December 16, 2006

The spread of bogus numbers in the meme pool

There's a sort of Darwinian effect of the media spotlight -- survival of the most sound-bitable. And some of the most rapidly reproducing sound-bites are quantified statements like "teenagers use just 20 words for a third of what they say", or "email lowers IQ by ten points", or "Eskimos have dozens (or hundreds, or thousands) of words for snow", or "women use three times more words than men do".

That's why PR types, journalists and public intellectuals create and spread a vast array of meaningless, mis-interpreted or just plain made up numbers. Sexy numbers prosper in the struggle for memetic hegemony. And unless there's some process to enforce honest signaling by penalizing over-interpretation, exaggeration, or flat-out lying, bogus numbers will dominate the meme pool, because for every truthful number, there's always a nearby bogus number that's sexier.

This afternoon, when I turned on the radio, I happened to hear a few seconds of an interview with Louann Brizendine ("Comparing Mars and Venus in Neuroscience"). The bit that I heard was:

Q: Now, I saw you quoted in the New York Times, speaking of pregnancy, that the female brain shrinks about eight percent during pregnancy? And doesn't return back to its normal size until about six months after delivery?

A: Yes, Debbie, that's a surprising study that uh has found eight percent shrinkage, even after you account for any increased water weight. And scientists don't know really why that happens, except that the female brain is doing all kinds of rewiring during that period, to get the mom ready to do maternal behavior. And also remember, the fetus is more like a parasite, and ((that)) it gets fed whatever it wants, and lots and lots of lipids and special fats exist inside the brain cells, and some scientists speculate that the fetus is sort of snacking on the mother's brain.

I noticed this same factoid in the NYT magazine interview ( "He Thought, She Thought") that we discussed here earlier, and I wondered whether it was true. In Brizendine's book "The Female Brain", the chapter "The Mommy Brain" discusses a brain shrinkage of unspecified size (p. 100):

Between six months and the end of pregnancy, fMRI [sic] brain scans have shown that a pregnant woman's brain is actually shrinking. This may be because some parts of her brain get larger as others get smaller -- a state that gradually returns to normal by six months after giving birth.

[There's a typo here, by the way -- fMRI stands for "functional magnetic resonance imaging", and it's a technique for measuring local changes in cerebral blood flow as a consequence of different sorts of brain activity, like looking at funny vs. unfunny cartoons. Measurements of brain size would use plain old structural MRI, with no f-for-functional involved.]

Anyhow, Dr. Brizendine has clearly learned the lesson about quantitative sound bites. The "mommy brain" shinks a bit: ...bore-ring... The "mommy brain" shrinks 8%: wow!

Now, 8% brain shrinkage isn't nearly as sexy as the business about men thinking of sex every 52 seconds, or women using three times as many words as men and talking twice as fast. But still, it's something.

And given previous experience, it occurred to me to wonder about the number. So I checked, back when I read the NYT article. As you'll see below, the number is somewhat bogus. But it's only inflated by 86%, which is a pretty small exaggeration compared to the tall tale about sexual thoughts every 52 seconds, which appears to be inflated by 23,736%.

So I decided not to write about it, since the content of the claim has nothing to do with speech and language. And if I tried to document every bogus statistic in the mass media, I'd never have any time for anything else. Just writing about Dr. Louann Brizendine's statements about sex differences in communication is starting to make me feel like the circus clown that follows the elephant around the ring with a shovel. However, after the "20 words for a third of what they say" business this afternoon, Geoff Pullum suggested by email that we should think more broadly about the public rhetoric of science. So here's some more raw material. Not as raw as some, but still.

Dr. Brizendine's endnotes refer the business about brain shrinkage in pregnancy to Oatridge, A., et al. "Change in brain size during and after pregnancy: Study in Healthy Women and Women with Preeclampsia", Am J Neuroradiol 23(1): 19-26, 2002. That article, I'm happy to say, gives a convenient table of the values that they measured, which I was able to copy as html directly from the online version:

Table 3: Absolute brain volumes (cm3) before, during, and after pregnancy

Subjects Before Pregnancy 15 Weeks’ Gestation 20 Weeks’ Gestation 25 Weeks’ Gestation 30 Weeks’ Gestation 35 Weeks’ Gestation Before Delivery (Term) 6 Weeks after Delivery 24 Weeks after Delivery 40 Weeks after Delivery 52 Weeks after Delivery

Healthy group
 1 1437.4 1494.5 1494.7
 2 1089.2 1109.7 1122.0
 3 1201.7 1245.4 1267.4 1252.5 1260.2
 4 1371.8 1383.6 1415.1 1420.2 1415.8
 5 1207.9 1237.7 1246.2 1253.6 1244.0
 6 1197.4 1195.5 1184.0 1187.3 1172.2 1163.0 1150.7 1180.1 1205.7
 7 1277.5 1265.9 1255.7 1238.0 1221.9 1268.8 1290.5
 8 1288.9 1247.1 1233.5 1241.5 1208.7 1248.3 1269.9
 9 965.5 953.9 946.2 975.2
Preeclamptic group
 1 942.7 981.7 981.9 977.3 975.5
 2* 1036.2 1070.3 1080.6 1043.1
 3 999.9 1027.5 1037.1
 4 1274.3 1303.9 1312.0 1315.2
 5 1183.4 1238.7

There were two women in the study, number 6 and 8, who were measured both before pregnancy and at term. Over that period, their brains shrank 4.06% and 6.6% respectively, for an average of 5.3%. There were eight women in the normal group whose brains were measured at term and 24 week (i.e. six months) after delivery. Their brains increased in size during that time by 4.0%, 3.0%, 5.5%, 3.2%, 4.8%, 5.6% and 5.1% respectively, for an average of 4.3%, with a 95-percent confidence interval of 3.4% to 5.2%.

How did this effect get roughly doubled to 8%? The only thing that I can think of is that the maximum value on the vertical axis of the paper's Figure 3A -- which is a graphical presentation of the same data as in Table 3 above -- is 8%:

I'm sure that this was an honest (though careless) mistake of memory. A dishonest author might have tried to inflate the effect still further -- would you believe 10%? 15%? Listen, it's a little-known fact that a woman's cerebral cortex completely disappears during preganancy! No, there really are some constraints on dishonest assertions in this arena. Even a journalist wouldn't believe that. Well, some journalists would, I guess -- the stuff in Leonard Sax's book about sex differences in sight and hearing is just about that far out. But they'd get teased about it later. I hope.

[I should mention in passing that there's nothing in the Oatridge et al. study about "allow[ing] for increased water weight", as far as I can see. They just measured brain volume in MRI scans. The values in the table above are their raw measurements, not adjusted for water weight or anything else.]

Posted by Mark Liberman at December 16, 2006 08:08 PM