For a few months, I’ve been measuring how well my brain is working using arithmetic problems. Each test session includes 100 simple problems (3+4, 7-0, 4*8) divided into 5 blocks of 20. I type the last digit of the answer as quickly as possible. I got the idea from Tim Lundeen, who got better on a similar task when he increased his DHA intake. My performance on an earlier version of this task was improved by flaxseed oil.

I’ve blogged about this. The virtues of this test include: 1. Fast. Takes only a few minutes. 2. Portable. Requires only a laptop. 3. Many possible answers (1, 2, 3, etc.). This reduces anticipation errors. 4. Many numbers (reaction times) per test. This allows me to get a measure of variability for each session and can correct for the difficulty of the problem. Aspects with room for improvement include: 1. Speed/accuracy tradeoff. Accuracy isn’t fixed. Depending on how accurate I want to be, I’ll go faster or slower. (I aim for 95% correct.) 2. No complex actions. The most enjoyable games have a motor-skill aspect that this task does not.

Here’s the data so far.

The big gap happened because I moved from Beijing to Berkeley. The most fascinating result, of course, is the sudden drop on February 2. Here is a close-up.

The drop was easy to notice. All of a sudden I was faster (and only slightly less accurate). The first test with better performance took place while my landlady, who lives upstairs, was practicing piano. Usually it’s quiet when I test myself. My first thought was that the music had caused the improvement. But it persisted so long after the music had stopped that the music couldn’t be the cause.

Part 2: I think I know what caused it. But there is a big problem with my explanation.

One reason for self-experimentation is very simple: To learn about the effects of a drug you are taking. Is it helping? The medical literature is unlikely to be unbiassed. Someone heavily involved in producing that literature wrote anonymously in the BMJ:

I also do a lot of ghost writing. Sometimes I report good quality studies to which I am proud to contribute, albeit anonymously. Yet, too often, I write so called reviews, amounting to mere panegyrics of the discussed drugs, or I report poorly designed and implemented “epidemiologic” studies, bearing gross biases. Many of the (paid) signing authors of these papers do not read the manuscript, let alone provide feedback. I am surprised at how easily such papers are accepted by some journals and how rarely their flaws are challenged.

Given the financial interests at stake, I do not see what recommendations or regulations will put an end to such long debated [meaning long-criticized] practices.

My first self-experimentation to have unexpected results involved an acne drug. I discovered it was ineffective — might have even been making things worse. Yet it was a standard treatment for acne. My dermatologist was surprised I had bothered to collect the data. “Why did you do that?” he asked.

To do useful self-experimentation here isn’t complicated. The main thing you would do would be to measure the problem before and while you take the drug. Before you take the drug, you’d want to measure the problem long enough so that you had some idea of what would happen if you didn’t take the drug.

If you’ve done this, I’d like to hear about it.