Humans near some types of pulsed radar systems have perceived individual pulses of RFR as audible clicks (without use of electronic receptors). This phenomenon, first investigated by Frey (1961), attracted much interest because it has been cited often as evidence that nonthermal effects can occur and because an initial hypothesis was that a possible mechanism for perception is direct stimulation of the central nervous system by RFR.

Many of the results support the hypothesis that a pulse of RFR having the requisite pulse power density and duration can produce a transient thermal gradient large enough to generate an elastic shock wave at some boundary between regions of dissimilar dielectric properties within the head, and that this shock wave is transmitted to the middle ear, where it is perceived as a click. Persons with impaired hearing are unable to hear such clicks, and experimental animals in which the cochlea (inner ear) has been destroyed do not exhibit brainstem-evoked responses.

Although this topic is presented here under "Studies of Humans," the results of experiments with animal subjects and nonbiological targets are also discussed, to avoid fragmenting the descriptions of several studies that involved both human and animal subjects.

Frey (1961) exposed human subjects to either 6-microsecond pulses of 1.3-GHz RFR at 244 pps (0.0015 duty cycle) or 1-microsecond pulses of 3.0-GHz RFR at 400 pps (0.0004 duty cycle). The ambient noise levels were respectively about 70 and 80 dB, but earplugs decreased the noise by about 25-30 dB. The mean threshold of average power density for RFR perception was about 0.4 mW/sq cm at 1.3 GHz for eight subjects and 2 mW/sq cm at 3.0 GHz for seven subjects. The corresponding peak power densities were about 270 and 5000 mW/sq cm. (No variances or other statistical data were given.)

The subjects were unable to match the RFR sounds to audio sine waves. With white noise controlled by a variable band-pass-filter, best match was obtained by removing all frequencies below about 5 kHz.

Four subjects with various degrees of hearing loss (for air-conducted and bone-conducted sound) were tested for perception of the 1.3-GHz RFR. Subject 1 with significant hearing loss of both kinds above about 2 kHz was unable to perceive the RFR sound at intensities 30 times above the threshold. Subject 2, who had bilateral severe air-conduction loss (about 50 dB) but moderate bone-conduction loss (about 20 dB), was able to perceive the RFR sound at about threshold level. Subject 3, with tinnitus and bilateral hearing loss ranging from about 10 dB at 250 Hz to about 70 dB at 8 kHz for air conduction, more severe loss for bone conduction, and who had been diagnosed as having neomycin-induced nerve deafness, was unable to perceive the RFR. Subject 4, who had normal bilateral air-conduction hearing to about 4 kHz but severe bilateral bone-conduction loss was also unable to perceive the RFR.

The author had concluded that a necessary condition for perception of RFR as sound was the ability to hear sound above about 5 kHz, but not necessarily by air conduction. Frey (1962), however, stated that some subjects with an audiogram notch around 5 kHz (i.e., with adequate hearing for frequencies above 5 kHz) did not perceive RFR sound. In this later study, the RFRs used were 425-MHz pulses of 125, 250, 500, 1000, and 2000 microseconds at 27 pps (respective duty cycles of 0.0034, 0.0068, 0.0135, 0.027, and 0.054) and 2.5-microsecond pulses of 8.9-GHz RFR at 400 pps (duty cycle 0.001). The ambient noise levels were 70-90 dB and the subjects wore Flent antinoise ear stopples, which diminished the ambient levels by about 20 dB from 100 Hz to 2 kHz and by about 35 dB at 10 kHz.

The average-power-density thresholds for perception of 125-, 250-, 500-, and 1000-microsecond pulses of 425-MHz RFR were 1.0, 1.9, 3.2, and 7.1 mW/sq cm, respectively; the corresponding peak power densities were 300, 280, 240, and 260 mW/sq cm (again given without statistical data). (The threshold for 2000-microsecond pulses of 425-MHz RFR was not determined because of inadequate instrumentation.) Thus, all four 425-MHz peak-power values were comparable to the 1.3-GHz threshold previously found (Frey, 1961), implying insensitivity to frequency in this range. The 3-GHz peak-power threshold was much higher, however, about 5 W/sq cm, and the 8.9-GHz RFR was not perceived for peak-power densities as high as 25 W/sq cm. The author suggested that the perception-threshold rise from 1.3 GHz upward was related to the dependence of penetration depth on frequency. The author, noting the high ambient noise levels, also suggested that the thresholds would be lower in quieter environments.

Frey (1962) also speculated about the possible sites and mechanisms of detection of pulsed-RFR, including RFR-induced changes of the electrical capacitance between the tympanic membrane and the oval window, detection in the cochlea, and interaction of RFR with neuron fields in the brain. The first possibility was discounted because of insensitivity of the RFR-hearing effect to head orientation relative to the RFR source. He indicated that the then-current experimental results were inconclusive relative to the other two possibilities.

Frey (1967) endeavored to resolve this point by studying the potentials evoked in the cat brain by exposure to 10-microsecond pulses within the range 1.2-1.5 GHz. One electrode was implanted in the brain stem of each cat, with tip in the nucleus subthalamicus, formatio reticularis, nucleus olivaris, or nucleus reticularis paramedianus. The electrode was of coaxial design to avoid RFR energy pickup. (The author stated that during extensive testing, the electrode showed no indication of energy pickup.)

The procedure after recovery from surgery (4-6 weeks) was to anesthetize the cat with Fluothane in oxygen, place it in an exposure chamber lined with RFR-absorbent materials, adjust the anesthesia to a depth that just prevented voluntary movement, and evaluate the status of the electrode by monitoring the pattern of brain electrical activity for a short time. The head of the cat was then exposed to the RFR from a horn 50 cm above, and the first 30-ms of brain-stem output after each pulse was recorded and averaged over successive pulses with a computer. Synchronization pulses at appropriate repetition frequencies were used to trigger the RFR generator and (with a 1-ms time delay) the oscilloscope, recorder, and computer used for assessing the evoked responses.

During experimental sessions, exposures of each cat to the RFR were alternated with 5-min rest periods. Interspersed with the RFR-exposures were runs with CW or pulsed acoustic stimuli at the repetition frequency of the RFR, but no details were given regarding how such stimuli were presented. The electrical activity in each brain region was recorded just before, and a few minutes after, euthanizing the cat.

The author indicated, without giving data, that the threshold average and peak power densities necessary to evoke brain-stem potentials were about 0.03 and 60 mW/sq cm. These values correspond to a duty cycle of 0.0005 and (for 10-microsecond pulses) to a pulse repetition frequency of 50. The results presented were representative averaged waveforms evoked in the four brain-stem regions by the RFR and pulsed acoustic stimuli. These showed that responses were evoked by both stimuli before but not after euthanization, indicating that the responses were not artifactual. The author also stated that no cochlear microphonic was apparent in response to the RFR, but this point could not be discerned from the waveforms shown.

Aset of four waveforms from the nucleus subthalamicus evoked by RFR at frequencies of 1.2, 1.3, 1.425, and 1.525 GHz showed the amplitudes for the two lower frequencies to be about the same, but that the amplitude was lower at 1.425 GHz and almost negligible at 1.525 GHz, results taken by the author to indicate that there is a broad range of optimal carrier frequencies consonant with calculations of RFR penetration in the head.

Frey and Messenger (1973) exposed humans to pulsed 1.245-GHz RFR at 50 pps in an RFR anechoic chamber. In one set of experiments, the average power density was held constant at 0.32 mW/sq cm and the pulse width was varied from 10 to 70 microseconds in 10-microsecond increments, yielding peak power densities from 640 to 91 mW/sq cm. In another set, the peak power density was held constant at 370 mW/sq cm and the pulse width was varied over the same range, yielding average power densities from 0.19 to 1.3 mW/sq cm. Four subjects with clinically normal hearing were each given 3 trials. The start of each trial was signaled optically. After a variable interval of up to 5 seconds, each subject was first given a pulsed RFR signal for 2 seconds, the perceived loudness of which was to be taken as reference level 100. About 5 seconds later, the test RFR signal was presented for 2 seconds and the subject was to signal its numerical loudness relative to the reference.

The results were presented as plots of perceived loudness vs peak power density and vs average power density (all on logarithmic scales). The point plotted for each test condition was the median value, without deviations, for all subjects and repetitions; no individual data were given. In the peak-power-density plot, the loudness rose sharply from about 3 at 91 mW/sq cm (70-microsecond pulses) to 60 at 125 mW/sq cm (50-microsecond pulses); at higher power densities, the sound level rose more slowly to a slightly rising plateau, i.e., to about 100 at 210 mW/sq cm (30-microsecond pulses), 120 at 315 mW/sq cm (20-microsecond pulses), and a slightly lower value at 630 mW/sq cm (10-microsecond pulses). The plateau indicated that there is an optimal band of pulse widths, within which the perceived loudness depends on the peak power density. The author ascribed the loudness diminution at 630 mW/sq cm to 10-microsecond pulses being shorter than optimal duration. The plot of median loudness vs average power density showed more scatter, with the points ranging from about 60 at 0.19 mW/sq cm (10-microsecond pulses) to a relatively flat maximum of 100 at 0.55 mW/sq cm (30-microsecond pulses) and diminishing to about 40 at 1.29 mW/sq cm (70-microsecond pulses). The author ascribed the dip for 70-microsecond pulses to this duration being longer than optimum.

From their data, Frey and Messenger (1973) calculated that the peak-power-density threshold for perception of RFR pulses is about 80 mW/sq cm, a value lower than those reported subsequently by Cain and Rissman (1978), discussed later. In the absence of information on scatter of the responses by each subject and because subjective judgments of relative loudness may be imprecise, the accuracy of the results of Frey and Messenger (1973) could not be evaluated.

White (1963) reported that when the surface of a body is transiently heated by RFR-absorption (or electron bombardment), elastic waves are produced by surface motion due to thermal expansion. This process was analyzed theoretically, with emphasis on the case of an input heat flux varying harmonically with time, to relate the amplitude of the elastic waves to the characteristics of the input flux and thermal and elastic properties of the body. Experiments with both electron impact and RFR-absorption verified the proportionality of the stress wave amplitude and the absorbed power density, and correlated well with the thermal and elastic properties of the heated medium.

Elastic waves were detected in all metals tested, in several carbon-loaded plastics, in water, and in a silver-coated barium titanate piezoelectric crystal. Mixing (production of beat frequencies) was observed when two pulses of different RFR frequencies were absorbed simultaneously. Comparison of the elastic-wave stress amplitude with radiation pressure showed that the former may be much greater than the latter, as demonstrated experimentally. When a barium titanate crystal was used to detect the elastic waves produced, heating by a single 2-microsecond pulse of electrons or RFR produced easily detectible signals at pulse power densities down to 2 W/sq cm, which corresponded to a computed peak surface-temperature rise of about 0.001 deg C and which produced piezoelectric-crystal voltages ranging from about 1 to more than 60 mV per kW/sq cm of absorbed power density.

Foster and Finch (1974) confirmed White's findings that RFR pulses can produce acoustic transients in water, and showed by calculation that for short pulses, the peak sound pressure is proportional to the energy per pulse, whereas for long pulses, it is proportional to the incident power density. Using 2.45-GHz RFR in several pulse-power-density and pulse-duration combinations and a hydrophone in saline (0.15-N KCl), they found that the transition between the two regimes occurs for pulse durations between 20 and 25 microseconds. The authors noted that the dependence of sound pressure on pulse duration and incident peak power density is consistent with the results of Frey and Messenger (1973) at 1.245 GHz. They also found that acoustic signals were not obtained in water at 4 deg C (where its thermal expansion coefficient is essentially zero) and that the polarity of the transient acoustic signal between 0 and 4 deg C was reversed from that for temperatures above 4 deg C. These results support the thermoelastic expansion hypothesis.

Sharp et al. (1974), in an experiment involving shielding regions of a subject's head from 1.5-GHz RFR pulses with RFR absorber, noticed that the apparent locus of the perceived sound moved from the head to the absorber. They then confirmed transduction of the RFR by the absorber into acoustic signals by using a sound-level meter to measure the delay times for acoustic propagation for distances of 0.3 to 0.6 m between the absorber and microphone. The pulses were 14-microseconds long and were randomly triggered at about 3 pps. By calculation, the power per pulse was 4.5 kW and that the pulse power densities were 7.5-15 kW/sq m (750-1500 mW/sq cm) for the range of separations above.

Subsequent tests by Sharp et al. (1974) showed that varying the carrier frequency from 1.2 to 1.6 GHz or using 2.45 GHz made little difference in the level or quality of the sound. In addition, detectable sounds could be produced with various sizes and shapes of absorber, including pieces as small as 4 mm square by 2 mm thick, and in various types of absorber and in crumpled aluminum foil. The threshold pulse power for audibility was about 275 W, yielding estimated pulse power densities in the range 0.46-0.92 kW/sq m (46-92 mW/sq cm). Tests were also done with constant pulse repetition rates up to 500 pps, with the finding: "The sound produced from the absorber seemed to track the repetition rate of the microwave pulses."

Taylor and Ashleman (1974) surgically prepared three groups of three cats, for recording potentials evoked by acoustic and RFR stimuli in three brain regions and determining the effect of cochlear disablement. Each cat was fitted on the dorsal surface of the frontal bone with a piezoelectric transducer for presentation of acoustic stimuli; it was removed for RFR-exposure. In the cats of group 1, the eighth cranial (vestibulocochlear) nerve was exposed, a glass pipette microelectrode filled with Ringer's solution was inserted into the nerve, and the round window on the same side was exposed. In those of group 2, a similar electrode was advanced into the medial geniculate to a location in the nucleus that yielded acoustically evoked potentials of appropriate characteristics, and both round windows were exposed. In the cats of group 3, a Teflon-covered carbon electrode was placed on the anterior ectosylvian gyrus of the primary auditory cortex and both round windows were exposed. Connections to the electrodes were made with carbon leads of high resistance.

The acoustic stimuli were produced by feeding 10-microsecond electric pulses to the piezoelectric transducer at 1 pps. The RFR stimuli were 32-microsecond, 2.45-GHz pulses at 1 pps fed via a directional coupler to a horn placed 10 cm posterolaterally from the cat's head at 30 deg relative to the sagittal plane. Incident power densities were measured with a bolometer.

Following post-surgical stabilization, the lowest piezoelectric voltage that yielded a response was determined and the voltage was increased to a level that was maximal in evoking activity. The transducer was then removed, the head of the cat was exposed to the RFR, and the minimum and maximal levels for evoking responses were determined. When responses to both acoustic and RFR pulses were clearcut, the cochlea was disabled by perforating the round window and aspirating the perilymph, and again the response to each stimulus was determined. The responses of the medial geniculate nucleus and auditory cortex were assessed after disabling each cochlea. When no evoked response occurred after disabling the cochlea, the stimulus was raised to the maximum available in each case.

The results indicated that cochlea destruction led to total loss of evoked potentials to both acoustic and RFR stimuli. Specifically, the eighth-nerve potentials were lost after unilateral cochlea destruction, the evoked-potential amplitudes from the medial geniculate and auditory cortex were markedly attenuated by aspiration of the contralateral cochlea, and disablement of the remaining cochlea resulted in total loss of the potentials.

From their results, the authors concluded: "We believe that the data strongly support the contention that the microwave auditory effect is exerted on the animal in a manner similar to that of conventional acoustic stimuli. Clearly, the elimination of the first stage of transduction affects the central nervous system response to both of these forms of stimulus energy in the same way."

Guy et al. (1975b) determined the power density threshold and modulation characteristics for the RFR-hearing effect in human volunteers, compared the potentials evoked in four successive levels of the cat auditory nervous system by RFR and acoustic pulses, used optical interferometry to quantitate the transduction of RFR pulses to acoustic energy in RFR-absorbing materials, and provided evidence that the RFR-hearing effect is due to direct conversion of RFR energy to acoustic energy in the tissues.

The back of the head of each of two human subjects was exposed to RFR at 15-30 cm from the aperture of a horn (in the near field) in an anechoic chamber at an ambient noise level of 45 dB, with RFR-absorbent material around the vicinity of the subject to eliminate reflections. The RFR consisted of 2.45-GHz pulses of duration that was varied from 1 to 32 microseconds. For each pulse duration, the RFR was presented in trains of three pulses per second, with the pulses in each train spaced 100 ms apart. The subject used a switch to signal perception of an auditory sensation. Standard audiograms taken prior to exposure showed that the hearing threshold of subject 1 was normal and that subject 2 had a deep notch at 3.5 kHz for both ears, with similar results for air and bone conduction.

The results for subject 1 showed that irrespective of pulse duration, the threshold for perception of the RFR was a constant peak energy density per pulse (product of peak power density and pulse duration) of 40 microjoules/sq cm. The corresponding incident average power density (for 3 pps) was 0.12 mW/sq cm. When subject 1 wore ear plugs, the threshold peak was only 28 microjoules/sq cm per pulse. Based on a spherical model of the head (Johnson and Guy, 1972), the threshold peak specific absorbed energy (SAE) corresponding to 40 microjoules/sq cm per pulse was 16 mJ/kg. The threshold for a pair of pulses within several hundred microseconds apart was the same as for one pulse with the same total energy as the pair. Similar results were obtained for subject 2 except that the threshold peak energy density was 135 microjoules/sq cm per pulse or about threefold (5 db) higher than for subject 1.

The authors noted that each pulse was perceived individually as a click and that short pulse trains were heard as chirps of tones corresponding to the pulse recurrence rate. Also, when the pulse generator was keyed manually, digital (Morse) code transmitted thereby could be interpreted accurately by the subject.

For the study of cats, each was fitted with a removable piezoelectric transducer to provide bone-conducted acoustic stimuli. A nonperturbing electrode consisting of a glass pipette filled with Ringer's solution was inserted surgically into the medial geniculate nucleus to a location that yielded acoustically evoked responses of the proper latency period. To minimize RFR pickup by the instrumentation, the electrode and ground connection were coupled with high-resistance carbon-loaded plastic leads through a low-pass filter to a high-impedance amplifier, oscilloscope, computer of average transients, and x-y plotter.

The acoustical stimuli consisted of pulses 1-30 microseconds in duration that were air-conducted from a speaker 17 cm to the right of the cat's head or were transducer-induced. The RFR stimuli were pulses, 0.5-32 microseconds in duration, of 918-MHz or 2.45-GHz RFR from a horn or aperture 8 cm from the occipital pole of the cat. Pulses of 8.5-10 GHz RFR were also used, as noted below. Each stimulus was presented at 1 pps. A noise generator provided background noise of up to 90 dB in the range 50-15000 Hz.

Representative response curves evoked by 20-microsecond air-conducted and bone-conducted acoustic pulses and by 20-microsecond pulses of 2.45-GHz RFR were displayed and were similar. The threshold peak energy-density per pulse for perception of 2.45-GHz RFR varied only from 17.8 microjoules/sq cm for 0.5-microsecond pulses to 21.6 microjoules/sq cm for 10-microsecond pulses, values about half the human threshold, but it increased more steeply with pulse duration, to 47.0 microjoules/sq cm for 32-microsecond pulses (except for 25-microsecond pulses, for which the threshold was only 15.2). The peak SAEs were determined by scanning infrared thermography. The values corresponding to 21.6 and 47.0 microjoules/sq cm per pulse respectively were 12.3 and 26.7 mJ/kg.

The threshold energy-density values for 918-MHz pulses were 22.6 and 28.3 microjoules/sq cm per pulse respectively for 10- and 32-microsecond pulses (with no dip for 25-microsecond pulses), and the corresponding peak SAEs per pulse were 16.0 and 20.0 mJ/kg.

The RFR thresholds above were obtained with 64 dB of background noise. Increases of the noise level to 80 dB (for pulses up to 10 microseconds) yielded insignificant changes in thresholds. However, the energy-density threshold for bone-conducted acoustic stimulation was about tenfold higher at 80 than 64 dB; for air-conducted acoustic stimulation, the threshold was a hundredfold higher for 70 dB of noise and was still higher for 80 dB, but by a factor of less than ten.

With 8.5-10 GHz, responses could be evoked only by exposing the brain through a large hole in the skull, with the RFR horn within a few cm from the hole. The threshold values were a peak incident power density of 14.8-38.8 W/sq cm per pulse, which corresponded to an average power density of 0.472-1.240 mW/sq cm (for 32-microsecond pulses, 1 pps) and an energy density of 472-1240 microjoules/sq cm per pulse.

In another series of experiments, Guy et al. (1975b) fitted cats with a piezoelectric transducer and inserted a nonperturbing electrode in the medial geniculate nucleus as before, and also surgically exposed the round window of the cochlea and attached thereto an electrode and leads, both of high-resistance material, for recording the cochlear responses evoked by acoustic and RFR pulses. The responses of one cat to an acoustic pulse from a loudspeaker and a 2.45-GHz pulse were displayed. The curve obtained for stimulation with the loudspeaker pulse showed the N1 and N2 auditory-nerve response and a cochlear microphonic similar to the pulse-induced decaying vibratory movement of the loudspeaker cone, which was determined with an optical interferometer. By contrast, the curve evoked by the RFR pulse showed the N1 and N2 response only, and no evidence of a cochlear microphonic. However, the cochlear microphonic for another cat stimulated by speaker pulse was of much lower amplitude (relative to the N1 and N2 response) and was absent in the response of another cat stimulated acoustically with the piezoelectric transducer. Thus, the absence of an RFR-induced cochlear microphonic does not rule out theories of the RFR-hearing effect based on transduction of RFR to acoustic energy.

In experiments similar to those of Taylor and Ashleman (1974), Guy et al. (1975b) also studied the effect of cochlea disablement. Cats were prepared surgically for recording evoked potentials in the eighth cranial nerve, medial geniculate nucleus, and primary auditory cortex. After establishing that appropriate responses were obtained with RFR and acoustic pulses, the cochlea was disabled, which resulted in total loss of all evoked potentials, even with the highest available peak acoustic and RFR powers and with computer averaging of larger numbers of signals.

Guy et al. (1975b) used an interferometer and a laser source to detect surface movements of several lossy materials induced by absorption of RFR pulses. (This device was also used to observe the speaker-cone movements noted above.) The results showed that surface displacement amplitude is dependent on the dielectric constant and loss factor of the material and on its density and elastic properties. An interesting result noted by the authors was the high amplitude obtained in a widely used RFR-absorbent material, a finding ascribed to its relatively low density and high compressibility.

Chou et al. (1975) studied the induction of cochlear microphonics (CM) in the guinea pig by pulses of 918-MHz RFR. For this purpose, they placed a fine RFR-transparent carbon lead against the round window and cemented it onto the bulla. An indifferent electrode was fastened to nearby tissue. Only preparations that yielded CM amplitudes exceeding 0.5 mV in response to 70-dB speaker clicks were used. The head of the guinea pig was placed within a section of cylindrical waveguide through a hole. The section was terminated with a sliding short, adjusted to yield maximum RFR absorption in the head. With this arrangement, the average SAE per pulse at 10 kW peak input power was 1.33 J/kg, or about an order of magnitude larger than the levels used in previous studies. The sound level near the animal's head was about 65 dB, mostly due to the noise from the pulse generator.

Each guinea pig was exposed intermittently for 1.5-min durations to 1-10 microsecond pulses of 918-MHz RFR, 100 pps, at various levels of peak power. The responses were amplified and recorded on a magnetic tape system that had a frequency response up to 80 kHz. After 3-5 hr, the animal was euthanized and recording of its response was continued until the physiological potentials disappeared completely. Recorded responses were processed off-line with a Computer of Averaged Transients.

The electrical responses at the round window of a guinea pig stimulated with single acoustic clicks showed that the CM preceded in time the N1 and N2 action potentials, and that when the polarity of the electrical pulses delivered to the speaker was inverted, only the polarity of the CM was reversed. Stimulation of the same animal with single RFR pulses yielded N1 and N2 potentials of about the same amplitude. In addition, barely discernible was an "electrical event" during the 200-microsecond interval immediately following the recording artifact caused by the RFR pulse. Time-expansion of this interval showed this event to be a 50-kHz oscillation of amplitude about 50 microvolts, which decayed within the 200-microsecond interval. This event, which was observed in five guinea pigs, was denoted as the RFR-induced CM response.

Comparison of the CM responses to 10-, 5-, and 1-microsecond, 10-kW RFR pulses, which corresponded to SAEs of 1.33, 0.67, and 0.133 J/kg per pulse, showed that the CM frequency remained the same but the amplitude dropped with decreasing pulse duration and SAE. Also evident was that the stimulus artifact masked the onset of the CM in each case, but that the latency for successive oscillations was about the same for the three pulse widths. These results support the conclusions that the CMs are physiological responses time-locked to the onset of the RFR pulses and are generated within the cochlea, specifically by hair-cell activation.

With death of an animal, the N1 and N2 responses to acoustic and RFR pulses disappeared, but the RFR-induced CM persisted for many minutes after the N1 and N2 responses had gone. The stimulus artifact remained after the CM had disappeared, indicating that the 50-kHz signal is a genuine physiological response.

The threshold SAE for producing the RFR-hearing effect in the guinea-pig head was 20 mJ/kg, about the same order of magnitude as for the cat head (10-16 mJ/kg) and the human head (16 mJ/kg). The authors surmised that previous failures to observe RFR-induced CMs in animals may have been due to use of SAEs below the threshold, masking by stimulus artifact, or use of amplifiers with pass bands that did not include 50 kHz.

The authors noted that guinea pigs can respond to tones up to 100 kHz, and suggested that the 50-kHz CM may be related to the size of the animal's skull. Based on this premise, the CM frequency would be within the range 15-50 kHz for cats and 5-18 kHz for humans.

Guy et al. (1975b) and Lin (1976a,b; 1977a,b,c) analyzed the postulated mechanisms for the conversion of RFR energy to acoustic energy in lossy dielectric materials. They concluded that pulsed-RFR-induced thermal expansion forces, which are proportional to the square of the peak electric field, are much larger than the radiation pressure or the electrostriction produced by the same RFR pulses and can generate in the head acoustic waves of the requisite magnitude for the hearing effect.

In Lin (1977c), equations developed for a spherical model of the head consisting of brain-equivalent material were used to obtain the acoustic resonant frequencies generated in the heads of guinea pigs, cats, and human adults and infants by exposure to RFR pulses. The results showed that the (fundamental and higher-harmonic) frequencies produced by RFR pulses are independent of the carrier frequency, but are dependent on head size, and specifically that the fundamental frequency is inversely proportional to the radius of the head.

Predicted from the equations was a fundamental frequency of 45 kHz for a 2-cm (guinea-pig) head, which was close to the 50-kHz experimental value found by Chou et al. (1975). For a 3-cm head, the predicted fundamental was 30 kHz, as compared with 38 kHz found experimentally for a typical cat. For humans, the predicted fundamental frequencies were 13 kHz for an adult and 18 kHz for an infant.

Chou et al. (1977), using the method described in Chou et al. (1975), recorded CMs from the round windows of guinea pigs and cats of various sizes induced by exposure to 10-microsecond pulses of 918-MHz and 2.45-GHz RFR. Exposures were done with horn applicators and the cylindrical waveguide system described above. In the guinea pigs, the CM frequency varied inversely with body mass; in the cats, however, there was no consistent variation of CM frequency with body mass. The authors noted that although head mass, skull mass, skull dimensions, skull thickness, and cerebellar-cavity dimensions all increase with body mass, the brain cavity and bulla dimensions increase only slightly. They found that CM frequency correlates well with the length of the brain cavity but not with the other dimensions of the head or skull. The average threshold energies per pulse for CM responses were 10 mJ/kg for adult cats, 2.5 for kittens, and 7.5 for adult guinea pigs.

Cain and Rissman (1978) used 3.0-GHz RFR pulses to study the auditory effect in two cats, two chinchillas, one beagle, and eight human volunteers. For the animals, surface or brainstem-implanted electrodes were used to measure the responses evoked by audio clicks from a speaker and the responses to 5-, 10-, and 15-microsecond pulses.

The threshold peak power densities were 2.2 W/sq cm for 5-microsecond pulses, 1.3 W/sq cm for 10-microsecond pulses, and 0.58 W/sq cm for 15-microsecond pulses for one cat, and respectively 2.8, 1.3, and 0.58 W/kg for the other cat. The corresponding threshold peak power densities for the beagle were 1.8, 0.30, and 0.20 W/sq cm. The values were 2.8, 2.0, and 0.58 W/sq cm for one chinchilla and 2.2, 1.0, and 0.50 W/sq cm for the other. Thus, for corresponding pulse durations, the beagle had the lowest thresholds and the lowest absolute threshold (irrespective of pulse duration).

The authors found that depending on pulse width, the range of threshold energy density for RFR perception was 8.7-14 microjoules/sq cm per pulse for the cats and 7.5-20 microjoules/sq cm for the chinchillas, and that the threshold averaged 5.0 microjoules/sq cm for the beagle. For 10-microsecond pulses, the threshold pulse power densities were 1.3 W/sq cm for both cats, 1 and 2 W/sq cm for the two chinchillas, and 300 mW/sq cm for the beagle.

The eight humans were given standard audiograms for both air-conducted and bone-conducted sound. In addition, because audiograms do not test hearing above 8 kHz, binaural hearing thresholds were determined for seven of the subjects for tone frequencies in the range 1-20 kHz. The RFR pulses were presented at 0.5 pps. Each subject wore foam ear muffs during exposure, to reduce the ambient noise level, which was 45 dB.

Subjects 1-5 could hear 15-microsecond pulses as clicks; their peak power density thresholds were respectively 300, 300, 300, 600, and 1000 mW/sq cm, and their energy density thresholds were 4.5, 4.5, 4.5, 9.0, and 15.0 microjoules/sq cm. Subjects 1-5 could also hear 10-microsecond pulses, with peak power density thresholds of 1800, 225, 600, 2000, and 2000 mW/sq cm, respectively, and energy density thresholds of 18.0, 2.3, 6.0, 20.0, and 20.0 microjoules/sq cm. Subject 1 was the only one able to perceive 5-microsecond pulses, with a threshold peak power density and energy density of 2500 mW/sq cm and 12.5 microjoules/sq cm. The other three subjects, 6-8, could not hear these pulses at the highest available peak power density but could perceive 20-microsecond pulses.

The authors found no correlation between the results and the standard audiograms. However, they did note that a strong correlation existed between RFR perception and hearing ability above 8 kHz as determined from the binaural thresholds. They also stated that their results are consistent with the hypothesis that an induced pressure wave in the human head in response to short RFR pulses (less than 20 microseconds) contains a significant portion of its energy at frequencies above 8 kHz.

In summary of these results with humans, only subjects 1-3 were able to perceive 15-microsecond pulses at a pulse-power-density threshold as low as 300 mW/sq cm (energy-density-threshold of 4.5 microjoules/sq cm); of this group, only subject 2 could hear 10-microsecond pulses, with 225 mW/sq cm (2.3 microjoules/sq cm) as the threshold; the thresholds for the other subjects were much higher than 300 mW/sq cm. The latter value of pulse power density can be taken as the nominal human threshold for the RFR hearing effect (e.g., for environmental assessments). It should be noted that the thresholds cited were for an ambient noise level of 45 dB and could be higher in noisier environments. It is also noteworthy that these investigators exposed the human volunteers to pulse power densities as high as 2,000 mW/sq cm without apparent ill effects.

Lebovitz and Seaman (1977) studied the responses in single auditory units of cats to acoustic clicks and pulses of 915-MHz RFR. For this purpose, the posterolateral aspect of the cerebellum was removed and a recording micropipette was inserted into the proximal portion of the eighth nerve. Acoustic clicks ranging from 25 to 200 microseconds in duration, but typically 70 microseconds, were presented to one ear at 10 clicks per second, with the contralateral ear stoppled. The durations of the RFR pulses were in the same range, the repetition rates were 10 pps or less, and the pulses were delivered at forward peak powers of up to 70 W with an applicator, yielding average power densities that never exceeded 1 mW/sq cm.

Medullary SARs were determined by euthanizing the cats after completing several experiments, letting the medullary temperature fall to about 30 deg C, exposing the heads of the cats to an appropriate level of CW RFR for periods of up to 80 seconds, inserting a thermistor into the medulla immediately before and after exposure, deriving the cooling curves, and using them to determine the linear relationship of temperature rise to exposure duration. From the slope of this line, 0.011 deg per second, and the effective forward power, 48.6 W, the normalized SAR was 0.94 W/kg per mW/sq cm. The energy absorbed per pulse was then calculated from the SAR and the peak power and duration.

Of 133 auditory units studied, 63 were responsive to both stimuli, and additional dose responses were obtained for the latter units as long as each showed stable responses. The apparent absence of response of 70 units to RFR was ascribed to the limited range of intensities available, about 10 dB as compared with 50-80 dB for the acoustic clicks. For a typical single auditory unit that did respond to RFR, the response was very similar to its response to acoustic clicks, differing primarily only in amplitude. The lowest RFR-energy-absorption threshold for a single unit was 4 mJ/kg. The latency interval between stimulation and response increased with decreasing acoustic or RFR stimulus intensity. The smallest latencies observed were 1.5-2 ms, with values up to 5 ms for near-threshold intensities.

The authors noted that the characteristic frequency (CF) of a unit is the frequency at which its response threshold is lowest and that the CF is determined by the mechanical properties of that part of the basilar membrane to which the cell is most directly related. Therefore, for responses to a click that show multiple peaks, the interpeak interval (i.e., the period between the first and second peaks) is about the same as the oscillation period of the basilar membrane and the former is a measure of the latter. The results showed high correlation between unit CFs for acoustic and RFR stimuli, an indication that the same mechanical factors within the cochlea are involved.

Chou and Galambos (1979) investigated the effects in 10 guinea pigs of external-ear blocking, middle-ear damping, and middle-ear destruction on brainstem-evoked responses (BERs) to both acoustic and RFR stimuli. The basic measurement technique was to record the amplitudes and latencies of the BERs to acoustic stimuli and RFR with a pair of carbon-loaded Teflon electrodes (Chou and Guy, 1979a), one of which was attached to the left mastoid process and the other to the skin.

The head of each guinea pig was exposed to 10-microsecond pulses of 918-MHz RFR at 30 pps in a cylindrical waveguide system (Chou et al., 1975) at energies ranging from 0.027 to 11.05 J/kg per pulse. All 10 animals were exposed to 0.1-ms acoustical pulses at 30 pps from a piezoelectric tweeter placed 15 cm from the nose (air conduction). For three animals, comparisons were also made between BERs to airborne and bone-conducted acoustic stimuli, using a piezoelectric transducer in contact with the frontal bone for the latter.

BERs were recorded after each of the following successive treatments: (1) blocking of the left external meatus with mineral-oil-soaked cotton balls, (2) alteration of the mechanical damping of the ossicular chain by filling the bulla with mineral oil, (3) destruction of the middle ear by cutting the ossicular chain and piercing the tympanic membrane, and (4) destruction of the cochlea by piercing the round window.

Treatments 1 and 2 reduced the airborne acoustically-stimulated BERs but not the RFR-induced BERs. Treatment 3 further reduced the airborne acoustic BERs, and also reduced the bone-conducted acoustic BERs and the RFR BERS to a lesser extent than the airborne acoustic BERs. After treatment 4 (cochlea destruction), no BERs were obtained from either acoustic or RFR stimuli.

These results constitute strong evidence that activation of the cochlea is necessary for auditory perception of pulsed RFR. The similar BERs obtained from bone-conducted-acoustic and RFR stimuli after destruction of the middle ear and the much lower BERs obtained for airborne-acoustic stimuli support the hypothesis that perception is due to transduction of RFR into acoustic waves that travel via bone conduction to the cochlea or are generated directly in the cochlea itself.

Chou and Guy (1979b) performed similar experiments with BERs induced in guinea pigs, to determine the RFR thresholds for BERs. The input-power thresholds for BERs induced by 918-MHz RFR were determined for pulse widths of 10 to 500 microseconds, and the values were divided by the cross-sectional area of the cylindrical waveguide (about 320 sq cm) to obtain the corresponding threshold peak incident power densities. Also derived was the incident energy density per pulse for each pulse width, and the pulse repetition frequency (30 pps) was used to calculate the incident average power density. Lastly, the threshold SAE per pulse was obtained from the incident energy density per pulse by dividing the latter by the mass of the animal's head.

The authors found that the threshold energy density for evoking BERs was essentially constant (1.56-1.87 microjoules/sq cm per pulse) for pulse durations of 10, 20, and 30 microseconds. The threshold SAE was 5 mJ/kg per pulse and the corresponding incident peak power densities were 156, 78, and 62.4 mW/sq cm, respectively. For pulse durations longer than 30 microseconds, however, the threshold SAE increased with pulse duration, and for pulses longer than 70 microseconds, the threshold peak power density for evoking BERs was essentially constant (90 mW/sq cm), with corresponding pulse-width related increases of energy density per pulse.

The waveforms of the RFR and acoustic BERs were found to be similar except for the longer latency of the latter (due to the longer sound-propagation path). Despite the large differences in pulse width, the latencies of the RFR-induced BERs were about the same, indicating that the evoked response is time-locked to the onset of the RFR pulse. Chou and Guy indicated that their experimental results agreed well with the predictions of the thermal expansion theory.

In a subsequent study, Chou et al. (1985a) exposed anesthetized rats to 2.45-GHz RFR pulses within a circularly polarized waveguide (Guy et al., 1979) in three orientations: (1) body along the waveguide axis and head toward the source, (2) body across the waveguide, and (3) body along the waveguide axis and head away from the source. The BERs were recorded with carbon-loaded Teflon electrodes, one at the vertex of the rat's head and another at either the left or right mastoid process (behind the pinna).

In one experiment, exposures were to pulses 10, 5, 2, and 1 microseconds in duration at 10 pps and a fixed peak power of 4 kW (spatially averaged peak power density of 12.3 W/sq cm) in the first orientation. The corresponding energy densities were 123.4, 61.7, 24.7, and 12.3 microjoules/sq cm. Representative BERs from one rat showed amplitudes that decreased with decreasing pulse duration or energy density. The responses were similar to those obtained from guinea pigs by Chou and Galambos (1979), but the latency of the peak BER was shorter in rats.

In another experiment, rats were exposed to RFR in each orientation, and the largest responses were obtained in the first orientation. In this experiment, exposures were to pulses of 1, 2, 5, and 10 microseconds at 10 pps and different peak powers, to yield various energy densities. The BER amplitudes at the four pulse durations and constant energy density were about the same, indicating that the response is dependent on energy per pulse and not on pulse duration. The threshold energy density (in the first orientation) was 1.5 to 3 microjoules/sq cm per pulse. Based on calorimetric data, the whole-body SAE was in the range 0.9-1.8 mJ/kg. The corresponding peak power densities for the four pulse durations were in the ranges 1.5-3, 0.75-1.5, 0.3-0.6, and 0.15-0.30 W/sq cm, respectively. The authors noted that these peak power densities were for exposure in the circularly polarized waveguide and that free-space exposure would require about threefold higher values.

Lin et al. (1979b) studied BERs induced in cats by acoustic and RFR pulses and the alterations of the BERs by the successive coagulative formation of lesions in several brainstem regions. Under anesthesia, the dorsal aspect of the skull of each cat was surgically exposed and several stainless-steel electrodes (100-250 microns in diameter) were advanced stereotaxically into the selected brainstem nuclei to locations that yielded maximal evoked potentials. In addition, a stainless-steel screw electrode was fastened to the skull at the vertex and a reference gold-pin electrode was placed near the lowest part of the right pinna.

Acoustic pulses about 70 dB above threshold sound level, generated by feeding currents 0.1 ms in duration into a pair of commercial earphones, were presented binaurally at 10 to 100 pps. Pulses of 2.45-GHz RFR, 0.5 to 25 microseconds wide and up to 10 kW peak, were delivered at 10-100 pps to the dorsal or frontal surface of the head with a small-diameter (15-mm), dielectrically loaded, direct-contact, diathermy applicator. The bioelectric activities at the vertex and at each brainstem location were fed through amplifiers having a passband of 80 Hz to 10 kHz and were summed with a signal-averaging computer. The first 10 ms of averaged responses were displayed on a video monitor and photographed.

Brainstem lesions were produced in succession at the tips of electrodes inserted in the inferior colliculus nucleus, lateral lemniscus, and superior olivary nucleus. The BERs were recorded after each lesion and compared with the prelesion BERs, as were microwave-evoked potentials (MEPs) and acoustically-evoked potentials (AEPs) recorded by the vertex electrode. At the end of each experiment, each cat was euthanized and its brain was fixed, removed, embedded in paraffin, and sectioned to ascertain the exact locations of the lesions and electrode tracks.

For each brainstem region, the pre-lesion RFR-induced BER was similar to the acoustic BER, but with no significant propagation delay for the RFR-induced BER relative to the stimulus pulses. The vertex MEPs showed four successive cycles, designated sequentially as Waves I, II, III, and IV. Of these, Wave III was often the largest and Wave IV the smallest. The sources of the waves (as well as for the AEPs) were thought by the authors to be the volume-conducted action potentials generated in the cochlea and auditory brainstem nuclei.

To determine the effects on the MEPs of changing the pulse repetition rate (PRR), 10-microsecond RFR pulses at 5 kW were applied sequentially at PRRs of 10, 25, 50, and 100 pps and then in reverse order. Minimal or no changes in latency were evident, but the amplitudes of the MEP waves were found to decrease with increasing PRR, a reversible effect.

Next, 10-microsecond pulses with peak powers of 10 to 2 kW were applied at 10 pps. Again, any changes in latency of the MEPs were minimal. The amplitudes of the MEPs decreased with decreasing peak power, but not in the same proportion for each wave. In the example presented (for one cat), the amplitude of Wave I at 10 kW was larger than of Wave II, but decreased faster with decreasing power than for Wave II, so the Wave-II amplitude at 4 kW was larger than that of Wave I, also a reversible effect.

The effects of pulse-duration changes on the MEPs were determined with 5-kW pulses of widths 2.5 to 25 microseconds at 10 pps. Once more, the latency changes were minimal. However, wave amplitudes increased with increasing pulse durations to a plateau for about 10-microsecond and longer pulses, and the temporal relationships among all of the waves were not altered.

Using 25-microsecond, 10-kW pulses at 10 pps, the effects of the lesions formed successively in the inferior colliculus, lateral lemniscus, and superior olive on the BER recorded by the electrode in each region were compared with the pre-lesion BER from that electrode. Also compared were the pre- and post-lesion MEPs recorded by the vertex electrode. The results for one cat were presented.

Each successive lesion yielded decreases in the BER amplitudes from all regions. The most pronounced effect was on the BER from the inferior colliculus nucleus following lesion formation in that region; the BERs from the other two brainstem regions and the MEP remained practically unchanged after producing a lesion in the inferior colliculus nucleus. Similarly, the lesion in the lateral lemniscus yielded the largest effect on the BER recorded by the electrode in that region, with minor changes in the BERs from the other auditory structures except for the inferior colliculus. The BER amplitudes from the superior olive were less affected by the lesions in the inferior colliculus and the lateral lemniscus, reflecting the distal location of the superior olive in the auditory pathway, but its BER amplitude was drastically reduced by the lesion in its nucleus. The amplitudes of the vertex MEPs were altered after each successive lesion, indicating that they were a function of the integrity of brainstem nuclei along the auditory pathway.

Tyazhelov et al. (1979) studied the qualities of the apparent sounds perceived by humans from exposure to 800-MHz pulsed RFR. The parietal area of the head was exposed to the open end of a waveguide fed from a 500-W source. The ambient noise level did not exceed 40 dB and was reduced by plugging the subjects' ears with stoppels or sound-conducting tubes. The pulse durations used ranged from 5 to 150 microseconds. The pulses were presented either continuously at 50 to 2000 pps (the latter for short pulse durations, to limit the average power density) or in trains of duration 0.1 to 0.5 seconds at rates of 0.2 to 2.0 trains per second. Each subject could be presented with sinusoidal audiofrequency (AF) sound waves independently of, or concurrently with, the pulsed RFR and could adjust the amplitude, frequency, and phase of the AF signal to match the timbre and loudness of the perceived RFR. Acoustic signals were presented to the subject by means of a pair of small hollow tubes extending from a speaker to the ears.

The high-frequency auditory limit (HFAL) of each subject for sinusoidal tones from 1 kHz upward was tested first. Three of the subjects had HFALs below 10 kHz and could not perceive 10-30 microsecond RFR pulses, results consonant with those of Cain and Rissman (1978). Of 15 subjects with HFALs above 10 kHz, only one could not perceive the RFR pulses.

All of the perceptive subjects reported that 10-30 microsecond pulses delivered at 1000 to 12,000 pps at peak power densities exceeding 500 mW/sq cm produced sound of polytonal character that seemed to originate in the head, and that the quality of the sound changed with increasing pulse repetition rate (PRR) in a complex manner. Loudness diminished sharply and became more monotonal as the PRR was increased from 6000 to 8000 pps. However, no more than three distinguishable tonal transitions occurred. Subjects with HFALs below 15 kHz were unable to distinguish between the sounds perceived from a 5000-pps and a 10,000-pps signal, and subjects with more extended HFALs reported that the pitch for a 5000-pps signal was higher than for a 10,000-pps signal.

The subjects were able to detect small (5%) shifts of PRR only in the 8000-pps region; at lower PRRs, the subjects erred on 100% of tests to detect the direction of PRR change, indicating that increases of PRR were often perceived as decreases in frequency. For pulses of constant peak amplitude, loudness was perceived to increase with duration from 5 to 50 microseconds, decrease from 70 to 100 microseconds, and increase again for 100 microseconds and upward. Such perceptual patterns were exemplified by plots of threshold pulse power (normalized to the 10-kHz PRR threshold) vs PRR for a subject with a 14-kHz HFAL and for another subject with a 17-kHz HFAL. These curves were roughly W-shaped, with central relative maxima at about 8 kHz. A plot of mean threshold pulse power (normalized to the threshold at 50-microsecond pulse duration) was also presented for subjects unable to perceive sounds for pulses longer than 50 microseconds. This curve was also W-shaped, with a central relative maximum within the pulse range 100-120 microseconds.

After subjects matched the pitch and timbre of a 2-kHz acoustic tone to the perceived sound of a train of RFR pulses at 2000 pps, they were asked to match the loudness of the acoustic tone with the loudness of the perceived pulses as the pulse duration was varied between 5 and 150 microseconds while the peak power was held constant. No actual data were given. Instead, the relationship between the ratio of acoustic signal amplitude to the pulse power for the subjects (both quantities normalized to their respective thresholds) were merged into a shaded area bounded by two straight lines through the origin on a graph that also displayed a straight line through the origin stated by the authors to represent the theoretical relationship between these quantities as predicted from the thermoelastic model. The entire shaded area was above the theoretical line, i.e., the ratios of acoustic amplitude to pulse power for all of these subjects were reported to be larger than predicted.

When acoustic tones above 8 kHz were presented concurrently with 10- to 30-microsecond pulses at PRRs slightly above or below 8 kHz, the subjects reported hearing beat-frequency notes. Also, for a PRR of 800 pps, similar beat frequencies were perceived when the acoustic frequency was set slightly above or below harmonics of the PRR. Moreover, when the tone and PRR frequencies were matched and the subjects were allowed to vary the phase of the acoustic tone, cancellation of perception of the two stimuli could be achieved. By proper phasing, subjects with HFALs below 15 kHz could also obtain cancellation between a 10-kHz acoustic signal and a 5-kHz train of pulses.

The authors also reported that the sensory characteristics (pitch and timbre) evoked by RFR pulses less than 50 microseconds in duration persisted when subjects' heads were lowered into seawater, with loudness diminishing roughly in proportion to immersion depth and vanishing entirely with total immersion. For pulses longer than 50 microseconds, even partial immersion resulted in loss of perception.

In their discussion, the authors suggested that many of their results are consistent with the thermoelastic hypothesis, but that others, such as the suppression of the perception of a 5000-pps train of RFR pulses by a 10-kHz acoustic tone, are at variance with that model.

Frey and Coren (1979) endeavored to detect surface movements purportedly induced in heads of animals by RFR pulses, using dynamic time-averaged interferometric holography. In this technique, a hologram of an object in vibratory motion is recorded for a long interval relative to one period of the vibration. Such a hologram will contain data on the spatial distribution of the time-averaged amplitude of motion of the object. Thus, nodal regions will appear brightest and antinodal regions appear darkest, with regions of intermediate intensity inversely related to their surface displacement.

Each animal studied was euthanized, placed with its abdomen on the surface of a vibrationally-isolated block of commercial RFR-absorbent material (or concrete in some tests), and exposed to RFR from above with a horn as soon as there was no detectable heartbeat or respiration. A set of 30 holograms was made for each animal, 15 each during alternating RFR- and sham-exposures. First, holograms during three RFR- and three sham-exposures were made after removing the hair from the dorsal surface of the head and the vicinity of the left pinna. Next, the skin over those areas was removed and six holograms of the musculature revealed thereby were made. Then, the muscle tissue was removed from the dorsal surface and mastoid area and six holograms were made of the skull. Six more holograms were made of the brain after removing the dorsal surface of the skull. The last six holograms were made of the base of the skull cavity (dorsal surface of a bulla) after removing the brain.

In the first of two experiments, 10 Sprague-Dawley rats were exposed to 25-microsecond pulses of 1.275-GHz RFR at a peak power density of 1.7 W/sq cm and a PRR of 50 pps. An additional set of holograms was made at 100 pps for five of the rats. In the second experiment, 16 adult guinea pigs were used. Of these, eight were exposed to 1.1-GHz pulses in a 2x2x2 factorial design with peak power densities 1.25 and 8.5 W/sq cm, pulse durations 10 and 20 microseconds, and PRRs 25 and 50 pps. The other eight were exposed to 1.2-GHz pulses in the same design. The holograms for three of the guinea pigs exposed at 8.5 W/sq cm showed that movement was engendered in the RFR absorber supporting the animals by the RFR pulses, so the support was replaced with a cement block.

The authors indicated that they could not detect any differences between the holograms obtained from each animal during RFR exposure and the holograms from the same surfaces of the same animal taken during sham exposure. (The comparisons were made on a coded and blind basis.) They concluded therefrom that the hypothesis of RFR transduction into elastic waves in the head and propagation of the sound to the cochlea by bone conduction, predicted by other investigators, is untenable. Instead, they suggested that the transduction site is more likely to be in the cochlea itself rather than elsewhere in the head.

The authors did not provide specific information about the appearances of the holograms or the differences sought between holograms of RFR- and sham-exposed surfaces, rendering it difficult to assess the validity of these negative findings. Also, the adequacy of the sensitivity of this holographic technique for detecting such movements was disputed by Chou et al. (1980a), with a response by Frey and Coren (1980). Based on the description of the holographic technique, one would expect that the brightness of a surface having non-uniform optical reflectance would appear non-uniform even if the surface were stationary. Also, a surface having uniform reflectance and moving as a unit, i.e., without motion of any area relative to another, would appear uniformly illuminated (but of lower brightness than if the same surface were stationary).

Amore fundamental question is whether or not the successive removal of skin, musculature, etc., would alter the characteristics of possible RFR-to-elastic-wave transduction wherever it occurred in the head. For example, suppose transduction takes place at the inner or outer surface of the skull with the skin and musculature intact (which may render the motion undetectable with this holographic technique). Would baring the skull by the removal of skin and musculature alter the characteristics of the transduction process significantly?

Olsen and Hammer (1980) used a hydrophone transducer implanted in a rectangular muscle-equivalent model to detect acoustical responses to exposure of the model to pulsed RFR. The model consisted of about 15 kg of polyethylene powder, water, sodium chloride, and gelling agent in proportions given by Guy (1971) and contained within an open rectangular polystyrene box. It was exposed to 0.5-microsecond 5.7-GHz pulses at 7-ms intervals (143 pps) at 5.5 cm from a standard-gain horn (about 0.07 of the conventional far-field distance) at an average power density of 120 mW/sq cm. The corresponding pulse power density exceeded 1.5 kW/sq cm. The authors noted that in the near field, the on-axis power density is a strongly oscillating function of the distance from the horn, and they found that the amplitude of the thermoelastic waves exhibited such behavior when the distance between the horn and hydrophone was slightly increased or decreased.

The SAR was determined calorimetrically at several depths within the model. The results were about 95 W/kg at 1 mm, 50 W/kg at 1 cm, and 5 W/kg at 2 cm, showing that most of the RFR energy was absorbed within the first 2 cm. The hydrophone used was directional and sensitive in the frequency range 50-700 kHz, and was inserted into the model on the axis of the horn at 15.24 cm from the exposed surface. The hydrophone output was fed to a broadband filter, amplified, and monitored with an oscilloscope. For comparison with prior studies, measurements with the hydrophone were also made in 1% saline (12 kg) in the polystyrene box.

The response of the muscle-equivalent model to the RFR pulses was a rapidly decaying thermoelastic wave lasting about 10 microseconds and a narrower RFR artifact, the latter serving as a convenient marker for measuring the time delay corresponding to the acoustic propagation speed of the thermoelastic wave. The delay between the thermoelastic response and the artifact was 85 microseconds. Using the depth of the hydrophone to calculate the propagation (group) speed yielded about 1800 m/s.

Asecond wave delayed from the first by about 380 microseconds was also evident. The authors ascribed this wave to two successive reflections, from the back and front surfaces of the model, for a total distance traveled (one round trip from the hydrophone) of twice the dimension of box parallel to the propagation direction, or 60.96 cm, a distance that yielded a propagation speed of 1600 m/s. They noted that transduction of the pulses into thermoelastic waves at the surface was tacitly assumed in the calculation of the higher speed, but that transduction actually occurs 1-2 cm from the surface. Taking 1.5 cm as the locus of transduction yields a speed of 1616 m/s for the 85-microsecond delay (with no change for the twice-reflected wave).

The amplitude of the twice-reflected wave was about 20% of the initial hydrophone response amplitude, which permitted the authors to estimate the acoustic attenuation within the model. Excluding reflection losses, estimated to be less than 10%, yielded an upper-bound loss of 14 dB for a 61-cm slab of muscle-equivalent material.

For studying transduced waves in saline, the hydrophone was placed 7.62 cm from the exposed surface. Unlike the rapidly decaying wave in the muscle-equivalent material, the RFR pulses yielded a ringing response by the hydrophone. The 61-cm-round-trip delay time between the initial and twice-reflected waves was 400 microseconds, which yielded a propagation speed of 1525 m/s, consonant with values found by others (Lin, 1978). Also prominent was a wave delayed by only 90 microseconds from the initial wave, apparently due to reflections from the hydrophone itself and from the front surface of the saline back to the hydrophone. (An analogous wave of intermediate delay time was barely discernible in the oscilloscope trace for the muscle-equivalent material.)

The authors suggested that the presence of ringing in the saline model indicated the induction of higher-frequency acoustic components and that the absence of ringing in the muscle-equivalent model may be because of a thinner RFR-absorption profile and/or greater high-frequency damping for the simulated muscle tissue.

Olsen and Hammer (1981) performed similar measurements in a rectangular muscle-equivalent model. However, the 0.5-microsecond, 5.7-GHz pulses were triggered at 760-microsecond intervals or twice the round trip time observed previously, to reinforce the thermoelectric waves. (Interpulse intervals of 380 microseconds could not be used because of equipment limitations.) An amplitude enhancement factor of about 3 was obtained at the end of a burst of four pulses.

Also studied by Olsen and Hammer (1981) was a spherical brain-equivalent model 10 cm in diameter. The model was exposed to 1.10-GHz RFR from an open section of waveguide, either as single pulses of 4-kW peak power and duration that was varied or as bursts of three such pulses with an adjustable interpulse interval. For a nominal 10-microsecond pulse, the SAR was 824 W/kg at the center of the sphere and 653 W/kg at the surface facing the source. A hydrophone was placed at the center of the model to detect thermoelastic waves.

Exposure of the model to single 14-microsecond pulses yielded ringing for each pulse, with a fundamental frequency of about 16 kHz and a time constant of about 500 microseconds, the latter said to be consistent with the attenuation obtained in the rectangular model. A plot of hydrophone response vs pulse duration over the range 10-60 microseconds showed maximum response for 20-microsecond pulses. Three-pulse bursts at burst frequencies ranging from 10 to 30 kHz yielded higher amplitudes than single pulses, with maximum enhancement at 16 kHz, as expected.

Olsen and Lin (1981) performed similar studies of spherical brain-equivalent models 6, 10, and 14 cm in diameter exposed to 10-kW, 1.10-GHz single pulses and bursts of three pulses from an open section of waveguide. To increase the amplitude of the thermoelastic waves in the 6-cm sphere, it was placed directly against the waveguide opening.

Aplot of peak hydrophone responses of the 6-cm sphere to bursts of 10-microsecond pulses vs burst frequency showed maximum response at 25.5 kHz. The corresponding data for the 10-cm sphere were the same as in Olsen and Hammer (1981). The response of the 14-cm sphere to single pulses was ringing at a fundamental frequency slightly above 10 kHz, and was maximum for 35-microsecond pulses. The responses of that sphere to bursts of 35-microsecond pulses had a peak at 11.5 kHz. Plots of the experimentally determined fundamental resonant frequencies for the three models were on the curve of resonant frequency vs head radius derived from the thermoelastic theory for a homogeneous brain sphere with stress-free boundaries, thereby supporting that theory.

In a subsequent study, Olsen and Lin (1983) surgically implanted disk hydrophone transducers 3.2 mm in diameter and 0.5 mm thick in the brains of rats, guinea pigs, and cats to measure the stress waves induced by RFR pulses. Connections to the transducers were made with coaxial cable 2.5 mm in diameter.

In the experiments with cats and guinea pigs, the hydrophone transducer was implanted about 1.5 cm deep in the brain of the anesthetized animal through a hole in the skull on the left side of the head near the top of the parietal bone. Next, the animal was taken to an anechoic chamber and exposed to 0.5-microsecond, 2 kW-peak pulses of 5.7-GHz RFR at 2 pps from a standard-gain horn, and to acoustic clicks. The output cable of the hydrophone was not connected during these stimuli; instead, metallic and nonmetallic electrodes (at unspecified head locations) were used to detect and compare brainstem potentials in response to each stimulus.

After the brainstem potentials were measured, the hydrophone output cable was connected to an oscilloscope and the animal was exposed to several series of 5.7-GHz RFR pulses at 14 pps. The animal was then removed from the chamber and placed next to a 3-kW-peak 2.45-GHz source, where its head was exposed to 2.5-microsecond pulses with a surface applicator. The output of the hydrophone was recorded during each exposure.

In the experiments with rats, brainstem potentials in response to 5.7-GHz pulses and acoustic clicks were measured before the hydrophone was implanted. After implantation, hydrophone signals were recorded during exposure to 0.5-microsecond, 5.7-GHz pulses in the anechoic chamber and to 5-6-microsecond, 2.45-GHz pulses with the applicator.

Representative hydrophone output waveforms for one cat and one guinea pig for the two RFR frequencies were presented, which showed that the shorter 5.7-GHz pulses stimulated vibrations having more of the higher-frequency components than the 2.45-GHz pulses. In addition, varying the 2.45-GHz burst frequency for the cat yielded maximum response near 40 kHz. Hydrophone output waveforms for six rats were also presented and were characterized similarly. In addition, a distinct vibration near 60 kHz, the computed fundamental mode of the rat brain, was discernible.

From their results, the authors concluded that RFR pulses do induce acoustic pressure waves in the brain, confirming previous predictions, particularly regarding the fundamental radial oscillation of the rat brain near 60 kHz. They also noted that the theoretically predicted frequencies are independent of heating patterns, but are functions only of the propagation speed of pressure waves and the size of the head. Open to question, however, is whether the use of coaxial cable and other metal leads introduced significant artifact.

Wilson et al. (1980) used C-14-labeled 2-deoxy-D-glucose (C-14-DG) to prepare autoradiographic maps of brain activity in 11 rats, to study the effects of exposure to RFR, acoustic clicks, and infrared radiation (IR) on the auditory system. In nine of the rats, the left bulla was opened, the ossicles were removed, and the bulla was packed with gelfoam. In the other two rats, one cochlea was destroyed by inserting a blunt probe through the round window. These operations, done prior to exposure, were to abolish or attenuate the response of one side of the auditory system to airborne sound.

Each rat was exposed to only one stimulus in a sound-isolation chamber for 45 min, while the rat was restrained within a cylindrical cage of RFR-transparent mesh. Just before exposure, each rat was injected with C-14-DG. On completion of exposure, the rat was euthanized and its brain was quickly removed, frozen, and sectioned in the frontal plane (30-micron slices). Autoradiographs of C-14-DG uptake throughout the brain were prepared and examined for differences in optical densities resulting from exposure to the various stimuli. Also, autoradiographs of representative sections through the auditory and vestibular nuclei were identified and enlarged, and the identified sections were stained with cresyl violet.

Two of four rats were stimulated with acoustic clicks at 87 dB SPL from a loudspeaker driven by 100-microsecond pulses at 10 pps. One of the remaining rats was exposed to IR from two heat lamps at a level stated to mimic the total thermal load induced by RFR exposure. Specifically, the voltage applied to the lamps was adjusted to yield a temperature-increase rate of a saline-filled beaker that matched the rate obtained from exposure to 918-GHz RFR at 10 mW/sq cm. (The authors did note that the spatial heating profiles for the IR and RFR were dissimilar.) The fourth rat was held in the sound-isolation chamber without stimulation.

Autoradiographs of these four rats (denoted as controls, i.e., not exposed to RFR) were qualitatively similar; all showed bilateral asymmetry in C-14-DG uptake by the inferior colliculus and medial geniculate body, with higher uptake on the side contralateral to the intact middle ear. As noted by the authors, this form of asymmetry was expected because most ascending pathways from one cochlea lead to the central nucleus of the contralateral inferior colliculus. Not clear was why the autoradiographs for two such different stimuli (acoustic clicks and IR) and those taken in their absence were so similar to one another.

In an initial experiment, one of the rats with left ossicles removed was exposed to 20-microsecond pulses, 10 pps, of 2.45-GHz RFR at an average power density of 2.5 mW/sq cm, for a peak power density of 12.5 W/sq cm. The autoradiographs of this rat from the inferior colliculus, unlike those for the control rats, showed bilateral symmetry in C-14-DG uptake, taken as indicating the utility of the C-14-DG method for demonstrating a known effect of RFR exposure on brain activity. The four remaining rats with left ossicles removed were then exposed to 918-MHz CW RFR, two each at 2.5 and 10 mW/sq cm, to identify possible effects of the CW RFR. The corresponding SARs in the midbrain were 1.1 and 4.4 W/kg, determined thermometrically with rat carcasses. For both CW RFR levels, C-14-DG uptake in the inferior colliculus was also bilaterally symmetric and the autoradiographs were "surprisingly similar" to those for the pulsed RFR.

The authors stated: "To exclude the possibility that CW microwave radiation produced this result by direct action on brain tissue, additional data were obtained from two animals in which one cochlea was destroyed. In both animals, the uptake of C-14-DG was greatest at the inferior colliculus contralateral to the intact cochlea. The degree of asymmetry at the inferior colliculus was, in fact, at least as great as that found in any of the control animals. This finding, coupled with the finding of a bilateral symmetry of C-14-DG uptake in the auditory pathways of animals with one middle ear ablated, demonstrated that CW microwave radiation acts at some site within the cochlea in eliciting auditory responses."

C-14-DG uptake in other structures of the auditory system, such as the lateral superior olive, medial superior olive, or cochlear nucleus, showed no bilateral differences except in the two rats with one cochlea destroyed. Autoradiographs from regions outside of the auditory system were bilaterally symmetric and showed no stimulus-related qualitative differences.

In their discussion, the authors suggested that the activity of the rat's auditory system in response to CW RFR, determined by integrated C-14-DG uptake during a 45-min period, is an effect distinct from the thermoelastic responses to RFR pulses, and that the CW interaction appears to occur somewhere within the cochlea. They estimated that a steady-state increase in intracochlear temperature of between 0.1 and 0.5 deg C would be induced in live rats exposed to 918-MHz CW RFR at 2.5 mW/sq cm, suggesting that such increases in temperature may be effective in altering auditory activity. In this context, it is reiterated that the average power density of the 2.45-GHz pulsed RFR used in this study was also 2.5 mW/sq cm and that the midbrain SAR was 1.1 W/kg.

In conclusion, the preponderance of experimental results indicates that auditory perception of RFR pulses is due to induction of thermoelastic waves in the head, rather than to direct brain stimulation by the RFR. Also, because individual pulses can be perceived, it is not meaningful to calculate average power densities for two or more widely spaced pulses and cite such values as evidence that the effect is nonthermal in nature. The response to CW RFR reported by Wilson et al. (1980) is an effect distinct from the thermoelastic responses to pulses and possibly is related to an intracochlear temperature rise of 0.1 to 0.5 deg C.


Cain, C.A. and W.J. Rissman
IEEE Trans. Biomed. Eng., Vol. 25, No. 3, pp. 288-293 (1978)

Chou, C.-K., R. Galambos, A.W. Guy, and R.H. Lovely
J. Microwave Power, Vol. 10, No. 4, pp. 361-367 (1975)

Chou, C.-K., A.W. Guy, and R. Galambos
Radio Sci., Vol. 12, No. 6S, pp. 221-227 (1977)

Chou, C.-K. and R. Galambos
J. Microwave Power, Vol. 14, No. 4, pp. 321-326 (1979)

Chou, C.-K. and A.W. Guy
J. Microwave Power, Vol. 14, No. 4, pp. 399-404 (1979a)

Chou, C.-K., and A.W. Guy
Radio Sci., Vol. 14, No. 6S, pp. 193-197 (1979b)

Chou, C.-K., A.W. Guy, K.R. Foster, R. Galambos, and D.R. Justesen
Science, Vol. 209, pp. 1143-1144 (5 Sept 1980a)

Chou, C.-K., K.-C. Yee, and A.W. Guy
Bioelectromagnetics, Vol. 6, No. 3, pp. 323-326 (1985a)

Foster, K.R. and E.D. Finch
Science, Vol. 185, pp. 256-258 (19 July 1974)

Frey, A.H.
Aerospace Med., Vol. 32, pp. 1140-1142 (1961)

Frey, A.H.
J. Appl. Physiol., Vol. 17, No. 4, pp. 689-692 (1962)

Frey, A.H.
J. Appl. Physiol., Vol. 23, No. 6, pp. 984-988 (1967)

Frey, A.H. and R. Messenger, Jr.
Science, Vol. 181, pp. 356-358 (27 July 1973)

Frey, A.H. and E. Coren
Science, Vol. 206, pp. 232-234 (12 Oct 1979)

Frey, A.H. and E. Coren
Science, Vol. 209, pp. 1144-1145 (5 Sept 1980)

Guy, A.W.
IEEE Trans. Microwave Theory Tech., Vol. 19, No. 2, pp. 205-214 (1971)

Guy, A.W., C.-K. Chou, J.C. Lin, and D. Christensen
Ann. N.Y. Acad. Sci., Vol 247, pp. 194-218 (1975b)

Guy, A.W., J. Wallace, and J. McDougall
Radio Sci., Vol. 14, No. 6S, pp. 63-74 (1979)

Johnson, C.C. and A.W. Guy
Proc. IEEE, Vol. 60, No. 6, pp. 692-718 (1972)

Lebovitz, R.M. and R.L. Seaman
Radio Sci., Vol. 12, No. 6S, pp. 229-236 (1977)

Lin, J.C.
J. Microwave Power, Vol. 11, No. 1, pp. 77-81 (1976a)

Lin, J.C.
J. Microwave Power, Vol. 11, No. 3, pp. 295-298 (1976b)

Lin, J.C.
IEEE Trans. Microwave Theory Tech., Vol. 25, No. 7, pp. 605-613 (1977a)

Lin, J.C.
IEEE Trans. Microwave Theory Tech., Vol. 25, No. 7, pp. 938-943 (1977b)

Lin, J.C.
Radio Sci., Vol. 12, No. 6S, pp. 237-242 (1977c)

Lin, J.C.
Charles C. Thomas, Springfield, IL, p. 108 (1978)

Lin, J.C., R.J. Meltzer, and F.K. Redding
J. Microwave Power, Vol. 14, No. 3, pp. 291-296 (1979b)

Olsen, R.G. and W.C. Hammer
Bioelectromagnetics, Vol. 1, No. 1, pp. 45-54 (1980)

Olsen, R.G. and W.C. Hammer
J. Microwave Power, Vol. 16, Nos. 3 & 4, pp. 263-269 (1981)

Olsen, R.G. and J.C. Lin
IEEE Trans. Microwave Theory Tech., Vol. 29, No. 10, pp. 1114-1117 (1981)

Olsen, R.G. and J.C. Lin
IEEE Trans. Biomed. Eng., Vol. 30, No. 5, pp. 289-294 (1983)

Sharp, J.C., H.M. Grove, and O.P. Gandhi
IEEE Trans. Microwave Theory Tech., Vol. 22, No. 5, pp. 583-584 (1974)

Taylor, E.M. and B.T. Ashleman
Brain Res., Vol. 74, pp. 201-208 (1974)

Tyazhelov, V.V., R.E. Tigranian, E.O. Khizhniak, and I.G. Akoev
Radio Sci., Vol. 14, No. 6S, pp. 259-263 (1979)

White, R.M.
J. Appl. Phys., Vol. 34, No. 12, pp. 3559-3567 (1963)

Wilson, B.S., J.M. Zook, W.T. Joines, and J.H. Casseday

Brain Res., Vol. 187, pp. 291-306 (1980)