Some may remember a paper from 2011 on the "reverse click" reaction, an interesting one where triazoles were pulled apart with mechanical force. This was an interesting system, because we really know surprisingly little, down on the molecular level, about what happens when bonds are physically stressed in this way. What do molecular orbitals look like when you grab both ends of the molecule and tug hard? Which bonds break first, and why? Do you get the reverse of the forward reaction, or do different mechanisms kick in (free radical intermediates, etc.)? (Note that the principle of microscopic reversibility doesn't necessarily apply when the conditions change like this).
Unfortunately, there seems to be trouble associated with this example. Science has an editorial "expression of concern" on the paper now, and it appears that much of it is not, in fact, reproducible (see this report in C&E News).
The paper was from the Bielawski lab at UT-Austin, and Bielawski is reported as saying that a former group member has confessed to manipulating data. But he also says that the conclusions of the paper are unchanged, which is interesting. My guess is that the "unclick" does happen, then, but nowhere as smoothly as reported. Someone may have sweetened things to make it all look better. At any rate, a correction is coming soon in Science, so we should get more information at that point.
This reminds me of the scheme I use to rate political and economic corruption. Stage I is paying someone off to do something they wouldn't normally do (or aren't authorized to do) for you. This happens everywhere, to some extent. Stage II is when you're bribing them just to do the job they're supposed to be doing in the first place. Many countries suffer from institutional cases of this, and it's supremely annoying, and a terrible drag on the economy. And Stage III, the worst, is when you're paying them not to harm you - a protection racket with the force of law behind it. Cynics may adduce examples from the US, but I'm thinking about countries (Russia, among others) where the problem is far worse.
Similar levels apply to fakery in the scientific literature. Here's how I break it down:
Stage I is what we may have in this case: actual conclusions and effects are made to look cleaner and better than reality. Zapping solvent peaks in the NMR is a perfect small-scale example of this - for organic chemists, solvent peaks are sometimes the training wheels of fakery. The problem is, once you're used to altering data, at what point do you find it convenient to stop? It's far better not to take that first step into matters-of-degree territory.
Stage II is unfortunately common as well, and there's a nice slippery path from Stage I that can land you here. This is when you're convinced that your results are correct, but you're having such a hard time getting things to work that you decide to "fake it until you make it". That's a stupendously bad idea, of course, because a lot of great results were never real in the first place, which leaves you hung out to dry, and even the ones that can be finally filled in don't have to do so in the way that you were faking them to happen. So now a real result is tainted by deception, which will call the whole thing into doubt when the inconsistencies become clear. And faked results are faked results, even if they're done in what you might think is a good cause. Many big cases of scientific fraud have started off this way, with someone just trying to fill in that one little gap, just for now.
Stage III, the bottom, is when something is faked from beginning to end. There was no question of it even working in the first place - it never did. Someone's just trying to get a paper, or a degree, or tenure, or fame, or something, and they're taking the shortcut. I think that there are two main classes of fakery in this category. In one group (IIIa?), you have people whipping up bogus results in low-profile cases where no one may notice for years, if ever, because no one cares. And you have IIIb, the famous high-profile cases (see Jan-Hendrik Schön, among too many others) where impressive, splashy, look-at-that stuff turns out to have been totally faked as well. Those cases are a study in human psychology. If you report a big result in superconductors, stem cells, cancer therapy or any other field where a lot of smart, competent people are paying very close attention, you will be found out at some point. How can you not be? We're in Bernie Madoff territory here, where someone comes into work every day of every week knowing that their whole reputation is a spray-painted scrim of deception that could have a hole punched through it any minute. How people can possibly live this way I don't really know, but people do. The self-confidence displayed by this sort of personality is a wonder of nature, in its way. IIIa cases are initiated by the desperate, stupid, and/or venal. IIIb cases, though, are brought on by people born to their task.
Update: as pointed out by several good comments, there are plenty of not-quite-fraud sins that neighbor these. Those are worth a separate post, partly because they're even more common than straight-up fraud.