The Probably Bogus Correlation between the Timing of the Shots and the BBN Position Estimates
Or, how the BBN attempted one of the Labors of Hercules, on a limited time budget.
First, off, when I say “Bogus”, a am not implying any lying or misrepresentation on BBN Part. Indeed, it seems to be a curious fact that BBN never brought up the subject of this “Correlation” of the Data. Instead, it has been brought up by others.
Now, there are many reasons to doubt the 1978 HSCA BBN Acoustic Study, and a few of them are shown below:
• The “Gun Fire” on the recording does not sound like gunfire, or any kind of a loud noise. [ Excuse: Recording system not designed to record gunfire. ]
• Loud Crowd noise not heard on the recording, as expected of a motorcycle in the motorcade. [ Excuse: Recording system not designed to hear distant voices, although it does record distant voice transmissions from nearby radios using Channel 2. ]
Both of these points could have easily been dealt with by the acoustic experts in their gunfire tests of August 20, 1978 in Dealey Plaza, using the still existing Dictabelt recording system still in use by the Dallas Police Department, but they failed to do so.
• There are not several minutes of sirens blaring, as expected of a motorcycle escorting the President to the hospital, but silence, followed by a gradual buildup of volume of sirens, and then their gradual fading away, over a period of about 36 seconds. As if the motorcycle was at the Trade Mart Center the whole time. [ Excuse: I haven’t heard a good excuse for this one ]
It is this point that, perhaps, deserves the most emphasis. • No motorcycle seen within a circle with a certain circle with a radius of 9 feet, as predicted by the BBN Acoustic Experts in the Altgens photograph. Yes, he could be somewhere within that circle if it is given a radius of 18 feet, as Dr. Thomas did in his drawings. But the motorcycle needs to be within 9 feet, not 18 feet, to match the 1978 prediction made by the BBN Acoustic experts.
• Hughes film, though it stops 1.5 seconds too soon, still makes it clear it was impossible for Officer McLain, or any other motorcycle, to reach the first circle, with a speed of 11 mph, by z175.
• The lack of a systematic pattern with the measured location of the target locations found by the acoustic tests, which does not seem to steadily progress up Elm Street, but seems to show most, or perhaps all of the shots were aimed at the same spot.
• It appears the so called “shots” occurred one minute too late, because of the phrase “Hold everything secure” was recorded at about the same time, evidently from a nearby motorcycle receiving transmissions from Channel 2.
But it does seem to have one strong point to be made for it. The Acoustic matches correspond to a systematic pattern, of a motorcycle moving steadily at 11 mph. This is subject this post will deal with.
In 1978, BBN Inc. was commissioned by the HSCA to do a study of the Dictabelt recording. It was known that a motorcycle had a stuck key. Somehow, the “Transmit” button got into the “On” position. When working properly, a patrolman presses a button, speaks into the microphone and gives his message. While holding this button, no one else can transmit a message. When done, he releases the button. From time to time, for some reason, the button gets stuck on the “On” position. This is known to have happened during a 5.5-minute period covering the time of the assassination.
The question is, was this motorcycle in Dealey Plaza? And if so, did it record the sounds of the gunfire? At first glance, it would appear no, because the recording has no sounds of gunshots, or bangs, or anything that remotely sounds like a gunshot. But the BBN was given this recording to check this out, just to be certain.
So, what did the Acoustic experts do? They closed down Dealey Plaza on August 20, 1978 to run some tests.
They set up an array of 36 microphones along Houston and Elm Street, at the locations where they guessed a motorcycle might be.
They set up 4 different “Target”, targets that would be safe for expert shooters to aim at and hit.
The idea, that each shot, depending on:
• The location of the rifle.
• The location of the ‘Target’.
• The location of the microphone.
would have a unique waveform, a unique “fingerprint”. By comparing one of these waveforms with one of the 6 waveforms from the Dictabelt recording, they could, hopefully, find a close enough match.
If no matches were found, that would be yet another indicator that the motorcycle was not in Dealey Plaza, along with the lack of the sound of gunshots and crowd noise, etc.
But, if matches were found, and the matches were significantly good, that would be an indicator that the motorcycle was in Dealey Plaza after all.
There are Dictabelt Impulses to be found on separate parts of the 5.5-minute section of the Dictabelt of interest. The ones that BBN choose to focus on were 6 Impulses that occur within a 10.1 second section.
The BRSW team then compared, manually, each of 432 test shot wave forms with all the parts of the 5.5-minute record that could reasonably have included assassination gunshot sounds. They used a binary correlation metric, with a ±6-millisecond window, applied to strip chart recordings of the relevant waveforms. For each suspected assassination shot and each test shot, the strip chart recordings were aligned for a best match and a score was obtained by calculating the correlation coefficient. The correlation coefficient was defined as the number of impulses (large peaks) in the wave form of the suspected shot that came within 6 milliseconds of an impulse in the test shot, divided by the geometric mean of the numbers of impulses available in the suspected shot and in the test shot. This correlation procedure does not make use of all the information available (impulses that barely resemble each other affect the score as much as impulses that match each other well), but it permits relatively easy computation of similarity. Only candidate shots that gave a binary correlation greater than 0.6 were studied further. Fifteen pairs, involving only six sets of impulses on the channel I recording, survived this screening.
In the shooting tests of 1978, they recorded 69 test rifle shots and 9 test handgun shots. Due to time constraints, they selected 12 of these test shots to compare with the Dictabelt recording. This means that, with Test 12 shots, each shot recorded on 36 different microphones, they would need to print out 12 times 36 or 432 strip charts, which would take some time. And print out 6 more strip charts for the 6 “shots” recorded on the Dictabelt. So that is the first thing the BBN would need to.
Then, for a through systematic search, they would have to compare each of the 432 Test strip charts with each of the 6 Dictaphone strip charts. For a total of 2,592 comparisons of pairs of strip charts. And to do this with 4 people (there are 4 names listed on the final report) all within 15 days. I think I see a problem here.
For each comparison they would want to make, they would:
• Manually align a strip chart from one of the “Test” recordings, with the strip chart from one of the 6 “Dictabelt” Impulses.
• Count the number of large peaks in the Test waveform (within 6 milliseconds) and divide by the geometric mean of the numbers of impulses available in the “Test” shot and the “Dictabelt” shot.
I’m not certain what all this all means, but it sounds like making some judgements about what is the best alinement, counting the impulses in the “Test” shot and doing some calculations. It sounds like each comparison might take 5 minutes, with all the measurements and calculations.
During this work, care would have to be taken to keep the strip charts were organized so they don’t get out of order or put in the wrong box. This is tedious work. It would be prudent to have each comparison done twice, so nothing is missed.
The firing tests were done on August 20, 1978. On August 30, 1978, Dr. Barger called the HSCA to report that the preliminary matches had been done and 15 matches had been found. So, BBN had at most 10 days to make these comparisons.
How long to compare each of the 12 “Test” Shots with each of the 6 “Dictabelt” Impulses? With 2,596 pairs of strip charts to compare and compute, and with 10 minutes per comparison, it comes out to 432 man-hours. Working seven days a week, 8 hours a day, with 4 men, there are only 320 man-hours available.
Working more than 8 hours is not advisable, since tired workers will tend to miss the occasional match. Frequent breaks would be well advised. It would also be advisable to do each comparison twice, by different people, at different times, so nothing, hopefully, is missed. Also, there is the initial time to print out the Test 432 strip charts, and more time at the end to organize and print out the final charts that were presented to the HSCA.
I would suggest that such a systematic search was not done, as it was impracticable, given the time restraints. So here is my scenario.
They selected one strip from one of the Dictabelt “shots”, I would suggest the one at 139.27, the third possible shot, later determined to be at z196 by BBN, and z205 by Dr. Thomas. They didn’t know which motorcycle might have had the stuck microphone, so they initially assumed it might have been one of the four right behind the President. After all, that group contained 4 of the 6 motorcycles that were in Dealey Plaza at the time of the first shot. They started the search somewhere forward, in the third group and worked their way back. It took a lot of searching until they found one of the somewhat “strong” correlation of 0.8 and considered it a good match. But a lot of time had been used by and they were running out of their 15 days and at this rate could clearly not check out all 6 “shots”. So, what to do.
They obvious thing to do is to search for the next “shot”, let’s say at 137.70 which occurred 1.57 second earlier. But where to search? Maybe their best bet is to assume a constant 11 mph speed of the motorcycle. Afterall, that was the estimated average speed of the limousine during the shooting. If the motorcycle maintained the same speed, that means the shot at 137.70 should be found somewhere Microphone 2 (5). And they found a match. And now they used the same procedure to check out all 6 “shots” and managed to complete their task in the 15 days.
There are signs that this was done.
1. The never checked out the second cluster of impulse patterns found about a minute after the first. True, they didn’t cover 5 seconds, but did anyone know for certain how many seconds the shooting lasted? The fact they did not run any test on those impulse patterns in the second cluster is a sign that there was some serious time pressure.
2. All the shots for a certain time, at 137.70, at 145.15, and at 145.61, where all on the same stretch of road. All except one. The one for 139.27, which ranged from 2(6) to 2(10) to 3(5). Exactly as to been expected as if they had to do a wide search for the first shot, they checked out, and then a more focused search for each of the other 5 possible shots.
3. The biggest reason though, is that if these tests were done much quicker than I would guess (and it is a guess, I have no idea how many seconds or minutes it takes to compare two audio strip charts), or they had many times more than 4 people to work on these comparisons, and actually did a through systematic search of the entire data set of 432 strip charts, then this would have been a powerful argument for their validity. The odds of them coming in the proper order alone would by 1 in 5 Factorial or 1 in 120. And to match the same systematic pattern, of a constant 11 mph, would be astounding. And yet the BBN never hammered this point home.
And on this last point, why would the BBN be presenting these lukewarm estimates, that there is a 50-50 chance that these results indicate four shots, or even a 95% chance that these results indicate four shots, when the pattern of where the “motorcycle” was detected fit such a strong pattern. If there is one thing these Acoustic experts were not, they were not stupid. They would have thought of and used this argument and come up with much higher probability estimates. Unless they knew there was some flaw in it. That there was no through systematic search. And that the resulting data matched this systematic pattern possibly because those were the only places they searched for matches, because of time constraints.
At the very least, particularly with all the other problems the Acoustic Data has, judgment should be withheld until the members of the BBN who did this study can be consulted and confirm or refute the notion that there was enough time for a through systematic study of the entire 432 data set for all 6 “shots”.
Note: September 26, 2020: Upon looking into the BBN material more thoroughly, I have adjusted my numbers, but believe the argument still holds.
