Experimental control

At minimum, an experiment must have at least 2 levels of some factor (the experimental variable). Otherwise, there is nothing to compare.

  Treatment     Result
Level 1       Outcome 1
Level 2       Outcome 2

Assigning subjects to conditions (levels of the independent variable)

Participants in an experiment are often referred to as "subjects." As the characteristics of individual subjects may influence the experimental outcome, it is very important to control for the extraneous variables, such as individual differences. There are 2 general ways to do this.

  1. Random assignment - use a table of random numbers to assign subjects to the experimental conditions (levels of the independent variable), More....
  2. Matched groups - assign subjects to the experimental conditions so that the 2 or more groups are alike on the characteristics that might influence the outcome (e.g., gender, intelligence, experience, age, etc.). More....

Controlling for experimenter characteristics

The characteristics of the person running the experiment might also influence the outcome. Ways to contol for experimenter error:

  1. Have the same experimenter run all subject groups.
  2. If more than one experimenter is required,
    a. provide training so that all behave in the same way
    b. have each run the same number of subjects
    c. randomly assign subjects to experimenter

Setting, equipment, procedure

The situation in which an experiment takes place is a source of extraneous variables. Features of the room, equipment, temperature, time of day, etc. should be held constant (kept the same) for all participants. Be sure that all the equipment works and that the procedure itself stays constant (except for the experimental manipulation).

Presentation: Reduce confounding

There are two ways to reduce confounding when presenting research materials (stimuli) to or eliciting responses from research participants: counterbalancing and random order of presentation.

1. Counterbalancing - the order of presentation of stimuli is reversed for one-half of the subjects. For example, in an experiment the information seen first might be remembered more easily. That is counteracted by having one-half of the subjects see the materials in reverse order. For example,
      Order in which stimuli are presented to subjects (e.g., word order in a memory study)
      Subject #1 Word 1 Word 2 Word 3 Word 4 Word 5
      Subject #2
      Word 5 Word 4 Word 3 Word 2 Word 1



Random order of presentation - stimuli are presented in random order to each subjects, or subsets of subjects. In this way, extraneous variables (sources of error) are randomly distributed across the treatment groups. A table of random numbers can be used to assign the order of presentation to individual subjects.
      Order in which stimuli are presented to subjects (e.g., word order in a memory study)
      Subject #1 Word 4 Word 2 Word 5 Word 3 Word 1
      Subject #2
      Word 3 Word 1 Word 4 Word 5 Word 2

Blind and double-blind procedures

A blind procedure is one in which the subjects are unaware of the treatment condition to which they have been assigned. They don't know whether they are getting the active ingredient or the placebo; or whether they are tasting Coke or Pepsi. The blind procedure helps to control for the effects of expectation or suggestion (for example, thinking they have consumed alcohol even when they have not can lead some people to report feeling high).

Expermenters are also subject to expectation effects. It is good practice, when possible, for the person running the subjects to be unaware of the specific treatment condition. In a taste trial, another person can be responsible for labeling the materials, keeping the specifics from the experimenter until afterwards. In a placebo study, the experimenter observing or collecting information from the subject should not know whether the subject is under the influence of an active substance or a placebo. This is termed a double-blind procedure, as both the subject and the experimenter are kept blind to the level of the independent variable.

Using subjects as their own controls

The same individuals are run under all of the treatment conditions (levels of the independent/experimental variable). Participants' performance can be assessed both before and after some treatment. In a study on the effects of math tutoring, the outcome variable would be the amount of improvement. The "after" score is compared to the "before" score. In placebo studies, the same subjects can be run under both treatment and control conditions. The treatment is alternated randomly with placebo, preferably under double-blind conditions (see above).

Self-test #2 - experimental control

Continue to Factorial design (it is short)