Basic Between-Subjects Designs

Between-Subjects Designs = different subjects in each group (or condition) of the experiment

One Independent Variable: Two group designs

1. Two independent groups = random assignment

a. Control group & experimental group design

b. Two experimental groups design

2. Two matched groups designs = subjects in the groups are matched on an extraneous variable that is highly related to the DV or can significantly influence the DV

a. precision matching = identical scores

b. range matching = pairs of scores must fall within a previously specified range

c. rank-ordered matching = adjacent scores in rank order are matched

An adequate test of some hypotheses can be made with a simple two-group design. However, an adequate test of other hypotheses may require more than two groups.

Basic experimental designs can still have one IV or factor, but with three or more levels. These designs may be:

1. between-subjects designs = different groups of subjects for each level of the independent variable

2. within-subjects designs = the same subjects may be tested at each level of the independent variable

We will initially deal with between-subjects designs.

The technical term for this type of design is:

Multiple-Groups Design = a between-subjects design with:

1. one IV

2. three or more levels of the IV. Often one of the levels is a control condition where the value of the IV is essentially absent or zero. [e.g., drug dose (placebo), levels of frustration (no frustration), peer pressure (no peer pressure)].

In situations where the absence of the IV is not practical or not possible, then one level is often what is the typical or normal [effect of light intensity on work performance (absence of light would make no sense), effect of anxiety level on test performance (almost impossible to alleviate all anxiety associated with a test)].

The most commonly used Multiple Group Design is the Multiple-Independent-Groups Design = subjects are assigned at random to the different groups (levels; treatment conditions; cells in design).

[Note that there are also within-subjects designs, which will be discussed in the next chapter]

Important considerations when using a Multiple-Groups Design.

1. How many levels of the IV are necessary to adequately test the hypothesis?

The answer is not always easy. Let a combination of (1) your hypothesis, (2) past research, and (3) common sense guide your selection of the number of levels.

Sometimes it may be fruitful or even necessary to conduct a pilot study - a mini-experiment - to test a couple of levels on a few subjects. May save time, money effort on testing unnecessary levels.)

At some point, adding another level will not gain much; a point of diminishing returns for the time, effort and money that must go into testing the hypothesis. Usually the best rule of thumb is to use the simplest design with enough levels that will make an adequate test of the hypothesis. [Effect of pay on performance - train subjects to detect planes on a radar screen against background blips - perhaps 4 levels (groups) will be sufficient for an adequate test = no pay, $.50, $1.00 and $2.00 for each correct identification of a plane]

2. What levels of the IV should be used?

Use IV values for the levels that are reasonably separated, yet extreme enough to detect differences; in other words, within a realistic range, not too extreme and not too close together. [e.g., in radar experiment, the money paid should not be $1,000 for each correct hit (this would be unrealistic), and it would not make much sense to test $1.00, $1.01 and $5.00] [Or if I wanted to see the effects of reward on test performance, I might use no reward and at least a couple of different levels of reward, but I would not tell one class that they will get 25 cents for getting an "A" and another class that they would get $25,000 for an "A". I'm sure that some of the students in the 25 cent group would tell me what to do with the money and the $25,000 group would all try to get an "A".]

3. What are the practical limitations on your experiment?

Every addition of another level to an IV requires more subjects, and sometimes more time, money, effort, logistics, etc. on the part of the experimenter.