HSCA 1978 Acoustic Study by BBN – Figure 367

[Joe Elliott]

The Multiple Matches Problem


This post is just focused on one question. If a scientific procedure finds multiple matches, when it should only find one match, is that a good thing?


For instance, let’s say we have an existing DNA testing procedure, Procedure A, that collects DNA from a crime scene, and compares it to a database of DNA of suspected criminals and looks for a match. And this procedure has proven good. Now, there is a new proposed Procedure B, that would be half as expensive. So, to test it, we take the DNA from known crime samples, each of which only contained DNA from one man (as proven by Procedure A) and try Procedure B on it. Procedure B may report:

Scenario 1: Only one match found, of a Mr. Jones.

Scenario 2: Two matches found, of a Mr. Jones and a Mr. Smith.

With this information alone, which scenario would bolster Procedure B best?

Clearly, it is scenario 1. Of course, to be fully bolster, we would have to check to see if Mr. Jones was the same match found when Procedure A was used.

But if it turned out that scenario 2 is what played out, that is bad for Procedure B. One of those results has to be false. We must have at least one false positive. And that is bad.

And if we had not 2 matches but 10 matches, that would be very bad. It would mean, if this was the typical result over many test samples, at least 9 in 10 of all matches are false positive. Maybe 10 of 10. Switching to Procedure B would clearly be unacceptable.

Clearly getting multiple matches is bad, because we have discovered that the proposed procedure generates false positives. That is never good.


We have essentially the same problem with the BBN tests of 1978. Looking at BBN’s Exhibit F-367, if we use the standard that 0.6 is good enough to be considered a match, then the results for the sound impulse at 145.15, we get a match for the following shots:

Test	Beginning Time of	Zap.	Zap.	Microphone Array	Rifle	Target	Correlation	Strong	Fluke
ID	First impulse on	Frame	Frame	and	Location	Location	Coefficient**
	Tape Segments (sec)	BBN	Thomas	(Channel Numbers)
L	145.15	304	313	3 ( 4 )	KNOLL	3	0.8	Strong
M	145.15	304	313	3 ( 7 )	TSBD*	4	0.7		Fluke
N	145.15	304	313	3 ( 8 )	TSBD	2	0.7		Fluke

So, we learn from this, that at 145.15, timed to the nearest one hundredth of a second, there was a shot from the Grassy Knoll and a shot from the TSBD. Actually, it appears there were three shots, two shots from the TSBD aimed at Target 2 and Target 4 and one shot from the Grassy Knoll fired at Target 3.

This is bad. We have at least two false positives. It means that there is at least a 2 in 3 chance that an acoustic match is a false positive.



Questions for Mr. Griffith, or anyone else:

Question 1:

Why would find a match for a shot at 145.15 for:

A shot from the Grassy Knoll at Target 3, recorded near 3 ( 4 )
A shot from the TSBD at Target 4, recorded near 3 ( 7 )
A shot from the TSBD at Target 2, recorded near 3 (  8 )

Be considered a superior result to only finding one match:

A shot from the Grassy Knoll at Target 3, recorded near 3 ( 4 )


Question 2:

Is finding multiple matches a good thing or a bad thing, when looking for matches? For both DNA and Acoustic tests on just one shot?


Question 3:

Do you deny that the BBN’s Exhibit F-367 generated false positives?


Question 4:

Is the correlation coefficient for the shot at 140.32 of 6 an honest result? Or did Robert Blakey pressure Dr. Barger to falsify the data that he reported to the BBN?


Question 5:

If Dr. Barger did not falsify his data, why shouldn’t the low correlation coefficient disqualify the shot at 140.32?



And indeed, there are many other several problems with the BBN data. I just used the sound impulse at 145.15 as one example.

Now, to be fair, most of these problems disappear if one demands a higher standard. If one says a correlation coefficient of 0.6, or even 0.7 is not good enough to be considered a match. But then we have to throw out the alleged fifth shot of Dr. Thomas because we have to be logically consistent. We can’t say a correlation coefficient of 0.6 is considered good enough to consider 140.32 to be Dr. Thomas’s fifth shot, while at the same time the stronger correlation coefficient of 0.7 to be considered not good enough for the shot 145.15.

And, I should stress, that adopting the strict 0.8 standard causes most “Multiple Matches” problems to disappear, but not all of them. That still leaves us with 6 matches for the 4 shots and so at least two of them have to be false positives.


Will Mr. Griffith dodge these 5 simple questions? Stay tuned to find out.

[Joe Elliott]

I make the following post reluctantly, because it may give Mr. Griffith a chance to dodge my five simple questions my “Multiple Matches Problem” post I just made. But I am making it so he can’t say I dodge his questions.

Oh, my! I see that you snipped the part of my reply where I proved that the 4-second impulse pattern is a dead end, a non-issue. You've been droning on and on about that pattern and about how it supposedly proved that the HSCA experts ignored other valid gunshot-like patterns on the dictabelt. And I've been telling you over and over that the pattern was rejected because it failed more than one of the initial screening tests in the preliminary analysis, but you kept saying, "Gee, as far I can tell, it was only rejected because it wasn't long enough." But now that I have proved from the BBN report that the pattern failed two of the five screening tests and that it contains no N-wave and muzzle-blast patterns, you go silent on the subject.

I stopped talking about it because it was clear we were making no progress on it.

You talk of this 4-second impulse sequence as being nothing like the 10.1-second impulse sequence.

You told me:

The 4-second sequence failed two of the screening tests, not just one. It failed the duration test, and it also failed the amplitude test.

Well, that sounds pretty definitive. I would accept a quote like that. Except there is one problem. That is not a quote from Dr. Barger. That is a quote from you.

The quote you from the BBN report you did supply:

The recorded outputs from both filters for the full 5 minutes were compared, examined, and plotted on a scale where 5 in. equals 1/10 sec. These plots revealed five impulse patterns introduced by a source other than the motorcycle. Upon closer examination, all but one of these patterns were sufficiently similar to have had the same source, and the impulses contained in these patterns appeared to have shapes similar to the expected characteristics of a shock wave and of a muzzle blast. The remaining pattern was sufficiently different in amplitude and duration as to have been caused by a different source. (8 HSCA 43)

This quote does not seem to be talking about the 4.0-second impulse sequence. Only the 10.1-second impulse sequence. What’s more, it seems to be saying that of the 5 impulses of interest, one of them, I assume 140.32, did not have the expected characteristics of a shock wave. So, they seem to be saying four of them are good, which I assume are 137.70, 139.27, 145.15 and 146.30 looked good to them. The fifth pattern, again, I assume, 140.32, had “sufficiently different amplitude and duration” to have been caused by a different source, something other than gunfire. And the impulse at 140.32 was, of course, the impulse than only had a correlation coefficient of 0.6, that Dr. Thomas insisted was the fifth shot.

So. it appears it is Dr. Thomas’s 140.32 “shot” that is “sufficiently different amplitude and duration” to be considered by the BBN, in 1978, to be a non-gunshot.

So, the quote you gave us, not only fails to contain any evidence discredit the 4.0-second impulse sequence, it seems to discredit the fifth shot that Dr. Thomas advocates.


Now, I should mention, I am not an expert on this acoustic evidence. So, I don’t know if the 4-second impulse sequence is similar to the 10.1-second sequence that the BBN focused on as containing the “gunshots”. But so far, Mr. Griffith has not provided quotes and links to show that BBN demonstrated it. All I have seen are quotes from Mr. Griffith (failed the amplitude test) seem to come from Mr. Griffith, not from Dr. Barger of the BBN. All I get from Dr. Barger is that the 4-second impulse sequence is too short to be the gunshots which were known to have lasted over 5 seconds. And that he stated it is probably caused by someone trying to transmit a brief message. But Dr Barger did not provide any technical details as to how he could tell. Were the amplitudes of these impulses too low? Mr. Griffith has not yet produced the telling quote from any of his statements or reports.

This is not a serious answer. Five shots is 25% more than four shots and 33% more than three shots. Plus, you omitted the fact that acknowledging the fifth shot would have also required admitting that there was another gunman firing from behind. So it is not at all "just stupid" that Blakey did not want to admit that five shots were fired. It was wrong and misleading, but it was not "stupid."

Blakey is willing to say there was a second gunman firing from the grassy knoll. But not that there was a third gunman who, like the first one, was firing from behind. That still makes no sense. That still makes Robert Blakely sound stupid, anyway you spin it.

This is similar to your silly answer when I made the factual point that the NRC panel had ample funding to conduct the same tests on the 4-second pattern that the HSCA did on the gunshot impulse patterns. Doing those tests would not have cost a lot of money, and the NRC panel had all the time in the world to get them done. But, they declined to do so. Why? Probably because they knew it would be a waste of time, because they, unlike you, had at least read the HSCA materials, and so they knew that the 4-second pattern had been rejected on entirely valid grounds.

Probably the 4-second impulse sequence was rejected because of the 15-day time limit they had to prepare for their September 11, 1978 report. And, perhaps, on the subconscious level, because they didn’t want to find a second cluster of shots, because that would discredit their first cluster.



But we are getting off on all sorts of tangents. Please first respond to my previous post, on the “Multiple Matches Problem”. That addresses the heart of my arguments on how BBN’s Exhibit F-367 does not support their contention that four shots were detected.

The worst result would be to find no matches.
The best result would be to find exactly 4 matches for 4 of the impulses.
Not the worst result but a pretty bad result is to find 6 matches for these 4 impulses, showing “False Positives” do occur.

[John Iacoletti]

And Dr. Thomas has been dishonest about the sirens heard on the Dictabelt, implying you can hear them loudly all the way to the hospital, when in truth, you don’t hear them, then they gradually get louder, and then fade away, as if recorded from a stationary motorcycle at the Trade Mart Center.

Can you provide a link to the audio recording where you are hearing this?

[Joe Elliott]

Can you provide a link to the audio recording where you are hearing this?

https://mcadams.posc.mu.edu/dpdtapes/

It is a 2:41 (two minutes, forty one seconds) recording at around 12:30.

Click on the little “speaker” and it may say you need to download Realplayer, which I did. It was fairly straight forward. I would prefer it was on youtube.

But yes, around 1:20 into the 2:41 recording, the sirens start up, build up and fade away by 1:57. I hear no more sirens from then on and the recording goes about another 44 seconds.

[Joe Elliott]

I'm currently looking at Fig 22 on page 63 where "false alarms" are marked with an X,

https://www.history-matters.com/archive/jfk/hsca/reportvols/vol8/html/HSCA_Vol8_0054a.htm

If you read on it's a simple fact they have have false alarms and they document them as they should.

In line with the fingerprint analogy proposed by Barger, for each acoustical fingerprint, they're now left them with several suspects, some of which are in fact false suspects. As I understand the explanations even more would have been acceptable and it's to some degree (lost reference) due to the spacing of the microphones. The further apart the worse the quality of detection becomes.

A pretty crazy example is #1, on top of page 65, at 137.70.

Even though the tape has already been screened and filtered they need to do further cleansing using additional "logic".

By the way (Otto can correct me if I am wrong) “Page 65” refers to the small “65” at the bottom of the page. One the same page, number “105” is on the top with a larger font. And for the blue index number along the top, clink on 107 to get to page 105/65.

Well, Mr. Griffith seems pretty reluctant to admitting all the “False Positives” in their data, but as you have noted, Dr. Barger is upfront and honest about them.

I like how they just blithely dismiss false positive after false positive.

Yes, we must dismiss the fourth entry for 139.27. And yes, we must dismiss the third entry for 139.27. But we won’t dismiss the first entry because that would destroy all our evidence for any gunshot at 139.27.

Or:

Yes, we must dismiss the fourth entry for 139.27 because that would indicate 130 feet in 1.6 seconds to gain that position. While there at it, why don’t they dismiss all the entries because the motorcycle would have to travel 164 feet in 1.5 seconds from the time the Hughes film is turned off to reach position 2 ( 5 ).

In the following, my phrases are in [[double brackets]].

However, the expected number of false alarms [[false positives]] to be found when testing four different impulses patters is 13 (see Appendix C), and only six [[false positives]] have been found. Therefore, it is not unreasonable to expect that there are seven more [[false positives]], although that would be the largest number possible since at least two of the remaining nigh are probable detections [[true positives]].

Ok. We found 6 false positives. We suspect there are more, maybe 7 more false positives, but we can’t tell which ones they are. But it can’t be as many as 11 more false positives because that would cause our case to go up in smoke.

[John Iacoletti]

https://mcadams.posc.mu.edu/dpdtapes/

It is a 2:41 (two minutes, forty one seconds) recording at around 12:30.

Click on the little “speaker” and it may say you need to download Realplayer, which I did. It was fairly straight forward. I would prefer it was on youtube.

But yes, around 1:20 into the 2:41 recording, the sirens start up, build up and fade away by 1:57. I hear no more sirens from then on and the recording goes about another 44 seconds.

Thanks.  Here's the direct link to the audio file for anybody else who might be interested.

https://mcadams.posc.mu.edu/dpdtapes/capture24.ram

I definitely hear the sirens.  Whether they "gradually get louder, and then fade away, as if recorded from a stationary motorcycle at the Trade Mart Center" seems to be a bit of an interpretive stretch.

[Joe Elliott]

Thanks.  Here's the direct link to the audio file for anybody else who might be interested.

https://mcadams.posc.mu.edu/dpdtapes/capture24.ram

I definitely hear the sirens.  Whether they "gradually get louder, and then fade away, as if recorded from a stationary motorcycle at the Trade Mart Center" seems to be a bit of an interpretive stretch.

I don’t think it is much of a stretch. But would be an incredible stretch to say this was recorded from a motorcycle accompanying the limousine to the Parkland hospital, but only recorded 36 seconds of sirens. We know McLain escorted the limousine to Parkland, because he said in an interview:

https://www.kenrahn.com/JFK/History/The_deed/Sneed/McLain.html

In any case, I caught up with and got in front of the limousine on Stemmons somewhere around Continental. The ride was wild! You know in your mind that you’re going way too fast, but if you slow down or fall, the cars behind are going to run over you. But you don’t think about those things, though, at the time; it’s all instinct.

We had to slow down when we got off Stemmons at Industrial. Along Industrial there was a railroad track which was located on a small incline some twenty to thirty feet before we were to hit Harry Hines Boulevard. Chaney, myself and another officer went airborne up the incline, hit the ground, and made the sharp left onto Hines.

When we arrived at the hospital, I parked my motorcycle and came back to the limousine about fifteen feet away. As the hospital orderlies approached to take him out of the car, Mrs. Kennedy was still laying over him, covering his head, and wouldn’t get up. So, I took it upon myself, reached over and caught her by the shoulder, pulled her and said, “Come on, let them take him out.” Somebody threw a coat over him just as she raised up, and they took him out on the right side of the car. She then stepped out on the left, stunned, and walked with me in a daze into the emergency room.

Looking at Google maps, catching up with the limousine around Continental meant that Officer McLain caught up with the motorcade within the first mile. Continental Avenue runs east and west, just north of the current 366 freeway. By the way, the name of Industrial Blvd has since been changed to Riverfront Blvd. (southern section) and Market Center Blvd. (northern section).

So, Officer McLain must have left Dealey Plaza immediately after the limousine left, caught up to it just north of 366 and the motorcade exited the Stemmons freeway at Market Center Blvd. (formerly Industrial Blvd.) and continued on to the hospital. And we know he went directly to Parkland because he was there to help Mrs. Kennedy out of the limousine. Clearly, we should be hearing the sirens for more than 36 seconds if he was escorting the limousine 80% of the way to Parkland.

The microphone being on a motorcycle at the Trade Mart Center is not too much of a stretch:

•   It is known that police motorcycles were waiting at the Trade Mart Center.
•   The limousine passed within 200 yards of the Trade Mart Center.
•   The phrase “Attention all units” heard on the Dictabelt was not broadcast by the Dallas police dispatcher, because it is not recorded on Channel 2. But it could have been the Dallas county sheriff dispatcher, and none of their vehicles were escorting the motorcade but some were waiting at the Trade Mart Center.
•   The sound of someone whistling does not fit with a motorcycle speeding off the Parkland, but does fit a motorcycle waiting at the Trade Mart Center.

[John Iacoletti]

•   It is known that police motorcycles were waiting at the Trade Mart Center.
•   The limousine passed within 200 yards of the Trade Mart Center.
•   The phrase “Attention all units” heard on the Dictabelt was not broadcast by the Dallas police dispatcher, because it is not recorded on Channel 2. But it could have been the Dallas county sheriff dispatcher, and none of their vehicles were escorting the motorcade but some were waiting at the Trade Mart Center.
•   The sound of someone whistling does not fit with a motorcycle speeding off the Parkland, but does fit a motorcycle waiting at the Trade Mart Center.

Where do you hear the phrase "attention all units"?

Where do you hear "the sound of someone whistling"?

How does whistling indicate Trade Mart anyway?

How do you know the 2:41 excerpt was continuously recorded and is complete?  I hear "clear 12:34" twice in the excerpt.  Once at 2:23 and then again softer at about 2:32.