Last week, I tried to measure again the time course of flaxseed oil’s effect on how well my brain works. As before, I used a letter-counting test. The test consists of trials where I see a four-letter display such as ECQZ and type as quickly as possible how many letters from ABCD are among them (in this case, 1). 200 trials per test, about one test per hour.

On Tuesday, about 3 pm, I drank 4 tablespoons of flaxseed oil. Here’s what happened:

The flaxseed oil seemed to reduce reaction time. The maximum reduction was 40 milliseconds, which happened 2-3 hours after drinking the oil. The effect was gone about 6 hours after drinking it.

The next day I expected my scores to be close to the pre-drink baseline. At 5 pm my score was much lower than expected. The difference from baseline was close to the effect of flaxseed oil; moreover, it disappeared at close to the same speed as the flaxseed-oil effect disappeared.

Although surprising, this had a plausible explanation: About three hours earlier, I had eaten three eggs from grass-fed (also called range-fed) chickens. (More precisely, I had had one egg in a smoothie at 11 am and two scrambled eggs at 2 pm.) Such eggs are believed to be high in omega-3. A 1992 paper compared the amount of omega-3 in supermarket eggs and eggs from a Greek farm, where the chickens ate “fresh green grass leaves and wild plants including purslane . . . fresh and dry figs, barley flour . . . insects of all kinds.” The supermarket eggs had little omega-3; the Greek eggs 10 times more.

My eggs came from the Bay Area Meat CSA, run by Tamar Adler (tamareadler a/t earthlink.net), a chef at Chez Panisse, who is looking for new members. Pickups in Berkeley and San Francisco.

In honor of this week’s BMJ cover story.

My current test to study omega-3s (letter counting)nvolves lots of typing. To avoid carpal tunnel syndrome, which I started to approach a few weeks ago, I use one hand to raise the other one, as these pictures show.

Since I started doing this, I haven’t had any problems. No discomfort. I usually put an hour or more between tests. (Each test involves 200 keystrokes — 50/finger — in a few minutes.)

During a recent trip to Los Angeles, I continued my self-experimental activities — three mental tests, which I did once/day. On the last day of the trip, my scores were much better than usual. There was an obvious explanation: I had taken my daily flaxseed oil (4 T) closer to the time of the test — 4 hours before rather than 12 hours before. This suggested that flaxseed oil has an effect that happens fast and diminishes quickly. Earlier observations had implied that the effect at least a few days to wear off completely.

Back home, I wanted to measure this effect. I started testing more often. With a two-answer (yes-no) test, I saw the short-lived effect a few times. But accuracy was relatively low (about 90% correct) due to anticipation errors. I switched to a new test that measures how fast I count letters.

After doing the new test about 70 times, my performance was fairly constant. I resumed trying to measure the short-lived effect. At 3 pm six days ago I drank 4 tablespoons of flaxseed oil. Here are the results:

The blue line shows when I took the flaxseed oil. Within a few hours, reaction time sharply decreased. The improvement slowly went away.

Two big conclusions: (1) Here is a new way to see the effect of flaxseed oil. My earlier experiments took a few weeks; this took a few days. (2) Low between-test variability. The cluster of points around the time of the first meal is an example. The one point below the cluster is a counter-example — I have no idea why it was low all of a sudden. But that is rare. Almost always erratic points suggest explanations. During the second meal I drank a sugar-sweetened drink, forgetting previous observations that these drinks lower reaction time for a few hours (no doubt because they increase blood glucose levels).

This experiment has one big flaw, which is that after I took the flaxseed oil I started making more frequent measurements. A year ago, I made the same mistake with my balance experiments. There is a kind of Heisenberg Uncertainty Principle here: The measurement itself — the test — causes learning. Learning lowers the baseline.

I’ll fix this mistake and a few others and do the experiment again.

At a reading, the novelist Dennis McFarland said that the hardest part of writing The Music Room had been after breaks in writing it. Before he could resume, he had to reread what he’d written so far. This became so painful that he forced himself to never stop.

Because of a break due to wrist problems, I’m going to backtrack a little. When my wrist started to hurt, I had been learning a new way to measure brain function. It’s a reaction-time task that I can do almost anywhere. On each trial I see four letters. For example:

The task is to respond as fast as possible how many of the letters are from the set {A, B, C, D}. In this case the answer is 4, so I would type “4″.

Here is another possible display:

The correct answer is 3. The possible answers are 1, 2, 3, and 4; I just leave my fingers resting on those four keyboard keys.

As soon as I respond to one display, the next appears. Each test has 4 blocks of 50 displays (= 200 trials) and takes about 4 minutes.

I slowly got better — faster and more accurate. This graph shows how my reaction times decreased:

When I started the task, I had to hit Enter after typing the answer (e.g., type “3″ then hit “Enter”). After 50 tests, I learned about an R function that got rid of the need to hit Enter after typing the answer. I could just type the answer (e.g., just type “3″).

This shows how my accuracy improved:

The points become more widely spaced around July 24 because at that point I started learning another reaction-time task. After I hurt my wrist I decided I was trying to do too much.

All excited about my two new reaction-time tests — one involving letter counting (if I see GADZ I type “2″), the other involving naming (if I see 8 I type “8″) — I did both of them in close succession this morning. Each has 4 blocks of 50 trials each. After the second test my left wrist hurt. Too much typing. Now I must reduce typing to a minimum for a few days.

I am collecting more self-experimental data than ever before. Partly because I am excited by the prospect of doing food-brain experiments that take just a few days (measuring effects of flaxseed oil and other foods that last a few hours) and partly because I learned how to get R to respond to single keystrokes. (Via the command getGraphicsEvent. Thanks to Greg Snow at Intermountain Healthcare.) This allows for much better reaction-time experiments; no longer do I need to respond and then hit Enter. Because the new method uses graphic windows, I have much better stimulus control.

I converted my letter-counting test (how many ABCD’s in GDKM? for example) to use the new command. Because the new command is so wonderful, I also used it to make a new test involving naming: The task is to type “1″ when I see a 1, “2″ when I see a 2, and so on. With eight possible stimuli (1, 2, 3, 4, 7, 8, 9, 0) and eight possible answers, there should be few anticipation errors. Accuracy should be high. The task takes advantage of the fact that I have already learned to type “1″ when I see a 1, which means there should be less problem with slow learning curves — learning (getting faster) continuing for a long time. The experiments I want to do need a steady baseline.

After running into Greg Niemeyer a few days ago, I realized it would help if I made these tests more game-like — then they would be more fun. I’m not sure how to do this so I hope to talk to Greg about it.