I looked up celiac disease and autism on pubmed the other day.  My mother’s secretary has a daughter recently diagnosed with autism, and diagnosed several years with celiac disease.  So we were sitting in the dining room and I’m snacking and looking up stuff on pubmed.

“There’s not much research, but the two studies I’ve found on autism and celiac did not find a link, except for this one quack guy,” I say, meaning Andrew Wakefield.

“Okay,” my mom says.  “So they haven’t done the research confirming it yet.”

“No,” I say.  “There are two existing studies that have looked, and they did not find evidence of a link.  I’m looking on pubmed, so if there were more studies they would very likely be there.”

“Ah,” my mom says. “So all we’ve got now to go on is anecdotal evidence.”

“No,” I say again.  “Studies looked.  They looked for a link, and they didn’t find a link, suggesting there’s not a link.”

I think at that point we detected mutually incompatible approaches to uncovering truth, and dropped the conversation.


It’s a normal human thing to figure out what you believe is true (often by assuming anecdotes are representative of overall reality), and then seek out social back-up to help convince others of it.  In this approach to truth (which drives scientists up the wall), if science is used, it is used to support one’s own truth claims to others.  If someone is not trying to make truth claims to others, then there is no need for science; it doesn’t tell you anything you don’t already know.

In contrast, the role of science as-generally-agreed-upon is to test what we believe to be true to see whether it really is true. In this approach to science, science can disconfirm anecdotes, and its role is to drive what people believe, not just to back what they already do believe.  So it’s much less useful for normal human goals.

Frustrating, that.   Also frustrating that we don’t have much research on a lot of things.  It could be true that the two studies on autism and celiac disease didn’t pick up on an actual connection…but the point is, two studies that show no connection are a lot more meaningful than no studies that show no connection.