How we change what others think, feel, believe and do

Types of experiment

Explanations > Social Research > Design > Types of experiment
True experiments | Quasi-experiments | Non-experiments | See also

When designing a piece of social research, there are three broad types of experiment, based on the experimental rigor used.

A simple division is as follows:

Randomized assignment

True experiment

Control group?
Y	Quasi-experiment
N	Non-experiment

True experiments

True experiments use randomized choice, selecting subjects and methods in a way that prevents bias in results.

This method is particularly important when seeking to demonstrate a cause-and-effect relationship.

In design notation, the letter R is used to show randomized assignment and a simple test with a control might look like this:

R	X	O
R		O

Quasi-experiments

Quasi-experiments do not use proper random assignment, typically they recruit people in a way that can cause bias, such as using 'people on the street'.

In compensation for this, other methods are used to increase the reliability and validity of the experiment, for example by using a control group.

In design notation, a test starts with 'N' for 'non-random' and could look like this:

N	O	X	O
N	O		O

Non-experiments

Non-experiments make no attempt to conform with experimental concerns such as randomized selection of participants or use of control groups.

Note that this does not make them invalid and many useful surveys are carried out in this way. In particular, descriptive research may use this approach as all it seeks to do is gather data and describe what is found.

For example a simple survey and analysis of this, in design notation, would look like this: