Chapter 7 Basic issues in experimental research
A well designed experiment has three essential properties:
1. The researcher must vary at least one independent variable to assess its effects on
participants’ responses;
2. The researcher must have the power to assign participants to the various
experimental conditions in a way that ensures their initial equivalence;
3. The researcher must control all extraneous variables that may influence participants’
responses.
Manipulating the independent variable
The logic of experimentation stipulates that researchers vary conditions that are under their
control to assess the effects of those different conditions on participants’ behavior. By seeing
how participants’ behavior varies with changes in the conditions controlled by the
experimenter, we can then determine whether those variables affect participants’ behavior.
This is a very different strategy than used with correlational research.
Independent variables
In every experiment, the researcher varies or manipulates one or more independent
variables to assess their effect on participants’ behavior. An independent variable must have
two or more levels. The levels refer to the different values of the independent variable. Often
researchers refer to the different levels of the independent variable as the experimental
conditions. Sometimes the levels of the independent variable involve quantitative differences
in the independent variable. In other experiments, the levels involve qualitative differences in
the independent variable.
Types of independent variables
Independent variables in behavioral research can be roughly classified into three types:
environmental, instructional, and invasive. Environmental manipulations involve
experimental modifications of aspects of the research setting. Instructional manipulations
vary the independent variable through instructions of informations that participants receive.
Invasive manipulations involve creating physical changes in the participant’s body through
physical stimulation, surgery, or the administration of drugs.
Experimental and control groups
In some experiments, one level of the independent variable involves the absence of the
variable of interest. Participants who receive a nonzero level of the independent variable
compose the experimental groups, and those who receive a zero level of the independent
variable make up the control group. Although control groups are useful in many experimental
investigations, they are not always used or even necessary. Researchers must decide
whether a control groups will help them interpret the results of a particular study. Control
groups are particularly important when the researcher wants to know the baseline level of a
behavior in the absence of the independent variable.
Assessing the impact of independent variables
Many experiments fail, not because the hypotheses being tested are incorrect but rather
because the independent variable was not manipulated successfully. If the independent
variable is not strong enough to produce the predicted effects, the study is doomed from the
outset. Researcher often pilot test the levels of the independent variables they plan to use,
trying them out on a handful of participants before actually starting the experiment. The
purpose of pilot testing is not to see whether the independent variables produce
hypothesized effects on participants’ behavior, but rather to ensure that the levels of the
independent variable are different enough to be detected by participants.
, In addition to pilot testing levels of the independent variable while designing a study,
researchers often use manipulation checks in the experiment itself. A manipulation check is
a question that is designed to determine whether the independent variable was manipulated
successfully. Although manipulation checks are not always necessary, researchers should
always consider whether they are needed to document the strength of the independent
variable in a particular study.
Independent variables versus subject variables
Researchers sometimes include other variables in their experimental designs that they do
not manipulate. These kinds of non-manipulated variables are not ‘independent variables’,
because they are not experimentally manipulated by the researcher. Rather, they are subject
or participant variables that reflect existing characteristics of the participants. Designs that
include both independent and subject variables are common and quite useful, but we should
be careful to distinguish the true independent variables that are manipulated from the subject
variables that are measured but not manipulated.
Dependent variables
In an experiment, the researcher is interested in the effect of the independent variables on
one or more dependent variables. A dependent variable is the response being measured in
the study. In behavioral research, dependent variables typically involve either observations
of actual behavior, self-report measures, or measures of physiological reactions.
Assigning participants to conditions
A strategy for testing the effects of independent variables on behavior makes sense only if
we can assume that our groups of participants are roughly equivalent at the beginning of the
study. If we see differences in the behavior of participants in various experimental conditions
at the end of the experiment, we want to have confidence that these differences were
produced by the independent variable. Thus, an essential ingredient for every experiment is
that the researcher takes steps to ensure the initial equivalence of the groups before the
introduction of the independent variable.
Simple random assignment
The easiest way to be sure that the experimental groups are roughly equivalent before
manipulating the independent variable is to use simple random assignment. Simple random
assignment involves placing participants in conditions in such a way that every participant
has an equal probability of being placed in any experimental condition. It ensures that, on
the average, participants in the groups do not differ.
Matched random assignment
Researchers sometimes try to increase the similarity among the experimental groups by
using matched random assignment. When matched random assignment is used, the
researcher obtains participants’ scores on a measure known to be relevant to the outcome of
the experiment. Typically, this variable is a pretest measure of the dependent variable. Then
participants are ranked on this measure from highest to lowest. The researcher then
matches participants by putting them in clusters or blocks of size k, where k is the number of
conditions in the experiment. The first k participants with the highest scores are matched
together into a cluster, the next k participants are matched together, etc. Then the
researcher randomly assigns the k participants in each cluster to each of the experimental
conditions.
Repeated measures designs
When different participants are randomly assigned to each of the conditions in an
experiment, the design is called a randomized groups design, also called a between-
