Association and Causation

I recently read a great cat annecdote in Temple Grandin's book Animals Make Us Human.

According to Grandin, someone's cat loved watching the water swirl around when a toilet was flushed. The cat couldn't figure out how to flush itself, but it had noticed that when there was paper in the toilet, it was more apt to be flushed.

So the cat tore up toilet paper and threw it in the toilet and waited expectantly.

This is a perfect example of the difference between association and cause. The cat correctly noted that toilet paper was associated with toilet flushing. But the cat incorrectly decided that toilet paper caused toilet flushing.

Of course, flushing didn't cause toilet paper any more than toilet paper caused flushing. It was a third factor, pulling the handle on the toilet, that caused the flushing.

Many of our scientific interpretations are like the cat's. If we see a fat person eating more than a thin person, most people conclude that overeating causes obesity. But perhaps obesity causes increased appetite, or perhaps a third factor that no one has discovered yet causes both obesity and increased appetite.

Whenever we see research that shows that two factors are associated, we should think of this cat story. Does the research provide any evidence that one factor caused the other, or is the researcher simply thinking like the cat because of a preconceived notion?

Gary Taubes is challenging the catlike assumption that because fat people tend to eat a lot and not exercise much, it's their behavior that is causing their obesity. Instead, he says it's insulin that is making the body store fat instead of burning it, and the resulting energy deficit makes the person want to eat more and exercise less.

The blogosphere is filled with people debating this theory. I won't go through it all here.

My point here is simply that we should keep this graphic cat story in mind as we evaluate evidence. Surely we humans can be smarter than a cat!