Yogurt and Nuts Correlate With Weight Loss: Why?

A new longitudinal study finds:

Despite conventional advice to eat less fat, weight loss was greatest among people who ate more yogurt and nuts, including peanut butter, over each four-year period. . . .

That yogurt, among all foods, was most strongly linked to weight loss was the study’s most surprising dietary finding, the researchers said. Participants who ate more yogurt [than the average for all participants?] lost an average of 0.82 pound every four years.

Why might this be?

Yogurt and peanut butter are both unusual foods. Yogurt is strange because unflavored yogurt has little or no smell. It tastes good for other reasons: strong sourness, creaminess, and coolness. People are also pushed to eat it not only by how pleasant it is to eat but by the thought that it is good for them. Most foods, in the form that we eat them, have a smell. I explain the yogurt results by saying that yogurt consumption replaced consumption of foods with stronger smells.

Peanut butter is unusual because when I was visiting publishers to sell The Shangri-La Diet, I met a woman who told me she had lost weight simply by eating peanut butter — that is, by adding peanut butter to the rest of her diet, making no other changes. I think she ate about 3 tablespoons per day. This predisposed her to think there might be something to my ideas. No one has ever told me such a story about another food. If peanut butter has a smell, it’s really weak. It’s pleasant to eat because of fat content and texture. When I was a boy, my mom made me peanut butter sandwiches (no jam) for school lunch. I never came to like them. This implies I never learned a smell-calorie association. The bread must have supplied a strong fast calorie signal so this implies that the peanut butter generated little or no smell signal.

Thanks to Eri Gentry.

Two Years of Weight Measurements

This shows Justin Wehr‘s weight over two years. He is 26 years old and 6 feet 2 inches tall — at 140 pounds, very thin. The record begins with a switch to a vegan diet. Over three months he lost seven pounds but gradually regained the lost weight, even though his diet didn’t change. In the middle he suddenly gained five pounds on a trip to Alaska and Seattle and then gradually lost it.

He describes his diet like this:

My diet was pretty average Midwestern meat and potatoes sort of thing prior to going vegan-ish, and I emphasize the ish. I’ve been vegan-ish since I started tracking weight, meaning that I don’t buy meat or dairy products at the store, but I’ll happily eat whatever sounds good off of a restaurant menu or whatever is being served when I’m eating at someone else’s place. I intentionally keep my diet very boring. I eat an absurd amount of peanuts and raisins, I estimate in the range of 600 – 900 calories per day. Besides peanuts and raisins, most of my calories come from lentils, frozen vegetables, and bread + olive oil. I drink almost exclusively water, with a few swigs of OJ most days, and have a glass of wine or a bottle of beer on occasion.

The features of this data that interest me are (a) weight loss when he changed what he ate (first three months) and (b) gradual regain of the lost weight (after that). Few theories of weight control can explain the regain. However, the theory behind the Shangri-La Diet can. It says that he initially lost weight because he shifted to foods with relatively weak flavor-calorie associations — weak because the foods were relatively new. As he ate them again and again, the flavor-calorie associations got stronger and this raised his set point.

Calorie Learning: Somewhat Better Design

I’ve done several experiments where I add a bunch of randomly-chosen spices to butter, spread the butter on saltines, eat the saltines, and then follow (or not follow) them with a piece of Wonder Bread eaten with my nose clipped shut.

In the previous experiment in this series, I made two spice mixes, A and B. After eating saltines spread with A, I ate a piece of bread. After eating saltines spread with B, I ate nothing. I did one trial per day, going back and forth between a trial with A and a trial with B. For example, Monday (A Trial), Tuesday (B Trial), Wednesday (A Trial), etc. Here are the results:

In an experiment just finished, I improved this design a bit: 1. I had trials with both A and B on the same day. (I made two new random spice mixes for this experiment.) 2. The first saltines of the day I rated “blind” — that is, without knowing what mix it was. Here are the new results superimposed on the previous results.

So far so good. The new results agree with the earlier results.

If Not Noseclips, Dark Sunglasses?

In this interesting video about losing weight, Paul McKenna, a British hypnotist, recreates a study in which people ate food blindfolded. In the study, they ate one-quarter less when blindfolded than when not blindfolded. This doesn’t impress me; nothing is stopping the blindfolded subjects from eating more at later meals. But it makes me wonder how not seeing your food affects flavor-calorie learning. It might make it stronger (you’re less distracted) or it might make weaker (the sight of food acts like glue to strengthen flavor-calorie associations — there is actually evidence for something like this).

While wearing noseclips while eating with others is too weird, wearing dark sunglasses might not be. And what about listening to music (for distraction) while you eat? My calorie learning experiments are continuing; eventually I should be able to test these possibilities.

Thanks to Gary Skaleski.

Calorie Learning: First Results

I made random flavors by mixing 4 randomly-chosen spice mixtures into butter. I spread the butter on 2 pieces of Wonder Bread. I used each mixture more than once — twice in a row (1st mixture), three times in a row (2nd mixture), four times in a row (3rd mixture). Each trial consisted of a flavor-calorie pairing — flavor from the spices, calories from the bread. Each trial also provided a rating, which measured the learning.

Here are the ratings of how good the bread tasted.

flavor-calorie learning results

This was good. 1. The ratings started near 50 (neutral) each time. I’d like to have a large supply of flavors so that I can start fresh each time. These results suggest that randomly mixing 4 spice mixes provides this. The 4 spice mixes were randomly chosen from 10 spice mixes — so there are a lot of possible combinations. 2. The learning per trial was substantial.

More in the category Calorie Learning.

Calorie Learning: Procedure

In the first experiment, I created a butter mixture with a random flavor and spread it on two pieces of bread, which I ate and rated.

To create the butter mixture with a random flavor, I took 1/2 stick of butter and heated it in a microwave until it was soft. Then I randomly selected four of the 11 Penzeys spice blends and mixed 1/2 tsp of each into the butter.

For each trial, I spread a thin layer of the butter on two pieces of the bread, which was at room temperature. I ate them in a minute or so. I rated how good they tasted on a scale where

90 = very good
80 = quite good
75 = good
70 = somewhat good
60 = slightly good
50 = neutral
40 = slightly bad

I’ve used this scale to rate food dozens of times. Sometimes the rating changes with successive tastes; if so, the assigned rating is the maximum.

After eating the bread I waited at least an hour before eating anything else or brushing my teeth in order to to eliminate interference from other smells.

I did no more than one trial per day. On most days I did a trial.

To find other posts in this series, use the Calorie Learning category.

Calorie Learning: Materials

These are the supplies I used in my calorie-learning experiments:

1. Wonder bread. I wanted bread with as little flavor as possible

2. Unsalted butter. Unsalted because the spice blends have salt.

3. Eleven Penzeys Spices spice blends. In particular, Baking Spice, Cake Spice, Chicken Taco Seasoning, Jerk Pork Seasoning, Poultry Seasoning, Mural of Flavor, Sate Seasoning, Southwest Seasoning, Sweet Curry (regular), Tuscan Sunset, Venison Sausage Seasoning. Each has 5-15 different spices. For example, Jerk Pork Seasoning contains paprika, allspice, ginger, cayenne pepper, sugar, nutmeg, black pepper, garlic, thyme, lemon grass, cinnamon, star anise, cloves, and mace. Baking Spice is a mixture of two kinds of cinnamon, anise seed, allspice, mace, and cardamom. Combining a few of them should produce a flavor unlikely to resemble any familiar flavor.

Calorie Learning: Background

The discovery of flavor-calorie learning (in rats) was no surprise. It was another example of flavor-consequence learning, which was well established. In the 1950s, John Garcia had found that if you make a rat sick after exposing it to a new flavor, it will avoid that flavor. Flavor-consequence learning belongs to the larger category of Pavlovian learning (also called classical conditioning), the sort of learning where an animal learns that an unimportant event (such as a bell) predicts an important event (such as food). Pavlovian learning belongs to the larger category of associative learning, which also includes action-event learning, such as a rat learning that bar presses produce food pellets. The action is pressing the bar; the event is getting a food pellet.

My Ph.D. was in the field of animal learning. Almost all animal learning research is about associative learning. When I taught introductory psychology, however, I found it hard to take advantage of my expertise because most of the research had little real-world relevance. The big exception was Shepard Siegel’s work on drug tolerance and craving. Tolerance and craving are due to Pavlovian learning, Siegel argued. Flavor-calorie learning, happening at every meal, might have been another exception had anything interesting been known about it — but nothing was.

The usual terminology is to say that in a Pavlovian-learning experiment, the animal learns to associate the CS (conditioned stimulus, such as a bell) with the US (unconditioned stimulus, such as food). In flavor-calorie learning experiments, the flavor source is the CS, the calorie source the US.

Calorie Learning: Introduction

In a series of posts, each titled Calorie Learning: [something], I’m going to use a blog to communicate self-experimentation. To see the whole series, look in the category Calorie Learning (under Self-Experimentation).

This research will be about how we (or at least I) learn to associate flavors with calories — more precisely, smells with calories. This learning is at the heart of the Shangri-La Diet, which derives from a theory that says the flavors of your food increase your set point if they are associated with calories. The stronger the association, the bigger the increase.

Why study this? 1. Maybe I can improve the diet. 2. It matters. It happens with every bit of food you eat. It controls what you eat and your appearance (assuming my theory is right). 3. Little is known about it. As I wrote in the appendix to The Shangri-La Diet, Anthony Sclafani has studied this learning extensively in rats. No one has studied it extensively in people. 4. The experiments can be simple and easy — or at least that’s what I think now.

A few weeks ago, a friend told me how much she liked those cellophane-wrapped white-bread sandwiches sold in delis and bodegas. Egg salad sandwiches, for example. They were addictive, she said. That sounded about right: White bread (and bread in general) is digested very fast, witness its very high glycemic index. Fast digestion means the calorie signal it generates in the brain overlaps a great deal with the flavor signal it generates in the brain. The more overlap of the two signals, the stronger the association created. The stronger a flavor’s association with calories, the more you like it.

Her comment gave me an idea: I can create a random new flavor by randomly combining many spices, mixing them into butter, and spreading the butter on white bread. The spices supply the flavor, which I can reproduce as often as I want by making a big enough batch of spicy butter when I start. Spice mixtures are cheap. I can easily and cheaply make a huge number of flavors that should taste entirely new. This means I can start fresh — which is where you want to start when doing a learning experiment — as often as I want. White bread is cheap, easily available, has little flavor, and provides a strong signal per calorie. If I want to increase the time between the flavor and the calories, maybe I can spread the butter on crackers, which have few calories, and eat the bread later.

Will it work? Stay tuned.

Science in Action: Flavor-Calorie Learning (another simple example)

At the heart of the Shangri-La Diet is the idea that we learn to associate flavors (smells) with calories. This learning was first shown in rat experiments. There’s some human evidence, but not much. If I could discover more about what controls this learning, I might be able to improve the diet. For example, maybe I could say more about what the flavor-free window should be.

My earlier self-experimentation on this subject – I used tea for flavor and sugar for calories — was helpful. To my surprise, I found that really small changes in flavor made a noticeable difference. If I switched from one canister of Peet’s Gunpowder Tea to a new canister, the ratings went down, although everything else stayed the same. From this came the notion of ditto food: Foods with exactly the same flavor each time are especially fattening. I hadn’t realized what a difference it would make if you kept the flavor exactly the same each time.

It’s been hard to learn more. After Christmas dinner, my mom gave me the leftover brandy (A. R. Murrow). I used it for a very simple experiment in which I learned to like it. I’ve never drunk brandy in any quantity and I started off not liking it. Every day for a few weeks, I drank one tablespoon. I drank it in a few sips over a few minutes. I didn’t eat anything else for at least 30 minutes. I rated how good it tasted on a 0-100 scale where 10 = very bad, 20= quite bad, 25 = bad, 30 = somewhat bad, 40 = slightly bad, 50 = neutral, 60 = slightly good, 70 = somewhat good, 75 = good, 80 = quite good, 90 = very good. The overall rating was the maximum of the ratings of the several sips. (The first sip usually tasted the best.)

Here are the results.

learning to like brandy

I’ve observed similar results five or six times. They are more support for the most basic conclusions: 1. The effect is very clear. One tablespoon of brandy has only 30 calories. 2. A really simple experiment is easy.

That’s a promising start but then it gets hard, or at least non-obvious. As a way to study flavor-calorie learning, this little example has several flaws: 1. Slow learning. 2. Expensive materials. 3. Little control of flavor. The best I can do is choose which liquor to buy. Soon I will run out of ones I haven’t used. 4. No way to separate flavor and calories in time. 5. No way to change the calorie source.

An earlier demonstration used a soft drink. It’s really Science in Inaction: I’ve made zero progress in a year.

Science in Action: Flavor-Calorie Learning (simple example)

The Shangri-La Diet is partly based on the idea that we learn to associate flavors and calories. A food’s flavors become associated with the calories in the food. This association makes the flavor more pleasant.

I would like to learn more about this associative process so I have been studying it. Here is a simple example. At intervals of a day or so between bottles, I drank 4 bottles of a lemongrass-flavored soda. I chose that flavor because it was unfamiliar. Each bottle had 50 calories of cane sugar. I rated how pleasant each bottle tasted on a scale where 40 = slightly unpleasant, 50 = neutral, 60 = slightly pleasant, and 70 = somewhat pleasant. I drank the bottles between meals — far away from other food.

Here are the ratings.

The flavor gradually became more pleasant.