Overview and History of the Acoustical Evidence in the Kennedy Assassination Case
by D.B. Thomas
In 1979 the House Select Committee on Assassinations (HSCA) reported that President Kennedy’s death was probably the result of a conspiracy. The primary basis for that conclusion was acoustical evidence of a gunshot from the Grassy Knoll. Not only has this evidence withstood serious challenge, it has now been amply corroborated. Nonetheless, it is evident from commentary on the web, television, and even in scientific journals, that the acoustics and its corroborative evidence are widely misunderstood. This essay attempts to explain the technical and non-technical aspects of the evidence and in that light to discuss the issues and controversies that have arisen since the original HSCA study. Most importantly, coupling the audio evidence with the video evidence provides us with a coherent reconstruction of the murder.
THE PHYSICAL EVIDENCE
Gray Audograph Disc Recorder like the one used by DPD to record channel two radio traffic.
On the day that President Kennedy was assassinated, the Dallas Police Department (DPD) was communicating over two radio frequencies, both of which were recorded. The primary channel, designated by them as channel one (Ch-1), was used for routine transmissions and was recorded by a Dictaphone belt recorder. An auxiliary channel, designated by them as channel two (Ch-2), was used for special events, in this case, for the police escort of the President’s motorcade. This channel was connected to a Gray Audograph disc recorder. Both machines were needle-in-groove type recorders in that each used a sharp stylus which cut an acoustical groove into a soft vinyl surface to make the recordings. The respective machines could then be switched to play back mode using the same stylus. During the Warren Commission era, the recordings were subjected to multiple playbacks for the making of transcripts and copies. Because they were not designed for multiple playbacks, the recordings suffered significant attrition at that time. Although both recordings still exist in the possession of our National Archives, neither can be safely played back today. However, in 1963, the officer in charge of the DPD communications department, James C. Bowles, made high quality electromagnetic tape copies of both recordings. Because of the poor condition of the originals, the Bowles tapes are considered to be the “best” record of the evidence on the recordings. During the subsequent HSCA investigation, new playback copies were made by the acoustical experts who analyzed the recordings, and in 1982, the FBI also made playback copies of the original disc and dictabelt. But because the disc and dictabelt were already aged and use-worn by that time, these copies contain artifacts and do not have the fidelity of the Bowles tapes.
Model AT2C Dictaphone Belt Recorder like the one used by DPD to record channel one radio traffic.
The Dictaphone machine used by the Dallas Police was a piggyback unit. When one belt became full, the machine would automatically begin to record on the second unit. Because a belt can only contain about 15 min of continuous recording a technician was constantly on hand to replace belts as they became full. To extend the time between belt changes, the machine was outfitted with a sound actuation switch which would stop the recorder during dead air (after about 4 sec), and automatically resume recording when a transmission was received. The Audograph disc resembled a more traditional phonograph record, except that this recording machine had a stylus on a fixed arm. The turntable is mounted on an axis which rides in a slot such that the axis is driven perpendicular to the stylus arm by a worm screw as the turntable rotates. Hence, unlike a traditional phonograph record, the Audograph disc is recorded (and played back) from the inside out, and playback is at linear track speed (inches per minute) instead of revolutions per minute. This arrangement prevents the problem common to floating stylus arms wherein the needle can be “stuck” in one groove until corrected manually. Also, it maximizes recorded message density in terms of signal per inch of acoustical groove, compared to the less efficient phonograph. Audograph Discs came in two sizes, 9 min and 30 min capacity. This machine was also outfitted with a sound actuation switch.
AUTHENTICITY OF THE RECORDINGS
Photo of the actual dictabelt - note the grease pen labeling.
Essentially all of the physical evidence in the Kennedy assassination has been compromised, and this includes the DPD recordings. The DPD in 1963 was particularly negligent in its duty to mark evidence or maintain proper chains of possession. Following the murder of the only suspect in the case, while in the custody of the DPD, police officers purloined most of the physical evidence for souvenirs. Much but not all of the evidence was retrieved by the FBI for the Warren Commission’s investigation. The recordings now in evidence were recovered from the private home of a police officer by the HSCA in 1978. The Dictabelt held by the National Archives has writing on its surface made with a white grease pen indicating that it is belt No. 10 from the date of 22 November 63. The DPD technician with the responsibility for operating the recorders in 1963 was able to identify the writing as hers. The problem is that in the transcripts made by the FBI during the Warren Commission era, the corresponding belt is identified as belt No. 5. The FBI may have used its own numbering system in making the transcripts and ignored the labels on the recordings. But typically the FBI evidence numbers begin with a letter Q. The two different numbers suggests, but does not prove, that there were at one time, two different belts, of which one would have to be a copy.
Mary Ferrell, who along with Gary Mack brought the DPD recordings to the HSCA's attention.
It is known that multiple legitimate and illicit copies of the recordings were made by the DPD, many for the souvenir hunters in the department. It is also known that possession of the recordings shifted between the DPD and the FBI without proper paperwork. It is thought that all or most of these taped copies originate with the first Bowles tape copy. The tape copy first provided to the HSCA by Mary Ferrell was determined to be a multiple generation copy. It is also known that Bowles rented a Dictaphone machine to playback the Dictabelt for the taped copy of the Ch-1 transmissions, as well as for the preparation of transcripts. Copies are often detectable because the recording process introduces a hum from the recording machine’s motor. Indeed, overlapping hums are evident on Dictabelt No. 10 (a background hum of 120 Hz resulting from two 60 Hz hums out of phase). James Barger, the lead scientist with the HSCA study has suggested that this secondary hum may indicate that it is a copy, rather than the original. Obviously, to make a copy of the belt one would require two instruments, one to playback the original belt and one to create the copy. Alternatively, Linsker et al. (2006) point out that the second hum only proves that there were two instruments on line when the recording was made and it isn’t known for certain what the second instrument was. Hence, the second hum does not prove that the belt in the archives is a copy, although it is consistent with that suspicion.
In 2004 it was reported that the National Archives had arranged with Lawrence Livermore Laboratory to make a virtual playback copy of the DPD Dictabelt using Laser technology. This technology has been successfully applied to produce playbacks of the acoustical grooves in the old Edison cylinder recordings which were in use before the invention of the phonograph machine. This was necessary because the original belt has become shrunken and brittle with cracks in the margins. Of course a true copy will also include any artifacts inflicted on the belt over the years, hence, for historical as well as acoustical studies, the best recording will remain the Bowles tapes which were made with high fidelity equipment when the belt was still in relatively new condition. The advantage of the Laser copy is that it will be an authentic copy of the evidence recording in the sense that it will be an untampered copy, which is not necessarily the case with the recordings available on E-Bay. However, as of this writing, the dictabelt is still in the archives and any plans for a laser copy are on indefinite hold.
CIRCUMSTANCES SURROUNDING THE RECORDING
The broadcast transmissions from the Presidential motorcade are recorded on the Audograph disc of Ch-2. Prior to the assassination most of the broadcasts originated from the Chief of Police Jesse Curry who was in the lead car of the motorcade. Curry would state the location of the motorcade as it wound through the streets of downtown Dallas. The first broadcast indicating that the shooting had happened was a shouted order by Curry telling the escort to proceed directly to Parkland Hospital. A transcript of these broadcasts over Ch-2 during the relevant time interval is provided here.
Table 1.- Transcript of DPD Ch-2 at approximately 12:30 p.m. 22 Nov. 1963
|Caller & Call Number||Broadcast|
|Lawrence ||I'm at the Trade Mart now. I'll head back out that way.|
|Fisher :||Naw, that's all right, I'll check it.|
|Curry :||[garbled] the Triple Underpass.*|
|Dispatcher:||10-4, 1 - 15 Car 2.|
|Dispatcher:||12:30 - KKB364.|
|Lawrence :||125 to 250.|
|Dispatcher:||15 Car 2.|
|Curry :||. . . to the hospital! We're going to the hospital officers! Go to the hospital! We're on our way to Parkland Hospital! Have them stand by!|
|*The first word in this transmission is not clear. An early transcript made by the DPD interprets the first word as "approaching."||Listen to Audio Recording|
Photograph of clock showing 12:30, the time of the gunfire in Dealey Plaza.
A significant aspect of these recordings is that as a part of radio protocol the radio dispatcher would append a time notation to his broadcasts. This was especially significant because recorded time is not actual time because of the sound actuation feature. It was also a part of radio protocol that the dispatcher was responsible for broadcasting the station’s call numbers at regular intervals. Because of adherence to the protocol we know that the assassination occurred very close to 12:30 local time, because the dispatcher made his 12:30 station identification just moments before Curry made his broadcast ordering the escort to go to Parkland Hospital.
Over on Ch-1, the primary police channel, a most unusual but fortuitous event occurred. For approximately 5-1/2 minutes the frequency is dominated by the sound of a motorcycle motor. Somewhere in Dallas the microphone on a patrolman’s radio had become stuck in the on position. Because the dispatchers on this channel (there were two) were also making time notations, we know that this motorcycle segment begins at approximately 12:28 and runs until approximately 12:34 local time. It thus overlaps the time of the assassination. The Ch-1 dispatchers were on opposite sides of a large radio console which provided them with the means of maintaining radio contact with the hundreds of patrol units simultaneously. Each dispatcher had his own digital clock for the purpose of making the time notations. The Ch-2 dispatcher had a separate console and his own clock. According to JC Bowles, who provides a detailed account of the communications department operation, these clocks were regularly synchronized with one another and with a master analog wall clock. The wall clock was synchronized to official time once a month. All of these clocks should have been within one minute of one another, but may have been as much as two minutes apart.
Because it happened during the assassination, at one time it was suspected that one of the motorcycle police officers might have deliberately held his microphone open in order to jam police communications. Ostensibly this would have facilitated the escape of the perpetrators by interfering with a coordinated police response mediated by radio communication. It is now accepted that the motorcycle segment was accidental. Firstly, the open motorcycle segment is only one of several that day, and not just at the time of the shooting. The microphone relay button worked by making a contact which when depressed closed the circuit so that the radio is in transmit mode. The switch had a spring which held it in the off position when not depressed. If the spring broke or came loose, the relay became free to slide between the contact and non-contact position. Hence, when the motorcycle was in motion the relay would slide making intermittent contact. The motorcycle broadcasts happened at least four times in the hour leading up to and following the assassination. Secondly, because the motorcycle segment ended only three minutes after the shooting there was little interference with any response the police might have made because in the immediate aftermath there was mostly confusion anyway. Thirdly, anyone wanting to jam communications to interfere with police action in response to the assassination would have jammed Ch-2, the motorcade channel, not Ch-1.
Based on the Ch-2 dispatcher’s time notation, the assassination occurred at approx. 12:30. At 12:33 in response to the motorcycle noise on Ch-1, the Ch-2 dispatcher made a crucial broadcast saying,
"There’s a motorcycle officer up on Stemmons with his mike stuck open on channel one. Could you find someone to tell him to shut it off!"
Listen to Audio Recording
This raised the crucial question, what made the dispatcher so certain that the motorcycle was on the Stemmons freeway? The most likely clue was the fact that at 12:32 one can hear sirens in the background over the motorcycle motor. Because the one emergency at the time was the assassination, and because the motorcade was at that moment on the way to Parkland hospital, and because the fastest road from Dealey Plaza, the scene of the shooting and Parkland hospital was the Stemmons freeway, the dispatcher had made the inference that the open microphone must be on Stemmons. If this inference is correct, given that there were 18 motorcycles assigned to escort the President’s motorcade, then there was a possibility, if not likelihood, that the unit with the open microphone might have been one assigned to the motorcade. That being the case then there was also the possibility that the open microphone could have been in Dealey Plaza with the motorcade when the shooting occurred. And that being so, then the gunshots could have been captured over the open microphone and might be detectable on the Ch-1 recording, somewhere in the background of the motorcycle noise.
ACOUSTICAL ANALYSIS OF THE DPD RECORDING
National Guardsmen and student protesters at Kent State in 1970. Acoustic analysis proved that three Guardsmen fired the first shots, and identified them by location.
Mary Ferrell and Gary Mack were among the first to realize the significance of the motorcycle sequence and brought this to the attention of the HSCA. Investigators for the HSCA recovered dictabelt No. 10 and contacted the Acoustical Society of America for advice on an analysis of the recording. The ASA provided a short list of three laboratories with the required expertise, at the top of which was the Cambridge MA firm of Bolt, Baranek & Newman (now BBN Technologies). This was the laboratory which had analyzed the Watergate tapes. More importantly, this was the lab which had analyzed the Kent State shooting tapes, the first forensic application of acoustics in a criminal case. The acoustical experts at BBN were able to show that not only had the National Guard soldiers fired on the students, contrary to their subsequent claims that they had only returned fire, but were able to pinpoint the three individual soldiers who had fired the first shots. All three, identified in photographs by the FBI, admitted in interrogation that they had discharged their weapons.
Vehicles equipped with Boomerang anti-sniper echo location system.
These same scientists were now asked to bring the same technology to bear on the DPD dictabelt recording. The principle involved is echo location; the same method by which a submarine is able to navigate without windows, and bats are able fly in caves. In fact the same principles were used by the same BBN scientists to develop the Boomerang technology now deployed by our soldiers in Iraq to instantly locate sniper positions.
The first step in the analysis was to determine if the gunshots are even on the dictabelt. This was not as simple as it sounds (no pun intended). A gunshot makes a sound classified as a white noise, which is to say that it does not have a characteristic frequency as do most of the ambient sounds in our environment. Sound is a disturbance in the air. The gun disturbs the air because the bullet in passing from the chamber to the muzzle causes the column of air in the barrel to collide violently with the air in front of the muzzle. Hence it is called a muzzle blast. Among the characteristics of a muzzle blast are that it is very loud and very brief. Only noises like rock concerts and jet engines are as loud, but those noises are not brief in duration. The problem is that loudness is a function of both the energy that goes into disturbing the air, its intrinsic loudness, but also the distance between the source and the listener. If one is a mile away from the gun, then the muzzle blast will not be very loud. Conversely, a clap of the hands next to the ear will seem very loud to the listener. The problem then is, if one finds a loud, brief noise on a recording, how does one know that the sound is a gunshot as opposed to some other sudden impulsive white noise with a source close to the microphone. The answer lies in the fact that an intrinsically loud noise made in an urban environment, such as downtown Dallas, will reverberate off of the buildings. Such echoes will be audible at considerable distances. By processing the motorcycle segment through an oscillograph, an instrument which produces a visible representation of the sound waves, the acoustical experts searched for high amplitude sound impulses that were in clusters that would represent the muzzle blast and its succeeding echoes.
A portion of the oscillograph showing one of the impulse patterns, in this case the one identified as the grassy knoll shot. The horizontal dotted lines show the cut-off level, which spikes must exceed in order to be counted as part of an impulse sequence.
There are many other technical details that the acoustical experts applied in this search, which can be found in their report to the HSCA. Most of these are not at issue so there is no need to repeat them here. Suffice it to say that in their oscilloscope screening of the 5-1/2 minutes of motorcycle recording, they located a sequence of sounds which on acoustical criteria could be the assassination gunfire. The sequence was ten seconds long, occurred almost exactly two minutes into the motorcycle segment (and thus almost exactly at 12:30) and contained five candidate impulse patterns. There is some confusion on this point because a sixth pattern was also suspicious. It was an attenuated pattern which was considered only because it was in close proximity to the others, a sort of guilt by association. It was assumed that the assassination gunfire were shots from a rifle. The presence of the attenuated pattern might indicate a pistol shot.
Because white noises are commonplace in radio transmissions, there are many potential explanations for the sound patterns on the dictabelt. They might have been nothing more than bursts of static from overhead wires, or shorts in the motorcycle’s electrical system. There might have been a lightning storm in the distance; it did rain in Dallas that morning. Anything that would produce brief impulsive noises in a cluster on a radio broadcast might look the same as a gunshot echo pattern. The laboratory study was a preliminary screen meant to determine whether or not such suspect sounds were even present on the recording. What was significant about the suspect sound patterns, was that they were grouped into a sequence as expected from the circumstances of the shooting and that they were deposited on the recording at or very near to the time of the shooting. Nonetheless, the definitive test would be to compare these suspect sounds with recordings of real rifle shots in Dealey Plaza. The suspect patterns on the DPD tape had the acoustical characteristics of gunshots in the generic sense. Test shots would show exactly what an echo pattern from Dealey Plaza would look like.
FIELD TESTS IN DEALEY PLAZA
Photo taken of microphones on Elm Street in Dealey Plaza during the HSCA's acoustical testing.
(view larger version)
In August 1978, with the help of the DPD, gunshots were fired and recorded on microphones placed along the President’s motorcade route through Dealey Plaza. When these test patterns were then compared to the suspect sound patterns on the dictabelt, they were found to match. That is, all five of the suspect sound patterns identified on acoustical criteria in the laboratory analysis were found to match to the echo patterns of test shots fired in Dealey Plaza. The odds of this happening if the sounds were not the assassination gunfire were remote. It would certainly be possible for a stray noise pattern to match closely to one of the sounds on the recording, especially if it were from a white noise source that tended to occur in clusters; but for all five to match was very unlikely (the suspicious attenuated sound that preceded the others failed to match to any of the test shots, even though the test shooting included firing a pistol on the grassy knoll and firing of a rifle with its muzzle withdrawn inside the book depository building).
Some details are provided here. Comparisons were based on echo delay time. Echo delay time is the time in milliseconds (thousandths of a second) between the arrival of the muzzle blast at the microphone and the succeeding echoes. Because there are five large buildings facing the motorcade route in Dealey Plaza, it is expected that a gunshot sound would reflect off the face of each building, but there would also be a refractive echo from the corner of each building. Hence, around ten large echoes would be expected, with the actual number depending on the actual position of the microphone relative to the buildings. Because of the dimensions of the plaza (about 500 ft across), and the speed of sound (about 1100 ft/sec), all of the echoes should arrive within a quarter to a half second (500 msec) following the muzzle blast. A match was scored if an impulse in the suspect pattern was at or close to the same echo delay time as an impulse in a test pattern. By close it is meant that they were within 6 msec of one another. The slack was due to the fact that the microphones were arrayed at 18 ft spacings. The degree of match was determined by calculation of a coefficient using a simple formula.
Where, N is the number of large impulses on the test pattern
n is the number of large impulses on the suspect pattern
and, M is the number that matched among them.
For example, if there were ten impulses in each pattern and nine of them matched, the value of the coefficient would be 0.8, or if 8 impulses matched then the value would be 0.6. Only comparisons which gave a score of at least 0.5 would be statistically significant (no more than a 5% likelihood to occur by chance), and this value served as the “detection threshhold.” Four of the five suspect patterns matched to a test pattern with a correlation coefficient of 0.8; one suspect pattern (the third) matched to a test pattern with a score of 0.6.
But there was an even more convincing aspect to the results beyond the matching of the sound patterns. It was the order in the matches.