Was the Zapruder Film Tampered with?

Almost since the Zapruder film emerged as a public record, suspicion has grown that it’s been tampered with. Among the original claims for tampering was one while the film was in the possession of TIME-LIFE. The putative change was in reversing the frames near 312-313 to dilute or negate the backward head snap. Later, the tampering hypothesis evolved to claiming that frames were removed in the vicinity of 311-314 in order to deliberately conflate two shots: one from the rear (TSBD) and one from the right front – striking the temple.

Not surprisingly, variant forms of the tampering hypothesis have made their way into the assassination research community – some with more heft and traction than others. One of the more enticing ones is offered by photo-analyst Mark Crouch, who proposes a subdued forging involving[1]:

“projection frame by frame onto the rear of a glass screen. The corrupted copy is then recorded frame by frame. When they reach the head shot frame they just add a little paint to the glass, shoot the frame, then clean the glass. They then project the next frame, do their little touchups to match perspective, then go on. They only have to do about fifty frames and the alterations are non-descript……it’s called aerial imaging photography”


This sounds like an awful lot of work to go through to obtain only an imperfect film or forgery, but Crouch maintains it was adequate to the task: in framing Lee Harvey Oswald for the killing. As he’s quoted[2]:

“The truth about the Z-film is that it was never meant to be micro-analyzed the way it was…The ‘real evidence’ of the Z-film is more subjective than objective. If you believe- as I do- that a trained and directed team of assassins was in the Plaza, then you must assume they observed Zapruder standing up there with his camera. The assassins would have no way of knowing if he innocently panned his camera toward them just before or during the assassination”

Crouch goes on to assert said assassins would thus have had to be “very concerned with what Zapruder’s camera recorded” …which would have necessitated the attention to manipulating the Z-film – and by extension, confiscating as many others as possible. (Since the Nix film was made from across Elm Street and may also have caught the assassins).

Author Noel Twyman asserts that he has “proven to his satisfaction” that frames were removed between Z-313 and Z-330[3]. He goes on to speculate[4]:

“I wondered what the Nix film would show. I reasoned that if it had not been altered in a similar fashion to the Zapruder film it would show a slowing down or stopping of the limousine if, as I believed, the testimony of eyewitnesses was correct”

These eyewitnesses reported the limo had “come almost to a complete stop at or near the time of the fatal head shot.”

Twyman then records he “obtained a set of slides of the Nix film and made prints on a Canon laser copier”, then proceeded to compute the rate of advance of the limo- thereby obtaining a graph he represents as Exhibit 12-3[5], showing a uniform speed. In other words, using the Nix slides he had he was unable to confirm the witness’ testimony of a near –stop.

I decided to test this using my own version of the Nix film – obtained from the Italian Documentary ‘The Two Kennedys’ (1976). Toward the end of the documentary, the entire Nix film (or what is described as such) is shown a number of times, including in slow motion. Even without the slo-mo shots I was easily able to confirm that the limo did make a near stop – just after the fatal head shot, before accelerating.

A further analysis allowed me to deduce something else: that frames had been manipulated (after Z-313) in the Z-film and other Nix film versions. Note that I didn’t say the frames were removed, at least not in my version[6]. We know that Orville Nix film (as he records in his interview with Mark Lane, ‘Rush to Judgment’) had detected frames removed on the film’s return from the feds, and there have been estimates of up to six different versions circulating. I don’t know how many distinct versions I’ve seen – but I have seen several on the net (You Tube). None of them shows what my version does. This suggests Twyman’s version as well as all the net versions are manipulated forgeries and not the original Nix film. When I alluded to tampering, what I meant is that somehow the manipulators had managed to “wipe out” a small white blob hurtling over the limo trunk – obviously a skull fragment. I suspect this was done using an “aerial photography” technique such as Mark Crouch earlier described, with touch ups done to the rear view glass to remove the blob, and then the frames re-filmed.

Why did this have to be done? Because a fragment detected hurtling over the limo trunk would clearly show a frontal shot and confirm a minimal conspiracy – e.g. at least two shooters, not just one in the Texas School Book Depository. So frames didn’t have to be removed from the Nix film but the incriminating hurtling “blob” did have to go – to excise the evidence for a head shot. (The law of conservation of linear momentum clearly dictates that a backward moving skull fragment couldn’t have been dislodged by a rear shot!)

How is it that I happened to have one of the few original copies of the Nix film? I can’t answer that, but one speculation I’ve heard is that when the CIA recruited Italian mob (or Sicilian) assistance (via Carlo Marcello), their (Mafia) operatives wanted proof of the kill shot to be delivered to them intact. This came via the Nix film, and at least this same version was obtained by the Italian film maker Giovanni Bisach for his (1976) documentary. Maybe he had connections with the Italian Mafia in the film making business, who knows?

According to a computer-enhanced analysis of the (Mary) Muchmore black and white film – by Jack White- at least one assassin was caught in outline on the grassy knoll- aiming his rifle. The flash of a badge discloses he was wearing a police uniform. The interesting aspect here, is that from White’s analysis (see in episode 6 of ‘The Men Who Killed Kennedy’) the shooter was within about 1-2 m of the position assigned using the acoustic testing evidence!

This brings up the issue of whether there is any way to assess the effect of direct frame tampering (or frame removal) on the acoustic impulse patterns. The premise here is, irrespective of what the plotters and their assistants may have done to the Z-film (or any other) it wouldn’t have severely affected or negated the rejection of the null hypothesis for the acoustic impulses – and in particular the p-values associated with the hypergeometric set: H{M,N.n,i}.

In his paper, ('Echo Correlation Analysis and the acoustic Evidence in the Kennedy Assassination Revisited, Science and Justice, Vol. 41, p. 21, 2001) D.B. Thomas treats both the test evidence and actual data from the date. The test evidence was obtained in August, 1978 when a test shot was fired in Dallas’ Dealey Plaza to provide a fiducial mark for the putative Grassy Knoll shot – such that it could be compared with the impulse record obtained on Nov. 22, 1963 and also how this mark lined up to events recorded on the Zapruder film. Thus, the test evidence (mainly in terms of echoes and echo delay times received via an echogram from a test shot- See Fig. 1) is essentially used to confirm the microphone recording & positions for the shots made on the actual date, by resort to microphones placed at the same (or approximately so) locations.

Using Thomas’ geometry (Figure 4) relating to the sound paths for the test and actual date patterns (associated with the muzzle blasts) it’s possible to estimate the effects of removing one or more Z-frames prior to Z-313, as well as to compute any altered p-value for an altered H{M,N.n,i}*.

For the test shot, using his sonar model and the muzzle velocity for a Norma 6.5mm Mannlicher-Carcano (which is very doubtful that the GK shooter used), one arrives at a 171 msec time for the shock wave to take the defined path to the microphone, from a distance of 28.3m. One must add the time taken for the putative bullet to travel the distance and add to that the time for the shock wave. In this case: 126 msec + 45 msec. The first parameter is relatively fixed since the distance from bullet to test microphone is fixed at 44.2m and the speed of sound defined for the sonar model was 350 m/s.

The second component will alter depending on the muzzle velocity. A German Mauser, with muzzle velocity » 760 m/s will traverse the 28.3 m distance in 37 msec, so that the time will now be: 126 msec + 37 msec or 163 msec, which is some 8 msec shorter. If the rifle used is now a Remington Fireball with muzzle velocity 825 m/s then the total time becomes: t1 + t2 = 126 msec + 34 msec = 160 msec, or 3 msec shorter still.

In the “analytically determined” schema (e.g. actual data obtained on 11/22/63), all the above values change slightly. For example, the distance to Kennedy (from the GK shooter) becomes 30.5 m, and the speed of sound is 342 m/s given the air temperature was 18 o C at the time of the assassination. The distance from the assassin to the nearest motorcycle was 67 m leading to the muzzle blast arriving at the motorcycle some Delta t = [67m/ 342 m/s] = 196 msec, after the shot. When air resistance is corrected for by + 11.5% (from the shooter location) the muzzle velocity resulting becomes 748 m/s (using a starting assumed bullet speed of 672 m/s). Because of a shooter “location uncertainty” of +/- 1.5 m the muzzle velocity uncertainty is at least +/- 32m/s, so one is left with a range of: 748 +/- 32 m/s. Assuming Thomas’ parameters are correct, then this excludes a Remington Fireball as a possible candidate weapon, but it does permit either a German Mauser or a .30-30 Winchester.

Using the above analytical values and assumptions (including the sonar model and muzzle velocity, uncertainty) separate computations enabled me to conclude that if as many as three Z-frames were excised – say between Z- 290 and Z-313, then the effect would be to alter the shooter location and uncertainty, to around +/-2.0 m. This would imply for the muzzle velocity: 748 +/- 44 m/s. More critically, it would mean that to compute the p-value one would need to subtract one degree of freedom– to re-evaluate the {N,n,i} set. The reason is that now we have a situation where the shooter is essentially allowed to “move” on account of the frame manipulation. This “motion” actually translates into one lost df, as per the original NRC deductive approach.

Thus, we need to make the change from Thomas' correct p -set: {86, 10, 8, 6} to {86, 10, 8, 5} and re-calculate the hypergeometric p. We obtain:

p= 0.000333

= 3.3 x 10^-4 or about 3 in 10,000 odds against the null hypothesis.

The conclusion here is that even gross tampering with the Z-film (at least 3 frames removed between Z-290- 313) in terms or eliminating frames doesn’t significantly impact the acoustic record or the rejection of the null hypothesis.



[1] Harrison Edward Livingstone: 1995, Killing Kennedy and the Hoax of the Century, Carroll & Graf Publishers, p. 160.
[2] Livingstone, op. cit., p. 140. Crouch also goes into extensive detail on the following page to do with how minor touchups” could have conceal the extent of head damage and also “conceal the evidence of a bullet striking the forehead at the hairline”. All this would be to assist in reinforcing the perception of Oswald as the perpetrator – while hiding the key evidence that the kill shot came from the right front.
[3] Noel H. Twyman: 1997, Bloody Treason, Laurel Publishing, p. 149.
[4] Twyman, ibid.
[5] Twyman, op. cit., p. 150.
[6] Clearly, however, frames would have to be removed from the Zapruder film – maybe five or six after Z-313, to eliminate a near stop. In this sense, I do concur with Twyman.

* Note: Thomas' parameters for his p-function had to be changed since his formula as shown doesn’t work, because the “population size” (which he assigns as M) is in the wrong position (as are several other parameters). His population size for M: the number of “windows” – spanning 190msec total time, at 2msec width per window; the “sample size” n is the number of spikes (assigned impulses) in the “evidence pattern” (taken from the Dallas PD microphone tapes on 11/22/63, and N is the number of spikes in the test pattern, based on test firing from the Grassy Knoll in 1978), while i denotes the number of “coincident impulses” – i.e. those which match from the test and evidence patterns. My p-function correction of Thomas' is shown in Fig. 1 (upper left).