The Stemmons Freeway Sign and the Lamppost


On Wednesday, November 27, 1963, Pete Barnes of the Dallas Police Crime Lab visited Dealey Plaza and took a series of photos with a Speed Graphic camera. He took a couple of his photos from the same pedestal Abraham Zapruder filmed the assassination from five days earlier. One of these photos is on page 60 of Gary Savage's book, FIrst Day Evidence. In his article, A Scientist's Verdict: The Film is a Fabrication, John Costella compares the Barnes photo to several Zapruder frames and also compares the Zapruder frames to each other. He declares the film a fabrication based on discrepancies he observes between the photo and the film frames.

We demonstrate below that these discrepancies are nothing more than the effects of perspective. In each case, the effect Costella observes is expected given the changes in camera position and pointing orientation.


The Sign Width Discrepancy

The first discrepancy to consider is the apparent difference between the Stemmons Freeway sign as seen in the Zapruder film and as seen in the Barnes photo. Since the photographers did not point their cameras in the same direction, Costella wrote software to manipulate each image to, as he claims, transform the perspective of each image so that the perspective is the same between them. He then aligned the backgrounds to produce the following overlay. The difference in the signs is clearly evident:

The most obvious discrepancy is the shift in sign position. The sign in the Barnes photo is to the right of the sign in the Zapruder frame. Costella points out this could be due to parallax, that is, Barnes could have taken his photo from a different place on the pedestal than from the place where Zapruder filmed. In that case, objects closer to the camera would shift more than those farther away. But, Costella claims, "most damning, rather, is the fact that it is too wide in the extant film, and that the angles of its edges are wrong."

Costella is wrong about this being a "damning" inconsistency. He does not fully appreciate the effects on perspective brought about by the change in camera position and orientation.


Perspective

Railroad tracks are parallel, yet when we look down them, they appear to converge in the distance. This effect is perspective and is the result light rays converging on our position. Or, stated another way, our field of view increases with distance and though the tracks maintain the same spacing with distance, they take up less of the widening field of view, causing them to appear closer and closer together with distance.

The great paradox of perspective is that we are constantly immersed in it, yet we often don't notice it. Sometimes we are surprised that the odd angles of, say, building edges in a photograph are not due to a distorting effect of the camera lens. Perhaps the most common mistake assassination researchers make in analyzing photographs is to underestimate the effects of perspective.


Where Pete Barnes Stood on the Pedestal

The shift in the sign's position made us suspect that Barnes stood at the rear of the pedestal. One of his other photos shows flowers resting on the front of the pedestal, leaving the pedestal rear as the only place to stand. When this issue first arose in the fall of 2002, Gary Mack mentioned there were two photographs showing Barnes on the pedestal taking pictures. These two photos, on the left below, are from the Dallas Times-Herald collection. Zapruder as captured in the Betzner photo is on the right for comparison.


Illustrating Perspective with a 3D Model

Given that Barnes was at the rear of the pedestal, we wondered what effects this would have on the apparent width and shape of the sign. To illustrate the effects of perspective, we built a simple 3D model in POV-Ray, a public-domain raytracing program. POV-Ray allowed us to position, rotate, and tilt the virtual camera so that it had the same perspective as the Barnes and Zapruder cameras had.

Wherever possible, we used the HSCA topographic map to locate Dealey Plaza features. The exact size, position, and orientation of the Stemmons Sign is not known, so we used a rectangular 8-foot wide by 4-foot high perfectly flat board about 51 feet from the pedestal to represent the Stemmons sign. Since Costella's claim is that particular effects such as width or sign edge angles cannot be due to perspective, our simple model is adequate for our demonstrations. We stress, though, that we do not offer this model as an exact model of features in Dealey Plaza. As such, our analysis here is qualitative rather than quantitative.

The graphic below shows the model as viewed with the virtual camera positioned at the rear of the pedestal. To accelerate rendering, we used green "sticks" to represent the edges of prominent features such as buildings on Houston, etc. that we used for alignment. After matching the Barnes field of view, we panned, tilted, and rotated, the camera until it pointed in the same direction the Barnes camera pointed. To within a small degree of error, this provided the same perspective that Barnes had, as evidenced by the similar angles of the green lines and the diverging building edge lines:

Note that John Costella used a different approach to check the perspective between photos. He wrote software to manipulate each image to remove perspective effects, e.g., after processing the Barnes photo and a Zapruder frame, say, the County Records building in each would look as if the camera were level for each. Supposedly, then, he could directly compare the Barnes photo to the Zapruder frame.

For this technique to work, though, the Zapruder and Barnes cameras would have to have been in the same location and Costella would have to make sure to account for all perspective effects. Technically, our method is more correct because in matching the camera viewpoints between a model and the image, we necessarily account for all perspective effects. Also, our method has the advantage that comparisons are made to the original photo and frames, unchanged except for lens distortion correction. Costella compares images which he further manipulated in different ways, possibly introducing discrepancies that are merely artifacts of his processing algorithms or his chosen parameters.


The Sign Width Discrepancy Explained

To illustrate how the sign can change apparent position and width between the two camera positions, we rendered an image with the virtual camera at the front of our model pedestal and again from the rear. The camera lens positions on the pedestal we chose for this demonstration are:

And here is a comparison of the two resulting images. On the left is an overlay of the two images with the column in the background aligned. The foreground shows the expected shift in sign position due to parallax. On the right, the signs are aligned to highlight the difference in apparent sign widths, with the blue Barnes sign clearly smaller than the brown Zapruder sign.

Why is the Barnes sign smaller? In this case, primarily because the rear of the pedestal happens to be farther from the sign than the front of the pedestal is. Perspective dictates that objects appear smaller with increasing distance. Other factors can affect the size of the sign such as the fact the sign is viewed from a different angle at the Barnes position where the sign presents less width to the camera. A comparison of the frame 193 overlay and the Barnes overlay shows that the change in sign position and dimensions correspond very well to the photographic images. That Barnes took his photo from the rear of the pedestal explains the shift in sign position, the change in sign size, and the change in sign edge angles.

Note that the Zapruder frames used here were corrected for aspect ratio and lens distortion by David Wimp.


The "Flip-Flopping" Sign Pole Discrepancy

The angle of the right sign post appears to change relative to the background between Zapruder frames 193 and 228. Costella highlights this change in the following graphic. Note that the book contains a typo in that the frame numbers of the images are reversed. We switched the images here to match the order of the labels, with frame 193 on the left and frame 228 on the right:

Notice that the wall corner in the background is vertical in Costella's perspective-transformed frames, but the angles of the sign pole and right sign edge are different. Costella claims that it is impossible for this to be a genuine effect of the film. Well, we added a right post to our model sign, assuming it to be perfectly parallel to the sign's right edge and rendered two images, matching the camera view for each frame. The camera positions differ by about 2.7 inches. Here are the results overlaid on the Zapruder frames:

The changes in the model sign pole and sign edge angles correspond to the changes seen in the Zapruder frames. Contrary to Costella's insistence that this is impossible, this is exactly the change we should expect to see.


The Pincushion Distortion Test

As a test of his hypothesis that the sign was pasted in, Costella checked the sign edges in the raw frames for the characteristic distortion inherent in the Zapruder lens, in this case, pincushion distortion. Such lens distortion causes straight edges in the scene to appear curved in the image. Because the right sign edge is at opposite sides of frames 193 and 228, the distortion effect should bend the edge in opposite directions.

Costella claims that the sign in frames 193 and 228 is "so beautiful and rectangular" and has a "constant and consistent shape," and thus supposedly verifies his fabrication hypothesis. He boasts, "No evidence of the fraudulence of the 'Zapruder' film is more unequivocal than this."

Our flip-flop graphic above shows that our model sign matches the distortion-corrected frames, demonstrating the signs are pincushion distorted in the extant frames. For another demonstration, we took the model signs produced for 193 and 228, applied a pincushion distortion transformation to them, and compared them to the raw Zapruder frames. In all cases, the signs in the Zapruder frames distort the same way as the distorted-model signs do. The shapes of the signs in the extant frames, in other words, are exactly what we should expect to see.

To see an animated comparison between frame 228 and a distortion-corrected version of it, see
David Wimp's distortion demo.

(Note that other photos seem to show that the left side of the Stemmons sign is twisted or tilted forward a bit relative to the straighter left side. Our model sign is a flat board and does not include any such twist. Thus, in the 193 Left graphic above, the angles of the edges in the model and the real sign differ slightly, but for the purposes of the pincushion distortion test discussed here, they both exhibit the same fairly straight shape.)


The Lamppost Discrepancy

In comparing his Zapruder panorama to his 1963 panorama, Costella found his Barnes lamppost differed from his Zapruder lamppost by 1.5 degrees.

We modeled the light pole by first referring to the HSCA map and then adjusting it its position to match the lamppost in the Barnes photo. We had to move the lamppost east on Elm 5.7 inches and then tilt it westward about 0.8 degrees to get a good match. A view that matches Zapruder's frame 263 shows the model lamppost at the same angle as the lamppost in the Zapruder frame:

Once again, there is no inconsistency between the Barnes photo and the Zapruder frame.


Putting it all in Perspective

John Costella can't tell us why the sign and the lamppost are the way they are in the Zapruder film. He shows us the results of his sophisticated digital processing, he shows us his extensive panoramas, but he can't tell us how the apparently different signs in the Barnes photo and Zapruder frames came to be. Following Jim Fetzer's lead, he offers us only doubt, mistrust, and some vague hints of a massive conspiracy lurking in the shadows.

The article's title, "A Scientist's Verdict," flags it as being done in the spirit of science. We think, then, that such an article should not leave the reader in the dark. We provide an explanation for why the sign and lamppost appear the way they do: Abraham Zapruder stood near the front of the pedestal and filmed the motorcade, moving his camera to the right as he tracked the limo. Several days later, Pete Barnes climbed up on the rear of the pedestal, on the only portion free of floral displays, and snapped a few photos. That's it. Everything else is just the natural consequence of perspective and of the effects of the camera lenses.


Resources

sign.pov POV-Ray 3.5 source for sign and lamppost model
sign31.pov POV-Ray 3.1 source for sign and lamppost model
sign.ini POV-Ray .INI file
www.povray.org Download POV-Ray
Distortion correction David Wimp's writeup on lens distortion

The following photos show the Stemmons Sign from different angles. The two posts are not at the same angle. In terms of east-west tilt, the south post is fairly straight, but the north one tilts eastward, apparently twisting that side of the sign. Our model uses a flat board as a reference although we did tilt it southward 2.4 degrees to better match the sign in the photos and frames.

signrickerby.jpg Rickerby, November 22, 1963
nuns112563.jpg Nuns praying, November 25, 1963
signfbi.jpg FBI recreation, May 24, 1964, photo courtesy of Gary Murr
signfbi.txt A history of the signfbi.jpg photo


Contributors

Joe Durnavich
Louis Girdler
James Gordon
Ron Hepler
Barb Junkkarinen
Craig Lamson
Josiah Thompson
David Wimp