Posted by: W. E. Poplaski | March 1, 2009

THE COMPLEAT EXPERIMENTER: 1. What is an Experiment?


~ ♦ ~      ~ ♦ ~      ~ ♦ ~

O, sir, doubt not that experimenting is an art; is it not an art to tease out the native hue of resolution?

~ ♦ ~      ~ ♦ ~      ~ ♦ ~

What is an experiment?

An experiment is a set of procedures for determining the effect that changing one variable (i.e., the explanatory variable) has on another variable (i.e., the response variable).  The explanatory variable is manipulated by the experimenter and its different levels are called treatments.  (See Data, Description and Experimentation for more about explanatory and response variables)

A typical simple experiment has two treatment levels of the explanatory variable.  One level is called the “control” treatment; the other level is the experimental treatment.  The control treatment has the explanatory variable set at what is considered normal or usual.  The experimental treatment is a change in the explanatory variable from what is usual.

The treatments are applied to experimental units (or, in human studies: subjects or participants) and then the response variable is measured on each experimental unit.  The sample size (n) is the number of experimental units that receive a treatment.  For example, an experimenter testing the usefulness of a new drug might use a placebo—a pill without the drug—as the control treatment for thirty subjects, and a pill that includes the drug as the experimental treatment with another thirty subjects.  The sample size is thirty and the only difference between the control and experimental treatments is the absence or presence of the drug in each set of thirty subjects.

Inferring causation means the experimenter concludes that change in one variable causes a change in another variable. Three principles of experimentation allow the experimenter to infer causation: (1) control, (2) randomization and (3) replication.

Control means that all important extraneous variables of the environment are held constant during the experiment.  For example, in an experiment to determine the effect of light intensity on the growth of ‘Purple Wave’ hybrid petunia plants, the extraneous variables would include temperature, fertilizer, water, and growing space.  So, it just wouldn’t do to grow the set of plants receiving the high light intensity treatment at 5º C while those treated with ‘normal’ light intensity are grown at 35º C—or worse, yet—to not even know what temperatures the two sets of plants were exposed to during the experiment.  Any changes in growth between the two sets of plants could be attributed to temperature differences just as much as to differences in light intensity, thwarting the experiment’s goal to infer causation.  Likewise, if the other extraneous variables are not controlled, they too would confound the results (confounding is a situation in which one is not able to distinguish between contributing effects). Therefore, all experimental units should be exposed to the extraneous variables in exactly the same way.

Randomization refers to randomly assigning the experimental units to treatment levels.  The experimental units themselves are never identical; each unit is unique in some, perhaps unknown, way.  For example, if the experimenter used seeds from two envelopes—produced in different batches by the nursery—then assigning the seeds in one envelop to the high light intensity treatment and the other to the normal light intensity treatment confounds the results.  Why?  Because the experimenter would be unable to determine if the plants grew better/worse due to differences in light intensity or because the seeds’ vigor differed for each envelop.  The solution is to randomly assign an experimental unit (seed from each packet) to a treatment level so that some seeds from each packet are assigned to each treatment level.  This works because randomization breaks up any unexpected associations between the experimental units and the treatment levels.

Replication refers to repeating the entire experiment—including the assignment of treatments to a new set of experimental units.  Depending on costs and resources, many experimenters choose to conduct an experiment with three replicates.  For example, the experimenter may question whether the light sources for the petunia experiment are very precise.  So, she conducts the experiment with three different devices for high light intensity, and likewise the control with three different devices (as well as three different sets of experimental units for each treatment level; her experiment now has three replicates).

[Note:  the term ‘replication’ also is often used, in another related sense, as a synonym for ‘sample size’.]

So, the idea behind an experiment is to manipulate—make a change in—the explanatory variable applied to experimental units, measure any changes in the response variable of those same experimental units, and then draw a conclusion about the relationship between the two variables.  The experimenter does this while paying attention to the principles of control, randomization and replication.


Note:  The explanatory variable is often called the ‘independent variable’ or ‘treatment factor’ (or simply factor).  The response variable is often called the ‘dependent variable’.


Next, The Compleat Experimenter: 2. Types of Mistakes.

For a more in-depth and rigorous look at experiments, see:


  1. […] The Compleat Experimenter: 1. What is an Experiment? […]

Leave a Reply

Please log in using one of these methods to post your comment: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s


%d bloggers like this: