WHY COMPUTER ANIMATIONS CANNOT "TRUMP" HARD EVIDENCE Some initial comments John P. Costella, Ph.D. Friday, 31 October 2003 The story released by the Associated Press describing the ABC network's upcoming special on the JFK assassination -- which apparently argues that the Warren Commission was right all along -- reveals that the "breakthough" which forms the cornerstone of the programme is the computer animation work of Dale Myers. We will have to wait until 20 November to see how many impressively slick animations will be used in the broadcast. But they will be convincing -- the viewer will feel like they are really there in Dealey Plaza, Dallas, on 22 November 1963. The quality should be better than Toy Story II. And indeed, animations like this were simply beyond computer technology the last time a major anniversary rolled around, ten years ago. But does this represent any sort of scientific breakthrough? It does not. Computer animation is not a new scientific, experimental technique, like DNA testing. The ability to create stunning graphics is recent, but it relies on the same laws of physics that were fully understood in 1963. It can allow researchers in many disciplines -- including forensic science -- to more rapidly, and more powerfully, explore the three-dimensional space of a reconstructed scene or object. It is a powerful tool, immensely helpful to the scientist and engineer. But it does not change the underlying science. Researchers of the JFK assassination have generally not had the use of such a tool. They have spent the best part of 40 years examining the extant films and photographs of the assassination in microscopic detail. To get a feel of the three-dimensional layout of Dealey Plaza, they have had to travel to Dallas and stand there for themselves. All of this would have been so much easier if they had had access to a precise virtual reality model of Dealey Plaza. But these people have put in the uncountable number of collective hours necessary to do it "the hard way". Bulking out an investigative report into the assassination with computer animations is the ultimate in "form over substance". To suggest that it lends greater credibility to the programme is akin to believing that a newspaper column can be made more credible by dressing it up in a nicer font. It may make the material easier to digest, and of greater "entertainment" value. But it also provides the means to deceive. Deceive? How can a visualisation tool deceive? As Jim Fetzer and others have astutely observed, a fundamental axiom of computer science is "GI-GO", namely, "Garbage In - Garbage Out". Any simulation, any virtual world, is a product of the information that its human creator has endowed it with. Small errors in this foundation information -- whether unconsciously or deliberately introduced by the creator's bias, or due to genuine uncertainty about the physical object being modelled -- can make huge differences if the virtual world is explored in sufficiently fine detail. Unfortunately, the excitement of immersing oneself in such a virtual world can cause one to overlook, or forget, these inherent uncertainties. For the incautious, the virtual reality can begin to take the place of "real reality" -- the extreme case being the belief that discoveries can be made in the virtual world that cannot be made in the real world that formed its basis. For sure, the experience can be more than the sum of its parts -- but its scientific content cannot be. How is this relevant to the ABC programme on the JFK assassination? Let's look at the story emerging publicly so far. We are told that these new computer animations "prove" that the Warren Commission got it right, and that Oswald acted alone. Clearly, that's an overstatement. The most that could possible be "proved", as a point of logic, would be that the Warren Commission's conclusions are CONSISTENT with the available evidence. Imagine that even that was true. You could put another gunman up on the grassy knoll, and have his shots miss, or his rifle misfire, and the physical evidence would not change. You could have Oswald accomplices assisting him in his activities after leaving the Book Depository, and it would not change the extant physical evidence. Or you could have someone else altogether firing from the Oswald window, while Oswald was eating his lunch on the second floor. "Proving" that Oswald acted alone is clearly ludicrous. So could these animations prove that the Warren Commission's conclusions are at least CONSISTENT with the physical evidence? Even a layman learning about the assassination for the first time will realise that, if such a "proof" is now needed, then clearly the physical evidence has NOT been consistent with the Warren Commission up to now. Can these new animations change those scientific results? Let's look at what all of these animations are based on. Firstly, the physical, three-dimensional layout of Dealey Plaza, Dallas, must be programmed into the system. Where does this data come from? Well, there were a number of surveys made of the Plaza following the assassination, and in the years and decades following. As we move further forward in time, away from November 1963, the uncertainty in our data becomes greater. Why? Because physical features of Dealey Plaza have changed in the intervening years. Street signs and lamp poles have been moved. The road surface has risen some six inches because new asphalt has been laid without removing the old. Concrete pillars and structures, and the buldings themselves, have been repainted and maintained. The grassy knoll itself has even been excavated in order to install an irrigation system. Trees have been planted and removed. The best methodology, then, is to go back to the professional surveys conducted immediately after the assassination. How accurate are these?* The Warren Commission's own Exhibit #595 is a professional survey created by Dallas surveyor Robert West on 5 December 1963, just thirteen days after the assassination. But Commission Exhibits 882 and 883 also represent a survey constructed by Robert West, on 24 May 1964. Do these two surveys agree with each other? At first sight they do -- but if they are carefully overlaid, it is found that the lane markings on Elm Street -- the "road stripes" are different. In the later survey, one set of stripes has been deleted altogether, right where the shots occurred, and the nearby sets of stripes have been stretched and moved to fill the gap. Are the road stripes important? In fact, they are, because the only professional photograph of the Presidential limousine taken while the shots were being fired shows the front wheel of the limousine right next to one of these stripes. The photo, taken by Associated Press photographer James Altgens, is crystal clear, and taken with a telephoto lens on professional equipment and film. There are numerous structures and features of Dealey Plaza that are visible. Given the clarity of the image, it should have been a straightforward task for the Warren Commission to determine exactly where both Altgens and the limousine were located at the instant the photo was taken. Indeed, Commission Exhibit 900 seems to represent a successful reconstruction of the layout (albeit with a vehicle that is not the Presidential limousine). But, again, if one overlays the two images in CE-900, one finds that the reconstruction is not perfect at all. In fact, it is terrible! (To see this most starkly, play the video clip at http://www.ph.unimelb.edu.au/~jpc/CE-900.mpg with playback set to 'repeat'. Note that the background objects are very nicely aligned, but the location of the limousine -- and the road surface and road stripe -- is grossly incorrect.) But does this inaccuracy really matter? How much of a difference does this make? If we were only worrying about pasting in an image of Clint Eastwood for the movie In The Line Of Fire, it would be good enough. But as a scientific reconstruction, it is dismal. Calculations show that Altgens' position would have to be incorrect by as much as TWENTY FEET to cause such a massive inaccuracy in the reconstruction. But is this degree of inaccuracy an isolated case? Let's go back to Commission Exhibit 883, the May 1964 survey by West. The data table on that survey was so small that it was introduced separately into evidence, as Commission Exhibit 884. This data table relates measurement points to frames of the Zapruder film of the assassination. But three of the frame numbers have been obviously, and carelessly, modified; and West's own field notes confirm that the frame numbers had been later changed. How could this be? The Zapruder frames had been counted off, selected frames enlarged and provided to West for the re-enactment, and used as key reference points. In changing Frame 168 to Frame 161, Frame 171 to Frame 166, and Frame 208 to Frame 210, the Warren Commission is telling us that the re-enactment and surveying process was inaccurate to the tune of seven frames. How much can things change in seven frames? Well, the limousine moves around six feet in that time. Whether it was the re-enactment or the survey that they considered inaccurate, we don't know. They never informed surveyor West of the changes -- nor, of course, any readers of the Report or the 26 volumes of Exhibits. We seem to be building up an idea of how unaccurate these crucial early surveys and reconstructions were: twenty feet, six feet. But this almost pales into insignificance compared to the measurements of the locations of the shots, also contained on these survey maps. Locating the shots is something that has not been agreed on in 40 years -- excepting, of course, the fatal shot that blew the President's head open, which clearly occurs at Frame 313 of the Zapruder film. So let's only concentrate on that one. Where is it located on the survey maps? Commission Exhibit 595, the 5 December survey, places it about 40 feet too far down the street. Commission Exhibit 875 refers to this shot occurring only four feet from a location marked on CE-595, that is about 35 feet down the street from Frame 313. Clearly, the survey and re-enactment data that is of greatest importance -- that obtained before changes began to occur to Dealey Plaza -- is only accurate to feet, or tens or feet, at best -- at least when they were matched up to the frames of the Zapruder film, or the Altgens photograph. Let's turn, now, to the Zapruder film. Let's assume that we can overcome these uncertainties in the fundamental data, and map out the trajectory of the Presidential limousine as it passes down Elm Street. A major obstacle to this data input process is literally that -- a visual obstacle. The Stemmons Freeway sign blocks our view of the occupants of the limousine at a crucial time. Worse, there are no other motion picture films extant which show the occupants in this time period. What, then, can we do? It will be interesting to see whether Myers has dealt with this problem honestly in his work for ABC. What does "scientific honesty" mean in this context? It means recognising the simple fact that there IS no data on the occupants while they are obscured by the sign. The simulation of the occupants should simply go black for this time interval. The limousine can still be animated -- we always see enough of it to know where it is. But the occupants should disappear. One way to sidestep this insurmountable obstacle is to simply INTERPOLATE the motions of the occupants. This simply means that one takes their last observed position before going behind the sign, and their first observed position after emerging, and joins the two with the simplest smooth trajectory that goes from 'A' to 'B'. Is there anything wrong in this? Absolutely! Imagine you went to a football game, and stood behind a sign during a minute of play. You then apply this trick: join each player up with smooth trajectories. What would it look like? Every player would drift slowly across the ground, from their starting position to their ending position. No football, no animation, no actions. Just a lot of ghosts drifting like clouds. In the context of the assassination, of course, we're talking about a fraction of a second, not a minute. Doesn't that make an interpolation reasonable? Well, if the occupants of the car were just driving down the road, enjoying the sunny day, it probably would be fine. But here we are talking about the very interval of time in which the President obviously reacts to a shot to his throat! A smooth interpolation will be absolutely misleading. And what would the result be? Naturally, as soon as the "real" action took over -- namely, that based on actual Zapruder footage -- the occupants will suddenly become very animated! In the football example, it is obvious what is real and what is a simulation artefact. But in the context of the assassination, this misapplication of science generates false "evidence" -- namely, it makes it appear as if both men react to a shot at the same time, as they come out from behind the sign. Anyway, let's wait for the 20 November programme. Surely Myers will be scientifically honest with us, and will have blacked out the occupants for those crucial fractions of a second. Let's now turn to the idea of tracing a trajectory. The idea is this: if we can locate two or more points on the roughly straight-line trajectory of a rifle bullet, we can trace that line back, and find out where it came from. The second part of this is just simple mathematics; it's the "if" that contains all the problems! The Zapruder film does not, itself, show the location of any entrance or exit wounds. If it did, this work would be much easier! Thus, we first have to try to determine WHEN shots impacted. Then we have to figure out WHERE they impacted on each person. Together with the motion of the limousine (if we can get that right), we can then locate these positions in the three-dimensional space of Dealey Plaza. Constructing the line through them is then just a mathematical problem. Determining either the WHEN or the WHERE is fraught with problems. Let's consider the WHEN. That's tied to the Zapruder film -- the only "time clock" we have of the assassination at all. The CIA had a go at this, at their National Photographic Interpretation Center, and came up with a half- dozen possible scenarios. The FBI had their own opinion. Governor Connally had yet another. Arlen Specter tied a couple of WHEN's together, but the Warren Commission ultimately didn't tie down just WHEN this magic-single-WHEN occurred. And just about every JFK researcher has their own version of WHEN's. Can the Myers simulations help us with the WHEN's? Hardly. It's based on the very SAME data that all of these OTHER scenarios are based on. If Dale Myers had some other film of the assassination, that the CIA and the FBI and Governor Connally and the rest of the research community had not seen, THEN he'd have a case for showing us something new! But in the absence of that sort of sensational revelation, it is both plainly and scientifically obvious that ANYTHING SHOWN IN THE SIMULATIONS THAT IS NOT SHOWN IN THE FILM IS AN ARTEFACT OF THE SIMULATION. In the computer, we might zoom in on the President's hand, read his palm, and check to see if his life-line was broken prematurely, if the simulation were fancy enough. But of course all that is just additional information input into the simulation -- "filling in the gaps". If it's not there in the fundamental scientific data, then it's not there! Now what about the WHERE? Do we know the exact trajectories of the bullets through the bodies of the President and the Governor? Well, surely the President's bullet wounds were traced through his body at the autopsy, and the bullet track through his brain determined by sectioning the brain. What? The wounds were not traced through the body? The brain is missing? Huh? What -- how? The autopsists didn't agree with the entry wound to the head shown in the photographs? What? There have, of course, been books written on the wounds and the medical evidence. Suffice it to say that it has been established beyond reasonable doubt that the President and the Governor were hit by bullets. Somewhere. But let's imagine that one was to take some small subset of the evidence that happens to be self- consistent, and to reject all the rest. How accurate would we need to be to reconstruct a trajectory? Well, imagine holding one of those small laser-beam pens that teachers and lecturers use these days. Attach it to the end of a "pointer". Now point the spot at something two hundred feet away. Move it around a little. That's exactly what is being done with the simulation! If it's one foot long, and you move one end sideways by an inch, then the end of the beam will move by 16 feet. Yowch! We have to do better than this if we're trying to tie it down to a single window! How accurately do we have the President's and the Governor's positions? Could they be out by an inch, two inches? ... Hmm. Gerald Ford moved the back wound up to the neck, by three or four inches. No -- too much uncertainty! The head wound was measured to an accuracy of four inches, according to the autopsists themselves -- yikes! If we had a track that went through the body, we'd then just have to worry about rotating them properly and lining them up ... but even Arlen Specter couldn't get that to work. The Warren Commission included some sketches -- but they are not scientifically accurate. The angles are all wrong -- that's obvious just from looking at the Zapruder film. And how can we be sure that one wound led to another, if no one bothered to trace the path of the bullet in between? Clearly, if you were to START with the evidence, and see what it tells you, then any scientist would conclude that the uncertainties are so great that the "uncertainty cone" would encompass entire buildings. And this, of course, PRESUPPOSES that the bullets are coming from the rear. The wounds themselves do not provide this information; they point to the opposite, in many cases. What if you were to then introduce into the model the possibility that there were MORE bullets? Clearly, you have many more points to join up, and tracing trajectories is essentially impossible. The only way to avoid this problem is to ASSUME that there were no more bullets. But you can't prove something by simply assuming it to be the case! There will, undoubtedly, be aspects of the Myers animations that I haven't anticipated here. But the general principles are the same: you can't get anything more out of simulation than you put into it. All you can get is a different view of it. John Costella --------- * Much of the information in the following paragraphs comes from the research of Chuck Marler and Tom Purvis.