Physicist Reputations Tarnished

ruszka writes "An article at PhysicsWeb goes over a growing concern in the physics community: their reliable image. This isn't a case of jumping the gun, as seen with cold fusion, but over fabrication in data results. Bell Labs and Berkeley are both recovering from cases where their own employees falsified data."

Comments

Re:Reproducible.

[Otter]

Biology in the US went through this a few years ago, at the peak of the Robert Gallo and David Baltimore "scandals". There were a handful of highly publicized cases that were blown up into a trend of "scientific misconduct is on the rise". Yeah, two instances make a trend. Glad to have thinkers like that in science.

Both of those cases turned out to be largely groundless, but not before all of us on NIH grants had to take mandatory ethics seminars, where we discussed ludicrous scenarios about authorship that bore absolutely no relation to reality. Now, ethics seminars have vanished and there hasn't been a high-profile case in years. (Some grad student in Francis Collins' lab falsifying results in 1998 or so.)

These physics scandals do seem to at least be genuine cases of fraud, and I imagine grad students will be herded through some seminars, no new cases will pop up and the whole thing will be forgotten in a year.

Most Aren't Reproduced

[Anonymous Coward]

This notion of independently reproducing results is very good, but that's often easier said than done.

One reason is that if the major factors that brought about the original experiments results are not well known (even though the original investigators thought they did), then REAL results maybe difficult to reproduce...even by those that originally did the work!

Also, one typically doesn't just reproduce results but tries to "reproduce and extend." But here's the rub, this attempt may alter poorly understood processes so much that a replication doesn't occur. A small change can even do this in "choatic / nonlinear systems.

When dealing with statistical results like testing drugs on human subjects. Some experiments show great promise but others don't, even when patient populations seem to be similar (i.e. you pre-blocked and randomized with say 1000 or more subjects and checked that the baseline groups all started the same) and treatments were standardized and blinding was adequate.

Populations can NOT generalize!

Much repeated testing may need to be done to repeat these results.

This and other things provide a sufficient loop-hole to even entirely make-up bogus data.

From personal experience, I've seen this happen and it does happen much more often than admitted by science or the press.