Via AiE&S, I found an article in the esteemed Chronicle of Higher Education discussing the prevalence of digital manipulation of figures in scientific articles. This is an issue of interest to be, because I’m something of an amateur digital manipulator myself.  For years I’ve taken part in online Photoshop contests, and even won small monetary prizes here and there for my efforts.  One of the reasons I do this (in addition to how much fun I have with it) is the potential for applying the skills I learn to my work.  Mostly by this I mean my web design hobby, but knowing your way around Photoshop really helps when making presentations or other science-related graphics.  The problem is that there is a very fine line between making something look nice and altering the scientific meaning.

Papers are starting to employ tools to look for digital tampering:

New tools, such as software developed by Mr. Farid, are helping journal editors detect manipulated images. But some researchers are concerned about this level of scrutiny, arguing that it could lead to false accusations and unnecessarily delay research.

I have to say I fall on the side of the “concerned researchers” here. In my experience, you tend to have two groups of people. There are those who understand image manipulation, its strengths and weaknesses, and when it can be appropriately implemented. Then there are others who don’t understand what’s going on and tend to think that if you even open your TIFF file in the program you’ve just sabotaged it’s scientific validity. I worry that we will end up with a situation of photoshop = bad, and that’s simply not true at all. As a matter of fact, the very description of the software being employed indicates that this is already the path we’re heading down:

The software looks for patterns in the digital code underlying an image. When files are opened and altered in Photoshop, for instance, codes are added that Mr. Farid’s software can detect.

I can hardly think of a single image I haven’t opened in such a program (I usually use The Gimp though).  Almost every figure needs at least to be cropped, have a label or two added, and saved out as different formats.  Does that mean every figure in my papers will be throwing up flags?

To be fair, it seems like the journals themselves are still taking a level hand for the most part:

So far the journal’s editors have identified 250 papers with questionable figures. Out of those, 25 were rejected because the editors determined the alterations affected the data’s interpretation.

10% rejection because of altered meaning seems to indicate that the vast majority of digital editing is not scientifically harmful.

And, as we so often see, Open Access may lend a hand in solving the problem:

One new check on science images, though, is the blogosphere. As more papers are published in open-access journals, an informal group of watchdogs has emerged online.

“There’s a lot of folks who in their idle moments just take a good look at some figures randomly,” says John E. Dahlberg, director of the division of investigative oversight at the Office of Research Integrity. “We get allegations almost weekly involving people picking up problems with figures in grant applications or papers.”

I’m not sure that I approve of the online witch-hunt oversight scenario this seems to set up, and this type of method won’t detect image tampering prior to publication (or at least online availability).

It’s a sticky problem, to be sure.  It’s clear that image manipulation programs allow for data falsification that is almost indetectable to the human eye, especially when done by an expert in the use of the software.  These same programs (and largely identical usage of the programs) can pull new scientific information out of otherwise useless images.  In my opinion, specific alterations should be mentioned in the figure captions.  I do this when I present figures at group meeting, for instance noting that I’ve adjusted contrast for easier viewing.  I wouldn’t mind in these cases a requirement to submit both images to the journal, so that they can judge for themselves the validity of the manipulation.  I just don’t want scientist to be discredited simply because they are trying to, with the best of intentions, improve communication of their findings.

I enter a few online image manipulation competitions from time to time. It’s fun, and I’ve learned a lot about Photoshop that has come in handy for other work, such as making graphics for presentations or the web.

I thought today I’d do a bit of a liveblog while I worked on an entry. Hope you enjoy

I’ll be doing an entry for the Woot weekly contest. They usually provide a theme, and the rest is up to the entrant. The theme this week is:

Show us a past or present woot brought down by scandal.

Let’s get this liveblog started.

8:00 - Brainstorming: This actually probably really started right after I saw the theme yesterday. This morning, however, I start sort of thinking a bit harder. At this time I think I want to do a play on the Teapot Dome scandal.

8:20 - Go back to contest description, look at already posted entries: Here I like to read the contest listing again, to make sure that I haven’t missed anything. Sometimes this gives me new inspiration or perhaps makes an idea I’ve come up with sound dumb. Not so much today. Also, I flip through the entries that are already posted to make sure that someone hasn’t already done something similar to mine. Thankfully, most of the entries so far are some play on drugs, Elvis, or both.

8:25 - Look up the Teapot Dome Scandal

8:27 - Start trying to figure out how to implement this cockamamie idea

8:33 - Look for source images: A source image can make or break a photoshop entry. I want to do an old newspaper, so off we go to the magical Flickr to search for them. After a bit of searching, I found a series of Yorkshire Evening Post snapshots from the 1940s, by Flickr user Mig_R. I really like the look of these, so let’s choose one as our starting point.

8:40 - Find the Font(s): Since I need to change the headline of the paper to a U.S. city, I’ll need that font. I use WhatTheFont along with a cropped image of the headline to identify a close match. It turns up Fette Gotisch, which is a fine font which unfortunately costs cash money. Time to do more searching. Finally I took the easy way out: since I only need a few words, I simply put those into one of the “try this font” boxes on a site trying to sell me the package. Nab a screenshot of the “test” and we are good to go. I used the clone stamp tool to erase “Yorkshire”, pasted the new city name in, and set the blending mode to “multiply” to remove the white background. After that I played with the hue/saturation, curves (drop in red, raise the green), and opacity to get it to match the rest of the text as much as possible. Finally I applied a layer mask and used a “grunge” brush in black to knock out little specks of the text.

To do the other bits of the masthead, I simply matched fonts from my machine and typed in what I wanted.

9:16 - Text work: Replacing all the text of the paper is quite a chore. You have to do a lot of careful work with the clone stamp tool so that the erased text doesn’t look too fake, then of course generate the replacement stuff. Here I decided that the scandal would be “Craps for wine”, in which people who wanted to get a Bag of Crap (Woot’s most famous offering probably) were required to send bottles of wine to the staffers, which they would then use to run Wine.woot.

10:00 - The image: To create the Bag of Crap image, I used the poster edges filter in order to get nice dark lines on the edges of the source. I set the blending mode to multiply again, dropped the opacity a tad, and played with the hue/saturation once again to get it to blend in with the vintage newspaper feel.

10:15 - More text work: Finally I did some ancillary text stuff to tie it in with the Teapot dome scandal (a minor occurrence in this paper’s reporting) and sort of wrap the whole thing up.

10:30 - I think I’m done, time to step away. Click the image for the full version:

Sometimes I’ll go back and make more edits later.  More often I’m sick of looking at the entry, and tend to purposefully delete the PSD file so I can’t fall prey to the endless tweaks.  I think we’ll go with the latter this week.  2 hours is about my limit for one entry.