A cat and email — even scientists can be fooled

There are so many studies, stories and researchers repeating the same fallacies of logic that it might be helpful to arm ourselves with a quick review. Let’s take the biggest fallacy of the day: “correlation is causation.”

Most people understand, at least intellectually, that correlation does not prove causation, no matter how strong the relationship may be. Wearing a bra has been associated with a 12,500 times greater risk for breast cancer, but no one would seriously believe that proves bras cause cancer.

In real life, this fallacy can sometimes be harder to see, especially if the correlation is something that’s popular and we’ve come to believe it to be a cause. There’s a term for this, too: confirmation bias, where we look for things that confirm what we believe. And the popularity of a belief — the more people who believe something — is often used as added confirmation of its truth. All of us fall victim to this fallacy on a regular basis, according to Dr. Steven Novella, a neurologist at Yale University School of Medicine.

It’s become popular to believe that eating too much, or eating “bad” food, causes obesity, so we only see and remember things that confirm our beliefs. But the sources of these beliefs are little more than correlations. “I saw a fat person eat a pink cupcake (or fill in the blank with whatever is the “bad” food or beverage of the day). — Ah ha! Proof that fat people are gluttons.” Only objective, systematic scientific investigations, as in well-designed, randomized, placebo-controlled intervention trials, can bring us to credible causes. But the science that disproves our beliefs is vigorously ignored.

A refresher of the correlation-causation fallacy can be found in an article, “Loopy Loops:”

You are about to learn of a beverage so dangerous, that we must ban or restrict its sales, or at least enact tax penalties on it to deter consumption. Here's what the research shows:

· Every American who drinks it dies.

· It's been linked to obesity: in fact, bigger people drink the most of it.

· It's associated with type 2 diabetes and all diabetics drink it in especially large amounts.

· All heart attack victims drink it and it's a known factor in heart failure.

There are been hundreds of studies finding these correlations — correlations so strong they make the evidence irrefutable. This is bad stuff....[rest of article here]

This fallacy is even not uncommon in medicine, as assumptions and preconceived beliefs about people or conditions confuse association with causation. This makes healthcare professionals prone to “attribution errors,” meaning symptoms and problems are attributed to the stereotype, explained Dr. Jerome Groopman, M.D., chief of experimental medicine at Beth Israel Deaconess Medical Center and author of How Doctors Think.

Epidemiological or observational “studies” are where we’re most likely to encounter misleading claims of causation, as reviewed here. It’s easy to forget that this line of research is only able to dredge up correlations. I call data dredge studies “Rorschach tests” of epidemiology, because their computer models can pull out patterns in almost unlimited combinations and conclude just about anything the researcher sets out to find. “Researchers are sometimes guilty of confirmation bias by setting up experiments or framing their data in ways that will tend to confirm their hypotheses,” said Robert Todd Carroll, author of The Skeptic’s Dictionary.

The biggest and longest-running Rorschach test is the Nurses Health Study — a huge quarry of questionnaires gathered since 1976 from over 120,000 nurses — that’s led to over 500 “studies” using its data. Others are the Framingham Heart Study, the Health Professionals Study, the Iowa Women’s Health Study and countless other databases that have been used to pull out select data to find all sorts of links. It’s easy to forget that the correlations they dredge up frequently contradict each other. And we’re likely to only hear about the ones they want us to hear: for instance that soda is linked to obesity, but not the link with whole grains, legumes and low-fat dairy products.

But this association = causation fallacy is increasingly not playing with the public anymore, as people are catching onto it. So now, more researchers are claiming that they can prove cause when they can show a “direction” to their associations.

In other words, the claim goes:

“If A happens before B, then A causes B.”

Of course, they are still just showing correlations which are still every bit as spurious for demonstrating causation alone. But, if A is something that sounds intuitively correct to us as a cause, then it’s easy to get taken in by this one ... even professionals who should know better.

But this is one of the most well-known fallacies of logic, called post-hoc ergo propter hoc. It is a Latin phrase that means “after this therefore because of this” and it’s committed when people assume that because something occurred after something else, it must have happened as a result. Just because one thing follows another does not mean that it was caused by it.

You can have endless fun with this one. Try this at home! I’ll start with a few favorites:

Killing turkeys cause winter.

Roosters crowing cause the sun to rise.

Last rites kill people.

Mouth washes cause cancer.

Think of all the nonsensical things that happened before the onset of the “obesity epidemic” that we can blame it on:

bell bottoms, platform shoes, Disco or Rap, mini skirts, penny loafers, Tony the Tiger or Captain Crunch, big hair and hairspray, health foods, pop or diet pop, disposable diapers, TV dinners, less air pollution, air conditioning, drive-through diners, the internet and email...

Peer-reviewed journals are even vulnerable to these fallacies of logic. The New England Journal of Medicine has done it again. Besides publishing an article last week by authors claiming to have shown that thoughts shared via email from fat friends across the country can make us fat, they published the story of Oscar the cat! It opens: “Oscar the Cat awakens from his nap, opening a single eye to survey his kingdom....” It seems Oscar is believed to be able to predict who will die. Even Dr. R.W. Donnell questioned why the Journal is publishing fiction without scientific rigor. Dr. Steven Novella at Neurologica examines the fallacies of logic behind uncritical stories like these, cautioning:

[B]efore we waste too much time trying to explain the nature of a phenomenon we should first confirm that the phenomenon exists. This is a chronic problem in the paranormal research community, who spend far too much time weaving explanations out of quantum mechanics for phenomena, like ESP, that probably don’t exist....

However, by printing the story they lend a great deal of credibility to the implicit claim that Oscar’s abilities are genuine. In so doing the NEJM has contributed in a small but real way to the scientific illiteracy of the public. The story reinforces the false notion that uncontrolled observations can be reliable and that seeming fantastical, even paranormal, powers are taken seriously by scientists. They missed the opportunity to teach the public about the pitfalls of bias in uncontrolled observation and the need for scientific study before concluding a phenomenon is real.

© 2007 Sandy Szwarc