To make an experiment of this, we decide to have a carrot and no carrot group. We find that everybody in the carrot group dies the next day. Unfortunately, we cannot attribute this to carrots because we have asbestos in our experimental lab, and only the carrot group was exposed, not the no carrot group. This is called a confound, and a better experiment would try to eliminate it.
So we bring both the experimental and control group into the lab, giving carrots to the experimental group and not to the control group. Unfortunately, everybody in the experimental, and nobody in the control group, dies the next day. However, we cannot blame the carrots yet--the mortality of the experimental group may have been due to the fact that we gave them something--not carrots in particular, just something. If we gave the control group something as well--let's say apples--then they might have died too. Apples in this experiment are what we call a placebo.
So we give the first people who show up apples, and give the last people who show up carrots. We find that everyone who got carrots died before those who got apples. However, we cannot yet blame the carrots. We know that old people are slower than young people, and therefore more likely to show up late. What we need is random assignment to experimental versus control groups.
In order to have an experiment, there must be (1) random assignment to groups, and (2) manipulation of an independent variable (in this case, carrots versus apples). These two features make it possible to draw causal conclusions from experimental results.
Correlation is a necessary, but not sufficient, condition for causation. A correlation can be attenuated (lessened) through either measurement error or multiple determination. Measurement error occurs, for example, when a participant is filling out a self-report questionnaire about achievement motivation and can't recall accurately that he or she doesn't usually study until the last minute. Multiple determination occurs, for example, when a participant is filling out a self-report questionnaire about achievement motivation as part of a job interview, and decides to say he or she usually studies for a test well ahead of time, even though he or she knows this is not true. (This attempt to look good is called a halo effect.)
The most important correlational method used by personality psychologists is factor analysis. Factor analysis is the basis of such trait approaches as the five-factor model and general intelligence (g).
Activity: Come up with two studies having to do with personality, one correlational and one experimental.