Rules/Slogans/Take-Home Lessons From Chapter 2

1. Most studies are not experiments: If participants are not randomly assigned to different treatments, the study is probably not an experiment.
2. If the study is not an experiment, it will not definitively establish the cause of an effect.
3. To establish that a factor causes an effect (finding out "what started it?"), you must establish much more than that changes in the alleged cause are related to changes in the suspected "effect." In other words, correlation is not causation.
4. Correlation between two variables is not causation because
   1. If we don't know which variable changed first, what we think is the cause may be the effect. (Magicians often deliberately fool us about the cause by fooling us about what came first [e.g., "I wrote your word down before you said it"] --but most do it much more successfully and less humorously than this guy).
   2. Both variables may be effects of some third variable--and because there are usually an almost infinite number of third variables--eliminating all those variables as causes is difficult. (Mother Nature, like a magician, often distracts us from the true cause of an effect).
5. No matter what you do, if you find a difference between your treatment group and your no-treatment group, that difference may have nothing to do with the treatment (In a sense, when you have two groups, you are always comparing apples with oranges .)
   
   However, if you use random assignment, you can say how likely it is that the difference you observed is due to comparing apples with oranges. To be more specific, if you use random assignment, you can use statistics to find out how often you would have found this difference if there had been no treatment effect. 
6. Nobody sees constructs. Constructs, if they exist (after all, constructs are "constructed"--that is, "made up") can only be inferred.
7.  Mind control and mind-reading are difficult, so be skeptical of claims about manipulating or measuring constructs.
8. Manipulations and measures of constructs are rarely 100% pure.
9. Generalizing your results to people or situations that you did not study may lead to overgeneralizing. Random sampling from the population and replicating results with different populations in different situations may increase your ability to make accurate generalizations.
10. The question of external validity is how well you can generalize from a study done on particular participants, in a particular place, at a particular time.
11. If you establish that the treatment changed behavior, you have internal validity. However, you may still be concerned about construct validity: what is it about the treatment that created the change in behavior--and what does the behavior mean? That is, the names (abstract labels) of the variables you claim  your treatment is manipulating and your measure is measuring may be wrong.
12. Human animals and non-human animals have rights.
13. Don't be a Nazi--provide informed consent before the study, allow people to withdraw during the study, and debrief after the study.
14. Never conduct a study without first obtaining approval from your professor and from your school's ethics board (if doing research with humans, that would be your school's Institutional Review Board (IRB); if doing research with animals, that would be your school's Institutional Animal Care and Use Committee (IACUC).