November 15, 2005

"Twice as long" is 50% longer, or maybe 42% longer, or was it 21% longer?

Simeon Yates and colleagues at Sheffield Hallam University have recently been reported to have done a study on gender differences in mobile phone text messages. I say that they "have recently been reported to have done a study" because I haven't been able to find a publication or preprint describing the study, so I'm forced to rely on reports in the mass media,, which I know from experience are often spectacularly untrustworthy. In this case, the information spread across the mainstream media is primarily an exercise in the expression of gender stereotypes. Some sample headlines:

BBC "Men write short, sarcastic texts"
Scotsman "Men keep phone texts short, sexy and full of swearing"
The Sun "Girls are going txt crazy"
Times of India "Women just can't keep it brief"

This would be less troublesome if the extraordinary quantitative illiteracy of the fourth estate were not so strikingly on display in the associated stories.

The Mirror's headline reads

"SEXES ARE DIVIDED ON JOYS OF TEXT: WOMEN'S text messages are twice as long as men's, new research reveals."

But a few sentences into the body of the article, we learn that

The study by Dr Simeon Yates of Sheffield Hallam University, says women enjoy in-depth texting while men use just a few words.

Dr Yates said: "Women's texts are 50 per cent longer. They text to communicate while men text to inform."

So suddenly "twice as long" is "50 per cent longer".

But it gets worse: The Sun's story gives us some (alleged) actual numbers:

Men’s texts average 60 characters, while women’s are 85.

That's more like 42% -- but I guess "50 per cent longer" sounds better, not to say "twice as long". And The Times of India tells us that

Dr Simeon Yates observed the mobile phone habits of hundreds of people. He found that female text messages typically go on for 82 characters, while men get their point across with just 68.

That's just 21% longer. Nevertheless, the same story leads with the trenchant observation that

As any long-suffering boyfriend or husband knows, some women just love to talk. Now research shows that females talk more than men even when communicating by text message.

That's a lot of interpretation to put on 82 vs. 68 characters, especially when another article tells us that men texting to women use an average of 80 characters. So even without knowing anything about the study's methodology, much less the actual facts found as opposed to the intoxicated exaggerations that seem to be the MSM's norm in reporting such stories, I'm reserving judgment on the other generalizations that are offered. These are things like

Women offer their friends love and support, while men often resort to sarcasm, sexual humour and swearing.

Men keep their texts short and snappy while women have seized on the mobile as a new way of expressing support and affection...

This certainly has the flavor of tabloid journalism, but it's not limited to the tabloids -- that last quote was from The Guardian. And some version of it might all be true -- or it might be journalistic froth that's quite different from what the study actually found, and also different from what Prof. Simeon Yates told the media. But as I observed a couple of months ago with respect to the ridiculous "email is worse than pot" story, scientists who go to the media with a story like this one -- or are taken up by the media without intending to be -- have a responsibility to make a responsible and factual version of their work available, either as a scientific publication, or at least as a responsible lay-oriented description on a web site somewhere. I looked at Simeon Yates' web site and found nothing; nor could I find anything on the web site of the journal he edits, Discourse Analysis Online, nor did searches on Google Scholar or Scirus turn anything up.

[By the way, I wonder: are copy editors expected -- or allowed -- to correct elementary errors in arithmetic?]

[Update: Linda Seebach writes

... yes, copy editors are expected to spot and correct errors in arithmetic, but actually at most papers (obviously they're not all the same) the line editor -- the primary editor who works directly with the reporter -- would have the chief responsibility. By the time the story gets to the copy desk, which may be hours later when the reporter is not available, it could be too late. If the copy editor sees "Eight is 40 percent of 73" he knows one of the three figures has to be wrong, but not which one.

It is also true that proofreading tends to focus on the local. Spotting inconsistencies in the text, perhaps widely separated, is harder than it would seem when you're looking for "Paris in the the spring" stuff.

That makes perfect sense.]

Posted by Mark Liberman at November 15, 2005 01:49 PM