Designing An Experiment: The Variables and the Groups

 
Psychology is not a science because it has all of the answers, but because of how it goes about answering its questions.  Any field of study that uses the "Scientific Method" is a science.  The scientific method is nothing more, nor less, than a set of procedures that we use to avoid certain common problems that would be more likely to occur with any other method of answering our questions. 

In truth, there is no one, single scientific method.  Several scientific methodologies have been developed and are commonly used.  Any Introduction to Psychology textbook will discuss the use of (for example) naturalistic observations, surveys, case studies, correlational studies, and experiments as scientific methods of investigation. 

Of all of these, the experiment is perhaps the most scientific of the scientific methods.  It is the only one that allows us to establish cause and effect.  For a scientific study to be an experiment, several conditions must be met.  One variable must be manipulated or changed, while another is measured.  Any other variable that could affect the measured variable must be controlled.  There must be at least two groups of subjects (or participants).  And the data must be analyzed to determine the likelihood of the outcome occurring simply by chance. 

The Variables in an Experiment 

The variable that is manipulated, or changed, is known as the independent variable.  The independent variable will be some sort of stimulus or condition.  The variable that is measured is called the dependent variable.  The dependent variable will be some sort of behavior or response. 

Any variable that could have an effect on the dependent variable (our subjects' behavior), other than the independent variable (the stimulus or condition that we want to learn about), is known as an extraneous variable.  In the simplest experiment, one would have a single independent variable and a single dependent variable, but there will always be many extraneous variables.  These extraneous variables must be controlled to keep them from affecting the dependent variable.  The logic is that if the only difference is in the manipulation of the independent variable, then any differences in the dependent variable must be due to the independent variable. 

Extraneous variables can be controlled in two ways.  The first is to hold constant while the second is to allow random variation. 

If we think that the time of day might affect performance on the dependent variable, we could run the experiment at 10:00 a.m. for all of the subjects.  If we hold the time of day constant, we might assume that whatever effect it might have will be the same for all subjects.  The problem with holding variables constant is this reduces generalizability.  If we only run our subjects at 10:00 a.m., the only thing we only know with any certainty is how they are affected at 10:00 a.m.  The more variables we hold constant, the more we reduce generalizability. 

To avoid reducing the generalizability of our findings, we might prefer to allow random variation when we can.  To control an extraneous variable by allowing it to vary randomly, we must meet two criteria.  First, the variable must be normally distributed.  See your textbook for a description of the normal distribution.  And, second, we must have a large enough sample that we have a high probability of getting subjects from all across the distribution. 

The Groups in an Experiment  

In order for a study to be an experiment, there must be at least one of each of two types of groups.  There must be at least one experimental group, and there must be at least one control group.  When we manipulate the independent variable, we produce at least two conditions.  The group that gets the unusual condition is, by convention, referred to as the experimental group.  The group that gets the usual or normal condition is referred to as the control group.  If we manipulate the independent variable in such a way as to produce more than two conditions, then we will have more than two groups.  We can have more than one experimental group, or more than one control group, but we must still have at least one of each. 

The purpose of the control group is to allow us to compare measurements of the experimental group's behavior, with an equivalent group.  That is why the extraneous variables must be controlled.  If we do a good job of controlling the extraneous variables in our experiment, the only difference between what the experimental group and the control group experience in our experiment is the independent variable.  Therefore, any differences we find between the behavior of our experimental group and our control group should be caused by our manipulation of the independent variable.  This is how we establish cause and effect: the effect of X on Y. 
 

Example:  
I want to find out: What is the effect of anxiety (X) on student performance on an exam (Y)? 

I would want to create two groups of subjects.  For one group (the experimental group), I will create a condition that leads to an unusually high level of anxiety in my subjects.  For the other group (the control group) I will try to create a condition that produces only the normal level of anxiety that students might experience in taking a test. 

Since I would want the two groups to be equivalent in every way except for level of anxiety (the independent variable), I must control any other variable that might affect student performance (the dependent variable).  Whether I hold sex, age, intelligence, amount of study, etc., (some of the extraneous variables) constant or allow them to vary randomly, in the end I want my two groups to be as equivalent as possible in terms of sex, age, intelligence, amount of study, etc., so that I will know that anxiety — and only anxiety — caused whatever difference in performance that I find between my two groups. 
 

Note:  Using two groups of subjects in this manner is referred to as using a between-subjects design.  An alternative way to produce the same effect is to use what is called a within-subjects design.  I could decide to test all of my subjects twice — once using a "high-anxiety" condition, and again using a "normal-anxiety" condition.  In effect, I am creating two "groups" composed of the same people, rather like making a "before- and after-" comparison.  Since I am testing the same people in both conditions, both "groups" will be equivalent. 
 
After setting up my groups, I would present a body of information, create the anxiety condition, and then administer the test.  After both groups have been tested, I would use descriptive statistics to summarize the performance of each of the groups.  I would use inferential statistics to determine whether or not any difference in the performance of my groups is statistically significant — that is, whether the difference is likely to have occurred by chance rather than because of my manipulation of the independent variable (anxiety).