Human experimental psychologists (also called cognitive psychologists) are in a curious position. Their subject — the human brain — is obviously the most complicated thing studied by any science. Its components (neurons) are not only very numerous and densely-connected they are also very inaccessible. Moreover brains soak up their environments in a way that other objects of study do not. It isn’t impossible to do experiments, but it isn’t easy. You can’t keep a supply of humans in your lab, for example. The difficulty of human experimental psychology is the main reason I decided to study animal experimental psychology. But the complexity of the brain is not only a difficulty but also an advantage: It means there is the most to be learned.
It is also easy to argue that human experimental psychologists study the most important subject of any science. Advances in understanding the human brain go “straight to the bottom line” — namely, human welfare and happiness — in a way that is true for few other sciences. Mood disorders and learning problems — not to mention obesity and poor sleep — cause a huge amount of suffering. Brain dysfunction is behind all of them. Norman Temple and I have argued that the sort of low-tech vary-the-environment type of research done by experimental psychologists is the most likely to produce useful results.
Given such a task, human experimental psychologists are lucky to be able to use an unusual tool: self-experimentation. Self-experimentation makes it remarkably easy to study topics with practical value (as I have). Because our world is already built around requirements of the human brain, self-experimentation requires no special equipment, no laboratory, and can be done at almost no cost. No other science, except human nutrition, has anything like this.
But human experimental psychologists don’t do self-experimentation! (There are a few exceptions, such as psychophysics.) The standard arguments for the avoidance don’t withstand scrutiny. Standard argument #1: Experimenter bias. The experimenter’s expectations may influence the results. Rebuttal: The human experimental psychologists who say this don’t practice what they preach: They don’t run their experiments blinded. Very few psychology experiments are run blinded. Standards argument #2: Lack of generalizability. You don’t know how general the results will be. Rebuttal: Yes you do. Typical psychology experiments have on the order of 8 subjects. They can be so small and still get reliable results because all the subjects change in the same direction. This means that if you know how one subject has changed you can predict how the others will change. In other words, the whole history of human experimental psychology — tens of thousands of experiments, a vast amount of data — shows that yes, you can safely generalize from one subject.
Because the stated reasons for not doing self-experimentation are so easy to rebut, the actual reasons may have more to do with human nature. My guess is that they are some combination of: (a) Fear of being different — different from other psychologists (the who-goes-first problem) and, especially, different from scientists in other areas, few of whom can self-experiment. (b) Desire for prestige. A large grant, a large lab, and activities in which others follow your orders are inherently more prestige-enhancing than doing something all by yourself.
This is why human experimental psychology, as currently practiced, really is science with one hand behind your back. Too bad it matters so much.
Thanks to Saul Sternberg for a thought-provoking discussion.