As promised in the previous post, some thoughts on superstition in science. This was originally posted in October 2004, and astute readers may note that my opening comments about sports went horribly awry not long after. I take this as proof of my point: talking about these things only screws them up.
Long-time readers of this site may have noticed a lack of sport-related posts in recent weeks, despite the fact that my teams are doing pretty well at the moment. This is not really a coincidence– I’m as surprised as anyone to see the Giants winning games (though you will note that they lost to the only really good team they’ve played…), but they’ve been winning while I haven’t been talking about them, so why screw up a good thing? The Woof Gods are mighty, and utterly without pity– just ask Curt Schilling (that “nothing better than making 55,000 New Yorkers shut up” comment didn’t work out so well, did it?).
Of course, this may seem a strange attitude for a scientist to have. We are, after all, supposed to be rational and completely above superstition, at least if you believe Hollywood. I have two responses to that: first, that sports fandom transcends science– that’s a given. The second response is that people tend to underrate the role of superstition in experimental science.
I’ve previously mentioned the Ti:Sapph laser I used in grad school, that every post-doc in the lab had a different can’t-miss trick for adjusting, none of which worked for me. We had a second laser of the same model, that served as the main trapping laser for the system, which had to be locked to a particular frequency each morning.
As a young graduate student, I had the lock procedure explained to me by the senior student who had built it. You would switch the laser controller to scan mode, turn the scan rate up to its highest setting, adjust the frequency until the laser was sweeping over the desired transition, then shut off the scan mode, use the fine control on the lock circuit to tune the frequency to the transition you wanted to lock to, and engage the lock. Then you had to turn the scan rate all the way down.
Now, there’s no earthly reason why the scan rate should make any difference at that point. The controller had been taken out of the scan mode, so the scan rate knob wasn’t really connected to anything. Matt swore that it was an important step, though, so I smiled, and nodded, and proceeded to ignore it for the first month or two that I was running the experiment.
And I’ll be damned if he wasn’t right. If you didn’t turn down the scan rate after locking the laser, it just didn’t stay locked as well as it should. There was absolutely no reason for it, but it was perfectly repeatable over a period of months. So I added it back to the routine, and it was an essential part of the start-up procedure from then on. A couple of years later, when I had to explain the system to a post-doc, he didn’t believe me, either, until he saw it for himself.
Since then, I’ve become accustomed to superstition in science. Things get turned on in a certain order, whether or not there’s a clear reason for it, because “it just works better that way.” The first couple of data points you try to take are always screwy in some way, so you might as well just skip a couple of shots, whether you have reason to think there’s something screwy about them or not. You always touch something metal before adjusting a laser.
Many of the little routines we go through have a logical basis– diode lasers are very sensitive to static electricity, so it’s a good idea to ground yourself before touching the laser, to prevent a static spark that could kill it. They tend to morph into something more reflexive, though– the last time I played with an argon laser, I caught myself carefully touching metal objects before touching the controls, even though static isn’t a concern there. It’s just become part of the process of laser adjustment for me. In a similar way, there’s obviously some logical explanation for the scan rate affecting the laser lock, back at NIST, but we never figured it out. It was easier to just take it as a fact of nature, and move on.
Talking about this process at lunch one day (specifically, the scan rate thing), a colleague described a study he’d seen in which a bunch of pigeons were put in a cage with a food dispenser and a light. After an initial period in which the light would flash before food was dispensed, they put a random interval between the light and the food. A pigeon who expected food when the light came on would be surprised to find that it didn’t come out for a second or two… right after it twitched its wings in a certain way. So the next time the light flashed, the pigeon would do a little wing twitch. And then a head bob. And after a while, whenever the light would flash, the pigeon would go into this complicated and entirely random dance, in the belief that it was essential for causing the food to appear.
There are days in the lab when I think we’re nothing but a bunch of dancing pigeons.