Seth’s blog

Yesterday I blogged about three new measures of mental function that I have recently started to use. Here are the first results.

1. Memory-scanning.

Each point is a mean; the error bars show standard errors. I expected a within-session warm-up effect (Trial 1 slower than Trials 2-5) but there isn’t one so I use variation across the 5 trials (100 digits each) to get a standard error.

2. Digit span.

The program increases the number of digits to remember by one when I’m right and decreases that number by one when I’m wrong. The test continues until there have been six reversals in direction of the number of digits (e.g., the sequence 5, 6, 7, 8, 7, 6, 7, 8, 9, 8 contains three reversals). Each pair of reversals is averaged to get an estimate of digit span; I use between-average variation to get a standard error.

Digit span slowly improves with practice, other researchers have found. The increase is slow, however — one digit for every two hours doing the task.

3. Speeded arithmetic.

The questions differ greatly in how long they take (7*9 much slower than 1+0) so I fit a model to remove obvious effects and use the variation of the residuals to get standard errors. I think these results are so erratic because I did the test in several different places and in some cases I corrected mistakes (which is very slow). Now I don’t correct mistakes.

My balance measurements, such as this, this, and this, have come close to convincing me that flaxseed oil improves my balance. If a nutrient improves one part of my brain, it will probably improve other parts, too. So I’ve added three more measures of brain function to my daily tests:

1. Memory scanning. A paper-and-pen task. After studying three digits (e.g., “3 7 9″) for a few seconds, I go through a block of digits marking each one “in” (equal to 3, 7, or 9) or “out” (not equal) as fast as possible. Each test consists of 5 blocks of 100 digits. Duration: 5 minutes. Pluses: Similar to a well-studied task (Sternberg’s memory-scanning task). Minuses: Requires a little bit of equipment (sheet of digits).

2. Digit span. I see a series of digits on my laptop screen then try to remember them. The number of digits goes up and down depending on my accuracy. Duration: 4-6 minutes. Pluses: A well-studied task. Quite different than balance, memory scanning, and speeded arithmetic (below). No special equipment. Minuses: Little computation involved, unlike balance.

3. Speeded arithmetic. I do 100 simple arithmetic problems (e.g., 4 + 8, 3 * 5) as fast as possible. Duration: 2 minutes. Pluses: Tim Lundeen found an effect of fish oil on this task. No special equipment. Measures long-term memory retrieval, unlike other tasks. Intense — the 2 minutes are full of mental activity. Minuses: No obvious ones.

One of these may emerge as a better way to study the issue than balance measurements. The biggest problem with balance measurements is strong practice effects. The more often I measured, the better I became. (The area of my brain devoted to the task seemed to increase. The tiny balance platform seemed to grow.) Perhaps practice effects will be less of a problem with at least one of these tasks. Perhaps one of them will show clearer effects of flaxseed oil.

In a comment on an earlier post, someone suggested using chess as a measure. A fun test would be a good addition. Chess has two big problems: 1. Openings are time-consuming and quite different from the rest of the game. 2. If you take longer to make a move you can make a better move. So the amount of time allowed per move must be fixed. Which is less fun.

The BBC is now running a four-part series on medical self-experimentation called Medical Mavericks. The examples have almost nothing in common with what I do. Here are some differences between what might be called “classic” self-experimentation (e.g., Weinberger, Marshall) and “slow” self-experimentation (what I do).

1. Classic SE: Improves medicine. Slow SE: improves my life.

2. Classic SE: few if any numerical data. Slow SE: Lots of numbers.

3. Classic SE: self-experimenter starts with strong opinion. Slow SE: self-experimenter starts without strong opinon.

4. Classic SE: lasts hours or days. Slow SE: May last years.

5. Classic SE: Demonstrative. Slow SE: Exploratory.

Long live both types.

My main posts about omega-3 are:

Background
why I became interested
historical background
the Israeli Paradox
fat and anesthesia
the clouded crystal ball
omega-3 and cancer
unclear effects on heart disease
prison experiment

Procedure
equipment to measure balance
new measures
how best to measure balance
a new test
better measures
hand placement
letter-counting test
circle test

Results
practice effects
balance experiment (high-omega-3 fats vs low-omega-3 fat)
arithmetic (guest post)
arithmetic (continued)
sleep
first results with new measures
time-of-day effect
what’s the best dose?
flaxseed oil vs. olive oil (balance)
flaxseed oil vs. olive oil (memory-scanning)
flaxseed oil vs. olive oil (digit span)
flaxseed oil vs. nothing (balance)
flaxseed oil vs. nothing (arithmetic)
a surprise
time course (1st try)
time course (2nd try, plus egg surprise)
time course (very short experiment)
time course (circle test)

Discussion
sleep effect?
placebo effect?
overall discussion
source of injury-causing falls?
what the results mean
gum surgery cancelled
dental health
sports injuries (part 1)
sports injuries (part 2)
sports injuries (part 3)
sports injuries (part 4)

Overall Summary
abstract of conference talk

Steve Levitt, co-author of Freakonomics, has done me the great favor of bringing my omega-3 self-experimentation to a wider audience in this post. He thinks my results might be due to my expectations. I posted this comment:

Thanks, Steve, for writing about this. Here’s why I think the balance improvements I’ve noticed are unlikely to be due to expectations:

1. I first noticed the effect putting on my shoes the morning after I started taking flaxseed oil. I had been putting on my shoes standing up for two years; until that morning, I had always had trouble. Every morning. (I had expected it to get much easier — practice effect — but it didn’t.) The sudden improvement was a complete surprise. I had never heard of such an effect. I had hoped that flaxseed oil would improve my sleep.

2. The sudden improvement I saw when I switched from 2 tablespoons/day to 3 tablespoons/day was also a surprise, although I realize this may be harder to believe.

3. When I switched from flaxseed oil and walnut oil to sesame oil, I expected my balance to get worse. It did, but not when I expected. (It took 2 days to see a change; I expected to see it on the first day.)

Which is not to say I’m sure. If the effects I’ve seen are repeatable, I’ll test myself not knowing what oil I’ve ingested.

And forgot to sign my name. Oops.

My reading of the data (such as this) is that placebo effects sometimes exist but are vastly overrated — like many dangers.

Addendum: Stephen Dubner, Levitt’s co-author, blogged today that

nearly everything we’ve written about, either in the book or our journalism or the blog, has some element of people worrying too much about something