Skip to content

← Thinking Clearly

Getting better anyway

Chapter 19 · Cause

In this chapter 4 sections
  1. Things that pass on their own
  2. The bad day doesn’t come back
  3. What a placebo does, and what it doesn’t
  4. The grandmother in the video

In 1961 a young psychologist called Daniel Kahneman came home to Jerusalem to teach, and some time after that he helped run a course for the instructors at the flight school of Israel’s air force. He told them what the research on learning said: praise works better than punishment. When he had finished, one of the most experienced instructors in the room put up his hand and disagreed. Many times, he said, he had praised a cadet for flying a maneuver cleanly, and the next time the cadet generally did worse. Many times he had screamed at a cadet for flying one badly, and the next time the cadet generally did better. “So please don’t tell us that reinforcement works and punishment does not,” he said, “because the opposite is the case.”

The instructor wasn’t making anything up. He had watched it happen, again and again. Kahneman later called that moment the most satisfying flash of understanding of his whole career, because he could see what the instructor hadn’t. A cadet who flies one maneuver beautifully is probably a good pilot who also had a good try. The next try is likely to be more ordinary. A cadet who flies one badly has probably had a bad try as well as being a learner, and the next is likely to be better. The shouting and the praise came straight after the unusual try, and the ordinary one came next, as it would have anyway.

Swap the cadet for someone who is ill, and the shouting for a remedy, and you have one of the oldest puzzles in medicine. People try something when they feel worst. From there, three things carry them toward feeling better whatever they take. Most illnesses end by themselves. A bad day tends to be followed by a more ordinary one, just as the instructor’s cadets did. And being looked after can make people feel better, and say so, even when nothing in the body has changed. Put the three together and nearly every remedy ever tried has seemed to work for somebody. That is how a tea that does nothing can collect a shelf of grateful stories, every one of them sincere.

Things that pass on their own

A cold is the plainest case. It comes on over two or three days: a blocked nose, a sore throat, a cough, that tired, heavy feeling. Then, in about one to two weeks, it goes. Colds are caused by viruses, so antibiotics don’t touch them. The body sees it off.

Now follow a cold through a family. On the first day it’s a sniffle, and nobody does much. By the third day it’s at its worst, and that’s when someone makes the drink your family swears by, or a neighbor brings a remedy of her own, or you buy something at the pharmacy. Over the next few days you feel better, and the last thing you tried gets the credit. It came just before the turn, because that is when people go looking for help.

A small pharmacy with shelves of medicine boxes behind two open doorways, under a big sign for soap, in a street in Patan.
Figure 19.1 A pharmacy in Patan in 2013. Whatever someone buys on the worst day of a cold, they are likely to feel better within a week or two. Photo: Gerd Eichmann, CC BY-SA 4.0

Plenty of everyday troubles have the same shape. Lower back pain is very common, and it usually improves within a few weeks, often by itself. Troubles like these rise, peak and fade. The peak is when people try things, and the fade comes next, so whatever was tried seems to have worked. Nobody gets to live the same week twice, once with the remedy and once without, to see what would have happened.

Not every illness passes, of course. Diabetes doesn’t. It is a long-lasting disease in which the body can’t keep the sugar in the blood under control. But even a disease that stays has numbers that move, and that is where the second reason comes in.

The bad day doesn’t come back

Any score is partly what’s steady and partly the luck of the day. Your mark in a class test depends on how much you know, and also on how well you slept, whether the questions happened to be the ones you’d practiced, and whether you misread one. Most of that luck is small, and it goes both ways.

So suppose a teacher picks out the ten lowest marks in a class of forty. Who is in that group? Some of the weakest students, certainly. But also some ordinary students who had a bad day, because a bad day is one way to land among the lowest ten. In the next test the weak students are still weak. The bad luck, though, doesn’t come back with the others. On average, then, the ten do better, even if the teacher did nothing for them at all. The ten highest marks, picked the same way, will on average do a little worse next time, for the same reason.

Try it

Pick the Worst

Forty made-up students, each with a steady skill and a day’s luck on each test. Pick the worst ten, or the best ten, on the first test, then change how much luck counts.

First testSecond test0255075100averagemarks out of 100
Pick:
The ten picked, first test
36
Same ten, second test
47
Change
+11

Nothing was done for them, yet the worst ten did better the second time. Their bad luck didn’t come back with them.

With no luck at all, the ten picked don’t move: everyone scores the same twice. Add a little luck and the worst ten climb, the best ten slip, and nothing was done to any of them. The more luck there is in a score, the bigger the swing back. This drift back toward the middle has a clumsy name, regression to the mean, and it happens wherever two measurements are only partly alike and someone picks out the extremes of the first.

The name comes from Francis Galton, a half-cousin of Charles Darwin, who noticed it in the 1870s in sweet peas. Seeds from very big seeds grew into plants whose seeds were smaller than their parents, and seeds from very small ones grew plants with bigger seeds. In 1886 he found the same pattern in people, with the heights of more than 900 grown-up children and their parents. Tall parents had tall children on average, but less tall than themselves. Short parents had short children, but less short. The children were about two-thirds as far from the average as their parents. Galton called it “regression towards mediocrity”, by which he meant only toward the ordinary.

Kahneman answered with a demonstration, which he arranged at once. Each person taking part threw two coins, one after the other, at a target behind his back, without seeing where the first had landed. They measured the distances. Those who had done best the first time mostly did worse on the second throw, and those who had done worst mostly did better. Nobody had been praised or shouted at, and nobody could have learned anything between throws. The swing back was all there was.

Doctors run into this every day. A person’s blood sugar varies a good deal from one day to the next: an illness, stress or recent exercise can push it up or down. So a single reading taken on a bad day is likely to be followed by a better one. The American Diabetes Association’s guidelines, which rate that day-to-day variation as high, don’t let one high test settle a diagnosis of diabetes unless the reading is unmistakable: a second abnormal result is needed. And a person who goes for a check-up because they feel unwell, gets a worrying number and starts something new that week is likely to see a better number the next time, whatever it was they started.

What a placebo does, and what it doesn’t

The third reason is the one people talk about most, and it gets more credit than it has earned.

A placebo is a dummy treatment: a sugar pill, an injection of salt water, or anything that looks like a treatment but has nothing in it that should work. In 1955 an American anesthesiologist, Henry Beecher, published a paper with a title people still quote, “The Powerful Placebo”. Gathering fifteen studies of different illnesses, he reported that about 35% of the 1,082 patients had been satisfactorily relieved by a placebo alone. Forty years later it was still the paper on placebos that people cited most, and the placebo effect had come to be treated as a scientific fact.

Look at what was counted, though. Beecher’s 35% were people who got a dummy and then got better. That number includes everything in this chapter so far: colds and pains that eased on their own, symptoms that come and go, people who came in at their worst. Two German researchers, Gunver Kienle and Helmut Kiene, went through his studies again in the 1990s. In their reading, the improvements could be put down to other things: illnesses that passed, symptoms that wandered up and down, regression to the mean, other treatments the patients were getting, and patients being polite to the doctors who asked how they were. They found no sign of a placebo effect in any of the studies, and concluded there most likely was none. Not everyone who studies placebos goes that far. But whichever reading is right, Beecher’s 35% was never a measure of what the dummy itself did.

The way to see what a dummy itself does is to compare people given a placebo with people given nothing at all. Two researchers in Copenhagen, Asbjørn Hróbjartsson and Peter Gøtzsche, gathered every study they could find that did this. By 2010 they had more than 200, covering 60 illnesses and conditions. Their answer was that placebos don’t have important effects in general. In things that can be measured, such as blood pressure or weight, they showed little or nothing. Where the result was what patients themselves said, especially about pain and nausea, a placebo often helped a little. In their first review, pain was eased by about 6 points on a scale of 100, on average, though it varied from almost nothing to quite a lot even among the best studies. And they couldn’t be sure how much of that was real relief and how much was people reporting more kindly after being treated.

One small study, run in Boston and published in 2011, shows the difference between feeling better and being better. Thirty-nine adults with mild or moderate asthma came in for twelve visits each, and on each one they got one of four things: an inhaler with real medicine, an inhaler with nothing in it, pretend acupuncture, or nothing at all. Each time, the researchers measured how much air they could breathe out in one second, and asked them how much better they felt.

Measured: more air breathed outReal inhaler20%Dummy inhaler≈ 7%Pretend acupuncture≈ 7%Nothing done≈ 7%Felt: patients’ own rating of improvementReal inhaler50%Dummy inhaler45%Pretend acupuncture46%Nothing done21%
Figure 19.2 The same 39 people with asthma, on four kinds of visit. Above, what a machine measured in their lungs; below, how much better they said they felt. The two are different measures, so compare the bars within each chart.

The real inhaler opened their airways by about 20%. The dummy inhaler and the pretend acupuncture did about as much as doing nothing, 7%. But when people were asked how they felt, the dummy inhaler and the needles scored almost as well as the real medicine, 45% and 46% against 50%, and far better than nothing, 21%. Feeling better and breathing better had come apart.

So a placebo is real, but it is small and particular. Being treated can change how people feel and what they report, mainly for symptoms like pain and nausea that live partly in how we notice them. In the asthma study it did no more for anyone’s airways than nothing at all, and in the reviews it did little or nothing to what a machine or a doctor could measure. How big the real part is, and where it matters most, is still argued over by the people who study it. But most of what gets called “the placebo effect” is the first two reasons, wearing a white coat.

The grandmother in the video

Now take the forwarded video, the made-up one in which हिमाली बुटी चिया cures a grandmother’s diabetes in 30 days. Suppose every word of her story is true. She drank the tea for a month, her blood sugar came down, and she feels much better.

Each of the three reasons could have done that. She probably started the tea when she was worried, after a bad reading, and the next reading was likely to be better anyway. Starting a new remedy often comes with other changes, too: eating differently, walking more, taking her doctor’s tablets more carefully. For type 2 diabetes, the NHS in Britain counts eating better, moving more and losing weight as part of the treatment, and says such changes can sometimes bring blood sugar back to a normal level, while warning never to stop diabetes medicine unless a doctor says so. And a month of hope, attention and a daily ritual could well make her feel better, whatever her blood sugar did.

Her story can’t say which of these it was, or whether the tea added anything at all. That isn’t because she is lying or foolish. It’s because nobody saw the month in which she didn’t drink the tea. Every person lives one life, and a story about one life, however true, has nothing to compare it with.

None of this means nothing works. The real inhaler opened those airways by 20%, a change the dummy couldn’t match, and that difference is the medicine. Real remedies show themselves in exactly that way: they do more than getting better anyway. The trouble is that getting better anyway is invisible. To see it, you need to know what happens to people like the grandmother who don’t drink the tea, and that takes a second group of people, and a great deal of care in how both groups are chosen.

Found a mistake? Please tell us.