Preface
Although this critique is written for other researchers and laypersons, the subject matter deals with the disciplines of acoustics and statistics. Therefore, I have had to compromise by sometimes using scientific wording unfamiliar to the layperson in order to strengthen or clarify a point addressed to the scientific community and yet other times oversimplify, draw analogies, or omit scientific notation. A thorough working knowledge of the basic facts of the assassination of President Kennedy and of the issue of the acoustical analysis done for the House Select Committee on Assassinations(HSCA) is a prerequisite for understanding this critique.
Summary
A police escort motorcycle's radio, which was stuck in the transmit mode while accompanying President Kennedy's motorcade through Dealey Plaza in Dallas, Texas on November 22, 1963, transmitted the sounds of the gunshots which killed JFK and wounded Gov. Connally to the Dallas Police Department where they were recorded on a Dictabelt. One of the shots originated from the grassy knoll. The Report of the National Academy of Sciences which refuted that fact is invalid due to bias, improper applications of method, careless errors, and deliberate misrepresentations of fact.
History
Under contract by the HSCA, the firm of Bolt, Beranek, & Newman(BBN) analyzed a tape recording of the JFK motorcade in Dealey Plaza and found 4 possible shots The most controversial was the shot from the grassy knoll, which would imply that a conspiracy was involved. Because BB&N could only state the probability for that shot was 50%, HSCA asked acoustics experts Weiss and Aschkenasy (W&A) to refine the data on that shot in order to reduce the uncertainty either way. W&A stated with a 95% confidence that the impulse on the DPD tape recording was a gunshot fired from the grassy knoll. The HSCA relied quite heavily on that conclusion in issuing its finding of a conspiracy.
The U.S. Justice Department (DOJ), instead of then investigating the conspiracy, attempted to discredit the HSCA findings. Its first attempt using the FBI bugging experts failed due to the lack of competence in the necessary disciplines. DOJ then had to assign the National Academy of Sciences (NAS) the task of discrediting the HSCA's acoustical findings. To be sure of the outcome, only government connected scientists, most with ties to the intelligence community, were selected and they worked in total secrecy without any challenge from independent researchers. It was hoped that their prestige would end the debate and their mandate was only to discredit the acoustical analysis of the HSCA, not to find the truth. The budget they were given was so limited that they could not do the necessary tests beyond those needed to discredit the HSCA. The DOJ was highly successful. It's been over a year since the issuance of the NAS Report and not one scientist has dared challenge it. It's a national disgrace for a country Which prides itself on freedom of thought that the challenge has to come from an ordinary blue-collar worker, rather than the scientific community.
Areas of criticism
I. Secrecy
II. Transcripts
III. Timing
IV. Voiceprints
V. Unanalyzed sounds
VI. Statistics
The NAS committee worked in total secrecy without
dialogue with critics. I
had written to them on December 1, 1980 suggesting several solutions
and points
that had to be addressed. Not only did they ignore them, they didn't
even
acknowledge my letter. Phone calls went unanswered. If this treatment
was
typical, it shows a bias that is unmistakable. Many of their
glaringly obvious
errors could have been avoided simply by accepting help from outside
researchers.
To me, the most shocking aspect of their report
is that no one had the
courage to accept responsibility. You'll notice that throughout this
paper, I
attribute all the committee's actions to Ramsey. This is not an ad
hominem
argument. This is primarily by default as he chaired the committee,
but also
because the members who actually did the work would not claim responsibility
for their contributions. To guess from appearances, it would seem that
the
entire committee met only a few times, pro forma, and that the bulk
of the work
was done independently by one biased, pro-Warren Commission, CIA-contract
consultant with the help of his blindly loyal students. I even doubt
that the
indlividual members took the time to read and understand the work.
We'll
probably never know due to the extreme secrecy imposed by the NAS.
Little more
of substance was accomplished than Luis Alvarez could have done on
his own. The
most amazing aspect of this charade is that it is almost unheard of
in the
scientific community, except in totalitarian regimes, that one biased
researcher publish his work unattributed under the cover of a
government-supported agency and then refuse to defend it or present
any
counter-arguments to its critics. This is not merely scientific arrogance,
it
is truth by fiat. I suppose that we should next expect the NAS to declare
that
the world is flat.
Another example of the secrecy is the fact
that Ramsey did not release
copies of the evidence along with the report so that someone else could
try to
duplicate their results. For instance, one of the major points of the
NAS
thesis is that Jesse Curry's message,"Go to the hospital." occurred
at least 20
seconds before the shots that were identified by BB&N/WA. They
based that on a
tape recording provided by DPD Capt. Bowles. Given the fact that NAS
and the
DOJ have kept the tape secret, how do we know that someone didn't splice
in 30
seconds of Lawrence Welk music in order to pad it? It certainly doesn't
match
the tape that I have and I doubt that it matches the tape that was
given to the
HSCA. Withholding of raw data is the first indication that the researcher
has
something to hide.
As other researchers have pointed out , there
have never been accurate
transcripts prepared of the DPD tapes. Ramsey made no exception to
that rule,
relying on DPD Capt. Bowles' latest disinformation. There is a wealth
of
information on the tapes themselves if anyone just takes the time to
transcribe
all the sounds on the tapes, even the silences. Most inaccuracies on
the Bowles
transcripts are simply negligence such as not bothering to mark repeats,
or
note the source of a transmission, etc. Sometimes Bowles intentionally
rewrote
a transmissions to correct grammar or omit pauses in order to keep
his fellow
officers from looking like country bumpkins. Most such editing is
inconsequential to the timing of the tapes. Yet there are some which
led Ramsey
to draw false conclusions.
On channel 2 there are several repeats due
to wear of the Audograph record
which added time to the apparent duration during playback. When comparing
timing between the two channels, several seconds must be deducted from
ch. 2's
timing (see timing chart ) to account
for those repeats. Between Jesse Curry's,
"Go to the hospital." and 190's, "You want me to still hold this traffic..."
which is crosstalked to ch. 1 and recorded simultaneously, there are
6 repeats,
only 4 of which were deducted by Ramsey. The other two he might have
missed,
because the transcripts did not accurately record them. Repeat 2 was
of the
dispatcher's message, "Parkland has been notified. 12:32.". Ramsey
did not
consider it a repeat, because only one word was repeated, "notified",
which
would not be of sufficient duration to be a period of rotation of the
record.
Yet, if Bowles had simply noted both times at which the "notified"
appeared, he
would have noticed immediately that there were 3.4 seconds between
them, which
is the correct time for a repeat. The mechanics of this repeat will
be
discussed in a further section.
Repeat 4 is the most blatant example of error
caused by Bowles
transcripts. Ramsey considered it not to be a repeat on the basis that
not
enough words were repeated to equal a period of rotation.
He based that conclusion on Bowles transcription of the messages between
Deputy Chief Lumpkin and Chief of Police Curry. In Bowles
version, Lumpkins' question ends with the words, "....men I have with
me?",
then Curry breaks in with, "Just go on to Parkland Hospital with me.
Just go on
to Parkland.". Period. Actually, the "with me." was the beginning of
a repeat,
which included part of Lumpkin's previous message ending with, "with
me?" and
continuing to repeat Curry's, "Just go on to Parkland Hospital." This
repeat
was also approximately 3.4 seconds in duration. So that's a total error
of 6.8
seconds or so in a very crucial area of timing, which was caused by
inaccurate
transcripts.
However, the most important error to me was
that Bowles left out and
Ramsey did not correctly include a message which I feel is the most
important
on the tape. In my letter of Dec.l, 1980, I pointed out that there
was a
transmission on the ch. 1 tape which proves that it was McLain's cycle
with the
open mike. About two minutes after the shots, a fellow officer yells
to the
cyclist with the open mike, "Take off, Buddy.", whereupon the cyclist
turns on
his siren and speeds off to catch up with the motorcade. I did not
expect
Bowles to include that, given the fact that the DPD has been a leader
in
anti-conspiracy propaganda and many of its officers have destroyed
or
manufactured evidence over the years. Yet, Bowles does include a highly
doubtful identification of the message, " hold everything", which confirms
the
NAS thesis of a cross-talk, when that message is not at all clear.
In his
report, Ramsey ignored the message completely. A simple examination
of the tape
at that point would have shown that the voice was speaking directly
into the
open mike, confirmed by the lack of a heterodyne, which results from
interference, and that the siren was on the open mike, confirmed by
the
presence of an interference arc. H.B. McLain was the only DPD cyclist
named
Buddy in the motorcade. † There
was a Buddy Walthers in Dealey Plaza, but he was
on foot directing traffic and stayed in Dealey Plaza. Ramsey made no
attempt to
find the officer who yelled to McLain, because then he could have testified
that it was indeed he who yelled to McLain, and then that proof that
it was
McLain's open mike might have required the NAS to actually look for
the shots
which would then have to be recorded somewhere on the tape. This was
unacceptable for scientists who were impanelled only to discredit the
conspiracy finding. Although there is no direct evidence of who is
the
unidentified voice, study of the movements of all the cyclists seen
in films
taken after the shots suggest fellow cyclists J.B.Courson, James Hargis,
or
Clyde Haygood as possibilities. I tend to think it was Courson as he
passed him
halfway down Elm Street. Haygood and Hargis parked their bikes on Elm
St. and
went up the grassy knoll in hot pursuit of the assassin. However, McLain
could
have slowed down to offer either one help and was yelled off. I do
not have the
resources to investigate this aspect more thoroughly, but the DOJ does
and will
not. It is hoped that someone will before more witnesses die. Ramsey's
omission
of this vital issue is a clear example of the cover-up of evidence.
A related problem is the fact that there are
so many versions of the tapes
themselves. The Bowles tapes, the Curry tapes, the FBI tapes, the Secret
Service tapes, the Warren Commission tapes, the HSCA tapes, etc. I
have tried
to get a copy from BB&N and from Ramsey. They both declined to
share it. When I
requested a copy under the Freedom of Information Act from the DOJ,
FBI, Secret
Service, and National Science Foundation, they all denied ever having
or
hearing of the tape. The HSCA records are locked up in the National
Archives,
safe from public access. Therefore, my studies are based on the version
I
bought from the Collector's Archives, whch I describe as the Canadian
copies. I
have no idea which other version they are identical to. If you wish
to
duplicate my research you can get copies from that source.
Ramsey's major thesis rests on the conclusion
that Curry's message, "Go to
the hospital." occurred before the purported shots. On this basis he
concluded
that there was no reason to look for shots. It's ironic that the NAS
panel was
called a "Committee on Ballistic Acoustics" when they didn't
perform one such
test or deal with the topic. They went to great lengths of propaganda
to just
avoid having to do any ballistic acoustics. Even if Curry's message
came before
the time that W&A found a shot, and Ramsey thought their and BB&N's
methods
flawed, there could still be shots on ch. 1 before Curry's message
that went
undetected. After all, if BB&N's method could not find shots, then
it might
have missed them. Ramsey did not look, for his mandate was only to
destroy, not
seek.
As I pointed out previously, Ramsey failed
to deduct time for 2 repeats on
ch. 2, amounting to app. 6.8 seconds. Further he based the timing on
tapes
which he and the government have kept secret. My Canadian tapes show
marked
differences. The timing between ch. 1 and ch. 2 can be compared by
identifying
any messages broadcast simultaneously over both channels. Ramsey spent
a great
deal of effort and money trying to do this for uncertain messages,
but ignored
an obvious one, Henslee's simultaneous broadcast to all emergency equipment.
This
one is unmistakable, well documented in transcripts, and needs no elaborate
tests to confirm. Comparing ch. 1 to ch. 2 based on that message as
the
benchmark and making the necessary adjustments for known factors produces
the
following results: Henslee's ch. 2 message at 324.5 minus Curry's at
32.7 =
291.8 secs. between messages. Deducting 31.1 secs. for the 8 repeats
= 260.7
secs. corrected. Two corrections must be made for ch. 1. First, 6 secs.
representing a break in recording must be deducted from the 432.5 yielding
426.5. Then the time must be corrected by .99 for the difference in
recording
speeds (confirmed by a comparison of the frequencies of Henslee's voice
during
the simulcast), yielding 422.2. The grassy knoll shot at 143.2 corrected
by .99
yields 141.8. Then 422.2-141.8 = 280.4. Then comparing the corrected
times,
260.7-280.4 = -19.7. This would tend to confirm that Curry's message
came almost
20 seconds after the shots, rather than a minute before as Ramsey has
concluded. These rough calculations were done easily with little expense
and no
need to rely on controversial tests. But, can more accuracy be obtained
by
using as a common reference point a message earlier than Henslee's
simulcast? I
believe that there is compelling evidence as presented by Ramsey to
justify the
opinion that the earlier message, "You want me to still hold this traffic..."
transmitted over ch. 2 was retransmitted via cross-talk on ch. 1 at
the same
time. Using this message as the common reference point, the comparison
produces
the results I have listed in my timing chart.
In order to fit both channels in the
same chart for easier comparison, I
have omitted some of the calculations used for correcting the times
listed as
correct. Arguments for those corrections and calculations used for
them are
found here and in other sections of this paper. In the interest of
simplicity,
no speed of tape correction was made to ch. 2, but rather to ch. 1.
I have
estimated that there is approximately a 1% difference in true speed
between the
two channels. That estimate is based on a comparison of the frequencies
of
dispatcher Henslee's voice during his simultaneous transmission, "Attention
all
emergency equipment...". After deducting 6 seconds for the break in
recording
to cassette, ch. 1 times were multiplied by .99. However, the difference
in
corrected times for "Attention..." at 426.5 and "You want..." at 318.6
is only
107.9. Ramsey's analysis of "You want..." proves that it is cross-talk,
so the
time for ch.2 of 109 secs. should be the same for ch. 1, amounting
to a
discrepancy of 1.1 secs. I have arbitrarily decided to add the 1.1
secs. to
ch.1 to correct for that. For one reason, it is possible that the 6
second
break in recording chopped off some of the original tape. Second, the
corrected
times fit the times given by BB&N. Third, the corrections to ch.
2 are already
complicated enough. Fourth, deducting more time from ch. 2 would tend
to take
the corrected timing out of fit with the dispatchers' time notations
more than
adding the 1.1 secs. to ch. 1.
The major correction to ch. 2 consists of
deducting time for repeats,
which have added time to the apparent length. Arguments are presented
for each
repeat:
Repeat 1 is the same as Ramsey identified.
Since it is a double repeat, it
is two times the period of rotation, 3.3 secs., for a total duration
of 6.6
secs. My stopwatch timing timed on the word, "secure" agrees with Ramsey's
corrected time.
Repeat 2 is the one that Ramsey missed entirely,
as I pointed out
previously. Now, because the stylus on a Gray Audograph moves from
inside
outwards( opposite of the normal phonograph ), the period of rotation
increases
with time. Therefore, this repeat measures 3.4 secs, which is confirmed
by
noting the difference in time between both "notified"s. An FBI technician
probably lifted the stylus to get past the known repeat.
Repeat 3 is the same as Ramsey identified.
Because some of the message is
unintelligible, it was impossible to tell if it was only one complete
repeat or
if someone had manipulated this repeat. The closest approximation by
timing
from different words was 3.8 secs. Although this is not a typical period
of
rotation, it agrees with Ramsey's uncorrected time.
Repeat 4 is another that Ramsey
misinterpreted, as I pointed out
previously. Starting the stopwatch at the break after "hospital" produces
a
time of 3.4 secs., which is a period of rotation.
Repeat 5 is the same as Ramsey
identified. Regardless of the fact that
the words in the repeat are unintelligible, the total duration is 3.4
secs.,
again a period of rotation.
This brings up an interesting
observation. Ramsey clearly labels this
repeat on the strip chart, Figure C-2, but if you look closely, the
decibel
level is very low. First, what could account for that, and second,
what
implications does that have. It appears to me that the most likely
cause is
skating or "soft mistracking", where the stylus rides up one wall of
the
groove, but does not break away and jump out of the groove. What are
the
implications? Profound! First, such mistracking can occur not only
during a
repeat, but also during sections without repeats. Thus, the fact that
this is a
mistracking can not,not,also prove that it is also a repeat.
Second, Ramsey tried to pad the
earlier part of ch. 2 with supposed
silences. One of the criteria he used in establishing the existence
of the
silences was that the strip charts showed periods where the signal
stayed below
his arbitrarily imposed threshold of 10 decibels below peak voltage
for more
than 4 seconds. He based that on the word of DPD Capt. Bowles, with
no hard
evidence. But even granting the possibility that there were hold relays
of
approximately 4 seconds, there is no firm evidence about the threshold
at which
they operated or that in fact that they were operating properly on
Nov. 22,
1963. Moreover, the fact that repeat 5 has a period of duration less
that the
estimated hold duration and the fact that its decibel level was much
lower than
the estimated hold thresholdwould suggest that such mistracking could
be
mistaken for a silence. Thus it is possible that none of the silences
identified by Ramsey were in fact silences. Further, it seems that
the signal
in every supposed silence remains higher than at repeat 5. Ramsey
then goes on
to arbitrarily add 46 seconds to ch. 2 to account for his silences,
justified
by the specious argument that perhaps ch. 2 wasn't used much at that
time. This
can easily be refuted by pointing to the fact that many officers were
trying
and unable to use ch. I and switched to ch. 2 to report that fact,
get orders,
or try to find out what was happening. Also, BBN's study showed that
ch. 2 was
running "nearly continuously" during that time.
As pointed out before, ch. 1 and
ch. 2 real times must be the same for
"You want..." and "Attention...". As you will remember, I arbitrarily
chose to
use 109 secs. for ch. 1. It could be slightly less, but not by much.
The
smallest corrected time was 107.9. Ch. 1 was recorded on a Dictabelt
running
continuously, in contrast to ch. 2's voice-actuated record. So there
is simply
no way that the time for ch. 1 can be reduced further. If anyone wants
to
quibble over the 1.1 secs., he can do the calculations the other way.
But if
the corrected time for ch. I is 109 secs., what impact does that have
on the
corrections for ch. 2? The uncorrected time of 324.5 minus 230.5 yields
only 94
secs., not 109. To make matters worse, 7 secs. must be deducted for
repeats 7
and 8 yielding only 87 secs. The only way to account for such a large
discrepancy is by the fact that there are 22 secs. of silence somewhere
during
that period on ch. 2. This is the reason for the NA instead of a corrected
time
and the reason for the 315.4 instead of 293.4. Without access to the
strip
charts which Ramsey has declined to release, there is no way to tell
exactly
how many, where, of what duration there are silences. If Ramsey had
done this
analysis, he'd have found the silences, analyzed their characteristics,
and
gone back to his previously suspected silences to do a more careful
study. Just
listening to the tape, it seems to me that there are much fewer messages
during
this period than the earlier one.
Repeat 6 is the same as Ramsey
identified. The duration, of 3.5 secs.,
the same as Ramsey's, is a period of rotation.
Repeat 7 was overlooked, because
Ramsey was only interested in the
period before "You want . .." . Like repeat 2, only one word was repeated
and
the time between each was a period of rotation. In addition, an analysis
of the
background frequencies between each word would show a complete match.
Repeat 8 is a very clear example
of a repeated phrase. The period of
rotation is 3.5 secs. It is so obvious that the only explanation for
Ramsey's
missing it is the bias he had for padding the earlier part of ch. 2.
Although I have no expertise in
this field, several errors appear
obvious even to a layperson. The most controversial finding by Ramsey
was his
validation of Barber's subjetive identification of a simultaneous transmission
of the words, "hold everything." I have never been able to hear any
distinct
words on ch. I at that time. Perhaps this is another example of the
differences
between various versions of the tape.
The correlation shown in NAS Figure
6 is misleading. Here Ramsey uses a
favorite old trick of changing the scale in order to create what Huff
calls a
"gee-whiz" graph. Notice that although the horizontal scale is the
same for
both the "You want..." correlation and "hold evcrything", the vertical
scale
for "You want..." is about 19mm/.1, whereas the "hold everything" is
about
28mm/.1. This exhaggerates the goodness of the correlation and distorts
the
spike in order to make it resemble the "You want..." spike. The actual
numbers
are more revealing.
The peak of the spike is about .518,
whereas even
background static reached .3, a difference of .218. BB&N's correlation
coefficient = 10¸Ö(12ˆ•˜ˆ•˜˜˜*•›ˆš••˜•ÿ14)
= .77. Yet Ramsey implies such a figure is
unimpressive. Then how could he accept .518 as significant? If there
is a
choice, logic demands that the source with the higher correlation be
chosen as
the true source. Incidentally, Ramsey criticized W&A for choosing
the
correction factor which gave the best match and went ahead to do the
same thing
himself for this correlation. Also, Ramsey criticized W&A for not
considering
alternative sources, but it would seem that the screening process and
Ramsey's
voiceprints rule out most other sources. Comparing the grassy knoll
shot
pattern to other portions of the tape which included static, voice
transmissions, etc., BB&N found that the correlation exceeded .6
only during
the 9 seconds where other shots were identified. Now Ramsey has found
a
correlation of only .518, which would not have passed BB&N's screening
process.
One weakness of his study, for
which he had the gall to criticize W&A,
was the failure to do a control. Perhaps any voice-transmission could
achieve a
correlation of .5, but not even a known simultaneous transmission could
achieve
a 1.0 correlation, due to the static or background noise. There are
some
examples on the tape where cross-overs from ch. 2 were recorded along
with live
sounds on ch. 1 from the open mike. Moreover, Ramsey does not offer
any
evidence that the message,"hold everything" is unique and identical
to Decker's
known transmission. If the message had been a very common one, like"l0-4."
or a
persistent one like"75.Signal 5?", there could be several such occurances
and
it would be difficult to prove which was the match. Ramsey failed to
look for
other occurances of "hold everything" and do a control correlation
for each. If
he had done the same comparison to Curry's "hold everything" message,
the
correlation would have been approximately .55.
The correlation for "You want..."
was much better,.766 at the peak of
the spike. Looking at the actual voiceprints , the two channels look
very much
alike, unlike the "hold everything"which look nothing alike. On hearing
the
two, it seems clear to me that they are the same message, though the
ch. 1
cross-talk sounds as if it has more background interference and that
the last
few words fade out or get chopped off. Ramsey also suggested the use
of Bayes
theorem, but gave no concrete examples, even though he had a perfect
opportunity here. Using Bayesian logic, he should have noted that although
the
"You want..." is clearly and audibly a known match, the highest correlation
he
could achieve was .766, which represents a probability of 1.0, amounting
to a
difference of .234 due to inaccurate methodology, technique, and/or
noise.
Therefore, the highest correlation one could hope to achieve might
be
approximately .8. Adding this loss factor to BB&N's.77 results
in an estimated
.97, which is quite significant.
Further, there is another simultaneous
transmission that should have
been analyzed. At the time of Henslee's simulcast"Attention..." there
was very
little noise. Why didn't Ramsey do a control on this message? Because
it might
achieve a correlation of only .8 or less, which would show how imprecise
and
subjective his correlation method and technique was, casting doubt
on the
identification of "hold everything". Second, it would have increased
the
probability of the grassy knoll shot.
In general the whole study was
biased and subjective, relying on the
FBI, Bowles, and the one member of the committee whose bias is well
documented.
Calling in outside consultants would not necessarily have guaranteed
elimination of bias, but perhaps might have made the study more objective.
As
a recent article in "Technology Illustrated" pointed out, there are
several
researchers at MIT, Carnie-Mellon University, and IBM doing advanced
work on
speech recognition by computers. Although they may have ties to the
intelligence community, it is assumed that they wouldn't tinker with
the basic
computer program, so that the computer could do the same study completely
objectively. Ramsey failed to use the best resources available.
Another area left uninvestigated was the analyzing
of additional sounds on
the tape, not just the ones Ramsey tried to use to prove Barber's thesis.
One
such sound is the message which I point out to the NAS. The fact that
Ramsey
did no study of the message proves how important it is. A thorough
study might
prove that the arcing is a result of McLain turning on his siren. This
would
put an unbearable burden on those who argue both that McLain turned
on his
siren immediately and that the open mike was stationary at the Dallas
Trade
Mart.
Also of interest is the nature of static and
interference from the Dallas
Electric radios. Could such interference resemble gunshots? Ramsey
criticized
W&A for not considering alternative sources, but failed to do any
study
himself. Ramsey also criticized W&A for not using the same method
as they did
on the grassy knoll shot on the other shots as a control, but
fails to point
out that the HSCA's contract was only for the detailed analysis of
the grassy
knoll shot. W&A were not asked to do the same work on the others
and were so
hard pressed for time that they could not have physically done it.
Since the
HSCA expired, none of their acoustical experts has shown any interest
in doing
follow-up work on their own, offer help to other reesearchers, or vigorously
defended their reputations. I have attempted to do so in my limited
way. I have
applied the method W&A used for the grassy knoll shot to the other
4 and
offered to share my latest calculations with Ramsey, but received no
reply. If
Ramsey thought that the method W&A used was flawed, he should have
shown that
by applying it to the other shots.
This is the most crucial and yet perplexing
aspect. No scientist involved
in any study to date can state with absolute certainty if there were
or weren't
shots recorded on the DPD tape. Everyone talks in terms of probability
and
likelihood. This is typical of scientists, but not comforting to the
layperson
who wants absolute truth. But most people are perfectly capable of
understanding
simple odds and probabilities associated with everyday life such as
games or
the weather. If there is a 99% chance of rain, you'd take your umbrella
or
postpone the picnic, whereas if there is only a 1% chance of rain,
you
wouldn't. That's no guarantee, but at least it's a good guideline on
which to
base decisions.
15 matches between test shot patterns
and DPD impulses had correlations
exceeding 0.6. BB&N noticed that the pattern of matches seemed
to fit the path
of the motorcade. Viewing the dots representing the time from
the first match
to the last on one axis and the distance along the microphone layout
on the
other as a scattergram, there appears to be a definite non-random sequence.
This chart is based on the raw data, which included obvious false alarms.
At
this point, there was no attempt to deduct them. Mikes had been placed
about 18
feet apart along the motorcade route as a compromise between the estimated
acceptance windows and the need to cover as much of Dealey Plaza as
possible
with the fewest necessary mike locations. The design of the test was
adequate,
given the constraints imposed. The pattern of matches can be as important
as
the total number. A cluster of several matches from neighboring mikes
matching
the same DPD impulse not only tends to confirm the impression that
the pattern
of the DPD impulse is a shot, but also gives clues to the real location
of the
policeman's mike. If an early DPD impulse pattern matches with mikes
2(5) and
2(6), the true location might be somewhere between them, but if it
matches
mikes 2(5) and 2(2), the true location is more likely between them.
If a DPD
impulse pattern matches widely separated mikes, say 1(4), 2(10), and
3(12),
that would suggest that 2 or maybe all 3 matches are false alarms.
To test whether there was any relationship
between the time and distance
coordinates, BB&N partitioned the matches into a 2 x 2 contingency
table and
used a generalized c² test with
the null hypothesis that matches were
independent so that the distribution in the cells should be random.
Using the
data 1,6,8,0 produced a c² = 11.4.
For 1 degree of freedom, BB&N stated that
the probability that this large a value could occur at random is less
than 1%.
Ramsey failed to criticize 2 important aspects of this calculation.
First, the
1% level may be too high, because BB&N did not use a table extensive
enough to
give an accurate percentage level for such a large value. The c²
value for the
1% level is 6.635. More extensive tables, such as the Biometrika, show
the
value for the 0.5% level is 7.879. A Fisher and Yates' table shows
the value
for the 0.1% level is 10.827. Thus BB&N's stated significance would
seem, to
the casual reader or layperson, to be not as significant as the value
really
is. Second, the c² distribution for
such a low total (N=15) tends to be skewed.
It might have been better to use an exact test. As Langley points out,
the
recommended test for N<50 is Fisher's Test(1934), which makes use
of the
hypergeometric formula: (n1!•n2!•n3!•n4!)¸(N!•a!•b!•c!•d!),
to derive the exact
probability. Although the computations can be more complex, because
of the
factorials, the fact that N is moderate and the importance of accuracy
necessitates using Fisher's Test. For the data 1,6,8,0 the exact probability
computed by this test is .0013986.
Ramsey attempts to criticize BB&N's highly
significant result by
arbitrarily and informally deducting 7 matches, saying that some of
the alarms
are dependent,because some of the microphone and rifle locations were
similar.
Why should independence be a criterion for inclusion when that is what
is being
tested? Ramsey cites no standard reference or test that advocates such
a
reduction in the individual cells. Even if that were a valid approach,
he'd
have to go through the table match by match and justify each deduction.
Then we
could challenge or debate each one. Due to the design of the test,
we would
expect patterns from neighboring mikes to be similar, but that does
not mean
they are dependent on each other. How close or how far away from each
other do
the mikes have to be for Ramsey to consider them "independent"? He
cites no
estimates for that. Ramsey's criticism that rifle locations were similar
seems
to have more validity and should be studied more carefully. BB&N
designed the
test so that rifles would fire at different targets, with the muzzle
first in
the plane of the window of the TSBD and then pulled back 2 feet inside
to see
if there was any significant difference in the patterns, and therefore
the
correlations. Perhaps the pattern would be different if the shockwave
trajectory were different. Or perhaps the inside wall and window significantly
reduced the decibel levels of the echoes if the gun were fired from
well within
the TSBD. If a DPD impulse matched one set of conditions rather than
the other,
that might provide vital clues to the real conditions of the assassination.
Evidence from other fields contradicts the
assumptions that Ramsey used to
calculate the probabilities of the other shots. Even Ramsey himself
conceeds
that his calculations may be too conservative, but fails to fully explain
what
assumptions he made. He deducted 6 false alarms as BB&N had from
the 15 matches
to get the 9 remaining. But one of those matches was incorrectly identified
as
a false match. BB&N stated that the one at 140.32 came too close
after another
more likely match to have been fired from the same rifle. Although
that is a
proper attempt to eliminate a false match based on corroborating evidence,
the
logic is flawed and the acceptance of the veracity of that evidence
is biased.
Perhaps an automatic weapon was used for those two shots, or one man
was firing
two weapons, or one shot came from the TSBD and the other came from
the grassy
knoll. The case that I believe most likely is that another man fired
the second
shot from a few windows down. There is some evidence and eyewitness
testimony
that there was a second man on the same floor of the TSBD and a different
weapon from Oswald's Mamlicher-Carcano. Ramsey's value of 7 is based
on the
assumption that only 2 shots were fired in Dealey Plaza, accounting
for 9
wounds, 3 points of damage to the limo, and 4 scars in the street.
That's how
he came up with 1-(7¸9)³. If
he had stuck to the old Warren Commission 3 shots,
then 1-(6¸9)³ = 0.7037037. The
trajectories of the wounds compared to the timing
suggested by the tape make 4 shots the minimum, even without a miss,
so
1-(5¸9)³ = 0.8285323. I have
found corroboration in all the other evidence for 5
sbots, which includes BB&N false alarm 140.32, so the equation
should be
1-(5¸10)³ = 0.875. Notice that
for these calculations, Ramsey doesn't complain that
the mikes were too close as he did for the chi-square. He wants to
have it both
ways. If he did, he'd have to deduct 5 of the matches as duplicates,
then 5
more false alarms, leaving only 5 matches for 5 possible shots, so
the equation
would be 1-(0¸5)¹ = 1.0.
The critical issue is BB&N's computation
of the probability that W&A could
obtain such a good match between the DPD tape pattern and the echo
pattern they
calculated for their estimated cycle-shooter locations. BB&N calculated
that
the probablity that such a match could occur at random was about
0.053, giving
them a 95+% confidence that the pattern on the tape was a grassy knoll
shot.
Some of Ramsey's criticisms of BB&N's computation are correct and
quite
important, but his only purpose seemed to be to increase the p above
0.05, not
to seek the true conditions, nor try for maximum accuracy. As W&A
pointed out
quite clearly, there is always a chance that the pattern is not a shot,
but a
random pattern due solely to chance.The thing which gives inversely
proportional confidence is the chance that it could occur at random.
If Ramsey thinks the randomness of the DPD
impulse pattern is an important
criterion, then the first thing that should have been done that none
of the
scientists did, would have been to test the pattern for randomness.
Using W&A's
Table 4, I counted the runs of signals and the runs of non-signals,
using 2 ms.
time windows starting at .1 ms. With the values: n1=22,
n2=163, N=185, and r=40,
I used Wald and Wolfowitz's formula with correction as shown
on Langley p.326.
The resulting z = -.0200981 doesn't even reach the 10% value of 1.64.
Thus this
test does not show that the pattern is not random. I then deducted
the 190 ms.
of silence between the two groups of impulses adtsted the randomness
with the
values: n1=22, n2=68, N=90, r=40. The resulting
z = .6626179 was slightly more
significant, but was still much greater than 10%. Thus even withV the
dedction
for the silence duration, non-randomness could not be shown. cause
this tet is
a modification designed to make calculations easier , I decided to
look for a more
complex formula. In Freund's Modern Elementary Statistics, p. 326
test of runs
r-m
is based on the formula z = —
s .
Calculation with the original values yields z = .2858912. Again deducting
the 95
time windows of silence this time yielded z = -3.0090982, which is
significant
almost to the .2% level. This shows a definite non-randomness.
Ramsey tries to criticize the 0.053 figure
with a ridiculous analogy to a
bridge hand having 3 Aces. That is like comparing apples and oranges.
As
Russell Langley points out in Practical Statistics, pp. 374-375, a
person might
throw a handful of coins into the air and notice the pattern they land
in. The
probability of their landing in those positions may be a billion to
one, but
that doesn't imply that it is due to anything other than chance. The
probability would be significant only if the pattern had been predicted
before
the experiment. Then we could say that it's extremely unlikely that
the pattern
is due to chance. The specific bridge hand was not predicted before
the deal.
W&A did predict an echo pattern that matched with the DPD impulse
pattern very
closely, as noted elsewhere.
Also, why would anyone suspect the dealer
of dishonesty? I dealt that hand
a few weeks ago and everyone had to pass. Why not suspect the opponents
who
shuffled and cut the cards? If a player were dishonest, he'd want to
arrange
the cards for maximum profit with the least suspicion. An excellent
card
manipulator can stack the deck or misdeal after the opponents have
shuffled and
cut to give himself a 7NT hand. But no one would double that 37 point
hand and
everyone would suspect him of card sharking. Most cheating at bridge
consists
of secret conventions, hand signals, or other relayingtechniques. Professionals
make use of ceiling peepholes and leg transmitters. Most average players
suspicions are not aroused by a few lucky finesses and psych bids,
but are when
the opponents get a 7 NT hand every time. If a card manipulator were
really
skillful, he might give himself a 10 card spade suit and two voids,
his partner
the 3 missing spades and a void in the fourth suit, plus giving the
dealer
those famous 3 aces and his side most of the remaining high cards,
thus fooling
him into taking some kind of action over his spade bid, so that when
he jumps
to 7 Spades, he makes the dealer guess whether to double or bid 7 NT.
Either
way would be diastrous , giving the manipulator maximum profit.
But Ramsey also makes simple errors in his
analogy. He states that p =.044
(actually .043841538), but that is cumulative including getting 4 aces.
For only
3 aces the p = .0412004786. He then goes on to compare the 0.053 which
he
considers a Poisson probability with 0.044 which should be considered
only a
hypergeometric probability. To make a valid comparison, he'd have to
take the
hypergeometric,and then calculate the chance of getting only 1 in 180
trials as
BB&N did. The expected number in 180 would be 7.41608615. Using
7, x = 1, the
Poisson Test shows that p = 0.006832038. So if you never get a hand
which
contains 3 aces out of 180 deals, you might get some sympathy.
Ramsey correctly pointed out BB&N errors
of logic and simple oversight. As
Ramsey pointed out, BB&N did make a simple oversight in stating
that N = 45
windows as used in the hypergeometric calculation. Each of the two
groups
containing impulses was 90, for a total duration of 180 ms., thus 90
windows
not 45. However, Ramsey doesn't tell you what the hypergeometric distribution
would then be, given N = 90. In fact, h(9;12,14,90) = 5.13843913 x
10-7 which is much
less than h(9;12,14,45), which is 3.12899277 x 10-4 . Note
here that Ramsey
fails to point out that BB&N had already made 2 conservative adjustments
to the
data. First, they deducted the 320 ms.(should have been 190 ms.) of
silence
between the two groups of impulses. If that duration had been included,
then
they would have started with a much smaller value for the hypergeometric.
Second, they deducted 1 match, because any match could be made simply
by moving
the time scale of either. Ramsey ignored this adjustment, but I feel
that it
must be questioned. Moving either time scale would inrease the sampling
space.
But given the fact that N is now so large, the difference may be minor.
On the
other hand,Ramsey makes the very important observation that 2 pairs
of
impulses, 19+20, 23+24 coincide in the same 2 ms. window and one pair
of the
predicted echoes coincide in the same window. That all depends on where
you
draw the boundary between time windows. In this case, it is natural
to start at
0.0 and make each 2 ms. Thus 312.4 and 313.1 fall in the same window.
Further,
#19 and #20 are both 283.7; they obviously fall in the same window
no matter
what boundary is set. So Ramsey deducts those coincident impulses to
reduce the
data to x=8, n=11, M=12, N=90. That's o.k., but then he goes on to
reduce those
by 7 to acunt for the so-called "free" parameters. Later he does note
that the
data tend to suggest non-randomness, except for his admittedly conservative
adjustment. However, he states plainly in the text that such an adjustment
of
deducting one match for each "free" parameter is a traditional method.
To me it
seems outrageous, and nowhere does he cite any reference or test which
advocates that approach. There may be similar methods applied to other
tests,
but it is a complete misuse to apply them to the hypergeometric arbitrarily,
with no experimental results or logical proof.
Yet even given his adjustments, Ramsey fails
to perform the same
calculations that BB&N did with that data. Specifically, the hypergeometric
h(1;4,5,83)=.206996006, not .223. Then the next problem arises when
trying to
compute the probability of getting such a match only once in the many
trials
possible just by moving the cyclist's position. Ramsey should have
noticed
W&A's error in estimating the width of the corridor travelled by
the cycle.
They derived the 18 feet length correctly by noting that the test pattern
from
only mike 3(4) matched, not the one before or after, so the cycle was
probably
in the range of mike 3(4). But when estimating the width, they erroneously
concluded that, because the mike was supposed to be in the middle
of the
street with no mike on either side, the cycle could have been anywhere
along
the width of the street. In making that estimate, they presupposed
certain
real-life conditions. They assumed that the cycle remained in the street,
otherwise the width could have been 100 feet or more. That is not necessarily
a
safe assumption, given testimony of witnesses that saw a cyclist try
to jump
the curb and drive up the grassy knoll. In this case they should have
examined
the evidence more carefully. The estimate of 180 Bernoulli trials is
very
important. W&A never actually performed 180 trials, but BB&N
estimated that
they could have and used that figure in the binomial calculation. There
are
several clues which they overlooked.
First, the cycle has been identified as McLain's
which was on the left
side of the motorcade at the side of the press cars. Also, the previous
matches
were with mikes on the left side of the motorcade route. For the cyclist
to
have been on the right side of the motorcade at the time of the grassy
knoll
shot, he would have had to cross the center lane through the press
cars. There
is absolutely no testimony that McLain did that. Further, there
is
photographic evidence that only one cycle, Courson, crossed through
the press
cars from the left to the right side of the street. These films,
including
Dorman and Couch, show that McLain stayed on the left. Also, the Altgens
6
photograph, taken at about the same time as Zapruder frame 255, just
a few
seconds before the grassy knoll shot shows several cars rounding the
corner in
the middle of the street. Calculating a smooth trajectory from his
known
positions in these films, the farthest to the right McLain could have
been
would be about 5 feet into the middle lane, and that is even doubtful,
given
the closeness of the press cars. We must assume that W&A would
remember that no
two objects can occupy the same space at the same time. If that space
was
occupied by a car, it is unavailable to the cycle as a possible trial.
So these
limits produce a corridor of 18 by 18 feet square, equal to 81 trials.
Ramsey's deduction of 7
parameters must be studied more in depth. For
example, he states that there are 2 parameters associated with the
location of
the shooter. If one is for the test shooter and the other for the second
assassin, then a more careful examination of the evidence would allow
one to
determine their positions within a very small margin of error. If both
are for
just one, representing the x and y axes, then why not a third for the
z axis?
Even then, the variation in recoil would leave an uncertainty of about
1 inch.
Would Ramsey still cite that as a "free" parameter when it has an impact
of
less than .001% on the calculations? If the 2 parameters are for one
shooter,
then maybe there should be 6 total, 3 for each shooter. Or what about
the cycle
position? Ramsey seems to imply that the cycle can be anywhere, whereas
I have
shown that the travel corridor was actually much narrower. Even such
minor
factors as the wind velocity, building shading, bullet airodynamics,
etc. could
be considered "free' parameters. The main problem is that Ramsey never
cites any
criteria for deciding what a parameter is and how "free" it is, nor
any
standard method of accounting for them. Some of the parameters he lists
are
dependent or off-setting. Others fall far below the built-in uncertainty.
In
addition, there are other factors which Ramsey did not consider. The
maps used
were only accurate to ±3 feet, so that's almost a 1% difference
for most echoes
right there. Or perhaps one temperature was measured with a dry bulb
and the
other with a wet bulb. What difference would that make? Perhaps the
shockwave
of one bullet or the explosion of the explosive bullet temporarily
rarified the
air through which the sound of the next echo travelled. What difference
would
that make? No logic seems clear. Any fool could make up a list of all
the
variable conditions that could have existed and then deduct one match
for each.
Perhaps the worst mistake Ramsey makes
is found in the next to last
sentence of his criticism of BB&N's computation. I tend to give
him the benefit
of doubt and assume it was a simple oversight or writing fatigue. If
not, it is
one of the most blatant misrepresentations of fact ever. Ramsey states
that the
level of p = 0.223 is not impressive in contrast to the claim of p
= 0.053. Again
Ramsey is comparing apples and oranges by contrasting the p of his
incorrect
hypergeometric with BB&N's binomial. In his attempt to raise the
p level,he
incorrectly plants the impression that the p level of the hypergeometric
is
what determines the significance. Here Ramsey went overboard the other
way. If
the h(x) is .206996006, then the expected occurances by chance in 180
trials
would be 37.2592811. Contrary to Ramsey's summary of BB&N's analysis,
one does
not obtain the binomial probability simply by multiplying the h(x)
by the
number of trials. That may be an approximation if the E(x)<x. But
by using a
standard binomial test for that data, the actual probability of only
one match
is 3.47957775 x 10-17 . Ramsey suggests using the Poisson
Test, but that may be
less accurate than a simple binomial test when the p is so large. For
the other
calculations where the p is small and n is large, the Poisson may be
a good
approximation. It is often used as an approximation, because calculations
are
done more easily than by the binomial. As an analogy, suppose you have
180
coins whose bias is known that approximately 37 heads will come up
when they
are all tossed. Then suppose that on the next trial only 1 head comes
up. Given
the expected value, l=37, the probability
of getting only 1 head is 3.15730616
x l0-15 . As it turns out, calculation is relatively easy
using the binomial,
because x=1 in all these cases so that nCr=n. This means that the calculation
of the combinations does not involve any factorials or logs of them.
Usually,
the variance should equal the mean to use the Poisson.
I believe that the actual situation
is h(8;11,12,90)= 9.0512911 x 10-7 .
Then the chance of 1 success in 81 trials= 7.3310083 x 10-5 .
That value of p
makes me 99.99% confident that the DPD pattern matches the predicted
grassy
knoll shot pattern, and given the fact that other shots were identified
by the
same method before and after this impulse convinces me that the DPD
pattern is
a shot from the grassy knoll.
As Ramsey pointed out, there are some
places in BB&N's report where the
wording is misleading. He missed some however and should have been
more
specific. For example, on p. 75, the following corrections should have
been
made: in the first paragraph, the first sentence should read, " 12
of the 22
predicted echoes would be..."; in the next sentence, "...that were
judged loud
enough..."; in the last paragraph, starting at the second sentence,
"We
observed that they obtained 10 matches, within ±1ms., out of
12 predicted
echoes, with 14 DPD tape impulses in a 370 ms. time span. We
note, however,
that the 12 predicted echoes were contained in two time intervals of
approximately 90 ms. each, for a total duration of 180 ms. These two
time
intervals were separated by a span of about 190 ms. in which no echoes
louder
than background noise appeared. Because an echo was counted if it occurred
within a 2 ms. time window, there were 90 possible windows in which
echoes
might have occurred, ignoring the 95 windows between the two time intervals.";
and on p. 76 BB&N should not say 9, or more out of 12. Instead
of pointing out
this error, Ramsey copied it. The calculation of the hypergeometric
is a simple
one, not cumulative. Only 10 matches were made, not 11 or 12. And the
number of
trials should be 81 as I pointed out previously.
Even though W&A's match between a DPD
impulse pattern and their predicted
grassy knoll shot was significant, Ramsey still criticized the conclusion
that
there was evidence that it was actually a shot. He claims that W&A
and BB&N
failed to consider alternative sources, but gave no examples. I have
always
thought that one of the basic tenets of science is that one cannot
attempt to
disprove a contention without offering an alternative explanation.
That would
be like saying that you think Einstein's Special Theory of Relativity
is wrong,
because other things could account for the observed phenomena which
it
accurately predicted, but offer no alternative theory or examples.
That's not a
scientific argument; it is demagoguery. In the case of possible shots
on ch. 1,
there are several obvious alternative sources. Ramsey dare not specify
or
consider them, because all of them can be easily disproven. I offer
some now
merely as strawmen to show why Ramsey couldn't risk specifying. If
static was
the source, matches should be found at other portions of the tape,
yet none
were found elsewhere and 15 were found within a 9 sec. portion of the
tape. No
attempt was made by Ramsey to show that any portion of the tape contained
static that was not random and then no explanation for the non-randomness,
such
as sunspots, RF interference from the Dallas Electric radios, natural
electrical discharge like lightning or static electricity, periodic
discharge
such as the motorcycle's ignition, or nearby cars without suppression
spark
plug wires, is offered, because there is no evidence that such sources
occurred
and were recorded on the tape, especially during the critical 9 secs.
However,
there are examples of mechanical sources, such as the fact that the
mike had
stuck open before the well-known 5 min. portion and occassionally after,
but
never has Ramsey tried to prove that the action of turning on or off
the
switch, or even the headlight or turn signal or emgency lights, or
the siren
could mimic a gun shot pattern. The mike turning on or off while other
radios
were trying to use the same channel should display a heterodyne( interference
tone ) of a specific pitch and the actual noise of the switch keying
in takes
less than .1 sec., whereas the detected shots had no heterodyne, were
recorded
while the mike was continuously open, and lasted for almost a half
second. Nor
did Ramsey attempt to compare the decibel levels of the keying in to
the shots.
Another mechanical event which Ramsey overlooked was the jarring of
the mike,
which should not be considered unusual given the fact that the mike
was on a
moving motorcycle, a fact that can be confirmed by the presence of
exhaus noise
and squealling tires while rounding corners. These may be what Bowles
calls the
"bonk-bonk" sounds heard near the time of the shots. For years I had
thought
these might be the shots, but at the closest time to the BB&N detected
shots,
there are only two of these bonks, spaced about a second apart, so
that if
these were typical shot sounds, and may be loud enough, there would
only be 2
shots to the assassination, separated by about 1 second. This definitely
does
not fit the known ballistics, medical, photographic, or earwitness
testimony.
JFK is clearly wounded by Zapruder frame 225 and struck in the head
at frame
313, a difference of about 4.8 secs. Any detected shots must
have at least
that separation. Some examples of jarring which Ramsey should have
tested are:
hitting the mike, running the bike over a bump, into a curb, etc.,
and kicking
the side while shifting gears or putting up the kick stand. These are
mechanical sources that also should be detected as impulsive sources,
but could
they mimic a shot? And how likely would it be that 5 such events occurred
only
within the 9sec.section and nowhere else.
Ramsey dare not even consider alternative
acoustical sources, lest he fall
into the same trap that the FBI did. Once you start with the premise
that the
sounds are impulsive, it doesn't matter what you chose as the initiator,
the
matching test will still identify the location of the source and the
open mike.
Then you can just argue what type of source could have made such a
loud sound
with echoes coming in after .5 sec. at 100 decibels. The FBI tried
to suggest a
stick hitting concrete, but when have you ever heard a stick hitting
concrete
inside a building as loud as 130 decibels? And if you try to use that
alternative source for all 5 shots, you are left with the improbable
conclusion
that professional assassins tried to kill the President of the United
States by
slapping sticks against concrete walls. The only sources capable of
producing
the high volume levels are explosive, so it really doesn't matter what
you call
the source, but that it's very loud and it's location is identifiable.
It could
be any type of firearm, even firing blanks, or even a loud firecracker.
But
again, the BB&N method would work and you'd identify the source
as a loud
explosive impulse. In the case of the TSBD, there is no physical evidence
of
anything other than a gun, and loads of physical and eyewitness evidence
of the
gun. On the grassy knoll, there is no evidence of anything other than
a gun,
and some circumstantial evidence that if there was a loud explosive
sound from
there, that it was due to a gun. There is no photographic evidence
of any other
source such as a firecracker or car exhaust backfire, but there is
photographic
evidence of a gun at that location. There is also the speculative
identification of a shockwave which would have to preceed the blast
of a loud
explosive sound. As far as I know, there are no supersonic blanks,
and there is
no evidence that a high explosive like TNT or RDX was set off above
the fence.
Other loud impulses such as jet plane sonic booms or thunderclap are
absurd
explanations. These few examples show why Ramsey dare not offer alternative
explanations.
Ramsey's improper methods not only impede
a proper investigation of the
assassination, but they also pose a danger to traditional science.
Take, for
example, his assertion that just because BB&N were able to reject
the null
hypothesis, that doesn't mean that their alternative hypothesis was
correct.
Ramsey introduces unnamed alternatives as equally likely causes. By
that same
reasoning, all statistical tests and analyses of experiments are invalid,
because the results may be due to some "magical" alternative sources.
Thus no
medical firm or researcher could claim success with a new drug or treatment
simply because the null hypothesis was rejected at a highly significant
level.
The basic defiition of success of an experiment would be invalid. Progress
would come to a complete standstill. I can understand why experts on
statistics
and acoustics are not interested in the assassination of a political
figure,
but they should rise up in anger when their science is degraded by
misuse.
Corrections to "Rebuttal to Ramsey"
Page 8: All calculations of the Runs Test were
incorrect. In each case,
the value for r was keyed into the calculator instead of the value
for N. I
have since expanded my computer's statistics program to include the
Runs Test.
Using the same data and tests, the results calculated by the computer
are as
follows: For n1= 22, n2= 163, N =185, r = 40,
z = .25964663 using Langley's version of the
Wald-Wolfowitz Test with Yates correction, whereas z = .0823970838
using the
simpler version found in Freund.
For n1= 22, n2= 68, N = 90, r = 40, z = 1.8024943
using Langley's version of the Wald-Wolfowitz
Test with Yates correction, whereas z = 1.65842282 using the simpler
version
found in Freund.
In order to reect the null hypothesis, that
the pattern is random, at the
5% level, the z score must be ³ 1.96.
In none of these calculations did the z
ever exceed 1.96, therefore there is no statistical evidence that the
DPD
impulse pattern is anything but random.
Although this result is merely peripheral
to the issue of the match
between the W&A predicted pattern and the DPD pattern, it does
confirm my
suspicion that prediction is much more important than randomness.
Addendum, Sept.1983: The
national security establishment
should also be alarmed that the work of this country's top acoustical
experts was dismissed so cavalierly. In addition to having analyzed
the Watergate tapes and proven who fired which shots in the Kent
State assacre, Bolt,Beranek and Newman is an important defense
contractor, especially to the Navy. If their basic understanding
of
science is as flawed as Ramsey suggests the implications are
frightening. In that case, they couldn't tell a Russian sub from
a
whale by sonar. No intercepted code could be decrypted, because
the
noise reducing or filtering algorithms would be suspect. As for
the
validity of Ramsey's work, it would be a travesty if someone
were
convicted on the basis of the same type of voiceprint analysis
he
did. If that work is accepted as a precedent for the growing
field of
speech recognitionthere will be a lot of dangerous and costly
misidentifications leading to cold leads and needless brinksmanship.
Addendum, Dec.1983:
Another piece of unanalyzed evidence which
I have recently come across is the live broadcast by KBOX at
the
Dallas Trade art. As reporter Ron Jenkins breaks in, he states
quite
clearly that he had just heard the broadcast about Industrial
Boulevard over the radio. Then the motorcade goes past; then
he and
the DPD motorcycles stationed at the Trade Mart join the race
to
Parkland Hospital. It is obvious from listening to the KBOX tape
that
none of the DPD cyclists at the Trade Mart had turned on their
sirens
until several seconds after the motorcade passed them. Thus it
is
impossible that the open microphone was at the Trade Mart, as
suggested by some critics of the HSCA acoustical studies. Sirens
are
clearly heard on Ch.1 at 12:32:56 ( 262.6 seconds after the mike
stuck open ) and on on Ch.2 at 12:32:37.3 (132.7 seconds after
Curry's "triple underpass" message). But the first mention of
Industrial Boulevard on Ch.2 was almost two minutes later, from
Unit#15-car 2, Capt. J.N. (Red) Souter's order to the dispatcher
to
have Unit#283 cut the traffic at Hines and Industrial followed
immediately by the dispatcher's order to 283 to cut traffic,
Hines
and Industrial. Furthermore, the patterns, types, and sheer numbers
of sirens you can hear at the Trade Mart joining the motorcade
sound
much different from the patterns, types, and much fewer numbers
on
Ch.l. This is proof that the sirens heard on Ch.1 were not at
the
Trade Mart. Therefore, they were part of the motorcade.
† I have since learned that one of the lead cyclists
named E. D. Brewer was
known by the nickname Buddy. So it may well have been McLain yelling
over the open microphone to his buddy to take off.
BIBLIOGRAPHY
U.S. House of Representatives, Hearings before the Select Committee
on
Assassinations (Investigation of the Assassination of President
John F.
Kennedy), 95th Congress, 2nd session, U.S. Government Printing Office,
Washington, D.C., 1979, Stock No. 052-070-04979-1
Ramsey, Prof. Norman et al., Report of the Committee on Ballistic
Acoustics,
National Research Council, Commission on Physical Sciences,
Mathematics and
Resources, National Academy Press, Washington, D.C., 1982
Adler, Henry L. and Roessler, Edward B., Introduction to Probability
and
Statistics, 5th edition, W.H. Freeman and Company, San Francisco,
1972
Freund, John E., Modern EIementary Statistics, 3rd edition, Prentice-Hall,
Inc., Englewood Cliffs, New Jersey, 1971
Hodges, J.L., Jr., and Lehman, E.L., Basic Concepts of Probability
and
Statistics, 2nd edition, Holden-Day, Inc., San Francisco, 1970
Hoel, Paul G., Introduction to Mathematical Statistics, 4th edition,
John Wiley
& Sons, New York, 1971
Koosis,Donald J., Statistics, A Self-Teaching Guide, 2nd edition,
John Wiley
& Sons, New York, 1977
Langley, Russell, Practical Statistics for Non-Mathematical People,
Drake
Publishers, Inc., New York 1971
Lindgren, B.W. and McElrath, G.W., Introduction to Probability and
Statistics,
3rd edition, The Macmillan Company, London, 1969
Naiman, Arnold et al., Understanding Statistics, 2nd edition,
McGraw-Hill, New
York
Maier, B.R. et al.,"Phonograph Playback", Popular Electronics,
November 1980,
pp.48-52
Schadewald, Robert, "The Speech Gap", Technology Illustrated,
June 1983, pp.
55-59