Seth’s blog

The most basic question about the Shangri-La Diet is how well it works. As the host of The Amazing Race might say, there are many ways to answer this question, each with their own pros and cons. Thanks to Rey Arbolay, we can answer this question in a new way: by looking at the data in the Post Your Tracking Data Here section of the Shangri-La Diet forums, where more than 80 people have posted.

The goal of SLD, of course, is weight loss. The problem with just plotting pounds lost as a function of time on SLD is that a loss of 10 pounds is quite different to someone who starts at 120 pounds and someone who starts at 300 pounds. To deal with this I have divided the results by starting weight. In the graphs below, the ranges of weights were chosen so that there would be roughly the same number of people in each graph.

Starting weight makes a big difference — at least, between the two extremes. Many of the people who aren’t losing very fast are in the lowest quartile. There are still a few outliers — people at the higher starting weights who are losing slowly if at all — but not many.

The next step is to do an analysis that estimates the effect of starting weight and somehow removes it. It might be better to divide people by BMI rather than starting weight, and maybe the y axis can be improved.

One of the most fascinating stories in Benzer’s oral history interview is about construction of the Salk Institute in La Jolla, California:

Benzer: Louis Kahn [the architect] asked Salk, “How much money have you got to put into the building?” And Salk said, “Ten million for endowment” — this was all from the commitment from the March of Dimes — “and a million dollars a year for operating expenses in perpetuity.” So Kahn went home and designed a building for $20 million. In fact, he bragged about this at some dinner he had in La Jolla. He talked about other buildings he had designed and he said it was always his policy to make the building for twice as much as the money available, because you could always count on the fact that people scurry around to find the extra money.

Salk went for that idea on the argument that later on it would cost much more to build it. That was absolutely true. But at the time it had the effect of liquidating the endowment. And everything suffered from then on. The institute . . . was always worrying about where the next buck was coming from.
. . .
Interviewer: So they liquidated their entire endowment to construct a more expensive building?

Benzer: Yes.

Kahn knew a general principal about human nature that I do not. Why do backers reliably “scurry around to find the extra money”? Something powerful is at work here.

I found a long interview with Seymour Benzer, a biologist at Caltech, who is one of my favorite scientists — lots of creative and important work. I was pleased to learn he is a foodie. During a 1956 trip to Japan he had sushi for the first time. “One of the greatest things about the trip,” he said in 1990 (when the interview took place), presaging a future in which every upscale American supermarket sells sushi. (For dinner tonight I made salmon tartare.) When I was a student at Caltech I knew the other students liked him, but I never met him.

Benzer began the use of fruit flies to study behavior. At Woods Hole I took a course called Neural Systems and Behavior with a fruit-fly segment taught by Laurie Tompkins. She had met Benzer at a party. When she told him she studied fruit-fly mating, Benzer asked if they have orgasms. Very early in his work on behavior, he gave a talk to Roger Sperry’s lab about his plans. After his talk there was a lot of debate about it. Some people thought it was very promising; others thought it was nonsense.

Interviewer: Why were people so skeptical?

Benzer: Why? A lack of imagination.

Excellent answer. I would have said: People are always skeptical.

I reduced the amount of omega-3 in my diet. I stopped taking flax-seed oil capsules (I had been taking 10 1000-mg capsules/day) and started drinking extra light olive oil (2 tablespoons/day) instead of walnut oil. I made the change at midnight: Tuesday high, Wednesday low. The graph below shows measurements of my balance.

From Saturday through Tuesday, and preceding days, my intake of omega-3 was high; on Wednesday and Thursday it was low.

My balance was worse Wednesday morning than expected by extrapolation, which supports the idea I started with: omega-3 affects my balance. The time course of the change (the impairment was clear in hours) resembles the original observation: I could put on my shoes while standing much more easily the morning after the day I increased my omega-3 consumption.

But, as you can see, there was a problem: My balance rapidly improved during the low omega-3 condition. Although the results support my original idea, they don’t support it as strongly as they might. A comment on a previous post was “Aren’t you worried that your expectation of worse balance will skew the results?” No, I’m not I thought when I read it. I had several reasons for not worrying about the effect of expectations, and now another has come along: Surprising results, which imply that expectations have little effect. I did not expect significant improvement from practice. I had believed that because I balance everyday for hours while standing and walking, there would not be a large practice effect. I was wrong.

Psychologists don’t know much about motor learning. There are few well-established empirical generalizations about what makes motor learning faster or slower. Another gap in our knowledge is about the nature of the underlying change. When you get better with practice, how does your brain change?

After I shifted to low omega-3, I was surprised not only by how much I improved but also by how quickly. Was my improvement due simply to more tests? I plotted my scores versus test number:

This graph suggests that I improved more per test (greater slope) during the low-omega-3 condition than during the high-omega-3 condition. I think it is a spacing effect: During the low-omega-3 condition, I tested more often. During the high-omega-3 condition, I did 14 tests in 3.5 days — 4.0/day. During the low-omega-3 condition, I did 11 tests in 1.2 days — 9.1/day. I tested more often because I wanted to track the decrease. I think this difference in test rate is the reason for the slope difference. This effect is the opposite of the usual spacing effect in learning experiments, in which close-together (”massed”) practice is less effective than widely-spaced practice.

Relevant to the theme of inspiration via self-experiment, these results and my experience gave me several new (at least to me) ideas about motor learning. One was the existence of this spacing effect. Another was that practice changes the brain by increasing how much of the brain is devoted to the task. (The areas used for other tasks shrink.) Practice increases accuracy because more neurons become involved. The output, the action, is an average from a larger sample. One reason I thought of this is that after lots of practice, and I became quite accurate, the circular area on which I was balancing seemed larger. The notion that the brain area used by the task gets larger helps explain the spacing effect. Spacing is important because the brain doesn’t care how often you have done something in the distant past; what matters is how often you are doing it now. Thus the spacing effect helps make efficient use of scarce resources (neurons). The spacing effect occurs because neural activity causes an increase in something (call it X) that slowly fades away. If later activity happens while X is above a threshold, neural rewiring occurs.

The big practice effects and the idea that practice is more powerful when more frequent should interest anyone who wants to improve their balance (and probably other motor skills), from athletes to the elderly. In a simple cheap easy safe way I got better quickly–too quickly, actually. What happened reminds me of Little League: My batting got much better when I started swinging a bat in my backyard.

The lesson for my experimental design is that I should reduce and keep more constant how often I test.

As I mentioned earlier, while measuring my balance I’ve been listening to a book called Cod: The Fish that Changed the World. Around Hour 4 of the book I realized it was related to what I was doing: fish, brain food. Duh!

Each test of balance consists of 5 warmup trials followed by 15 regular trials. Each trial generates one number, a duration: how long I stand on one foot before the other foot touches the floor. It’s is a bit like surfing–balance, balance, balance, balance, balance, wipe out. (Surfers, skateboarders, skiers, snowboarders, gymnasts . . . this may interest you.) I enter the stopwatch times directly into my laptop. Each test lasts about 12 minutes. Because of the book, they’re pleasant.

I made several more baseline measurements of my balance with two changes:

1. To reduce fluctuations in the concentration of omega-3 in my brain, I did my best to take the flaxseed oil capsules as evenly spaced as possible. The general rule was to take 1 every 2.4 hours (= 10 per day). I didn’t take the capsules with me when I left home but I did follow that rule when I was home (not waking up to take them, however).

2. To make the distribution of (log) balance times more Gaussian (normal), I raised the maximum possible time from 30 seconds to 60 seconds. Previously I had stopped the test at 30 seconds; now the cutoff was 60 seconds. The problem was 30 seconds was too common — my balance was too good. Before the change, 3% of baseline measurements (6 out of 210) were 30 seconds. After the change, 12% of measurements (25 out of 210) were between 30 and 60 seconds and <1% (1 out of 210) were 60 seconds.

The graph below shows results (mean & standard error) for 28 sessions.

The early problem (first 10 tests), discussed in my previous post, was that the means were fluctuating too much. A one-way ANOVA, with each test a different level, gave F(9, 140) = 2.6, p = 0.008. This is why I started trying to evenly distribute the flax capsules over the day. This seemed to work. For the last 18 tests, F(17, 252) = 1.0, p = 0.4. Unfortunately there is obviously an upward trend but that is okay because the change I am going to make — much less omega-3 — should if anything impair balance.