INTRODUCTION
At dusk on a summer’s evening, a still quietness descends upon the countryside. The silence broken only by the occasional, distant hoot of an owl, the stillness by the silent fluttering of a small dark shape cavorting around the treetops.
No bird this dark, dusk flier, but a bat. Silent to our ears but nonetheless shouting at the top of his voice, and on the basis of the echoes of his shouts pursuing and catching tiny insects which we only detect by the bites they leave on our arms and necks.
It was this ability of the bat to find its way about and to hunt in total darkness that intrigued the Abbe Spallanzani in the late 18th century. After many long and painstaking experiments, he found that the bat had some sort of sixth sense, requiring neither sight nor touch, to manoeuvre in the dark, but that its flight performance was severely degraded if the ears or mouth were blocked up.
Contemporary opinion, however, scorned the idea that a bat might ‘see with its ears’, and it was not until the early 20th century that echolocation was suggested as a mechanism by which the bats might navigate (Maxim 1912), although the proposed sound source was the low-frequency sounds produced by the bats wings rather than vocalisations.
It was Hartridge in 1920, who first suggested that ultrasound was a more likely medium, but he made no attempt to verify his suggestion experimentally. Indeed the use of ultrasound by animals was first established amongst the insects by Pierce, who had designed a ‘supersonic’ receiver for the purpose. Griffin then suggested pointing the receiver at a bat, and it was found that bats did indeed produce ultrasonic sounds.
Over the succeeding years, it was established that these ultrasonic sounds were being used for navigation and for hunting, and the structure of the sounds was elucidated in more detail.
The task is not a small one, however, for the bats constitute one of the largest order of mammals, second in number of species and individuals only to the rodents. With more than 850 known species the bats constitute over one fifth of all known species of mammals. They are also a very widespread order and the only indigenous mammal to New Zealand.
Their ways of life have also spread to fill almost every niche available to a nocturnal flying mammal. The majority of species are insectivorous since the bats were originally derived from insectivore stock, but many species are frugivorous or nectarivorous, and there are also carnivorous, sanguivorous and even piscivorous species.
The order is split into two sub-orders, the Megachiroptera or fruit bats, and the Microchiroptera or insectivorous bats. The names can be a little misleading, however, for the small Megachiropterans are not as large as some Microchiropterans, and dietary habits of the two groups are by no means clear-cut.
The Megachiroptera are found only in the Old World, and while most are frugivorous many live off flowers or pollen or are specialized for drinking nectar. With a few exceptions, these bats do not use echolocation, but have large extremely sensitive eyes, giving them probably better night vision than any other animal. In total darkness however they are helpless and they, therefore, tend to live in trees rather than in caves. One genus of fruit bats, Rousettus, has taken to living in caves and has developed a system of echolocation independently from the Microchiroptera. Instead of using the vocal chords to produce the echolocation sounds they click their tongues, producing short clicks very similar to the clicks used by the echolocating birds Steatornis and Aerodramus.
The role of the fruit bats in the New World has been taken over by one of the five super-families of the Microchiroptera, the Phyllostomoidea. This super-family is split into three families, the Phyllostomidae with about 140 species, of diverse habits, and diets including fruit, nectar, pollen, flowers, insects and vertebrates, the smaller Desmodontidae specialized for drinking blood, and the Mormoopidae. All the bats in this super-family have nose leaves, although the size and shape are very variable.
In the old world nose-leaves are worn by the super-families Megadermatoidea with two families, and the Rhinolophoidea, also with two families. The majority of these bats are insectivorous, although a few, especially the Megadermatids, are also carnivorous.
There are two super-families of simple-nosed bats, the smaller and purportedly more primitive group being the Emballonuroidea with four families, of which one family was only discovered in 1974 (Hill 1974). The Vespertilionoidea consists of seven families of diverse habits and ranging from a single species of fairly rare bats to a family with 280 species of ‘common’ bats.
(NB (note added 2018) – since the above was written the classification scheme for bats has been extensively revised as a result of DNA testing. The order Chiroptera is now divided into the Pteropodiformes (or Yinpterochiroptera) and the Vespertilioniformes (or Yangochiroptera). The former group comprises the Pteropodidae, Rhinolophidae, Hipposideridae, Megadermatidae, Craseonycteridae and Rhinopomatidae, while the remaining families fall into the Vespertioniformes. This division raises interesting questions as to whether laryngeal echolocation was established before the two groups spli and the ability was then lost in the Pteropodidae, or whether it developed independently in each branch).
The echolocation signals of the bats are also quite diverse, but it is not generally possible to correlate a particular type of echolocation behaviour with a particular way of life or taxonomic grouping, although there do tend to be affinities between the types of signals used by a particular family.
Basically, there are two general types of echolocation system. A short frequency modulated pulse, with a wide bandwidth can be used for measuring range, or a long constant frequency signal can be used for measuring velocity using the Doppler effect. In practice, different bats tend to use combinations of these, or they may use different types of signals at different times, and the number of harmonics present is also variable.
Myotis may be considered as the archetypal FM bat, a typical cruising pulse covering a frequency range of about an octave from 80 to 40kHz, and with a duration of about 2-3 ms. When a target is detected and the bat goes into an interception manoeuvre the pulse duration decreases and the frequency may drop. The repetition rate is also increased since the bat is concentrating on the prey and can emit another echolocation signal as soon as it has processed the last echo. At the moment of interception the instantaneous pulse rate my be as high as 200 per second, and the pulse duration may be as little as 0.1 ms. This increase in repetition rate during an interception manoeuvre is characteristic of bats in general and is termed a ‘catch buzz’, due to the sound it makes on a bat detector. With some of the lower frequency bats, such as Taphozous, the buzzes may sometimes be heard by the unaided ear.
The complement of the classic FM bat is the classic CF bat, Rhinolophus. It is probably true to say that the echolocation system used by Rhinolophus ferrumequinum has been studied in more detail than that of any other species of bat. It is generally characteristic of all the Rhinolophus, and is also used by some other families such as the Hipposideridae. The description given here will be based on Rhinolophus ferrumequinum.
The basic cruising pulse is a long constant frequency signal, of about 80 ms duration at a frequency of about 82 kHz. The ‘nominal’ frequency for a particular bat will be quite constant, but there is some geographical variation. British bats, for example, tend to use a slightly higher frequency than German bats, and Rhinolophus ferrumequinum (nippo) uses tone of about 77 kHz. Frequently the cruising pulses will be terminated with a downward frequency sweep of about 10-12 kHz, and by altering the resonances in the vocal tract the bat can choose whether most of the energy will be contained in the CF portion (sweep-decay pulse) or in the FM portion (sweep-peak pulse). There may also be a vestigial frequency up-sweep at the beginning of the pulse.
As with Myotis, when the bat is intercepting a target the pulse duration decreases and the repetition rate increases, but the CF portion of the pulse is never entirely eliminated. Also, the CF frequency of the pulse does not drop as it does in Myotis, although the depth of the sweep may be increased.
If the emitted frequency of a Rhinolophus is examined carefully it can be seen that there is a change in frequency from one pulse to the next, although each pulse is very accurately controlled. It has been shown that the bat is compensating for the Doppler shift on the echoes of the transmitted signals. The frequency of the next pulse is altered so that the echoes will all be very close to a ‘preferred’ frequency, which is slightly higher than the ‘resting’ frequency of the bat.
When the hearing sensitivity of this bat is examined it is found that there is a sharp peak in sensitivity, with an extremely high Q, just at the bat’s preferred frequency. Additionally, the proportion of the cochlea devoted to the frequencies in this region is disproportionately large and densely innervated, forming an ‘acoustic fovea’.
The problem of interference from the signals of other bats is significant for both CF and FM bats. The FM bat can average the echo ‘image’ he receives over several pulses; the image due to his own signals will be stable while interfering signals will be incoherent and will therefore just contribute background noise. Targets whose range is changing can be allowed for by predicting their new expected position extrapolated from previous range measurements as long as the acceleration is not too great. The task is made easier by the fact that the bat has an extremely good spatial memory, on which he will often rely in preference to contradictory echolocation information, so that the static background environment will provide a constant reference system.
The CF bat can do the same averaging trick in the frequency (or velocity) domain, but loses the advantage of a static reference system (except possible when in flight). Several ‘tricks’ however have been developed to reduce the problems. Some species of Hipposideros, which use echolocation signals just like those of Rhinolophus although of shorter duration and higher frequency, will all lock on to the same emitted frequency when in a crowded environment such as the roost. They can then extract useful information from any echo that they hear, whether from their own or another bats signal since they know the frequency at which it was transmitted. In one species the bats within a ‘colony’ will lock onto one of two different frequencies which are separated by an amount equal to the Doppler shift which would be generated by a moving wing tip of a bat in flight. This provides optimal isolation between the two groups since CF bats tend to ignore doppler shifts from receding targets.
An alternative technique of time gating has been adopted by Pteronotus parnellii, which uses an echolocation signal with the same shape as that of the Rhinolophids, but in which the third and fourth harmonics are also present. As with the Rhinolophids, the main energy is in the second harmonic and the fundamental is suppressed. Functional areas in the auditory cortex have been found in which neurons respond to echoes of the FM portion of the second and/or third harmonics if, and only if, they are preceded by the FM portion of the fundamental. Different neurons are found to code for different fundamental-harmonic delays. Since the fundamental is emitted at such low amplitude it seems likely that this system reduces interference from other bats echolocation signals. CF coding areas are also found which respond only to a combination of the fundamental and one or more of the harmonics, but in these cases the best response is obtained when there is no delay between the different frequency components.
Rhinolophus and Pteronotus parnellii are obligate CF bats, which means that it is feasible to study their echolocation systems in the laboratory, and Doppler compensation can be demonstrated by such techniques as supplying a swinging pendulum as a target or providing electronically doppler shifted echoes. However several species of bats, such as the Pipistrelle, will use short CF pulses of about 20 ms duration when flying in the open, but will switch to FM pulses when flying in a confined environment, or when intercepting a target. It is possible that the CF pulses in these cases are used as low resolution ranging pulses designed to maximise the chances of detecting a target at some distance, since it is possible to put more energy into a long pulse than a short one and signal to noise ration can, therefore, be maximised. Pye, however, has shown that even pulses of quite a short duration can contain useful Doppler information and it would be strange if this information was discarded by the bats. Demonstration of doppler compensation would be sufficient evidence that the velocity information was being detected by the bat, but when recording bats flying in the open there is a Doppler shift on the recorded sounds due to the speed of the bat relative to the microphone, and this Doppler shift will be of the same magnitude as the changes in frequency that the bat would be producing for doppler compensation. Without knowledge of the relative flight speed of the bat the two effects cannot be distinguished.
A solution to the problem was found in the use of a small portable Doppler radar (Halls 1975), which was designed and built for the purpose by Marconi. With a microphone attached to the radar, the radar signal and the bat's sonar signal could be recorded simultaneously, and the Doppler shift due to the speed of the bat relative to the microphone could be compensated for. Plots of the frequencies emitted by a pipistrelle in flight and during hunting manoeuvres showed that the Pipistrelle is deliberately changing frequency in flight, but it was found that interpretation of the reason for the changes was difficult without a knowledge of the target in which the bat was interested (Pye1980).
A similar problem arises with regard to the new world bat, Saccopteryx. This bat uses short CF pulses, but successive pulses alternate between two different frequencies, the separation between them being equal to the Doppler shift that would be produced due to the flight speed of the bat. It has been suggested that the bat may be using one of the tones as an uncompensated ‘ground-looking’ sonar for navigational purposes, while the other may be pre-compensated for the bat’s flight speed as a forward-looking hunting system.
It was suspected that the closely related African bat Coleura afra might also use a two-tone echolocation system and we set out to verify this in 1975. A two-tone bat was soon discovered hunting in the evenings, but identification of a wild, flying bat at night is difficult! Specimens of Coleura taken from the roost would only produce FM signals when flying indoors, or when held in the hand and bats recorded as they left the roost would use single frequency CF, albeit at the expected frequency. Confirmation that these bats did indeed use a two-tone system was eventually obtained by recording captive bats as they were released into the wild. At first, they would emit single frequency CF pulses but as they faded into the distance they could be heard to switch into the two-tone mode we had observed previously.
Bats, of course, are not the only users of ultrasound. Many of the insects upon which the bat's prey have developed the ability to hear the ultrasonic sounds of an approaching bat, and some can produce ultrasonic signals of their own, apparently in an attempt to jam the bat’s sonar. The jamming signals consist of trains of wideband clicks, but the repetition rate is such that the spectrum contains several frequency sweep components. Other insects such as lacewings which can also hear the approaching bats merely indulge in evasive manoeuvres in an attempt to avoid capture.
(There ought to be something in here about u/s communication by insects)
My involvement in the ultrasonic bio-acoustics began in 1974 when I was appointed to the post of research assistant to Prof. J.D. Pye, with the primary task of systematically analysing his recordings of the echolocation cries of over 100 different species of bats.
In the course of this work, I became interested in the non-echolocation ‘squawks’, especially common in the recordings of hand-held bats, and in their relationship to the echolocation signals. This led to a joint programme, with Glenis Long, to make a series of recording from inside a roost of Greater Horseshoe bats. While I was interested in the range and structure of the ‘communication’ calls, Glenis was particularly interested in the ontogeny of the echolocation signals. The roost chosen was one which has been studied in great detail by Roger Ransome, and this meant that we not only had easy access to the roost but that we could predict the behaviour pattern of the bats through the year.
It was also at about this time that work began on a new type of bat detector. This instrument, which translates ultrasonic signals into sounds which we can hear, is one of the fundamental tools for work in ultrasonic bioacoustics. Two basic modes of operation are possible. A broadband detector will pick sounds of any frequency but gives no information about the frequency content of a signal. The basic form of broadband bat detector was a simple envelope detector, which gave a poor response to long, pure tone signals. The alternative form was a tuned instrument. Basically, a long wave radio with the aerial replaced with a microphone. This overcame the problem of long signals by using a BFO, and by virtue of being tuned gave information about the frequency content of a signal. However, it was easy to miss a passing bat simply because the detector was tuned to the wrong frequency. Pye had combined the best of both type of instruments in tuned/broad-band detector, in which the broadband mode was produced in the same way as tuned operation, but with the local oscillator replaced with a spike train, which has frequency components at 10 kHz intervals over a range of at least 200 kHz. Use of a spike train, however, necessitated the use of a direct conversion technique which has certain technical disadvantages.
I then suggested that a super-heterodyne tuned/broadband detector could be constructed if the local oscillator were swept linearly at a rate of 10kHz. Such a signal will have a line spectrum, with the lines spaced at the repetition rate of the sweep, and with nearly equal amplitude over the range of frequencies through which the oscillator is sweeping. A prototype instrument was built, and plans were made for its commercial production. However a bat detector can only be as good as its microphone, and I, therefore, undertook to examine the characteristics of the existing types of microphone, and if necessary to develop a new one for use with the new instrument. At the same time Peter Ward was designing a new headstage amplifier and the combination of the new, simplified microphone and new headstage led to an overall increase in sensitivity of 24 dB. In tuned mode, the new instrument could detect a 40kHz tone of only 11dB SPL.
As a contrast to the sophistication (and expense) of the new S100 bat detector, we also developed a small Mini bat detector, based on a converted long wave pocket radio, fitted with a commercially produced electret hearing aid microphone. Although only providing tuned operation this instrument gave excellent value at only one-tenth the cost of the larger machine, and with the aid of detachable horn gives comparable sensitivity.
In 1967, Cahlander had described the construction of an analogue computer for the generation of sonar ambiguity diagrams and had used it to analyse several pulses from the interception sequence of Myotis lucifugus. The ambiguity diagram is a three-dimensional function, derived from an echolocation signal, with horizontal axes of range and velocity. The height of the function at a given point is the probability that a target will have the corresponding range and velocity coordinates. Thus a function with a single peak will give unambiguous determination of range and velocity, the width of the peak providing an estimate of the accuracy with which they can be measured. If there are several peaks of equal amplitude then it is equally likely that the (single) target is at any of the range/velocity co-ordinate pairs of those peaks. Sidelobe structure is also important since this determines the ability to resolve between different targets. If there are significant sidelobes then these could obscure the presence of a weaker echo at those range/velocity coordinates.
Clearly, this type of computer is invaluable for the analysis of echolocation signals, for while it makes several assumptions about the nature of the bat’s signal processor, and only applies to the information available from a single pulse, it does give some estimate of the sonar properties of the signal. Given the diversity of bats, and of their echolocation signals the benefits of this type of analysis are very large.
The ambiguity function is generated by taking two copies of the signal to be analysed, multiplying them together, and then taking the integral of the product. This is then repeated with one of the signals shifted slightly in time, until a complete slice of the diagram along the range (time delay) axis has been produced. One of the two signals is then given a slight Doppler shift and the process is repeated for each slice.
The two crucial steps in the process are the ability to delay one of the signals by successive increments and the other is to be able to generate the Doppler shifts. A means is also required for storing the signals while they are being processed, and of plotting out the final results. Cahlander used a rotating magnetic drum, upon which the two signals could be recorded and repeatedly replayed as the drum revolved. Doppler shifts were generated by copying one of the signals onto the drum with the tape recorder run at a slightly different speed each time, and the delays were produced by gradually moving one of the drums replay heads around the circumference of the drum. The necessary hardware for this process is quite complicated, especially the moving head required for the time delays.
Beuter, in Germany, approached the problem from a different direction and used a digital minicomputer to generate ambiguity diagrams and auto-correlation functions (equivalent to the zero doppler slice af the ambiguity function) of several different species of bats. Minicomputers sufficiently powerful to deal effectively with the large volumes of data involved are however very expensive. In the early 1970’s a suitable machine would have cost between 20 and 30 thousand pounds.
I realised that we had much of the equipment needed already to hand. An old Sonagraph had a magnetic drum which would enable a copy of the signal to be replayed repeatedly, and a Biomac 1000 signal averaging computer could be used to store each point in the function as it was calculated and plot each slice on an x-y or y-t plotter. When I pointed out that the, then new, charge coupled analogue delay lines could be used to generate the time delays we were almost there, and David Pye then suggested that by linearly varying the delay in the delay lines as the signal went through them, we could also generate the Doppler shift.
The detailed design and construction of an ambiguity diagram computer based on these principles formed the basis of a further appointment as research assistant to Prof. Pye. and echolocation and communication signals of several species were analysed on it. Since then other workers have improved and extended the machine for more detailed studies of certain species.
In 1980, after I had left Prof. Pye’s research group, I was asked to undertake a field trip to Thailand to carry out a survey of the very rare bat, Craseonycteris thonglongyai. This bat was first discovered in the Kanchanaburi province of Thailand in 1974 and was found to be not only the smallest mammal known (a full grown adult weighs less than 2g), but also represented a new family. This was the first new family of mammals to be discovered this century. At first found in three caves their numbers rapidly declined, until by December 1980 there were only 50 known individuals from a single location. Far fewer than the number of preserved specimens in museums. It was hoped that with the aid of bat detectors it would be possible to assess the existing numbers with a minimum amount of disturbance, and hopefully to find additional colonies of these bats.
In addition to an S100 and a Mini bat detector, I also designed a new style of instrument, especially for the trip. The existing bat detectors did not provide any means of recording the detected signals in a way which retained frequency information, except by using a high-speed instrumentation tape recorder, which was not available. Miller, in Denmark, had produced a bat detector which converted the signal into a square wave and then divided the square wave frequency by 10 to give a low-frequency square wave output, which could be recorded on a simple cassette recorder for later analysis. Although information about harmonic content was discarded the frequency of the dominant component could be determined. However, this instrument suffered from very poor sensitivity. This was largely due to the fact that the output signal was always a constant amplitude square wave, and it was, therefore, necessary to set the threshold for the front-end squaring circuit well above the background noise level. Otherwise one would hear full amplitude noise all the time. Also, the microphone used was an audio electret hearing aid microphone, similar to that used in the Mini bat detector, which was not ideally suited for use in broad-band detectors.
I modified the principle of the Miller count-down device to overcome the threshold problem by extracting the envelope of the original signal and using that to modulate the low-frequency output signal. It was then possible to set the front-end threshold well below the noise level, giving a marked improvement in sensitivity. I also used this circuit in conjunction with the S100 bat detector, thus reaping the benefits of the very sensitive front-end of that instrument.
My modified count-down circuit has since been incorporated into the S100 bat detector, replacing the old broadband mode, and is now sold as the S200. It proved to be very useful in the field, and recordings made in Thailand were later used to confirm that bats seen hunting in the evening are almost certainly Craseonycteris.
References
Cahlander, D. A. (1964). ECHOLOCATION WITH WIDE-BAND WAVEFORMS: BAT SONAR SIGNALS. Retrieved from http://www.dtic.mil/docs/citations/AD0605322
Cahlander, D. A. (1967). Discussion of Theory of Sonar Systems. In R.-G. Busnel (Ed.), Animal Sonar Systems: Biology and Bionics Vol II (pp. 1052–1081). Jouy-en-Josas.
Griffin, D. R. (1958). Listening in the Dark.
Hartridge, H. (1920). The avoidance of objects by bats in their flight. The Journal of Physiology, 54(1–2), 54–57. https://doi.org/10.1113/jphysiol.1920.sp001908
Hill, J. E. (1974). A new family, genus and species of bat (Mammalia: Chiroptera) from Thailand. Bull Br Mus (Nat Hist) Zool, 27, 301–336.
Maxim, H. (1912). The sixth sense of the bat. Sir Hiram Maxim’s contention. The possible prevention of sea collisions. Scientific American, Suppl Sept, 148–150.
Pierce, G. W. and Griffin, D. R. (1938). Experimental Determination of Supersonic Notes by Bats. J. Mammal., 19, 454–455.
Pierce, G. W. (1938). The Songs of Insects. Cambridge, Mass.: Harvard University Press.
Pye, D. (1980). Adaptiveness of echolocation signals in bats Flexibility in behaviour and in evolution. Trends in Neurosciences, 3(10), 232–235. https://doi.org/10.1016/S0166-2236(80)80071-3
