to its remarkable avoidance performance obvious. But to settle the matter finally and hopefully forever, Don and I decided to put together a flawless group of simple experiments worthy of the problem’s long history and wide interest. Our first step was to divide an experimental room into two halves by hanging wires spaced 12 inches apart from ceiling to floor. The small diameter of the wires Don chose to hang was already known to make them difficult for bats to detect.
We used this wire barrier to test how well a given bat avoids obstacles by counting the number of times it contacted a wire while flying back and forth through them up to hundreds of times. We then retested the animal after depriving it temporarily of hearing, vision, or the ability to make vocal sounds. Here are samples of the percent hits made by Myotis lucifigus, one of the three bat species we tested in this way.
- Normal (n=129 tests): 30% hits
- Eyes covered (28): 24%
- Ears plugged (29): 65%
- Mouth tied shut (8): 67% .
That 24% suggests eyes are a hindrance rather than a help when a bat avoids objects in daylight, but no one has ever confirmed the finding, so far as I know.
Next we turned on the Pierce ultrasonic detector, pointed its microphone toward the wires, and heard, recorded, and measured the cries both normal and sensory deprived bats emit. Cries typically last a few ms; normal animals emit them at a higher rate as they approach and detect the wires; and the rate returns to the lower cruising value on the other side. The abnormal behaviors of sensory deprived animals were invariably matched by predictable deviations in their cries. For instance, deaf bats never change the emission rate, and along with the mouth tied shut group, they blunder about, randomly hitting wires or flying directly into a wall.
To summarize. Two near-novice graduate students solved an interesting old problem using three independent methods.
- The chronologically first method demonstrated physiologically that bat ears respond to sound frequencies several octaves above the upper limit of human hearing, which means bats are equipped to hear echoes of any very high frequency cry they might emit.
- The second method compared the obstacle avoidance performance of normal animals before and after being deprived reversibly of vision, hearing, and the ability to utter cries. Only deaf and silenced bats are helpless when flying around obstacles.
- The third method used a then-new device that converts ultrasonic into audible sounds to demonstrate that flying bats emit cries we cannot hear when they avoid obstacles, and that successful avoidance is highly correlated with where and when the emission rate is highest.
Conclusion: The production and reception of ultrasonic cries simply and adequately explains the phenomenon of obstacle avoidance by flying bats. Don and I took sound and silent moving pictures showing normal and sensory deprived animals flying back and forth through the wires to convince the skeptics.
Figure 6. Summary of acoustic imaging process in echolocation.
(A) The bat emits a sonar sound that travels outward (green arrow) to impinge on an insect at some target range (r). Reflections return to the bat from various insect body parts—e.g., wing (dark blue arrow) and head (red arrow) separated by a small difference in range (Δr). These sources of reflections are called glints, and the insect is depicted here as a 2-glint target.
(B) Sound-pressure waveforms of the bat’s transmitted sonar sound (green) and the reflections from the insect’s glints—wing (dark blue) and head (red). The transmitted sound is heard directly by the bat to initiate processing by establishing a zero-time origin for reception of subsequent echoes (long vertical dashed line across pictures). Echoes return to the bat’s ears after a delay (t) related to target range (r) at rate of 5.8 milliseconds/meter. Reflections from the 1st and 2nd glints add together at a small time separation (Δt) to interfere with each other when they form a combined echo arriving at the bat’s ears. For each transmitted sound, the acoustic stimulus received by the bat is not just the echo but a pair of sounds (purple)—the outgoing broadcast received directly at the moment of emission followed by the returning combined echo comprised of however many reflections are returned by the target’s glints.
(C) Spectrograms—or time-frequency plots—of the transmitted sound (left) and combined echo (right) showing the FM sweeps. The broadcast signal contains two sweeps—a 1st harmonic from 45 to 22 kHz and a 2nd harmonic from 80 to 45 kHz—that appear as sloping ridges, one over the other. The echo also contains these two harmonics, but the sloping ridges for the combined echo spectrogram are not as smooth as those in the broadcast spectrogram because interference between the overlapping reflections from the 1st and 2nd glints creates alternating peaks (p) and notches (n) that appear as ripples or undulations. The notches are locations where energy in the echo has been cancelled by interference; their locations are the principal defining characteristics of the ripples because their frequencies (f1, f2, and f3) are related to the delay separation (Δt) of the reflections from the glints (see Fig. 2A). The bat’s inner ear transforms the spectrograms of broadcasts and echoes into neural spectrograms composed of spikes that register the FM sweeps in terms of the tuned frequencies of different neurons and times, or latencies, of the spikes (see Fig. 4).
(D) Diagram illustrating neural spectrograms of a broadcast and echo. The spectrogram is composed of on-responses in neurons tuned to different frequencies (vertical axis). For downward FM sweeps (as in C), excitation in the bat’s inner ear moves rapidly from high-frequency (basal end of cochlear spiral) to low-frequency (apical end of cochlear spiral). As this excitation rushes past receptors, it triggers spikes (red circles) in auditory-nerve fibers from any given location (see Fig. 3A,B). (Conventional spectrograms from Fig. 1C are shown here as shades of gray.) The on-response (1st spike) is sharply coincident with the onset of this rapidly-moving excitation, and it registers the time-of-occurrence of the neuron’s tuned frequency in each sweep. At any given frequency, the total duration of excitation is very brief—only a few hundred microseconds in duration—because the sound sweeps rapidly away to lower frequencies. Consequently, there is only time enough for the on-responses to occur reliably. The whole volley of on-response spikes (red circles) spread across neurons tuned to different frequencies traces the shape of the FM sweep for the broadcast and then for the echo. In the auditory nerve, if excitation is strong, cells often produce one or more additional spikes that follow the on-response, but they are less sharply synchronized to the onset of excitation, and cells exhibit spontaneous activity, too, that has no relation to stimuli. These secondary spikes are stripped out of the volley as the spikes pass through the cochlear nucleus, so that only the on-responses shown here are used to determine echo delay. Note that the frequency tuning of the neurons segregates the traces for the 1st and 2nd harmonic sweeps in each sound. Because the echo contains closely-spaced reflections of the broadcast that were returned from different parts of the target (Fig. 1A), interference occurs, and some frequencies are cancelled out to create notches in the spectrogram. Auditory neurons tuned to frequencies of notches fail to produce on-responses, so there are gaps or “holes” (unfilled circles around f1, f2, and f3) in the neural spectrogram for the echo. In auditory nerve fibers, generation of spikes takes place on the neurons at a site very close to the organ of Corti (Fig. 3B), and the spikes have a latency of about half a millisecond. The proximity of the spike-generation site to the receptors, and the proximity of the bat’s cochlea to its brainstem (see Fig. 3A,D) may make spike initiation particularly stable in time and minimize the distance spikes have to travel to reach the first synapse, where synchrony of spikes across closely-spaced auditory-nerve fibers is used to sharpen temporal registration of the spectrogram before it is distributed upward throughout the auditory pathway for further processing.
(E) Hypothetical traces depicting two possible versions of the image of echo delay perceived by the bat to represent target range. The horizontal axis (r) is a psychological scale showing the perceived value of delay. Although this now refers to events inside the bat, such values nevertheless can be measured in behavioral experiments. The two possible versions of the image differ according to how information in the spectrograms is used to estimate range from delay. Does the bat perceive an “image” corresponding to the delay of the echo spectrogram representing the overall range (r) to the target, shown here as a blue distribution for the combined echo as a whole? In this case, the underlying representation is a single numerical delay estimate with a potentially broad time-width corresponding to the integration-time (ti) of the spectrogram (see C). Both glints are subsumed into one delay or range perception. Or, does the bat separately perceive the glints themselves (1st glint dark blue; 2nd glint red) even though they are separated by a small interval of distance (Δr) so that their reflections are separated by an equivalently small amount of time (Δt) compared to the integration-time (ti)? If the bat perceives the first kind of image, then target range (r) most likely is estimated from the auditory equivalent of rightward displacement of the spectrogram of the transmitted sound until it superimposes on the spectrogram of the echo. Although the range difference between the glints (Δr) is not itself perceived as such, the presence of interference notches in the echo provides the bat with indirect information about the target’s structure. However, if the bat perceives the second kind of image, in which the two glints are recognized as being at different ranges, then the means by which the bat determines the delays of the reflections must be different for the two glints because their manifestation in the combined echo spectrogram (C) is different. First, the overall delay of the combined echo most likely is estimated from the auditory equivalent of rightward displacement of the spectrogram of the transmitted sound to determine the overall range (r). Second, the delay separation of the reflections from the two glints most likely is estimated from the frequencies of the interference notches in the combined spectrogram to determine the range separation (Δr). Note that this latter procedure explicitly involves transforming numerical values of frequency (for notches) into an equivalent numerical value of time (for delay separation).
(F) Expanded diagram of perceived ranges (r, Δr) to use as a guide for interpreting results obtained in behavioral and neurophysiological experiments with 2-glint stimuli (examples of real perceived ranges in Fig. 2C-F). These results distinguish between the two possible types of images perceived by the bat. The most important experimental finding from behavioral studies is that the numerical value of the range separation (Δr) is perceived on the same psychological scale as the numerical value of overall range (r), so that the bat’s images indeed do depict the locations of both glints. The most important finding from neurophysiological studies is that FM stimuli composed of artificial transmitted sounds and two-glint echoes at different delays (t) and delay separations (Δt) are represented jointly in terms of two orthogonal neuronal response parameters—the overall delay of echoes from response latency and the delay separation from the occurrence or nonoccurrence of spikes that encode the frequencies of interference notches.
- Galambos, R, and D. R. Griffin. Motion pictures of bats avoiding obstacles: two demonstrations. Anat. Rec. 1940. 78:95.
- Galambos, R. Cochlear potentials from the bat. Science, 1941. 93(2409): 215.
- Griffin, D. R. and R. Galambos. The sensory basis of obstacle avoidance by flying bats. J. Exp. Zool., 1941. 86: 481-506.
- Galambos, R. The avoidance of obstacles by flying bats: Spallanzani's ideas (1794) and later theories. Isis, 1942. 34: 132-140.
- Galambos, R. and D. R. Griffin. Obstacle avoidance by flying bats: The cries of bats. J. Exp Zool. 1942. 89: 475-490.
- Galambos, R. Cochlear potentials elicited from bats by supersonic sounds. J. Acoust. Soc. Am. 1943. 14: 41-49.
- Galambos, R. Flight in the dark: A study of bats. Scientific Monthly, 1943. 56:155-162.
- Galambos, R. The production and reception of supersonic sounds by flying bats. Pages 42-44 in Harvard University Summaries of Theses, 1941, Cambridge, Published by the University, 1945.
- Galambos, R. The 1939-40 bat experiments that validated Jurine's claim. Le rhinolophe. 1995. 11: 17-95.