Self-experimentation can be an effective way to improve one's health*. One of the problems with diet self-experimentation is that it's difficult to know which changes are the direct result of eating a food, and which are the result of preconceived ideas about a food. For example, are you more likely to notice the fact that you're grumpy after drinking milk if you think milk makes people grumpy? Maybe you're grumpy every other day regardless of diet? Placebo effects and conscious/unconscious bias can lead us to erroneous conclusions.
The beauty of the scientific method is that it offers us effective tools to minimize this kind of bias. This is probably its main advantage over more subjective forms of inquiry**. One of the most effective tools in the scientific method's toolbox is a control. This is a measurement that's used to establish a baseline for comparison with the intervention, which is what you're interested in. Without a control measurement, the intervention measurement is typically meaningless. For example, if we give 100 people pills that cure belly button lint, we have to give a different group placebo (sugar) pills. Only the comparison between drug and placebo groups can tell us if the drug worked, because maybe the changing seasons, regular doctor's visits, or having your belly button examined once a week affects the likelihood of lint.
Another tool is called blinding. This is where the patient, and often the doctor and investigators, don't know which pills are placebo and which are drug. This minimizes bias on the part of the patient, and sometimes the doctor and investigators. If the patient knew he were receiving drug rather than placebo, that could influence the outcome. Likewise, investigators who aren't blinded while they're collecting data can unconsciously (or consciously) influence it.
Back to diet. I want to know if I react to wheat. I've been gluten-free for about a month. But if I eat a slice of bread, how can I be sure I'm not experiencing symptoms because I think I should? How about blinding and a non-gluten control?
Procedure for a Blinded Wheat Challenge
1. Find a friend who can help you.
2. Buy a loaf of wheat bread and a loaf of gluten-free bread.
3. Have your friend choose one of the loaves without telling you which he/she chose.
4. Have your friend take 1-3 slices, blend them with water in a blender until smooth. This is to eliminate differences in consistency that could allow you to determine what you're eating. Don't watch your friend do this-- you might recognize the loaf.
5. Pinch your nose and drink the "bread smoothie" (yum!). This is so that you can't identify the bread by taste. Rinse your mouth with water before releasing your nose. Record how you feel in the next few hours and days.
6. Wait a week. This is called a "washout period". Repeat the experiment with the second loaf, attempting to keep everything else about the experiment as similar as possible.
7. Compare how you felt each time. Have your friend "unblind" you by telling you which bread you ate on each day. If you experienced symptoms during the wheat challenge but not the control challenge, you may be sensitive to wheat.
If you want to take this to the next level of scientific rigor, repeat the procedure several times to see if the result is consistent. The larger the effect, the fewer times you need to repeat it to be confident in the result.
* Although it can also be disastrous. People who get into the most trouble are "extreme thinkers" who have a tendency to take an idea too far, e.g., avoid all animal foods, avoid all carbohydrate, avoid all fat, run two marathons a week, etc.
** More subjective forms of inquiry have their own advantages.