ABSTRACT
The opercular complex of amphibians functions to enhance perception of airborne environmental sounds below 1 KHz.
The columella and tympanum of frogs function in the perception of acoustic information above 1 KHz.
The opercular complex and amphibian papilla comprise the general hearing mechanism in amphibians.
The tympanum, columella and basilar papilla, present in totality only in frogs, are concerned with reproductive communication.
Interaction of the two systems in frogs provides a mechanism for enhancing input signal to noise ratio during chorusing.
Experimental determinations of acoustic perception in amphibians must be controlled for anaesthesia and forelimb disposition. The presence of anaesthesia or unnatural limb position can affect the animal’s ability to perceive low frequency sounds by preventing normal opercularis function.
INTRODUCTION
Seasonally, frogs are among the most soniferous of vertebrates. Biologically, vocalization is a major feature of the anuran adaptive radiation, for it is primarily by vocal means that the diverse species maintain their reproductive isolation. Thus, frogs would be expected to have a well developed sense of hearing, capable of dis-criminating between the various calls important to their behaviour. Early studies (van Bergeijk, 1957; Geisler, van Bergeijk & Frishkopf, 1964) examined the ear of the frog as if its function were analogous to that of mammals: to provide discrimination of all environmental sounds within a given sensitivity of frequency and intensity. The resultant model, though theoretically feasible, has not been substantiated and remains unsupported.
In 1959, Lettvin.et al. demonstrated that the eye of a frog did not provide a complete visual image of its environment but processed at the receptor (the retina) only events significant to the immediate well-being of the animal. This teleological approach, i.e. investigating how biologically significant stimuli would be processed, was then adopted in studies of anuran ear function, primarily in the bullfrog, Rana catesbeiana (see Frishkopf, Capranica & Goldstein, 1968, for a review of these studies). These investigations produced evidence that there are two classes of fibres in the auditory nerve of the bullfrog: (1) ‘complex’ units which respond to vibrations and low frequencies and are inhibited by the presence of certain other frequencies, and (2) ‘simple’ units which are spontaneously active, respond to higher frequencies, and cannot be inhibited. On anatomical evidence, the complex units have been associated with the amphibian papilla and the simple units with the basilar papilla (Frishkopf & Goldstein, 1963; Sachs, 1964; Frishkopf & Geisler, 1966). In R. catesbeiana, the best frequency response characteristics of these two groups of units were shown to correspond to the frequency bandwidths of maximum energy in the so-called mating call. This was interpreted to indicate that detection of significant sound was accomplished by the simultaneous activation of the two groups of units, i.e. the reason that frogs have two papillae is to process significant sounds at the receptor (analogous to the visual processing in the retina) (Capranica, 1965). To date there has been no investigation of how the frequency response characteristics of each of these sensory papillae are obtained.
The presence of two papillae is unique to amphibians. The basilar papilla and its associated accessory structures have been regarded as the precursors of the amniote cochlea, although the structure is simple in amphibians. The amphibian papilla is found only in amphibians and is generally larger and more complex than the basilar papilla (Paterson, 1960; Geisler et al. 1964). Further this is the only papilla in many salamanders (Lombard, 1971).
A second unique feature of the auditory system of most frogs and salamanders is the presence of two ossicles in the fenestra ovalis. One of these is the proximal expansion or foot plate of the columella and the other, lying more posteriorly, the cartilaginous or bony operculum. The columellar foot plate is connected to the tympanic membrane in frogs by the columella proper and its distal cartilaginous projection, the extracolumella (Col fp, Tymp, Col, Ex col, Fig. 1). In salamanders, which lack a middle ear cavity and tympanum, the columella attaches to the palatoquadrate of the suspensorium or squamosal bone. The columella, with its foot plate and extra-columella, is a derivative of the hyomandibular of fish and is homologous to the columella or stapes of amniotes. The operculum provides origin for a muscle which proceeds postero-dorsally to insert on the ventral face of the suprascapula (Op, Op m, ss, Fig. 1). This muscle is evident in all frogs and salamanders which have a terrestrial phase in their life-history. The operculum, a de novo structure in amphibians, is thought to be a freed portion of the otic capsular wall (Kingsbury & Reed, 1909), and has no homologue in other vertebrate groups.
While the function of the columella and tympanum are fairly obvious, the functional significance of the operculum – opercularis muscle complex is more obscure. Two general functions have been proposed: (1) that the complex functions as a link in an acoustical transmission route for perception of acoustic disturbances in the substrate via the forelimbs (Kingsbury & Reed, 1909), and (2) that the complex functions in the maintenance of balance via the vestibular system by monitoring body position (Ecke, 1934). The former proposal has been widely accepted with little criticism and no evidence for many years and is presented as fact in many texts (Romer, 1970; Goin & Goin, 1971; Porter, 1972). The later hypothesis, though less widely accepted, has received some supportive evidence (Baker, 1969). In this paper we report on the role of the opercularis muscle and other middle ear structures associated with it.
MATERIALS
Specimens of the following species were used: Hyla regida, 6; H. versicolor, 2; H. cinerea, 3, and Smilisca baudinii,, 2, of the family Hylidae; and Leptodactylus melanonotus, i, of the family Leptodactylidae. The H. regilia were collected from Catalina Island and San Dimas Canyon, Los Angeles Co., California. The H. versicolor specimens were taken from east Texas populations, the H. cinerea from the vicinity of Tampa, Florida, and the Smilisca from the Yucatan Peninsula of Mexico. The Leptodactylus was collected in Baja California, Mexico. Accurate field data for all animals have been recorded in personal files.
METHODS
Surgical procedures
All animals were anaesthetized for surgery by immersion in an 0·5 % solution of urethane. After surgery, full recovery from anaesthesia was generally allowed before physiological recording was attempted. For single unit experiments the eighth nerve was exposed at its entry into the brain case by dissecting away a small region of the oral epithelium and connective tissue and a portion of the braincase floor from the roof of the mouth. The nerve sheath was then dissected off. For recording from the midbrain, a triangle of skin, with the apex midway between the eyes and the base slightly behind the suprascapula was folded back over the body, thus exposing the roof of the brain case. By drilling out the bone with a dental drill or dissecting off the cartilage, the dorsal surfaces of the optic tecta were exposed. The dura was split medially and pulled off to both sides. Circulation to the brain could be monitored in the medial cerebral vein, which was always left intact. In animals used repeatedly, the flap of skin was returned to the normal position between experiments. Post-operative frogs showed no discernible impediments.
Experimental surgery was performed simply by: (1) severing the opercularis muscle at its attachment to the suprascapula, (2) severing the nerves to the muscle peripheral to branches to other structures, or (3) dissecting out the tympanum from the surrounding tympanic annulus. Three additional experimental manipulations, not involving surgery were performed during threshold/frequency determinations: (1) adpression of the forelimbs, (2) keeping the animal under anaesthesia, and (3) isolating the animal from the substrate. Anaesthesia was maintained by covering the animal in an absorbent tissue soaked in 0·5 % urethane. Isolation from the substrate was accomplished by placing the animal on a 5 cm thick cotton pad. This experimental series was designed to counterfit possible traumatic effects resulting from surgery in the experimental procedures. It also enabled us to record the overall response of the receptor while making changes in the peripheral structures without having to repeat the procedure many times and sum the responses from many experiments. In addition, advantage could be taken of the apparent total crossing of eighth nerve projections to the tori. This enabled unilateral surgery so that the animal could provide its own control under experimental conditions.
Care was taken at all times to ensure that the animal had fully recovered from surgical anaesthesia. Any degree of anaesthesia during threshold-frequency determinations often resulted in loss of the low frequency sensitive region. Two control curves determined under anaesthesia are shown in Figs. 7 and 8. In both the low frequency sensitive region is absent or poorly developed.
For all midbrain recordings the unanaesthetized animals were restrained only by a pin through the nasal region. By keeping the animals moist and running the experiments in the dark, little problem was encountered with movement. Finally, the position of the forelimbs was maintained in a normal position at all times except under certain experimental conditions. Random or uncontrolled placement of the forelimbs often resulted in the loss of the low frequency sensitive region as shown in Fig. 14. For the single unit recordings where the animal was on its back, the forelimbs were manipulated until a strong low frequency response was achieved. They were then held in position with pins. In this experiment the animal was restrained by a wet tissue blanket as well as pins.
Experimental set-up
All physiological procedures were conducted in an acoustically, vibrationally, and electrically isolated chamber measuring 1·25 × 0·75 × 0·75 m internally (Fig. 2). The walls were laminated wood and acoustic tile and lined with fibre wool insulation. Over the operating frequency and intensity ranges used no difficulty with echo interference or standing waves occurred. A loud speaker (J. B. L. LE8T) flat (± 1·5 dB) over the operating range 1–5 KHz was centrally mounted in one end. A heavy metal stand (1, Fig. 2) isolated from the floor of the chamber by a metal-rubber-metal vibration absorbing sandwich (2, Fig. 2) was fixed in the middle of the chamber. Fixed to this stand was a dissecting microscope, a micromanipulator (Narashigi M3) (Elec, Fig. 2) and monitoring microphone (B and K 4131) (Mic, Fig. 2). The microscope was fitted with an ocular grid which was used to centre the frog in a constant position relative to the speaker and to position the microelectrode in the midbrain. With the electrode holder kept at a constant angle, we were able to obtain accurate and repeatable electrode placement. A rotatable paraffin platform attached magnetically to the stand enabled orientation of the experimental animal (Pit, Fig. 2). The microphone was placed as close to the animal as possible and equidistant from the speaker (about 60 cm).
A block diagram of the experimental apparatus is shown in Fig. 3. Sound signals were produced by gating a pure tone from a sine wave generator (Hewlett Packard 5D6) through an audio keyer. The duration and repetition rate of sound pulses could be varied but in all experiments a pulse of 20 msec duration repeated every 0-5 sec was used. Rise and fall times varied with frequency to produce a smooth transient free of ‘popping’. The signal was fed through the audio amplifier of a Crown 8000 series tape recorder to the loud speaker and to one channel of the master oscilloscope (Tektronix 502d) for continuous monitoring.
The microelectrodes were insulated tungsten with a 1–3 μ diameter tip for recording compound action potentials in the midbrain and with a 0–5 μ tip for recording single units extracellularly from the eighth nerve. The electrode was followed by a pre-amplifier (Bioelectric BFI) and a differential amplifier (Tektronix 122) which then fed into the master oscilloscope. The signal could be monitored by audition through a speaker and further analysed by a computer of average transients (C. A. T. Instrument Co.).
The monitory microphone was connected to a precision sound level meter (B and K 2003) and a band pass filter (B and K 1613) for recording sound intensity levels at the animal’s ear during experiments. The sound chamber was maintained at a constant temperature (19 °C) during the experiments.
Experimental procedure
Experimental studies of amphibian middle ear components are complicated by the major difficulty in maintaining the mechanical auditory system intact while recording neurophysiological responses in the auditory pathway. The eighth nerve is not accessible dorsally in most frogs without surgical damage to part of the otic capsule. Normally, to avoid this problem, a ventral approach to the auditory nerve is employed. However, this involves positioning the animal on its back and exposing the nerve through the roof of the mouth. Since the opercularis complex can be affected by the relative disposition of the skull and shoulder girdle (see below), such an unnatural posture is better avoided in functional studies.
In examining the neural response to acoustic stimuli, three major techniques have been applied in frogs: (i) recording single units from the midbrain (Potter, 1965), (2) recording single units from the eighth nerve (Frishkopf & Goldstein, 1963), and monitoring midbrain slow wave evoked responses (Loftus-Hills & Johnstone, 1969). Our technique is a combination of those of Potter and Loftus-Hills & Johnstone. To measure overall response, we used the compounded potentials at a group of units rather than single units or the slow wave summation. Potter found that units in the torus produced simple responses to sound stimuli. The midbrain is readily accessible dorsally after removing part of the calvarium without any damage to the functional components of the middle or inner ear. The animal can then be positioned to reflect the normal disposition of the middle ear components. To determine the overall response of the entire system, summation of responses from many single units is required. We found that in the torus we could record either single units (as extracellular action potentials) with fine electrodes (0·5 μ) or multiunit extracellular responses with coarser electrodes. Even with the coarse electrodes, at low stimulus intensities the number of units in a multiple recording was reduced. Near threshold, such recording became indistinguishable from single unit recording made with fine electrodes. Potter (1965) found that single units vary considerably in threshold, latency and peak potential such that specific units could be sorted out from a multiple recording. Since we were primarily interested in overall response, the multiplicity of units was always maximized for each electrode placement.
To see if this measure was comparable to single unit recordings in more peripheral levels of the pathway, we conducted a series of experiments to determine ‘best frequencies’ obtained by responses in the brain and compared these to ‘best frequencies’ obtained from single units in the eighth nerve using H. cinerea. The experimental procedures for both determinations were the same except for the orientation of the animal (on his back for single unit recording) and size of electrode. H. cinerea is a much larger and more robust animal than our principal experimental species H. regilla and is capable of enduring a longer series of experimentation. The results (Fig. 5) show that our measure of ‘best’ frequency is indeed equivalent to that obtained from summation of single unit responses in the auditory nerve and has the advantage that the full range of the audiogram can be used to detect changes in sensitivity due to experimental manipulations.
After surgery and gross electrode placement, a search stimulus of either normal and pulses of an appropriate frequency (if the response characteristics were already known) or of pulses of white noise was presented at relatively high intensity (70–80 dB re 0–0002 dynes/cm2). For midbrain recording the visual cortex and ventricular space were then penetrated to gain the auditory region of the torus semicircularis. When a strong compound action potential (or an extracellular action potential of a single unit in the nerve) was encountered, the search stimulus was switched to the experimental series which ran from 1 to 5 KHz in 0·1 KHz intervals. At each frequency the sound intensity was raised until the response was evident, then gradually lowered until it disappeared. The sound intensity as measured in decibels (re SPL 0·0002 dynes/cm2) by the sound level meter at this point was recorded as threshold for each frequency. Disappearance of the response was judged by visual and auditory means. To remove subjective bias, replications were always made with alternate observers. This procedure was tested against the more time-consuming, but potentially more accurate method of Loftus-Hills & Johnstone, i.e. reducing the sound intensity 2 dB at a time and averaging 50 responses at each intensity level with the computer of average transients. With a good signal to noise ratio, we found this procedure to be unnecessary for accuracy or repeatability, and it was used only in rare cases when clear signals were not available.
The response signal (Fig. 4) was essentially a temporal summation of the midbrain single units described by Potter. The initiation of the response had an average latency of the order of 14 msec. Depending on the intensity of the stimulus and degree of electrode isolation, the length of the response was from 1–2 msec up to 40 msec. The maximum amplitude of response exceeded 500 μV and at threshold was about 50 μV. Background noise levels were usually about 35 μV.
Statistical treatment was performed after converting dB values to dynes/cm2. The means were then converted back to dB values for graphic display.
RESULTS
The threshold/frequency curves determined show, in control and experimental situations, consistent major features in the animals examined. Though some variation in shape and absolute threshold is evident when comparing details even in the curves from animals of the same species (for example, H. regilia, Figs. 6, 9, 13, 14), we do not intend to examine them. Instead, focus will be maintained on the general aspects of curve shape, both inter- and intraspecifically. No differences in ability to hear were evident in the data from right to left ears in a given animal. This general finding is shown in Fig. 5. In all subsequent graphs the data from both ears have been lumped. No differences in ability to hear at any frequency were evident in situations where the animal was isolated from the substrate (right ear determination in Fig. 5). This experiment was performed with the other species used in a random manner. At no time were any differences in perceptive ability noticed. Data from these determinations were integrated with that from substrate contact determinations for presentation.
Control threshold/frequency curves
In control situations two broad regions of maximum sensitivity are seen, one centred below 1 KHz and the other centred higher, around 2–3 KHz (solid lines, Figs. 5–14). Exceptions to this pattern can be found in Figs. 7 and 8, where the lower sensitivity region is missing. These discrepancies will be discussed in a later section. A general feature of these two ‘peak’ regions of sensitivity is a relatively lower threshold in the lower frequency region. This is seen in Figs. 5, 6, 9, 10, 11 and 12. This feature is not evident in Figs. 7 and 8 due to the absence of a low frequency region. In Fig. 13 the minimal threshold is lower in the high frequency region, and in Fig. 14 the thresholds are equivalent. It is evident from the stippled range limits about the solid lines that a fair degree of consistency was possible from one determination to the next. For this reason only the curves determined by the mean response at a given frequency will be discussed.
H. cinerea, Figs. 5, 8 and 11. In H. cinerea the low frequency sensitive region lies between 0·3 and 1·6 KHz (Figs. 5 and 11). The higher frequency sensitivity region lies between 2·3 and 3-5 KHz. There is a broad region of lessened sensitivity between 1·6 and 2·3 KHz. From the regions of ‘peak’ sensitivity the animal’s ability to hear falls rapidly at both the high and low ends of the spectrum. In Fig. 8, this general pattern is confused by the absence of a broad low frequency sensitive region.
H. regilla, Figs. 6, 9, 13 and 14. In H. regilia the low frequency sensitive region lies between 0·1 and 1·0 KHz. The higher frequency sensitive region lies between 1·0 and about 3·3 KHz. The higher frequency sensitive region can be further subdivided into two subordinate sensitivity regions. The lower of these lies between about 1·5 and 2·2 KHz and the higher between about 2·4 and 3·0 KHz. From the regions of ‘peak’ sensitivity the animal’s ability to hear falls rapidly at both the high and low ends of the spectrum.
H. versicolor, Fig. 7. The curve presented for H. versicolor shows a higher frequency sensitive region between 1·0 and 3·4 KHz. Sensitivity appears at its maximum between 1·6 and 2·1 KHz. A lower frequency sensitive area is not apparent in this curve. In other H. versicolor studied but not presented here, a lower frequency sensitive region occurred between 0·1 and 0·9 KHz. In these cases the animal’s ability to hear falls off rapidly at both the high and low end of the spectrum. This same sensitivity ‘fall-off’ is evident at the high frequency region of Fig. 7.
L. melanonotus, Fig. 12. In L. melanonotus the low frequency sensitive region lies between 0·1 and 1·5 KHz. The higher frequency sensitive region lies between 0·1 and 1·5 KHz. The higher frequency sensitive region lies between 1·8 and 3·4 KHz. In the high region, sensitivity appears at a maximum between 2·6 and 3·4 KHz. Though the sensitivity ‘fall-off’ from the low frequency region toward lower frequencies is rapid, that of the high frequency region toward higher frequencies is less pronounced. Indeed, some minor areas of relatively greater sensitivity centred at 3·7 and 4·3 KHz are evident.
S. baudinii, Fig. 10. In S. baudinii the low frequency sensitive region lies between 0·1 and 1·1 KHz. The higher frequency sensitive region lies between 1·6 and 2·7 KHz, with maximum sensitivity at 2·1 KHz. There is a region of lessened sensitivity between 1·1 and 1·6 KHz. From the region of ‘peak’ sensitivity the animal’s ability to hear falls off rapidly toward lower frequencies but more gradually toward higher frequencies.
Experimental threshold/frequency curves
Single units
Fig. 5 shows a comparison between threshold/frequency data obtained from the tori semicirculares and single unit summations from the right eighth nerve of the same animal. There is good correspondence between the number of single units found maximally responsive to a given frequency, and the sensitivity of the animal, determined at the midbrain, to the same frequency. Where the eighth nerve evidenced a greater number of single units responsive to a given frequency, the midbrain showed maximum sensitivity. Where the eighth nerve produced fewer or no single units, the midbrain indicates a lesser sensitivity. Regions of maximal sensitivity by either method occur at 0·5, 1·0 and 3·0–3·4 KHz.
Tympanum
Threshold/frequency curves before and after bilateral removal of the tympana are shown in Fig. 6 for H. regilia and Fig. 7 for H. versicolor. Threshold/frequency curves from a H. cinerea with a small, malformed left tympanum are shown in Fig. 8. In all three cases the maximal reduction in sensitivity occurs at higher frequencies, generally above 1 KHz. Though sensitivity loss does occur below 1 KHz, it is irregular, in-consistent and less pronounced.
Comparison of the control and experimental extremes at a given frequency below i KHz in Fig. 6 (vertical bars: experimentals; stippled envelope: controls) shows a fair degree of overlap. This is also the case in H. cinerea (Fig. 8) where there is very little difference in the curves below 1 KHz. In H. versicolor (Fig. 7), the loss appears to increase irregularly from about 0·2 KHz but does not reach a consistently high level until 1 KHz. In all three animals the loss in sensitivity is fairly constant once a maximum level is achieved. Comparison of the control and experimental curves in a given animal shows that once a maximum differential occurs the two curves are to the same shape, i.e. both contain the same ‘peaks’ and ‘valleys’, even in detail. In general the pattern shows three phases: (1) an irregular small loss in sensitivity below 1 KHz, (2) a transitional zone between 1 and 2 KHz where sensitivity loss increases fairly regularly, and (3) a zone > 2 KHz where the loss is maximal and consistent.
Opercularis muscle
Comparative threshold/frequency curves, before and after experimental interference with normal opercularis function, are shown in Figs. 9–14. In all cases, maximum loss of sensitivity occurs below 1 KHz after ablating normal muscle action (dashed lines). Loss of sensitivity above 1 KHz is small and irregular, if present, and not significant. In general, the sensitivity loss is maximum with experimental manipulation in the region of ‘peak’ sensitivity under non-experimental conditions. The sensitivity loss also generally falls off rapidly on either side of the maximum loss. In S. baudinii (Fig. 10) and H. cinerea (Fig. 11), however, the loss in sensitivity decreases much less precipitously from the maximum towards the higher frequency side.
(a) Muscle transection
Figs. 9–12 show the results of surgically cutting the operodaris muscle. In H. regilla (Fig. 9) the operation was bilateral and in 5. baudinii, H. cinerea and L. melanonotus (Figs. 10–12) the operation was unilateral. In all cases the sensitivity loss was at a maximum below 1 KHz (dashed lines). Also, maximum loss occurs in the region of optimum low frequency sensitivity. Loss above 1·5 KHz is minimal and not significant. Between 1 KHz and (1) 1-5 KHz in H. regilla (Fig. 9) and L. melanonotus (Fig. 12) and (2) 2 KHz in 5. baudinii (Fig. 10) and H. cinerea (Fig. 11), there is a zone of decreasing sensitivity loss.
In those cases where the muscle section was unilateral, only recordings from the contralateral torus semicircularis indicated a sensitivity loss (dashed lines, Figs. 10, 11 and 12). Recordings from the ipsilateral torus in each case indicated no deviation from normal (dotted lines Figs. 10 and 11). The mean ipsilateral response is seen to lie almost entirely within the envelope of extremes of non-experimental conditions at all frequencies. Ipsilateral recordings for H. regilla (Fig. 9) are not shown because the animal expired during that phase of the experiment.
(b) Nerve transection
The effect of bilateral surgical section of the opercularis nerve supply is shown in Fig. 13. Normal disposition of the operculares muscles was not altered during surgery. The sensitivity loss is at a maximum below 1 KHz (dashed line). The maximum loss also occurs in the region of optimum low frequency sensitivity. Loss of sensitivity above 1 KHz is irregular and insignificant. Above 1 KHz the experimental threshold/frequency curve lies almost entirely within the envelope of extremes found under non-experimental conditions. The region of maximum loss conforms to that found with experimental section of the opercularis muscle (Fig. 9), and positional alteration (Fig. 14) in the same species.
(c) Postural modification
The effect of postural modification (bilateral adpression of the forelimbs) on the threshold/frequency curve of H. regilia is shown by the dashed line in Fig. 14. The sensitivity loss is at a maximum below 1 KHz and occurs in the region of optimum low frequency sensitivity under non-experimental conditions. Sensitivity loss above 1 KHz is very irregular and difficult to generalize. In most instances the experimental curve of means lies within the range found under normal conditions. Reversal of the postural modification eradicates the sensitivity loss (dotted line). The threshold/frequency curve after regaining normal posture is, with a couple of minor exceptions, completely coincident with the range of pre-postural change curves. The region of maximum loss under experimental conditions conforms to that seen using other methods of interfering with normal opercularis function in the same species (Figs. 9 and 13).
(d) Anaesthesia
The effects of anaesthesia on the threshold/frequency curves of H. versicolor and H. cinerea are shown in Figs. 7 and 8. In neither case is a low frequency sensitive region clearly discernible. A low frequency sensitive region is present in unanaesthetized animals of the same species (H. cinerea, Figs. 5 and 11, H. versicolor not shown).
DISCUSSION
Functional and evolutionary interpretations
Controls in all species examined showed a clear-cut division of the auditory response into two basic components as had been found previously (Capranica, 1965; Potter, 1965; Loftus-Hills & Johnstone, 1969). There is a low frequency zone below 1 KHz that has been associated with the amphibian papilla, and a high frequency zone, above i KHz that has been associated with the basilar papilla. That is, low frequency sounds and vibrations pass through one receptor and high frequencies through another. Our results show that the middle ear components also function differently in respect to each of these ‘channels’.
Interference with the tympanum – columella system does not appreciably affect sensitivity (as measured in the auditory centre of the brain) to low frequency but does significantly lower the sensitivity to high frequencies. Conversely, there is a marked drop in sensitivity to low frequencies when normal functioning of the opercularis system is prevented, but response to frequencies above 1 KHz is hardly affected. Capranica (personal communication) has reported that sensitivity (recorded as micro-phonics in the basilar papilla) is unaffected by severing the opercularis muscle, confirming that the high frequency ‘channel’ is not dependent on the opercularis system. A property that might be expected of these systems is that they should operate at peak efficiency at a particular frequency. This frequency would depend on their force transfer efficiency and time constant characteristics as determined by physical dimensions. This would provide some degree of frequency tuning that is characteristic of anuran ears. Our experiments show, however, a consistent loss of sensitivity in the higher frequency part of the audiogram with tympanum – columella complex interference. If there was tuning, removal of the system would be maximally effective at the tuned frequency and lower at frequencies away from this point. In the low frequency portion, affected by manipulations of the opercularis system, the best frequency of hearing is maximally affected and the degree of change falls off sharply at the lowest frequencies tested. In fact, the very low frequencies were almost un-affected, and as far as we could tell without quantitative measure, neither was the response to vibrations. If this can be verified quantitatively it will likely show that the opercular system is not necessary for vibration detection (as suggested by Smith, 1968). Rather, the complex functions only to provide a low frequency airborne sound channel with peak efficiency between about 0·3 KHz for large frogs such as R. cates-beiana and 0·7 KHz for small species such as H. regilla.
The direct connexion between an intact opercularis system and normal response in the low frequency-vibration zone would appear to support the hypothesis of Kingsburg & Reed (1909): that the opercularis muscle provides a mechanical link between shoulder girdle and inner ear for the transmission of low frequency vibrations from the substrate by way of the forelimbs. However, animals isolated from the substrate by a non-transmitting pad showed no alteration in their response at any frequency. Also, when making single unit recordings from the auditory nerve; i.e. when the limbs were not in contact with the substrate, sensitivity to low frequencies was the same as that for normal postures.
If the opercularis complex is not a direct transmission line for sounds then it must function in some other fashion to enhance the reception of low frequencies at the inner ear. Fig. 1 shows a reconstructed opercularis complex and columella demonstrating the interlocking nature of the operculum and the columellar foot plate. Tension in the muscle would pull the operculum into locked position with the columellar foot plate producing a coupled plate occupying the entire oval window. The fact that removal of the tympanum and the distal portion of the columellar shaft does not severely reduce sensitivity to low frequencies indicates that low frequencies impinge on the oval window with about equal force whether or not there is direct coupling to the tympanum. Under this condition, the efficiency of force transfer of the sound pressure to fluid displacement in the inner ear appears directly proportional to the surface area of the plate in the oval window. In addition, the mass of the plate also increases with coupling, thus producing an increase in inertia of the now compound plate. Finally, the tension of the opercularis muscle would conceivably enhance this increase in inertia by ‘stiffening’ the middle ear complex. Since displacement acceleration is slow at low frequencies the compromise of increased efficiency of force transfer for slower reaction time can be met. As the frequency increases so does displacement acceleration and at some point loss of responsiveness due to inertia outweighs the gain due to high force transfer efficiency. Also, as the frequency increases the sound pressure at the oval window decreases if no mechanical link to the outside is provided, as indicated by high frequency loss on removal of the typanum. Therefore if the oval window plate were a single unit (of area equivalent to the two elements) it could not serve both low and high frequency transmission with equal efficiency.
By uncoupling the two elements (as represented in our experiments by severing the muscle or otherwise preventing its normal state of tension), the columellar foot plate is free to move independently of the operculum. The mass of the moving plate is decreased, thus reducing inertia and allowing responsiveness to greater displacement acceleration (i.e. higher frequencies). Also, the system gains mechanical advantage of the order of the ratio of the tympanum area to the area of the columellar foot plate, which is of little consequence when the area of the inner plate is not small relative to tympanic area (as when the two elements are coupled). It may be, thus, that the apparent near equality of force transfer at the oval window and without this mechanical link is due to the fact that when the oval window elements are coupled there is little mechanical advantage derived from the link. Thus the middle ear components (tympanum-columella system and the opercularis complex) provide alternative mechanical efficiencies to the transmission at low or high frequency sounds, either being brought into play simply by a change in tension in the opercularis muscle. Reception at low frequencies is enhanced by coupling the operculum and columellar foot plate into a single oval window plate of large surface area which has high force transfer efficiency but slow reaction time. Removal of opercularis muscle tension uncouples the two oval window elements, allowing the efficient transmission at high frequencies through the tympanum-columella system, which has a faster reaction time. The structures involved are so small in the animals themselves that empirical tests of the relationships between force transfer efficiency and area of the oval window plate, frequency, inertia, and mechanical advantage cannot easily be made directly on the animal. We are at present engaged in both comparative studies and experiments with dynamic models to more directly deal with this problem.
Our functional interpretations can be tested indirectly by comparison of the relative development and distribution of the various functional components in different groups of amphibians that inhabit different major adaptive zones. The role of the high frequency tympanum–columella–basilar papilla system in anuran communication has been well documented (see Straughan, 1973). The role of low frequency opercular complex-amphibian papilla is not so clearly defined. Because of its response to vibratory stimuli and its presence in terrestrial salamanders (sensu lato), which in general do not communicate acoustically, its primary function is likely monitoring environmental sounds in terrestrial situations. A best low frequency response (0·2–0·5 KHz) is common to both frogs and salamanders (Lombard, in preparation) and corresponds to the frequency bandwidth of potentially important ambient sounds (Hardy, 1956). In salamanders this response cannot be associated with vocalizations because they are largely mute (Maslin, 1950). In an evolutionary sense, we believe that in terrestrial situations the primary function of the amphibian auditory system is the detection of general environmental sounds and the sound communication has been a later derivation in the Anura. This interpretation is supported by the ontogeny and relative distribution of this system in the amphibia. Temporally, the appearance of a complete opercular complex is essentially a metamorphic event. Thus, no amphibian larvae possess this system. This supports the notion of Kingsbury & Reed (1909) that the opercular complex is an adaptation for terrestrial perception. In addition, all paedogenetic urodeles and some totally aquatic members of derived families which have no terrestrial phase fail to develop the opercular complex (Kingsbury & Reed, 1909; Dunn, 1941; Monath, 1965). Among the Anura all members of the Liopelmatidae and Pipidae (Wagner, 1934; Stephenson, 1951) and some aglossids (de Villiers, 1932; Sedra & Michael, 1959) lack an opercular complex. All of these animals are primarily adapted to an aquatic existence. There are apparent exceptions to these generalities. Many frogs and salamanders which spend a great deal of time in aquatic situations such as Rena and Ambystoma retain an opercular complex. Where this is true, however, the animal also spends a fair amount of time on land where an opercular complex would be adaptive.
In anurans which have secondarily lost their high frequency communication channel, as in some species at high elevations, many aquatic (McDiarmid, 1971), and some burrowing forms (Emerson, 1970), the corresponding functional components – the tympanum and columellar shaft-become reduced or lost. In two such species examined, Bufo periglenes and Kalula pulchra, the low frequency threshold was equivalent to other ‘vocal’ species of the same size, but high frequency response was completely lacking (Straughan and Lombard, unpublished data).
A parallel situation exists in the Caudata where terrestrial acoustic communication has never been demonstrated. Completely terrestrial adults are not universal in this order but at least two independent terrestrial lines have evolved (Wake, 1966). The presence of an apparatus sensitive to environmental sound stimuli in air would be an adaptive advantage to these groups. Since they have independently arisen from different primarily aquatic ancestors (which would have no need for such a device) each line has developed an opercular complex, similar in basic form but differing in detail. There has thus been a convergence towards the development of the same functional system. The major difference seen involves the derivation of the opercularis muscle. In the Plethodontidae the opercularis is a derivative of the cucullaris. In other terrestrial forms the muscle is a derivative of the levator scapulae (as it is in the Anura) (Dunn, 1941; Monath, 1965). This strongly supports the notion of the adaptive importance of this system to terrestrial amphibians. In salamanders, where there has been no development of an alternate sound transmission channel (such as a typanum), the system is open to evolutionary modification resulting in a single uncoupled plate suitable for low frequency transmission only. On the evidence available (Monath, 1965) it would appear that such an evolutionary modification has been the rule in terrestrial salamanders.
Finally, presumptive interactions of the two perception systems as they exist in frogs needs comment. During reproductive chorusing an ability to inactivate the low frequency system would confer a distinct advantage. With a decrease in sensitivity to ambient environmental sound, the ability to perceive specific reproductive information would be relatively enhanced. That is, the total system would gain a more favourable input signal to noise ratio. Our experiments show that a 20–30 dB loss in sensitivity results at the optimum low frequency when the opercular complex is not functioning This represents a considerable drop in sensitivity. The higher frequency region, which is matched to preponderance of energy in the call, is not affected by inactivation of the opercular complex. The animal thus has a system favourable to ‘focusing’ on species specific information during reproductive interactions.
ACKNOWLEDGEMENTS
The bulk of this research was supported by grant 71-2112 from the AFOSR administered by the University of Southern California. Terminal stages of the work were supported by grant-SOi-RR-05367-11 from the USPHS administered by the University of Chicago. The technical help of L. G. Bishop, D. R. Dvorak, L. Masuoka and C. Hillary, the University of Southern California, is gratefully acknowledged. The efforts of R. McDiarmid and R. Harris in sending specimens of H. versicolor, H. cinerea, S. baudinii and L. melanonotus to us are gratefully acknowledged. The illustrations were prepared for publication by M. Oster, Department of Anatomy, University of Chicago. Special thanks are due D. B. Wake and J. A. Hopson for critical reading of the manuscript.