of Experimental Design Terms
of this glossary is to define some of the fundamental concepts of research design with
which you should be familiar before you sign up for your first experiment. The
definitions may also be helpful for the part of the course dealing with research methods
covered in January when we discuss these and other related concepts more thoroughly. Note
that an introductory discussion of concepts related to research appears in Chapter 1 of
the Gleitman text, pages 13 - 37. See also Gleitman's general glossary
of terms, pages B1 - B 25.
proposition or an assumption that one attempts to verify (or refute) through
experimentation or observation. An example of an hypothesis might be: "Students study more effectively
in quiet than in noisy environments."
experiment, the experimenter deliberately manipulates one or more variables (factors) in
order to determine the effect of this manipulation on another variable (or variables). An
example might be measuring the effect of noise level on subjects' memorization performance of a list of
standard nonsense syllables (such as ZUP, PID, WUX, etc.).
...the "treatment" variable that the
experimenter hypothesizes "has an effect" on some other variable. (See Dependent
Variable, below). In the example above, the independent variable would be the level of
noise (in this case with three levels: low, medium, high). In an experiment, the
independent variable is directly manipulated by the experimenter. But in an observational
study, or when naturalistic observations are used, the independent variable is
not directly manipulated by the experimenter, and the levels of the independent variable
occur naturally and are already given when the study begins.
Group and Control
experiments, the levels of the independent variable consist of only two: a
treatment-present condition and, for comparison purposes, a treatment-absent, or
no-treatment, condition. The group receiving the treatment-present condition (one of the
two levels of the independent variable) is called the experimental group, and the group
receiving the treatment-absent, or no-treatment, condition (the other level of the
independent variable) is called the control group.
variable that the experimenter hypothesizes is "affected by," or "related to," the independent
variable. It is the "outcome" or "effect" variable,
usually a measure of the subjects' performance resulting from changes
in the independent variable. In the example, above, the dependent variable might be the
number of nonsense syllables recalled correctly.
In an experiment, the independent variable is likely not the
only variable potentially responsible for observed changes in the dependent variable. Many
other possible variables, such as the time of day, the mood of the subject, recent news
events, or the weather, might also affect outcomes in an experiment. These error
variables, often unnoticed, unknown, or unmeasured, may be responsible for observed
variations in the dependent variable even within any one group in the experiment in which
the level of the independent, or treatment, variable remains constant. Thus, error
variables account for all the individual differences in responding not accounted for
specifically by changes in the independent variable. Error variables must be controlled,
for example by randomization.
Confounding Variable (or Confound):
A particular error variable whose possible effects on the
dependent variable are completely consistent with the effects of the independent variable.
The presence of a confounding variable precludes being able to ascribe the changes in the
dependent variable exclusively to the independent variable. The changes could also be due
to the confounding variable. Confounding variables must be controlled for, for example,
through randomization or by holding them constant.
One can prevent the effects of a specific, identifiable error
variable from clouding the results of an experiment by holding this error variable
constant. For example, if all subjects are the same age, then variations in age cannot act
as an error variable. A variable that is thus held constant is called a control variable.
(Of course one can then no longer generalize the results to those of ages other than the
age selected for the experiment.)
Null Hypothesis and Alternative Hypothesis:
The null hypothesis refers to the statement that changes in
the independent variable have no effect on the dependent variable and that
therefore whatever difference was found between the experimental and control groups simply
occurred by chance through the influence of random error variables. The alternative
hypothesis refers to the statement that changes in the independent variable really have
an effect on the dependent variable and that the difference in performance between the
experimental and control groups was greater than what would be expected by chance through
only the influence of random error variables. (In the example, above, the null hypothesis
is that changes in noise level have no effect on subjects' recall scores. The
alternative hypothesis is that changes in noise levels have an effect on subjects'
experiments, it is not desirable that participants know the exact nature of the hypothesis
being tested. (There is evidence in certain kinds of experiments that if participants know
what the hypothesis is, they might, either consciously or unconsciously, respond in order
to try to prove the hypothesis correct -or perhaps
than respond "naturally"
In such experiments the experimenter arranges to keep participants temporarily "in
the dark" about the
hypothesis until after the necessary data have been collected. Immediately thereafter,
however, the experimenter is obligated to inform participants about the true nature of the
hypothesis, why the experiment was designed as it was, and what previous investigators in
the relevant areas had found. The experimenter might also ask participants to report to
what extent they "saw through" the experiment and what they thought the
experimenter was trying to find out. (If an experimenter has other participants yet to
test in such an experiment, the experimenter might ask you to keep the information
revealed in the debriefing session confidential until the conclusion of the whole
experiment. Please cooperate with experimenters in this regard.)
[Return to Table of Contents]