Last week we reported on a photograph of self-proclaimed Tea Pary founder Dale Robertson holding a sign featuring a racist (and misspelled) slur. In an exclusive interview with Mediaite’s Tommy Christopher, Robertson claims that the photo was a fake. I was asked to take a look at the now-infamous photo of Tea Party leader Dale Robertson to assess the likelihood that the word ‘niggar’ was added at a later time.
First, some credentials. I used to be a designer at Adobe Systems, Inc., makers of Photoshop. Among other things, I worked on the Photoshop user’s manual and designed chapters for the Photoshop Classroom in a Book. As for experience with this sort of forensic analysis – I did spend about a week as a juror having signature differences explained to me, for what it’s worth.
On to the analysis.
The original image
I’d been sent a copy of the image in a fairly low-resolution format, so I decided to seek out as close to the original as possible. I found a high-resolution version still on the Houston Tea Party website, HoustonTPs.org.
Each time a modern digital camera takes a picture, it stores certain information along with the image. This data, called EXIF data travels along with the photo unless it’s stripped out as part of a photo editing process.
In the case of the original photo, the EXIF data revealed that the original photo was taken on February 27, 2009, without a flash. It was saved out of Photoshop CS3 for Windows on March 4th of last year – apparently the last time the original file was edited. It is possible to falsify EXIF data, so this is not conclusive, but if accurate, that’s well before the furor over the photo erupted.
Looking closely at the original also revealed other information.
First, the majority of the lettering seems to have been done in dry erase marker, the letters (particularly in ‘slave owner’) bearing signs of having lost color by being brushed against.
Second, the word ‘niggar’ was written in regular marker on a piece of posterboard or other stiff paper which was then attached to the larger sign. You can see, in the enlargement above, the upper-left corner of the paper. Below, the lower right corner, including that the bottom of the added sign was covered with duct tape.
At the top of the additional sign, above the second G in ‘niggar’, is what appears to be a piece of tape, indicating that the writing may be on the reverse of another sign of some sort.
Next, I compared the lettering for the primary sign with that of the posterboard addition.
At left, below, are the letters from the word ‘niggar’; at right, the same or similar letters from elsewhere on the sign.
Stylistically, the lettering for the N, A and R are consistent with lettering elsewhere on the sign – but, importantly, are not exact duplicates. That is, the letter A in ‘niggar’ isn’t a copy of one of the other A’s, though it’s clear that they are very similar in style. (In the case of the A’s, it appears the general outline of the A was drawn first, then a triangle in the middle, then the entire letter filled in.)
The G’s are harder to determine, since the only other G is partly covered at the top of the sign. But it’s obvious that they aren’t copies of that G, nor are they copies of the C in ‘congress’.
Evidence of manipulation
I then looked at the evidence within the image data itself, focusing on two things: compression artifacts and fill.
Artifacts. The original image was saved in the JPEG format, which is a “lossy” format, meaning that every time the file is saved, it loses data in order to reduce file size. Part of this process means that artifacts, or compression fingerprints, remain in JPEG images, as you can see faintly above. The discolorations in the red circles are compression artifacts.
Fill. If this image is forged, and there was originally another word where ‘niggar’ now appears, that word would have to have been covered up in Photoshop. (If there weren’t another word there originally, why was he holding the sign?) There are two ways to do this – either by filling the entire area with a solid color, or by using the clone tool (or its sister, the healing brush) to cover the original lettering with copies of blank areas of the posterboard. More on this in a moment.
To assess the likelihood that the image was forged, I made a forgery. Presenting my more powerful political statement:
In this image, you can clearly see that the letters P, O, N and Y come from elsewhere on the sign, though they’ve been manipulated and darkened. Note, too, that the area behind ‘pony’ has been filled to cover ‘niggar’.
In creating the ‘pony’ version, I first covered ‘niggar’ with a solid color which, to the naked eye, makes the paper appear blank. I then saved the image out as a JPEG and re-opened it to compare compression artifacts and fill. The result:
The area of the fill, while invisible to the naked eye, is obvious when the color of the image is adjusted. Note the subtle line moving from the left edge up and to the right, and the right edge of the fill area about three-quarters of the way across the color-adjusted version. Most importantly, note the evenness of the color in the fill area. In the original, there’s no such evenness; the area between letters is mottled and splotchy in the color-adjusted version. To the naked eye, that’s not the case. (Also, note how prominent the compression artifacts are in the color-adjusted version.)
Adding ‘pony’ and saving as a JPEG results in the following:
There are additional artifacts, but note the flat, even color between the letters.
Then I filled the area by cloning. In the cloned sign, the filled area is blockier and less even than in the original, albeit only slightly. The differences here are more subtle and open to interpretation.
It’s obvious that if the original image in this case was forged, the creator of the forgery paid particular attention to covering the original wording of the sign.
The fundamental issue, though, is this. Someone with the appropriate Photoshop chops would need to have:
1) falsified the EXIF data,
2) created a sophisticated mottling to cover the original word (when a fill would do), and
3) carefully matched the styling of the original letters without duplicating them.
All to impugn the signholder with a word that could only have been slightly more offensive than other alternatives – after all, not many words fit where ‘niggar’ is now.
But the final piece of evidence worth considering this: the original, high-resolution file I downloaded came from the Tea Partiers website. Why would they upload and host a forged image of one of their leaders if it contains a word they’ve disavowed?
There’s no evidence that the photo was altered in any way. For it to have been altered convincingly requires a level of sophistication that doesn’t seem to be worth the pay-off in this case.
Further assessment – this took longer than I should have spent on it.
Have a tip we should know? firstname.lastname@example.org