During the tenure of the House Select Committee on Assassinations, Acoustical experts determined that the sounds of the Kennedy assassination had been captured by a Dallas Police radio microphone and recorded by their communications recording system. It was further determined that among the sounds determined to be gunshots was one which had been fired from the Grassy Knoll . Because of the considerable historical significance of this evidence another opinion was sought and the findings of the first laboratory were confirmed by a second . These findings led to the official conclusion of the Assassinations Committee that there probably was a conspiracy behind the death of President Kennedy . A formal request was made to the Department of Justice to reopen the case for the purpose of identifying the perpetrators. The Justice Department chose instead to commission yet further study of the acoustical evidence. This third study, conducted by a panel organized by the National Research Council found that the HSCA conclusion was invalid on the grounds that the sounds alleged to be the assassination gunfire had been deposited on the Dallas Police recordings at a point in time after the assassination was over . The purpose of my presentation today is to address this issue and to explain how one might arrive at a different conclusion.
On the fateful day that John F. Kennedy visited Dallas in November 1963, the Dallas police were recording their radio transmissions over two channels. A frequency designated as Channel One (CH-1) was used for routine police communications. A second channel, designated Channel Two, (CH-2) was an auxiliary channel, which on this day was dedicated to the President's motorcade. Up until the time of the assassination, most of the broadcasts on this channel consisted of Police Chief Jesse Curry's announcements of the location of the motorcade as it wound through the streets of Dallas (James C. Bowles, head of the DPD communications department in 1963 has published transcripts of the police tapes  and all excerpts cited herein follow his interpretation). Chief Curry was in the motorcade's lead car. President Kennedy's limousine was the second car in line. The context of the broadcasts on CH-2 allows us to fix the approximate time of the assassination. In accordance with established radio protocol, the dispatcher on each channel regularly announced the time, frequently appending the notation to his calls. Fifteen seconds after the dispatcher on CH-2 had noted the time at 12:30, Curry began shouting into the microphone,
"Go to the Hospital, We're going to the hospital,
Officers, Parkland Hospital."
This was the first of a series of calls concerning the assassination and allows us to place the shooting at between 12:30 and 12:31 by the CH-2 dispatcher's clock. Meanwhile, over on CH-1, a most fortuitous event occurred. Beginning at about 12:28, by the CH-1 dispatcher's clock, and lasting until about 12:34, a microphone on a police motorcycle radio stuck open. For about 5-1/2 min, the CH-1 broadcast was dominated by the sound of the motorcycle's motor. Unable to communicate effectively, police officers began to switch over to CH-2. At 12:33 the CH-1 dispatcher broadcast,
"There's a motorcycle officer up on Stemmon's with his mike stuck open.
Could you send someone up there to tell him to shut it off!"
A relevant question is, what made the dispatcher so certain that the motorcycle with the open microphone was on the Stemmons Freeway? An important clue is that around 12:32 the open microphone had captured the sound of sirens. The only emergency in Dallas at that moment was the assassination and at 12:32-33 the President's motorcade was on the Stemmons Freeway en route to Parkland Hospital. There were 18 motorcycles escorting the President's motorcade and evidently the dispatcher had inferred from the sirens that the unit with the stuck microphone was among them.
If this inference was correct, then there was a strong possibility that the motorcycle and its open microphone was one of the 13 police motorcycles that were in Dealey Plaza at the time of the assassination, and if so, it could have captured the sounds of the gunfire on the police recording system. Eventually acoustical experts did find suspect sounds with the characteristics of gunshots recorded during the motorcycle sequence. The suspect sounds occur in a nine second interval that begins approximately two minutes into the stuck microphone sequence. The NRC panel concluded that this nine second interval was not coincident with the time of the assassination.
If the suspect sounds truly are the assassination gunshots they would have been deposited on the police recording system coincident with the murder. If they are not the assassination gunshots then it is much less likely that they would be coincident. Thus, an analysis of events on the separate police channels provides us with an independent test of the hypothesis that the gunfire has been correctly identified. Several broadcast events are common to both channels during or close to the motorcycle segment. Recall that officers were switching between channels because of the problem on CH-1. This channel surfing, combined with the urgency of numerous callers trying to reach the dispatcher with information related to the incident at the same time, resulted in several instances of crosstalk. Crosstalk is a phenomenon that happens when a broadcast over one channel is picked up by a live microphone on a radio tuned to the other channel. The result is a simulcast; the simultaneous broadcast of one call over both channels.
There are five such simulcasts during the motorcycle segment. The crosstalk instance which occurs closest to the time of the assassination, as judged by the context of the broadcasts on CH-2, was a broadcast by Deputy Chief Fisher. I will now play a three minute section of the CH-2 recording, ending with Chief Curry's call "Go to the Hospital." A transcript is provided (Table 1) to aid in following along with the recording at the time of the last calls.
Table 1.- Transcript of CH-2 DPD Radio Broadcasts at around 12:30 p.m.
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: "At the Triple Underpass."
DISPATCHER: "10-4, One. 15 Car 2.”
DISPATCHER: "12:30, KKB364.”
LAWRENCE: "125 to 250."
DISPATCHER: "15 Car 2."
CURRY: "GO TO THE HOSPITAL. WE'RE GOING TO THE HOSPITAL, OFFICERS."
Part of Fisher's broadcast on CH-2, the phrase, "I'll check it," crossed over to CH-1. I will now play the CH-2 phrase followed by the gunfire sequence on CH-1 so that you can hear for yourselves the crosstalk under discussion and the putative gunshot sounds that follow two seconds later.
Table 2.- CH-1 DPD Radio Broadcasts near 12:30 p.m.
FISHER: "I'll check it."
[STATIC LIKE SOUNDS ALLEGED TO BE THE GUNSHOTS]
DECKER: "Hold everything secure..."
Two seconds after the Fisher phrase on CH-1 the gunfire sequence begins. Two seconds after the Fisher phrase on CH-2 one finds Curry's announcement that he is, "At the Triple Underpass." Ergo, the first putative gunshot sound was recorded synchronous with a time when the lead car of the motorcade was in a position near the Underpass. By referring to this aerial photo of Dealey Plaza we can get some perspective on the topology of events. The Triple Underpass is the railroad bridge at the east end of Dealey Plaza. Chief Curry must have been on Elm Street somewhere in the vicinity of what is now called the "Grassy Knoll" when he made his broadcast. The question is, where was the President's limousine at this time? In this photograph, a polaroid taken by witness Jack Weaver, we can see both the lead car and the President's limousine in Dealey Plaza moments before the assassination. The lead car is approximately 150 ft ahead of the limousine on Houston Street at this time. Another film in which the two cars appear together is in the famous Zapruder film, which shows the two cars arriving together at the Triple Underpass after the shooting. The proximity of the two cars shown by the filmed evidence demonstrates that the President's limousine must have been on Elm Street when Curry made his call and hence, the juxtaposition of the broadcasts relative to the alleged gunfire sequence, and the juxtaposition of the cars on Elm Street relative to the real gunfire are found to give a close match.
There is a problem with this scenario, however, as straightforward as it may seem. The issue is whether or not one can rely on the crosstalk instances to provide an unequivocal timeline of events. For example, the NRC panel relied on a different instance of crosstalk, a broadcast by Sheriff Decker which occurs at least one minute after the assassination, or at least one minute after Chief Curry's call to "Go to the hospital." On CH-1 a fragment of Decker's broadcast, "Hold everything secure," crossed over from CH-2 and is found at the end of the gunshot sequence. The NRC panel relied on the Decker crosstalk as evidence that the putative gunshots were actually recorded one minute after the assassination. The problem is that the elapsed tape time between Fisher's and Decker's broadcasts on CH-2 is 98 seconds, but the elapsed tape time between Fisher and Decker crosstalks on CH-1 is only ten seconds. Clearly there is a serious time discrepancy involved. Moreover, there are five simulcast events between the two channels at the time of the motorcycle segment, but the time intervals between any pair is different on the separate channels (Tables 3 & 4). As a consequence, the crosstalks misalign regardless of which pair of simulcasts one chooses as the tiepoint. This is demonstrated in the graphic representation of the timelines where I have used the two close crosstalks by Sergeant Bellah as the tiepoint.
Therefore, in order to have confidence in the synchronization of events on the two channels we must first resolve the time discrepancies between the crosstalks. To do that we require two things: firstly, we need to understand the phenomena that are responsible for the time discrepancies, and secondly, we must have an independent source of information on the real lapse of time between events.
12:29:50 "I'll check it" Fisher
12:31:28 "Hold everything secure" Decker
12:33:38 "Check those radios" Bellah-1
12:33:50 "You want me ... Stemmons" Bellah-2
12:35:21 "Attention all ... vehicles" Henslee
*Times anchored to notation of 12:30 on FBI 60 Hz adjusted playback.
Fisher to Decker 10 98
Decker to Bellah-1 158 130
Bellah-1 to Bellah-2 15 12
Bellah-2 to Henslee 114 91
Fisher to Henslee 297 331
*Times from 60 Hz adjusted playback reported by Michael O'Dell.
The crosstalk misalignments result from three primary factors: recorder stoppage, stylus displacements, and warps in playback speed of the recordings.
A) Playback Speed.- Speed warps during playback cause measured time intervals to differ from real time. For those of you who still remember phonograph records, this problem is exactly the same as one would get by playing a 33 rpm record at 45 rpm. In this case the speed warp resulted from using a playback instrument different from the instrument that made the original recordings. Captain James C. Bowles, a supervisor with the police communications department in 1963, rented a dictaphone machine to playback the police dictabelt for copying and transcription. Because the DPD dictaphone machine was in constant use, Bowles had to use the rented machine to make his copies of the CH-1 dictabelt. The problem with dictaphone machine is that there is no set recording or playback speed. It had a knob for adjusting the belt speed (presumably as an aid to transcription) but there are no increments or standard settings. By using a different machine for playback it was inevitable that a chronic speed warp would be introduced. The problem was not as large or as obvious as playing a 33 record at 45 speed, but only about a 5% difference. But even a 5% difference becomes significant. Over a 6 min segment of recording the timeline error would be about 18 sec, which is too large to ignore. The degree of warp was measured by the HSCA's acoustical experts by using the 60 Hz power hum from the motor that drove the dictabelt. It was found that when Bowles made his tape recorded copy of CH-1 his rented dictaphone machine drove the belt about 5% faster than when the original recording was made. Thus, time intervals on the Bowles tapes are shorter than real time and need to be decompressed by a factor of 1.05 to correct for the difference in the instrument settings. The NRC panel confirmed this warp when it measured the interval from HOLD to YOU on the Bowles CH-1 tape at 171 sec, but when the dictabelt itself is played back at a speed set to match the 60 Hz power hum the same interval is 178 sec, equivalent to a warp factor of 1.045. The time intervals shown in Tables 3 & 4 have been corrected using the 60 Hz power hum.
B) Stylus displacements.- Both original recording instruments recorded and played back with a stylus which etched an acoustical groove or tracked an acoustical groove, and as in the familiar case of a worn phonograph record, the recordings were susceptible to stylus displacements. Displacements could be forward or backward resulting in, respectively, skips or repeats, both of which are readily detectable in the instances when they take place during voice transmissions, but which are not readily detectable if any occurred during silences. Stylus displacement was a particular problem during playback of the audograph disc. As a result of the wear and tear from the transcription process, the disc would not playback without jumping. Thus, the Bowles tape of CH-2 contains multiple repeats as well as missing broadcasts, and in fact, this is the reason that every bootleg version of CH-2 differs from each of the others. Because a skip will reduce playback time, and a repeat will add playback time, one would certainly want to resolve the number of skips and repeats in order to
arrive at an accurate measure of original recording time. As it turns out, the problem is not as serious as it might have been because the audograph mechanism tended to compensate for skips. The stylus on this machine is fixed - not free to jump as in a phonograph player. In the case of a phonograph record, for example, a scratch could induce a needle jump such that a record could play for an hour, constantly repeating the same section of groove. The audograph machine was designed so that such couldn't happen. Unlike a record player, the audograph disc plays from the inside out. Moreover, the turntable is driven by a screw drive which pushes the axis of the turning disc in a direction perpendicular to the fixed stylus during play. As a consequence the tension on the stylus induced by displacements tends to cause the stylus to skip in the opposite direction if and when it next displaces. The stylus is inexorably driven towards the outer rim of the disc by the screw drive as the disc turns. The compensatory skip may not occur immediately but it becomes highly unlikely that there will be an accumulation of significant added time. Of course, with regard to any given time interval there is a possibility that a repeat has added time and that the compensatory skip hasn't yet occurred, and therefore any given time interval is likely to be off by about 3-1/2 sec or so (the rotation time of the disc).
C) Recorder Stoppage.- The recorders on both channels were sound actuated and would stop recording when there was dead air. The actuation switch was programmed to shut off the recorders anytime there was more than about four sec of dead air. However, Jim Bowles states that line noise was often sufficient to keep the recorders running. On CH-1 this was not a problem because the motorcycle motor noise kept the recorder active. On CH-2 however it is clear that the recorder did stop. Those of you familiar with these recordings know that during the time period before the assassination, from about 12:15 to 12:30, by the dispatcher’s time notations, there is only about 5 min of recorded tape time. It is problematic that recorder stoppage cannot be detected directly, nor is there is a direct way to measure the duration of the stoppage when it occurred.
Dispatcher Time Notations
An important aspect of the police radio protocol was that the dispatchers regularly noted the time. Each dispatcher had his own clock. These were analog clocks which were synchronized regularly to keep them within one minute of one another. Typically the dispatcher would append a time notation to one of his regular calls. Thus, there would often be more than one time notation during a minute of time if there were many calls, or there would be no notation during some minutes if no call was made. Because the time notations were not meant to be precise to within seconds the intervals between notations will not be exactly 60 sec. However, over segments of time of many minutes, the time interval between notations will average out to 60 sec if there is no interruption in recording time. That is why the time notations are critical in resolving the discrepancies in the time intervals between the crosstalks. The time notations give us an independent measure of actual time passage between events on the recordings.
Reconciliation of Misalignments
The misalignments between the crosstalks can be resolved if one pays attention to the factors that cause the time warps and the passage of real time denoted by the dispatchers announcements. For example, note the small discrepancy between the two Bellah crosstalks, an interval of about 15 sec on CH-1 but only about 12 sec on CH-2. When one factors in the 5% warp in tape speed the actual discrepancy is 3.4 sec. That amount matches the rotation time of the audograph disc and therefore the discrepancy can be explained as a stylus displacement; a skip of one groove. Because the amount of time involved is so small it isn’t necessary to prove that our explanation is true, but only that the discrepancy is susceptible to explanation by the phenomena that we know are causing warps
in the timelines. Moreover, one cannot add 3.4 sec to the timeline because of the compensating effect of the audograph playback. This skip may well have compensated for an earlier repeat.
The next largest misalignment occurs between the second Bellah crosstalk (You want me...Stemmons) and Dispatcher Henslee’s deliberate simulcast (Attention all emergency vehicles). The 60 Hz adjusted interval is 114 sec on CH-1 but only 91 sec on CH-2. The difference, 23 sec is easily explained by a recorder stoppage on CH-2. But just because there is a simple explanation is not evidence that the explanation is correct. In theory the discrepancy could be due either to lost time on CH-2 or added time on CH-1 or a combination of both. The corroboration for the resolution is found in the dispatcher’s time notations. The time notation sequence is shown in Table 5. Regression analysis is a mathematical test of linear relationships. In this test we are comparing the actual timeline (recorded time) and the corrected timeline by adding the hypothetically missing 23 sec to the actual time lapse shown by the dispatchers time notations.
12:30 0 0 0
12:31 60 95.8 95.8
12:32 120 121.2 121.2
12:34 240 212.8 212.8
12:35a 300 268.9 268.9
12:35b 300 300.3 300.3
12:36a 360 329.9 352.9
All Notations Slope .902 .944
[Minus 12:31] Slope .947 .998
Regression statistics for Regression Statistics for
uncorrected time notations corrected time notations
intercept a = .1876 intercept a = -5.5
slope b = .947 slope b = .998
s.e. of b = .0472 s.e. of b = .0606
Correlation r² = .988 correlation r² = .982
t = 20.15, df = 5, p = 2.797 x 10-6 t = 16.46, df = 5, p = 7.56 x 10-6
F = 483.2, df = 1,6 p = 5.792 x 10-7 F = 325.5, df = 1,5, p = 9.61 x 10-6
mean x = 240, mean y = 227.4 mean x = 240, mean y = 234
In the regression test a slope of 1.0 would be perfect agreement; there would be a one to one agreement between recording time and real time. The actual slope is .902, a fair agreement, but far from perfect. Part of the problem is with the 12:31 time notation. This notation is way off line, coming 36 seconds later than it should. But there is a logical explanation. Sixty seconds after the 12:30 time notation Sheriff Decker was transmitting orders to his men. As soon as Decker stopped transmitting the dispatcher made the 12:31 time notation. Radio protocol included making regular time notations but it did not include interrupting broadcasts for that purpose. If we delete this aberrant notation the agreement between actual and expected increases to .947, essentially a ninety-five percent agreement. If we then add the hypothetically missing 23 sec between the second 12:35 and first 12:36 time notations, we not only increase the agreement, but arrive at a near perfect slope of .998. Thus, the dispatcher’s time notations provide us with independent corroboration that the hypothesis of recorder stoppage to explain the 23 sec discrepancy is correct.
By making this correction we achieve a coherent timeline for CH-2 because there is now good agreement between recorded time and real time. Most importantly, the timeline is achieved with evidence completely independent of the juxtaposition of the putative gunshot sounds on CH-1 and of the broadcasts that establish by context the time of the assassination on CH-2. Using this rational timeline the period from the Fisher crosstalk to the Henslee crosstalk on CH-2 is 354 sec. This brings in to focus the real problem with the DPD recordings because the same time interval on CH-1 (Fisher to Henslee) is only 297 sec.
Aligning the two timelines (CH-1 and CH-2) using the Bellah-1, Bellah-2 and Henslee simulcasts as anchors, the Decker and Fisher broadcasts still fail to align. Moreover, it becomes evident that the reason for the misalignment is that about a full minute of time is missing (57 sec) from CH-1 between the Fisher and Bellah-1 broadcasts. The missing time is too great to be attributed to recording speed, and besides, we have already measured and applied this correction. Similarly, recorder stoppage cannot explain the missing time because the motorcycle broadcast would have kept the recorder active. The one phenomenon that works is a stylus displacement. There are some conspiracy minded folks that envision sinister forces at work. However, the fact is that the dictaphone machine had to be serviced about every half hour or so because of the limited amount of recording space on the dictabelts, and if the person changing the belts (or someone else) accidentally bumped the instrument, this could have induced the displacement. Under this scenario, at approx. 57 sec in real time after Fisher's broadcast, the stylus could have displaced backward landing in a track 31 sec before the Fisher broadcast and from there continued to over-record until the point where the stylus originally displaced, in the process superposing the crosstalked Decker broadcast 10 sec after the Fisher broadcast on this channel. The needle jump hypothesis explains the 88 sec misalignment between the Fisher and Decker broadcasts while simultaneously explaining the 57 sec misalignment between the Fisher and Bellah crosstalks. No other phenomenon does both.
Ultimately, the misalignments can be explained and resolved by the factors that we know to be causing time displacements on the recordings: recorder stoppage, speed warps and stylus displacements, establishing timelines independent of whether or not there are gunshots on the recordings and independent of the juxtaposition of the suspect sounds to the broadcasts concerning the assassination. That being so, the timeline evidence does support that the putative gunshots are synchronous with the time of the shooting by reliance on the crosstalk closest to the assassination, the Fisher broadcasts, and remain synchronous with the time of the shooting using all of the crosstalks once the discrepancies are resolved.
Any evidence is subject to different interpretation and that is the case here. Many of these crosstalk instances are difficult to hear, which is why I have played them for you so you can make up your own minds and believe your own ears. Different listeners have reported different interpretations of the broadcasts, even the more audible CH-2, as witness the various transcripts that have been published. In part because they were captured over a second intervening microphone and in part because the motorcycle motor on CH-1 is contributing unwanted noise, the crosstalks are anything but loud and clear. A digital voiceprint analysis has not been applied to the Fisher broadcasts, as it has to the Decker and Bellah crosstalks, so caution is advisable. Bowles identification of the call on CH-1 as a fragment of Fisher’s call might be mistaken. Moreover, given some of the problems with the aforementioned voice-print analyses, such a study still might not be conclusive. Whether or not the Fisher broadcast crossed over on to CH-1, the time warps are real and this makes any synchronization based on common signals subject to some doubt, regardless of how rational our corrections to the timelines seem. The point I wish to make in conclusion is that the NRC panel’s reliance on a single instance of crosstalk, the Double Decker, does not establish asynchrony between the sounds identified as gunshots and the time of the assassination. As a consequence, the acoustical identification of the assassination gunfire on the Dallas Police recordings has yet to suffer a substantial challenge.
I am indebted to Jim Barger, Gary Mack, and Michael O’Dell for comments, information, and some of the photographs that contributed to this presentation.
1. Barger, J.E., S.P. Robinson, E.C. Schmidt & J.J. Wolff. 1979. Analysis of recorded sounds relating to the assassination of President John F. Kennedy. Bolt, Barenek & Newman, Inc. Cambridge MA. Proceedings of House Select Committee on Assassinations Vol. 8. p. 116.
2. Weiss, M.R. & A. Aschkenasy. 1979. An analysis of recorded sounds relating to the assassination of President John F. Kennedy. Dept. Computer Sciences, Queens College, City University New York. Proceedings House Select Committee on Assassinations Vol. 8. p. 32.
3. U.S. Congress, House of Representatives, 1979. House Report Wo. 95-1828. Select Committee on Assassinations. Final Report, p. 76. U.S. Gov’t Printing Office. Washington D.C.
4. National Research Council. 1982. Report of the Committee on Ballistic Acoustics. Prepared for Dept. of Justice, Washington D.C. Report No. PB83-218461, p. 16.
5. Bowles, J.C. 1993. The Kennedy assassination tapes: a rebuttal to the acoustical evidence theory. Pp. 313-410 in: G. Savage, JFK: First Day Evidence. Shoppe Press, Monroe, LA.