experimentation in psychological research

 

definition

 

Experiments are tests of hypotheses in which investigators manipulate an independent variable in order to determine a cause and effect relationship between the independent and dependent variable.

variables involved

independent variables

In any experiment, the variables that an experimenter manipulates are called independent variables.  Independent variables are manipulated when an experimenter randomly assigns participants to experience different levels of the independent variable.

 

Changes in the level of an independent variable cause changes in a dependent variable if the experiment has high internal validity.  Internal validity is the extent to which we can assume that changes in the dependent variable are due to the independent variable and not to some extraneous/confounding variable.

 

There are two different types of independent variables:

 

1.                   Between-subject variables are variables in which participants experience only 1 level of the independent variable.

2.                   Within-subject variables are variables in which participants experience all levels of the independent variable.

 

DEPENDENT VARIABLES

 

The variables that are measured by an investigator to determine if the independent variable had the intended effect are dependent variables.  Dependent variables are so named because they depend on the behavior of the participant, which is hopefully affected by the independent variable. 

 

Changes in the dependent variable are caused by changes in the independent variable if the dependent variable has been measured sensitively.  A sensitive dependent variable has no range effects.  A range effect occurs when all scores on the dependent variable occur within a limited range, either at the top of the scale (i.e., a ceiling effect) or at the bottom of the scale (i.e., a floor effect).  Range effects are detrimental because they mask relationships between variables.

 

control variables

 

A control variable is a potential independent variable that is held constant during the experiment.  In a properly designed experiment (i.e., one with high internal validity) all variables other than the independent variable are controlled.

 

CONFOUNDING/EXTRANEOUS VARIABLES

 

A confounding variable is a variable that offers an alternative explanation for the results of the experiment.  

 

In a well-designed experiment (i.e., one with no confounds, and therefore high internal validity) we can be fairly certain that our dependent variable was caused by our independent variable.  In a poorly designed experiment, we cannot be sure if the dependent variable was caused by the independent variable or some other variable (i.e., a confounding/extraneous variable). 

 

Confounding occurs when an influence other than the independent variable (i.e., an extraneous variable) has a differential effect between the groups on the dependent variable in an experiment.  Confounding does NOT occur when an extraneous variable affects all groups equally.

 

properly designed experiments

 

REQUIREMENTS

 

Experiments may occur in the laboratory or in a field setting.  Experiments conducted outside the laboratory are called field experiments.  In either setting, there are two essential requirements for any properly designed experiment:

 

1.      Potential confounding variables must be controlled.  There are two ways to control extraneous variables:

 

  1. All potential confounding variables are held constant.  That is, they are made into control variables (see above).
  2. All potential confounding variables are randomized.  If a confounding variable can't be controlled, its effect must randomized across all levels of the independent variable.

 

2.      Manipulating an independent variable by randomly assigning participants to groups must occur.  Manipulating an independent variable controls for the direction of causation and third variable problems associated with a relational design by equating the groups in the study.

 

It is much more difficult to establish control in field experiments versus laboratory experiments, however field experiments could be considered a more realistic setting than the laboratory.

 

FORMS OF WELL-DESIGNED EXPERIMENTS

 

Well-designed experiments can take two forms:

 

1.      Post-test only design.  In a post-test only design, participants are randomly assigned to a level of the independent variable, experience the manipulation, and then respond to the dependent variable.  

 

Scores between the groups on the post-test are then compared to determine if the independent variable had an effect.

 

2.      Pre-test, post-test design.  In a pre-test, post-test design, participants are measured on the dependent variable, randomly assigned to a level of the independent variable, experience the manipulation, then respond to the same dependent variable a second time.

 

Advantages to the pre-post design include:

 

  1. Change scores (i.e., the difference between pre-test and post-test scores) can be calculated in addition to comparing scores between the groups on the post-test.
  2. Equal groups assumption is testable.
  3. Allows for identification of participants of interest.
  4. Mortality is identifiable.  Mortality refers to the loss of participants though the course of the experiment.

 

Disadvantages to the pre-post design include:

 

  1. They are time consuming.
  2. Participants may become sensitized to the materials (i.e., the dependent variable), leading to hypothesis guessing.

 

the advantage of experiments

 

The most important advantage the experiments have over relational and descriptive research is the ability to uncover cause and effect relationships.

 

disadvantages of experiments

 

There are several disadvantages to conducting experiments.  They include:

 

1.      Reactivity refers to the idea that an experimenter can change behavior merely by observing it.  Individuals who know they are being watched may behave differently than they would otherwise.

An apprehensive subject feels uncomfortable about being evaluated in an experiment due to evaluation apprehension.

 

Deception, honest experiments and unobtrusive measures are often used to counteract reactivity.

 

2.      Demand characteristics are those cues in an experiment that participants may use to guess the hypothesis, and therefore alter their behavior.  This altered behavior may take different forms:

 

            i.      The good subject will do anything necessary to confirm what (s)he believes the hypotheses to be.

          ii.      The negativistic subject will attempt to sabotage the experiment.

 

3.      Generalizibility.  The particular participants in the experiment may be different than the population to which the results are to be applied.  The setting of the experiment may be greatly different than the real world; in such a case the ecological validity of the experiment may be in question.

 

4.      Artificiality.  Because many experiments occur in the laboratory, many critics argue that the atmosphere is to far removed from real-life scenarios (i.e., too artificial).  There are two different types of realism:

 

         i.      Mundane realism occurs when the setting of an experiment is very similar to non-laboratory settings.

       ii.      Experimental realism occurs when participants are psychologically involved in the experiment.