The Science News Cycle
« previous post | next post »
According to Jorge Cham at PhD Comics (click on the image for a larger version):
This is pretty accurate, except that it's a bit unfair to blame "Random Dude" posting on "the Internets", since in the cases we've documented over the years, participants at every other step of the process bear a much larger portion of the guilt — especially the university (or journal) PR Office, and most especially the MSM reporters and editors.
Q. Pheevr said,
May 19, 2009 @ 10:04 pm
It also makes my grandma out to be a much more credulous person than she ever was.
Zubon said,
May 19, 2009 @ 11:04 pm
If you publish at p=0.56, you deserve what you get. :p
MJ said,
May 19, 2009 @ 11:09 pm
I got a smaller version when I clicked the image. Or maybe the same size, it was tough to tell.
[(myl) You must have a small screen. The underlying image is 600 pixels wide, while only 475 or so will fit in our WordPress display column. You can see the original size at the linked PhD Comics site — or try right-click>>view image (or your local OS/browser equivalent).]
bfwebster said,
May 19, 2009 @ 11:41 pm
Ah, that's priceless. And painfully accurate. At at a confidence level much higher than p=0.56. ..bruce..
AJD said,
May 20, 2009 @ 12:35 am
It's not p = 0.56; it's rho, the correlation coefficient. Rho is 0 if there is no correlation between A and B, and 1 if A and B are perfectly correlated.
KCinDC said,
May 20, 2009 @ 12:38 am
That's not p, the confidence level (yes, 56% would be ridiculous). It's ρ, the correlation coefficient.
Murray Jorgensen said,
May 20, 2009 @ 4:24 am
Not a good idea to use \rho = .56 rather than r = .56 as Greek letters are normally used for population parameters rather than sample-based estimates which this presumably is. It would have been good to include a p-value or sample size as well. None of this would have made any difference to subsequent steps in the loop!
Lance said,
May 20, 2009 @ 5:45 am
Oh…I also misread ρ as p. Next on my list: rereading my old statistics textbook.
Stephen Downes said,
May 20, 2009 @ 7:27 am
Actually, the circle leaves out the 'internets' step altogether – in reality, it goes dir3ectly from 'news wire organization' to 'cable news'. The 'internets' stage is epiphenomenal on the 'cable news' stage.
Zubon said,
May 20, 2009 @ 8:15 am
Gah, caught by the two-pixel difference between characters. I guess you do need to click for the larger version.
Mark P said,
May 20, 2009 @ 8:27 am
I could insert my standard "reporters are among the least well educated of all professionals" rant here. It's hard enough for those in technical fields who aren't really up on all their statistics. For someone like a reporter, it's like trying to understand a foreign language. I suspect that university PR people and reporters often come from the same pool. In fact, a reporter I worked with became a PR person at the same university I attended when I left my newspaper job.
But if a researcher is ever burned by this process, he's going to have to share the blame if it happens again.
[(myl) Yes — after years of adopting "blame the journalist" as my rule of thumb, I've come to the conclusion that the people who write news releases are a crucial nexus of nonsense; and that researchers ought to feel more responsibility for monitoring the PR presentation of their work. See the discussion of one recent case in "Debasing the coinage of rational inquiry", 4/22/2009.]
Chris said,
May 20, 2009 @ 9:11 am
Using p and rho to denote importantly different variables seems like a bad idea to me. Wouldn't they get confused all the time? There's hardly any difference.
In this example, the journalists were the ones substituting causation for correlation, which is a pretty big step in the propagation of error.
Of course, anyone getting the information thirdhand or worse and passing it along without checking the sources has some of the blame, too. That is, if the source is actually available to the public…
Ulrike said,
May 20, 2009 @ 3:32 pm
> Not a good idea to use \rho = .56 rather than r = .56 as Greek letters are normally used > for population parameters rather than sample-based estimates which this presumably is.
Both r and \rho are used to denote correlation coefficients. r is usually Pearson's coefficient and \rho Spearman's coefficient (see their Wikipedia pages).
As for the confusability of \rho and p – their publishable values are pretty far apart (p should be very low and \rho approach 1 or -1) and they always are (or should be) mentioned together. Given context, I don't think there's much room for confusion.
Nathan Myers said,
May 20, 2009 @ 5:18 pm
Part of the problem is that what scientists see as important is often very different from what the public is, and should be, interested in. Scientists usually find it hard to comprehend anyone legitimately perceiving results differently than they do, and attribute such differences to ignorance or to active stupidity. (This is not to say there is not plenty of ignorance and active stupidity at work!)
Most simply, to a scientist, what is important about a result is generally how it differs from what their immediate rivals have most recently asserted. To the public, the meaning of the result includes all the context leading up to the experiment. When Naish and Witton published their blockbuster analysis of Azhdarchid feeding strategy (concluding that the enormous pterosaurs lived like storks, not gulls), some colleagues saw them debunking a widely-published absurdity, but the public (and many colleagues) saw a fascinating glimpse of life gone by.
That the news accounts often identified azhdarchids as "dinosaurs" incensed many paleontologists, but the technical distinction between the pterosaur and dinosaur lineages is rightly of little interest to most public readers. Pterosauria and Dinosauria alone comprise Ornithodira, which might just as well have been folded a little differently, making the azhdarchids legitimately "dinosaurs", so the distinction is largely a matter of words. News accounts identifying azhdarchids as ancestors of birds deserves legitimate scorn, as birds had already supplanted pterosaurs in practically every other ecological niche by the time of the azhdarchids' heyday.
[(myl) It would be nice if the worst problems with MSM science reporting were cladistic carelessness and insensitivity to intra-disciplinary arguments. In fact, I'd be inclined to view these as hardly rising to the level of problems at all. What I object to is the invention of major, often socially-relevant conclusions out of over-interpretation, misinterpretation, or sometimes out of nothing at all. Unfortunately, the sequence depicted in this cartoon is much less extreme than what the real thing is often like. ]
Pix said,
May 21, 2009 @ 7:05 am
I'm sure there's a stage missing.
"No link between A and B, says Government"
Yeah, that's the one.
iakon said,
May 21, 2009 @ 10:11 am
I once ran across (in a newspaper) the phrase 'hacks and flacks' meaning journalists and publicists (propagandists). Presumably this is their own jargon for themselves. And yes, they both get the same 'higher education', leading to Batchelor of Communications (whazzat?) and Master of Journalism, which is why they flit back and forth between hacking and flacking.
» The Science News Cycle :: Granite Geek :: NashuaTelegraph.com said,
May 21, 2009 @ 8:30 pm
[…] http://languagelog.ldc.upenn.edu/nll/?p=1442 […]
Dave Brooks said,
May 21, 2009 @ 9:05 pm
Don't forget to blame the Sagan Syndrome: the universal opinion among researchers that incomprehensibility is the necessary (and too often sufficient) mark of True Science, with resulting discredit attached to anybody who lowers himself to communicate with non-PhD peons out there in the *shudder* general public.
This clubbiness almost guarantees that the outside world will misunderstand what you do, so it is disingenuous to complain when that misunderstanding occurs – especially when the complaint is always coupled with "why don't you ungrateful people give me more money to do my research?"
(I'm a newspaper reporter, so you can dismiss this is sour grapes if you wish. )
Cecily said,
May 22, 2009 @ 7:05 am
The latest phdcomic continues the theme, with "Research topics guaranteed to be picked up by the news media":
http://www.phdcomics.com/comics/archive.php?comicid=1175
Zubon said,
May 22, 2009 @ 9:41 am
"I've come to the conclusion that the people who write news releases are a crucial nexus of nonsense"
Quite. Recent example from my work: there was a AAA study recently on aggressive driving. The headline of the news release was written so as to give the impression that aggressive driving caused the majority of car crashes. This helps to get the story picked up. Reading the full thing, you see that the majority of crashes involved at least one behavior that, if there were three or more behaviors from that category, would be defined as aggressive driving in Delaware. (I won't get into the assumptions for why they used that methodology.)
Moderate Backyard Rant and a Flat Out Lie « Robot Pirate Ninja said,
May 12, 2011 @ 8:43 pm
[…] it takes about two second to find what happened here….again….it's just like this "normal" science reporting scenario but with a hardcore profit motive. Here's what the actual scientist said about this version […]