Is this photo manipulated?

January 28, 2007 2:40 AM

Scientificfraudfilter. Does this photo look like it's been fabricated or at least significantly manipulated?

It's of a nucleic acid blot, which is the end result of an experimental method used in molecular biology. I have reason to believe the results suggested by this blot are misleading, and after fooling around with the photo in Photoshop I've started to suspect that it may have been manipulated/altered/fabricated.

What do you think?

posted by shoos to science & nature (18 comments total)

Hmm, it showed up in preview...

Try this or this.
posted by shoos at 2:45 AM on January 28, 2007

It's really difficult to tell. There's too much in the way of JPG artifacts to really make any judgment.
posted by Steven C. Den Beste at 3:15 AM on January 28, 2007

I haven't dealt with nucleaic acids blots before, so I can't deal with it from that angle, but I can say that when I play around with it in Photoshop...I can't see much at all for all the dreadful JPEG artifacts!

Do you have a better quality version of the image? Is this the version you're working on? I don't know what sort of forgery you're looking for, but if you think you can see square "cut marks" around the blots, I'm more inclined to think these are JPEG compression artifacts.
posted by Jimbob at 3:19 AM on January 28, 2007

Yeah, what the Beste Man said.
posted by Jimbob at 3:20 AM on January 28, 2007

I'm afraid I have to agree with SCDB and Jimbob. The image has too much JPEG compression to make any reasonable inferences. However, if you told me for a fact that it had been manipulated and asked me to speculate I'd say that I see evidence that the middle blot has been lengthened and that its original unmodified length was approximately that of the third blot.
posted by RichardP at 3:37 AM on January 28, 2007

Here's an article about JPG artifacts that may clarify what we're talking about. Here's another.

Another problem with JPG is that if images are recoded (i.e. a JPG file is reconverted to JPG with a different compression setting) the artifacts cascade. Multi-generation JPG files can end up look truly monumentally awful.
posted by Steven C. Den Beste at 3:51 AM on January 28, 2007 [1 favorite]

Yea, what Jimbob and SCDB said. Here is a decent NYT article with some thoughts about scientific photoshop forgery that I thought was interesting.
posted by gemmy at 6:21 AM on January 28, 2007

I agree that there is not enough information in this image to make a determination.

I think your best option is to try and duplicate the experiment! If you can't duplicate the results according to the paper, then it has failed peer review.
posted by clord at 6:48 AM on January 28, 2007

Is that picture upside down? If it isn't, it is weird that it looks like there is nuclease contamination but that the trailing ends of the bands look so freaking clean (if I get a leading smear, I always have a trailing smear as well). Also, I'd say that someone took an eraser to the area above the middle and right bands.
posted by The Bishop of Turkey at 6:56 AM on January 28, 2007

If I saw this in a figure in a manuscript, it wouldn't set off any red flags. While it certainly could have been fabricated, it is totally reasonable that it's kosher. The Bishop of Turkey may have a point about the top of the bands. I think it's possible that this is a small portion of a larger blot. But again, it doesn't set off any immediate uh-ohs.
posted by manchengo at 9:32 AM on January 28, 2007

The only thing that looks suspicious to me is the differnece between the amount of space between the first and second lanes, and the second and third.
posted by greatgefilte at 10:28 AM on January 28, 2007

This is the highest quality version of the photo that's available to me. It's from an online article, and wasn't authored by anyone I know. I'm by no means any graphics expert, so I may be way off base. I also can't perform this exact experiment, since I don't have the same reagents, but I do strongly suspect that the results are misleading.

The reason the photo set off bells to me was that after I adjusted the levels to see if I could pick up any faint signals, it looked a bit peculiar, with straight vertical lines, not just jpg artifact boxes, on the side of two of the bands (for the uninitiated, "bands" are the large dark spots). It also revealed that the background noise abruptly changes at the edges of the bands. When I've compressed my own blot images into jpg, I've never seen anything like this.

I ran a couple of these "forensic actions" on the photo, and they also highlighted some unusual (to me) features. I put the action-processed versions of the photo on my Flickr page, and also added a blot photo which I know is legit and looks very typical in my experience.

On preview: greatgefilte, I noticed the same thing, and those vertical lines I mentioned are consistent with the notion that the right two lanes weren't originally next to the left lane.
posted by shoos at 11:29 AM on January 28, 2007

There isn't nuclease contamination, or if there is then there isn't very much. It looks to me like all three lanes are seriously overloaded with lane 1 the most overloaded (hence it appearing to be closer to 2 than 3 is to 2). The picture doesn't appear to be manipulated in photoshop to me if that's what your asking. The person who took the picture played around with the black, white and gamma settings on the gel dock to get the second image. I have oodles of images like that except that I'm usually trying to get a faint band darker when I start mucking around with the settings. Why they did it in this case is beyond me unless they were looking for a different band than the ones that are shown. The thing that bothers me is the lack of a marker.
posted by LunaticFringe at 11:34 AM on January 28, 2007

Is this some kind of electrophoresis? Regardless of what technique is used, you obviously can't cite something like this in your research (if that's your intent). You'd have to contact the source directly, verify the details of his/her experiment and you'd probably be able to ascertain the reliability of it then. That's the best idea -- try to find out who authored the image and contact them. They'd probably be more than happy to talk to you about it.
posted by Aanidaani at 1:13 PM on January 28, 2007

Grabbing an image from an online publication and adjusting brightness contrast is very unlikely to pick up unseen bands. The data set available in that image has probably gone through multiple conversions both on the part of the authors AND the online publishers. The image probably changed formats through the publication process multiple times - its simply horrible data to work with if you're trying to look for something subtle that wasn't highlighted in the original publication.

Because of this, suspecting scientific fraud based on anecdotal photoshopping of the published image is a big assumption to make. There are thousands of non-deceptive reasons why that reported result may not be what your group might expect (i.e. different reagents, techniques, clones of cells, etc.)
posted by dendrite at 4:58 PM on January 28, 2007

I'm by no means an expert. Actually, consider me the opposite. However, if you color correct an image in Photoshop (play with the levels) aren't there always a lot of straight vertical lines? Try adjusting the red and green levels for any photo and then looking at the RGB. Is that similar to what you're talking about. I apologize in advance if this post is totally stupid.
posted by xammerboy at 11:32 PM on January 28, 2007

I can't help you with this particular image, but Science had a feature on (unethical) image processing in an issue last December (the one with Breakthrough/Poincare on the cover). If featured an example very similar to your example.
posted by swordfishtrombones at 5:02 AM on January 29, 2007

I didn't realize it was swiped from a publication, I thought it was the asker's data.
posted by LunaticFringe at 5:30 AM on January 29, 2007