In some scientific experiments in, say, medicine or biology, you have a lot of control over your subjects and your independent variables--you can randomly assign subjects to certain conditions and select the treatments you want to give them. For example, if you're studying the effects of Valium on memory, you get to pick the doses of Valium that you give to people.

In a pseudoexperiment, however, you have little or no control over your subjects and independent variables. For example, suppose you're instead studying the effects of mild head injury on memory. You don't get to choose who has a head injury and who doesn't--you just have to take whoever comes.

Why do you care? Well, it's harder to draw conclusions from pseudoexperiments; in particular, it's generally difficult--even impossible--to make claims about causation. Let's say that in the example study above, you find that people with a history of head injuries do more poorly than control subjects in school. Now, maybe the head injury caused the poor school performance. Then again, maybe these head-injured people were generally dumber to begin with and had a tendency to do dumb things in every aspect of their lives. Then too, maybe these people weren't dumb; maybe they were ADHD types who can't sit still in school but who get thrills from participating in dangerous activities. It's hard to tell.

Okay, so given the problems with pseudoexperiments, why do them at all? Well, as I suggested above, sometimes they're the only way you can address an interesting question or problem. If you're comparing men to women, African-Americans to Caucasians people, children to adolescents, college students to elderly adults, or crack addicts to non-addicts, you don't get to pick who goes in what group--but it's still useful to know if there are differences between them, and careful experiments can help rule out possible causes of those differences.

What this node touches on is a fundamental dichotomy that exists in the scientific method, as is practised today. In science, there are generally two ways of examining a system of interest: by way of mensurative or manipulative investigation. The conclusions that can be drawn from the study (and the statistical methods available to the researchers) vary depending upon which approach is taken.

What leighton describes as a "proper scientific experiment" is in fact the manipulative approach to investigation. As he states, the researcher has considerable control over the subjects (they need not be patients or animals, in passing -- they can be any conceivable object) and can control the system of interest (read: control over independant variables). The advantage to this type of approach is that the results, if they are significant, can be readily attributed to the manipulation. Thus, if treat one group of patients pre-disposed to heart conditions with aspirin and give the others a placebo, the observed effect can be attributed to our manipulation of the system (dosing with AAS). In general, in these types of experiments, the experimental design and subsequent analysis of the data follow the model of an ANOVA.

What leighton describes as a pseudoexperiment is in fact a mensurative experiment, wherein the researcher does not manipulate the system of interest, but instead collects information about the system and attempts to control (statistically) for confounding variables in order to determine whether or not a hypothesized dynamic exists. This is not only a perfectly acceptable method of scientific inquiry, but often is the only one that is available in many fields. For example, epidemiologists, astronomers and ecologists often have no option but to simply observe the system of interest and attempt to draw conclusions from the patterns within the data. Experimental design, in a mensurative experiment, is considerably more complex (at times), and the statistical methods employed are more likely to be correlative or predictive (regression) in nature.

I thought that this needed to be added, given the pejorative nature of the term pseudoexperiment. Many scientists spend their lives and careers working very hard uncovering truths without evern performing a manipulative experiment

Log in or register to write something here or to contact authors.