It's a good time to give a brief update about the experiments that I'm running at work. (The last time I spoke about them was here.) For those who haven't seen one of these posts, I've been chasing an odd idea for months now, on and off. So far, the tests that I've put it to have come up negative - or at least not positive, which is often something different. I've continued to refine my thinking about what's going on while searching for a better system to work with.

The other thing that newcomers to this site need to know is that I can't tell you exactly what this idea is. Now there'ssome great blog material, eh? But I'm hoping that folks will understand. If I were in academia, I wouldn't talk because I would be afraid that someone would scoop me and get priority by publishing first. Since I'm in industry, I worry both about that and the proprietary advantages that this stuff might bring. (If it works out well, there certainly could be some.) Actually, come to think of it, academics think that way too these days, don't they?

I'll try not to be too reminiscent of the South Sea Bubble scam artist that Charles MacKay talks about - the one who advertised partnerships for sale in "An undertaking of great advantage and no man to know what it is." He took in a large sum within a couple of days, and as MacKay tells it, he "was philosopher enough to be content" with his profits, bailed out to France and was never heard from again. At least I'm not selling shares.

On to the news. In my last, er, thrilling installment, I was about to set up (with the help of a colleage over in a pharmacology lab) a new experiment which would be the best shot yet. Well, it took a while to get this one analyzed, during which time the solutions sat patiently in the freezer. I had to remake a standard compound for my colleagues in the analytical group, and they had to work out good methods to run the samples with, which needed something else first, and so on. But we got things together last week, and got the data. And, well. . .the thing might have actually worked.

I was running four separate variations on the same system, and all four did some of what I wanted, apparently. They did it to different degrees, and in a pattern that (if it's real) is quite suggestive. But the whole thing was getting close to the limits of detection - not so close as to make the results totally suspect, fortunately, but everyone involved would be more comfortable with a higher signal/noise ratio.

These results have immediately suggested some follow-up experiments, which I'm going to do with the most promising of those four variations. There are three more components of the system that I'd like to vary. One of them should (helpfully) lead to a stonger signal, which should vary as a function of the component we're adding. That'll be a useful check. And the other two experiments will go the other way and wipe out the effect completely, each by a different route, but only if things are happening by the mechanism I'm postulating. These experiments will carve great swaths through Explanation Space, blanking out whole regions of potential false-positives.

And if these runs go according to predictions, well, I'm going to be in the position of the dog who finally caught a car. Scientists can relate to those dogs, you know. What makes us different from the dogs is, we all know just how we would drive the thing if we caught it, and just where we'd want to go. Woof.