The acoustic evidence in the
Kennedy assassination

Michael O'Dell

In 1978 the House Select Committee on Assassinations (HSCA) concluded that there was probably a conspiracy to kill President Kennedy⁽¹⁾. This conclusion was primarily based on acoustic evidence contained in Dallas Police Department radio recordings. An NRC panel later disputed the HSCA conclusions. In 2001 D. B. Thomas⁽²⁾ published a paper that rehabilitated the original findings. This report demonstrates that the prior reports relied on incorrect timelines, and made unfounded assumptions that when corrected do not support the identification of gunshots on the recording.

 

History

In 1978 the House Select Committee on Assassinations (HSCA) concluded that there was probably a conspiracy involved in the assassination of President Kennedy⁽¹⁾, a conclusion primarily based on the acoustic evidence contained on the Dallas Police Department radio recordings.

On the day of the assassination the Dallas Police Department (DPD) operated two radio channels. Channel I was for normal police radio traffic and channel II was assigned for the use of the presidential motorcade. Each channel was recorded by a different device. Channel I was recorded on a Dictabelt and channel II on a Gray Audograph. Both machines worked by engraving a track into a plastic medium. The Dictabelt used a rotating cylinder and the Audograph used a flat disk, similar to a phonograph record. Both machines were transmission actuated.

An unknown motorcycle tuned to channel I had a defective microphone button that caused it to continuously transmit over a five minute period during which the assassination took place ⁽³⁾. If this motorcycle had been part of the motorcade it might have picked up sounds of the gunshots. If true, those sounds could be used to determine how many shots were fired, their timing, and using echo location methods, where the shots came from.

Working for the HSCA a team (BRSW) from Bolt, Beranek and Newman Inc. (BBN), headed by James E. Barger, studied the recordings. BRSW performed a series of test shots in Dealey Plaza and used recordings of these shots to compare with signals on the DPD recordings. BRSW concluded that channel I contained impulses probably caused by the gunshots, with a 50% probability that one shot came from the grassy knoll in front of the President ⁽⁴⁾. Because of the level of uncertainty in this finding the HSCA asked another team try to raise the confidence level of the results. Weiss and Aschkenasy (WA) used an acoustic modeling method and concluded there was a 95% probability of a shot from the grassy knoll ⁽⁵⁾.

Having reached the conclusion of "probable conspiracy", the HSCA asked the Justice Department to pursue the case. The Justice Department requested that the National Research Council (NRC), part of the National Academy of Sciences, review the acoustic evidence. In 1982 the NRC Committee on Ballistic Acoustics, headed by Norman Ramsey, released their report ⁽⁶⁾ and a version of that report was published in Science ⁽⁷⁾. The NRC report disputed the statistical significance of the HSCA findings, and following a suggestion from Steve Barber, proved that there is an instance of crosstalk on channel I at the time of the alleged shots ⁽⁸⁾. Crosstalk occurred when sounds from one channel were picked up by a microphone tuned to the other channel. This instance of crosstalk ("hold everything secure...", referred to as the Decker crosstalk) occurred during the alleged shooting sequence specified by BRSW and WA. Through spectrographic comparison the NRC confirmed that the faint voice that could be heard on channel I was a fragment of speech from channel II that happened about a minute after the assassination. Since this crosstalk happened at the same time as the supposed shot impulse patterns, the impulse patterns could not be of the assassination gunfire. The NRC report also contained a timing chart of events on both channels and included correction factors for tape speed and other effects ⁽⁹⁾.

Thomas

In March 2001 D. B. Thomas published an article in the peer-reviewed journal of the British Forensic Science Society ⁽²⁾. This article received considerable media attention and revived the debate over the acoustic evidence. Thomas concluded that the NRC made statistical errors of their own and the probability for a grassy knoll shot was actually greater than 96%. He also concluded that by using a different instance of crosstalk to align the channels the shots could be correctly placed in time at the period the assassination actually happened.

When the Thomas paper was published I saw the news reports and was intrigued by the promise that the question of conspiracy in the assassination might be settled scientifically. After reading the article I sought out other informed persons to discuss it with. One of the persons I was fortunate to meet had a multi-generational copy of the DPD recordings and a copy of the original report from the Committee on Ballistic Acoustics. Receiving copies of these from him, and with good audio software (Sonic Foundry's Sound Forge 6), I set out to examine Thomas' findings and the prior work. Later Dr. Ramsey provided me digital CD copies of the recordings from his files. Selections from those CDs are now available on a National Academy of Sciences web site ⁽¹⁰⁾.

The paper by Thomas contains several arguments intended to refute the findings of the NRC committee and which helped him conclude that the HSCA acoustic findings were valid. One of the most important of these arguments is an alternative crosstalk synchronization that allowed Thomas to place the shots at the time of the assassination.

Thomas argued that if a later instance of crosstalk, ("you want me...") were used to synchronize the two channels then the shots did not appear too late. The NRC explained this apparent difference by pointing out that channel II was sound activated and could easily have stopped between the two crosstalk instances. Thomas did not accept this explanation, believing that a regression chart from the BRSW report proved that channel II did not stop during the interval. If the channel II recording had been continuous then one of the crosstalk instances had to have been displaced on the recording by a jumping recording stylus, or some other cause. Because such a displacement appeared to be the only explanation reconciling the timing data, the synchronization of the Decker crosstalk with the "shots" could no longer rule out the possibility that they were the real shots.

Thomas relied on the timing data published in the NRC report ⁽⁹⁾. Early on, I attempted to reproduce the timing of channel II that appears in that report. The effort was unsuccessful for various reasons. Several different copies, with different histories, were compared and each was found to run at a slightly different speed. None of the copies I had matched the times printed in the NRC report, and no adjustment for tape speed alone corrected the problem.  Two of the recordings used by the NRC panel are known as the Bowles version and the FBI version. James Bowles, the Dallas Police Department Communications Officer at the time of the assassination, made tape recordings of both channels from the original media in 1964. In 1981 the FBI made a new tape recording from the original channel II media using a phonograph player as the playback device.

Timing

One of the timing problems with channel II involves the mechanics of the Audograph machine. By comparing the Bowles version of channel II against the FBI version I found a number of forward skips on the Bowles version that weren't accounted for in the NRC report. Herman Chernoff (NRC panel) suggested a mechanical cause ⁽¹¹⁾, and I was then able to acquire some photographs of a machine ⁽¹²⁾.

The NRC panel made a reasonable, but wrong, assumption about how the Audograph machine worked ⁽¹³⁾. The tape recording made from the Audograph by Bowles contained numerous repeated sections. On playback the needle would sometimes get stuck in a groove and repeat that groove, sometimes twice. Table C-1 of the NRC report subtracted the time that these repeats took from the elapsed time. This seems to make sense, but due to the mechanical construction of the Audograph machine it was the wrong thing to do.

The needle assembly on the Audograph doesn't move at all. The disc is mounted on a spindle and moves horizontally under the needle as recording progresses. The horizontal motion of the spindle is a result of being mechanically geared to a worm gear, such that, as the spindle rotates, the center of the disc moves away from the needle assembly. There is no freedom of motion in this mechanical system, except for the tip of the needle. The repeats happen because the needle tip has some flexibility and may get stuck for an additional rotation or two, but the horizontal movement of the disc never stops and the needle must eventually catch up to where it should be.  When the needle does catch up it will skip over sections of the recording. In the long run any repeats will be roughly matched by forward skips where the needle jumps over grooves.  (See Audograph photos.)

It was a simple matter to notice the repeats but forward skips were not noticed. It would have been more accurate if the repeats were not subtracted at all. When the NRC subtracted the time for repeats they reported the elapsed time as shorter than what it really was.  There are certain other uncertainties in Table C-1, involving the speed correction factor, the "gap", and the choice of event markers. All of these factors combined made reproducing the prior time studies very difficult. Because of these uncertainties, and with the problem of the overlooked skips, I looked for a reliable way to accurately time events on channel II.

Another copy of channel II was made by the FBI on a high-quality phonograph, instead of using the Audograph machine. The tape of channel II made by the FBI for the NRC panel appears to be a complete recording without skips or repeats in the period in question. The NRC report states, "[the original Gray Audograph was] transcribed, as described in Appendix C, onto tape, with care taken to minimize the 60 Hz hum that was added to the signal and to ensure that no skips or repeats were introduced in the tape recording of either channel. No break interrupted the Channel II recordings as was the case for the Bowles tapes." ⁽¹⁵⁾

Careful examination of the recording during the six minutes after the shooting confirms that there are no obvious repeats in it. A side by side comparison revealed the forward skips in the Bowles version that the NRC didn't account for, and also would have revealed skips in the FBI version had they been there, and had such skips not been precisely mirrored on the Bowles tape.

The FBI tape is a perfect candidate for performing timing studies, except for one problem. As the NRC report says, "The Gray Audograph disk (Channel II) could not be played on an original Gray playback unit without introducing skips and repeats. It was possible to play it successfully without either of these artifacts being introduced by using a phonograph turntable and phonograph arm, cartridge, and stylus. However, phonograph turntables operate at a constant rpm, whereas the Gray equipment's rpm reduced as recording progressed. Moreover, the Gray Audograph records from the inside out, whereas normal records begin at the outside. Thus, when the tapes are played back, there is a speed distortion that causes material at the beginning of the tape (the inside of the record) to be slowed down (time intervals between events are longer and the frequencies are lower than those originally recorded) and material at the end of the tape (end of the record) to be speeded up relative to true speed." ⁽¹⁴⁾

Timing studies on this tape first require a method of correcting for the speed distortion. One such method is described in the NRC report in Appendix C ⁽¹⁶⁾. They used the 60Hz hum introduced into the original recording by the Audograph and measured it at multiple points along the tape. When these points were plotted and a linear least-squares fit line drawn the amount of speed distortion at any point on the recording could be determined. The NRC only used this method to time a few events on channel II.

Because a sonogram does not provide a highly precise measurement I chose a direct method of measuring wavelength. The audio software allows time markers to be placed on the waveform. The marker resolution is equal to the sample rate, 44,100 samples per second. A band-pass filter between 40Hz and 80Hz was applied and the hum measured where it was prominent. Cycles were measured where the waveform intersects the central zero line. Time markers were placed at the proper points to measure multiple cycles. The total duration of the measured cycles divided by the number of cycles is the wavelength at the midpoint of the markers.

From a series of such measurements along the length of the recording, I plotted the power hum on a graph, and drew a least squares fit line (Fig. 1). The measured frequency is plotted as a ratio to the correct 60 Hz. This results in a correction factor for a given point equal to the amount the apparent playback speed is faster or slower than real time.

Fig 2.  Ratios of Audograph power hum frequency to 60hz with a least-squares fit.

The least squares line is described by the linear equation: R = ST + Y, where R is the correction ratio, S is the slope of the line, T is the measured elapsed tape time and Y is the Y-axis intercept.  The slope of the 60Hz ratio line is 0.000406 and the Y-axis intercept is 0.601909. To calculate the correct elapsed time for any given segment of the FBI tape we use the following equation:

C = (0.000406 ((T₁ + T₂)/2) + 0.601909) (T₂ - T₁) [Eq. 1]

C is the corrected length of a segment, T₁ is the beginning time of the segment and T₂ is the ending time. The correction is done referencing the midpoint of the segment ((T₁ + T₂)/2) and the resulting ratio is multiplied by the segment length (T₂ - T₁) to arrive at the corrected segment length. To determine the correct time for a point on the tape the same equation is used with T₁ set at zero.

Using this equation it was possible to produce correct time measurements for events on channel II. Some of these are listed in Table 1.

Channel I was recorded at a fixed speed and does not suffer from the same time correction problem as channel II, and it also does not contain skips and repeats that need correcting. It is only necessary to establish the correct playback speed, and then align events on the two channels.

On the CD is a track (Disc 1, Track 5) made by the FBI from the original Dictabelt ⁽¹⁷⁾. The speed of this tape was set by the FBI using the 60 Hz power hum on the recording. This track was used as a speed reference, and I set the playback speed of Disc 2 Track 6 to match it. Disc 2 Track 6 was used because Disc 1 Track 5 is a smaller segment of the Dictabelt recording and did not contain some events I wanted to include.

Synchronization

Although the NRC's timeline ⁽⁹⁾ is inaccurate, mostly due to the misunderstanding about the Audograph mechanics and missing the skips, the corrected timeline still supports their conclusion ⁽¹⁸⁾ that the impulses occurred after the shooting.

The NRC panel demonstrated that the phrase beginning, "hold everything secure..." was cross-talked onto channel I at the same time the suspected third and fourth shots occur ⁽¹⁹⁾. The first suspected shot occurs about 6.5 seconds before that.⁽²⁰⁾

Table 2 contains the relative times for these events on both channels, using the "hold everything secure..." phrase as the synchronization point. Fig. 2 illustrates these figures on parallel timelines. Since "GOING to the hospital" occurs 67.5 seconds before "HOLD" on channel II, and the first suspected shot occurs about 6.5 seconds before "HOLD" on channel I, that shot must be at least 61 seconds after the shooting.

Fig. 2.  Channel synchronization timeline using Decker crosstalk.

Thomas has said that we should be able to use any instance of crosstalk to align the channels ⁽²¹⁾, implying that we can use any instance to locate the impulses. This is not correct. We should naturally want to use a tie-point nearest the event we are interested in. The Audograph was transmission activated and would stop after a period of silence. Since any stoppage would compress the apparent time between events on channel II, the further away from the impulses we select a synchronization point the more compression there is likely to be from the machine stoppage. The effect on the synchronization from using points later on the timeline is to artificially move the impulses on channel I to earlier times relative to channel II.

Using later synchronization points would only be justified if we knew that channel II did not stop in the intervening period. In his paper Thomas makes an argument for this ⁽²²⁾. To prove that the Audograph did not stop he relied on a chart ⁽²³⁾ from the BRSW report. The chart plotted the dispatcher's time annotations on channel II against elapsed time on the tape, and drew a least-squares fit line. Since the line supposedly had a slope of almost one ⁽²⁴⁾ Thomas believed that the Audograph did not stop for any appreciable amount of time in the six minutes after the alleged shots.

It must be emphasized that even if a correct plot did have a slope of nearly 1.0, it could not prove that no time had been lost due to stoppage. The standard deviation of the slope estimate (0.05) is high ⁽²⁵⁾, allowing a large window of possible values, and Bowles' description ⁽²⁶⁾ of the procedures the dispatchers followed undermines any confidence in the accuracy of the annotations.

Fig. 3.  Least-square regressions of channel II dispatcher time annotations.

Using the corrected timing data, I plotted the dispatcher's time annotations (Fig. 3), just as had been done by BRSW. The slope of the line for the six minutes beginning with the 12:30 annotation is 0.918, not nearly 1.0 as reported by BRSW, and relied on by Thomas. A slope of 0.918 would be consistent with the machine having stopped for about 30 seconds in those six minutes, even if the dispatcher annotations were reliable time markers.

The crosstalk alignments with channel I are also evidence that the channel II machine stopped. Beginning with the "hold everything..." instance, we can align the next three instances. If both machines ran continuously there should be equal time intervals between the same events on both channels. If channel II stopped then there should be less elapsed time in the interval in which it stopped.

Between "hold everything..." and "check all these..." there is 28 seconds more time on channel I than on channel II, indicating channel II stopped for that long. Between "check all these..." and "you want me..." there is a difference of 2.6 seconds, and between "you want me..." and "attention..." there is a difference of 24.8 seconds.

If, despite the evidence, one were to argue that channel II did not stop and the crosstalk alignments were caused by a needle skipping or some other cause, it would be necessary to postulate not only one such skip, but one between every crosstalk instance. A much more reasonable explanation is that the machine did indeed stop, as it was designed to do.

Because the timing evidence indicates that the machine did stop there is no justification for using later synchronization points to align the impulses. For the sake of completeness I will address the alternative synchronization argument offered by Thomas in his paper.

Thomas proposed an alternative synchronization based on the "you want me..." instance of crosstalk ⁽²²⁾. He believed that this synchronization was in conflict with the "hold" synchronization and that if this alternative were used the putative shots would be placed before Curry's instruction to "Go to the hospital."

The purported conflict arises because he cited previous studies as reporting that "you want me..." occurs 179 seconds after the shots on channel I and 180 seconds after "Go to the hospital" on channel II. If this synchronization were valid it would place the shots near the time of Curry's instruction instead of a minute after it. Thomas acknowledged that this could be accounted for in two ways ⁽²²⁾. Either the "hold everything secure..." crosstalk was misplaced on channel I, or the channel II recorder paused between the two events.

On the newly corrected timeline there are 181 seconds between the first shot and "YOU" on channel I and 210 seconds between "going to the hospital..." and "YOU" on channel II. The alternative "you want me..." synchronization still places the shots at least 29 seconds after they had to have happened.

Echo Matching

Although the timing evidence does not favor the identification of gunshots on the recording, some may think that the significance of the echo-location match is so great that invoking a speculative and unlikely set of events (i.e. dubbing, repeated needle jumping) is justified to explain away the timing.  It must be remembered that simply finding a match, in and of itself, is not meaningful. WA were free to move the microphone and shooter locations, and make other adjustments within certain parameters, until they found a match. The question is, how meaningful is their match?

WA wrote, "the odds are less than 1 in 20 that the impulses and echoes were not caused by a gunshot from the grassy knoll" ⁽²⁷⁾. Even if there were no dispute over their math or statistical methods that statement was an error. What they had supposedly found was a 5% likelihood of the match being found in random noise ⁽²⁷⁾. This tells us that the match probably detected something, but it does not tell us it was a gunshot. WA treated the gunshot hypothesis as the sole alternative to random noise, perhaps because they presumed the impulses really were echoes, and did not consider alternatives to a gunshot.

If some other real-world cause generated a pattern similar to a gunshot then finding a match would be more likely than finding it in a random distribution. As Thomas' article states, "If the source of the impulse patterns was some non-white noise the distribution of peaks might be non-Poisson and therefore have a different probability. However, the NRC panel offered no evidence nor suggested any non-white noise phenomena that might account for the impulse patterns". ⁽²⁸⁾

There is an alternative cause that makes patterns similar to the gunshot pattern WA looked for. What neither BRSW, WA, Thomas or even the NRC (explicitly) acknowledged is that the cross-talked Decker speech itself is a non-random sequence of impulses that occurs at the precise position of the alleged grassy knoll shot on the recording. ⁽⁸⁾

This speech signal must have an appreciable effect on the impulse pattern segment, otherwise we could not measure it or hear it. Peaks from speech patterns are typically clustered in distinct syllables separated by sections of relative quiet. See Fig. 4, the waveform of Decker's word "secure" from the channel II recording. The impulse pattern that WA matched also exhibited such grouping.⁽²⁷⁾ In fact, WA described the impulse pattern as being grouped into two sections divided by a 190ms window of relative quiet. Note that marked on Fig. 3 is a period of 190ms dividing the cluster of impulses in the second syllable of "secure" and the first syllable of "until".

Figure 5 from the BRSW Report ⁽²⁹⁾ is a spectrogram of the channel I section in which the impulse patterns were located. The accompanying text notes, "Just after 144 sec, a single loud click occurs, followed by a region of very faint speech..." ⁽³⁰⁾ It is interesting that BRSW also noted the speech here but never seemed to recognize its later significance to WA's echo matching. More importantly for this study, the spectrogram provides a landmark by which we can locate the alleged grassy knoll impulse pattern on the recording.

The "single loud click" near 144 sec. is unmistakable and marks the time when the "hold everything secure" phrase becomes audible. By using the scale on the Fig. 5 spectrogram and measuring the distance between the 144 mark and the click I determined that the click occurs at 144.05. BRSW's suspected grassy knoll shot was found on the same recording at 145.15 ⁽³¹⁾. WA adjusted the beginning time of the impulse pattern they studied by about 0.2 seconds, putting it at 144.95 ⁽³²⁾. The beginning of WA's pattern is therefore about 0.9 seconds after the loud click.  The audio software allows me to place markers and listen to sections between markers. By putting a marker at 0.9 seconds after the click, and listening to regions before and after the marker it is clear that the marker coincides with the "K" sound in the second syllable of the word "secure".

Fig. 4.  Decker's speech from channel 2, which occurs at the instant of the alleged grassy knoll shot on channel 1.

The 190 ms. window of relative quiet between two groups of larger impulses that WA described occurs where there is a speech pattern with the same parameters. The similarity underscores the significance of WA's failure to consider other possible causes. It is persuasive evidence that WA simply identified a match to the impulse pattern created by the combination of Decker saying "secure" and motorcycle noise, which was further transformed by frequency filtering. This alternative to the gunshot hypothesis is a non-white noise, non-Poisson explanation for the distribution of peaks in the impulse pattern. WA's statistical procedure does not provide any reason why we should choose the gunshot hypothesis over this speech hypothesis, nor over other possible hypotheses.

Fig. 5.  A comparison of three waveforms.

Speech sounds commonly make waveforms that look similar to the gunshot echo patterns used by BRSW and WA.  See Fig. 5, a comparison of three waveforms.  The first two (W1 and W2) are taken from the WA report.  W1 is one of the test recordings of a grassy knoll shot.  W2 is the dictabelt match to W1.  W3 is my voice speaking a "K" sound.  The similarities of the clear voice pattern in W3 to the dictabelt pattern are remarkable.  W2 contains two distinct groupings of regular peaks, a large peak at the beginning and even a feature that could easily be confused with a shockwave or "N-wave".  These are all features attributed to the dictabelt pattern that supposedly help identify it as a grassy knoll shot.

"K" sounds are not the only speech sounds that make patterns like these.  Other sounds, like "D", "T", "S", in the right combinations, will make similar patterns.  They are common enough features of speech waveforms that in a test like BRSW's where they were scanning a segment looking for possible matches to numerous test shots, if that segment contained enough speech it would be surprising if they did not find some matches.  As we know, this segment of the dictabelt did contain speech, and not just "hold everything secure".  The crosstalk instance that begins with "hold everything secure" goes on to include nearly the entire phrase, "hold everything secure until a homicide or other investigators can get there".  BRSW's matches at 145.15 (the "knoll shot") and 145.61 coincide with the words "secure until".

There is an earlier and fainter instance of crosstalk just a few seconds before.  It is so faint that it is very difficult to distinguish what is said, but a human voice is barely discernable.  We know from channel II that Decker was speaking for 10 seconds before he uttered the words "hold everything secure" so it's reasonable to think the earlier instance is an earlier part of Decker's speech.  BRSW's matches at 139.27 and 140.32 coincide with this other faint crosstalk instance.

The existence of speech at the locations where BRSW/WA found matches seriously undermines the significance of those matches.  BRSW's procedure detected many false positives and produced matches to the same impulse patterns from both shooting locations.  Given the speech patterns that exist and the tendency of the procedure to produce false positives, the BRSW findings can't be given much weight.

Although WA's procedure was more rigorous it produced no data that would distinguish between finding a gunshot pattern or a voice pattern.  One might suggest there is a gunshot pattern and the voice pattern at the same time, but there are problems with this idea.  The known voice pattern alone is sufficient to explain the data so there is no solid foundation for such a claim.  It also ignores the way waveforms combine.  The voice pattern and the gunshot pattern are enough alike that there is no way to distinguish which peaks might have come from the voice and which from a hypothetical gunshot, and even if it were there the voice pattern would have modified the gunshot pattern from its original form.  Even if both existed at the same time WA's match would be to the combined signal, not to the gunshot alone.  Finally, the idea that both exist requires us to accept the highly unlikely proposition that at the same instant where WA found a grassy knoll shot, just by chance, there is another unrelated signal with a similar appearance and description.  There is no reasonable basis to believe that WA's pattern is of a gunshot.

Conclusions

1. The timeline relied on by the NRC report and by Thomas is inaccurate.

2. Both the "hold everything" and the "you want me" crosstalk alignments demonstrate that the suspect impulses happen too late to be the assassination gunshots.

3. There is no evidence that the Audograph machine that recorded channel II ran continuously in the first few minutes after the shooting, and evidence indicates that it did stop. Because the Audograph stopped, later instances of crosstalk cannot be used to align the suspect impulses on channel I.

4. There is no statistical significance of 95% or higher for a shot from the grassy knoll. There is persuasive evidence that BRSW/WA simply found a match to the speech pattern that exists at the same location on the recording.

Acknowledgements

I am grateful to Herman Chernoff, Richard Garwin, Norman Ramsey and Paul Horowitz for their explanations, advice and assistance; Steve Barber for his help and his ear; Don Thomas for his encouragement and openness. Paul Hoch was very helpful with the manuscript.  James Barger, Charles Rader, Mitch Todd and Anthony Marsh took time to offer ideas, information or answer questions.  Thanks to John McAdams for giving this work a home on the internet.

Appendix

Figures A-1 thru A-3.  Photographs of an Audograph machine.

Table A-1. Background Power Hum Measurements of DPD Channel II 

Table A-2. Events on DPD Channel II 

Table A-3. Events on DPD Channel I 

References and Notes

1. US Congress, House of Representatives, House Report Wo. 95-1828, Select Committee on Assassinations. US Government Printing Office, Washington DC. 1979 (hereinafter referenced as HSCA). Final Report, pp. 76, 94. [http://www.history-matters.com/archive/jfk/hsca/report/contents.htm]

2. Thomas DB. Echo Correlation Analysis and the Acoustic Evidence in the Kennedy Assassination Revisited. Science & Justice 2001; 41: 21-32. [http://www.forensic-science-society.org.uk/Thomas.pdf]

3. Barger, JE, Robinson, SP, Schmidt EC & Wolf, JJ. Analysis of Recorded Sounds Relating to the Assassination of President John F. Kennedy. Bolt, Baranek & Newman, Inc., 1979, (hereinafter referenced as BRSW Report). HSCA Proceedings Vol. 8, p. 70. [http://www.history-matters.com/archive/jfk/hsca/reportvols/vol8/html/HSCA_Vol8_0019a.htm]

4. BRSW Report. HSCA Proceedings Vol. 8, p. 107.

5. Weiss, MR & Aschkenasy, A. An Analysis of Recorded Sounds Relating to the Assassination of President John F. Kennedy. Dept. Computer Sciences, Queens College, City University New York, 1979 (hereinafter referenced as WA Report). HSCA Proceedings Vol. 8, p. 32. [http://www.history-matters.com/archive/jfk/hsca/reportvols/vol8/contents.htm]

6. National Research Council, Report of the Committee on Ballistic Acoustics. Prepared for Department of Justice, Washington DC. Report No. PB83-218461, 1982 (hereinafter referenced as NRC Report). [http://www.nap.edu/books/NI000372/html/]

7. Committee on Ballistic Acoustics, National Research Council. Reexamination of Acoustic Evidence in the Kennedy Assassination. Science, Oct. 8, 1982.

8. NRC Report. p. 25. [http://books.nap.edu/books/NI000372/html/25.html]

9. NRC Report. Table C-1. p. 62. [http://books.nap.edu/books/NI000372/html/62.html#pagetop]

10. The National Academies Press, Audio files of JFK Assassination Recordings,  http://stills.nap.edu/html/JFK_audio

11. Herman Chernoff, E-mail correspondence. 10/4/2001.

12. Gray Audograph, Gray Manufacturing Company, Hartford, CN. Model BIC-4 Master. Serial number M160134.

13. NRC Report. p. 61. [http://books.nap.edu/books/NI000372/html/61.html]

14. NRC Report. p. 67. [http://books.nap.edu/books/NI000372/html/67.html]

15. NRC Report. p. 28. [http://books.nap.edu/books/NI000372/html/28.html]

16. NRC Report. p. 68. [http://books.nap.edu/books/NI000372/html/68.html]

17. The National Academies Press, Audio files of JFK Assassination Recordings,  http://stills.nap.edu/html/JFK_audio/tr5_128.mp3

18. NRC Report. p. 34. [http://books.nap.edu/books/NI000372/html/34.html]

19. NRC Report. p. 23. [http://books.nap.edu/books/NI000372/html/23.html]

20. BRSW Report. HSCA Proceedings Vol. 8, p. 101.

21. Thomas DB. Hear No Evil. The Acoustical Evidence in the Kennedy Assassination. http://pages.prodigy.net/whiskey99/hearnoevil.htm

22. Thomas article. p. 29.

23. BRSW Report. HSCA Proceedings Vol. 8, p. 72. Fig. 8.

24. BRSW Report. HSCA Proceedings Vol. 8, p. 70.

25. Herman Chernoff, E-mail correspondence. Mar 7, 2002.

26. Bowles, JC. The Kennedy Assassination Tapes, A Rebuttal to the Acoustical Evidence Theory, 1979. [http://www.jfk-online.com/bowles1.html#ref]

27. WA Report. HSCA Proceedings Vol. 8, p. 32. [http://www.history-matters.com/archive/jfk/hsca/reportvols/vol8/html/HSCA_Vol8_0018b.htm]

28. Thomas article. p. 26.

29. BRSW Report. HSCA Proceedings Vol. 8, p. 66. Fig. 5. [http://www.history-matters.com/archive/jfk/hsca/reportvols/vol8/html/HSCA_Vol8_0035b.htm]

30. BRSW Report. HSCA Proceedings Vol. 8, p. 65. [http://www.history-matters.com/archive/jfk/hsca/reportvols/vol8/html/HSCA_Vol8_0035a.htm]

31. BRSW Report. HSCA Proceedings Vol. 8, p. 101. Table II. [http://www.history-matters.com/archive/jfk/hsca/reportvols/vol8/html/HSCA_Vol8_0053a.htm]

32. BRSW Report. HSCA Proceedings Vol. 8, p. 50. [http://www.history-matters.com/archive/jfk/hsca/reportvols/vol8/html/HSCA_Vol8_0027b.htm]