. 5
( 115 .)


assesses to determine the
on the participantʼs behavior. The essence of experimental research is to manipulate an
in¬‚uence of the independent
independent variable (or two or even more independent variables) and look for related
variable on the participants™
changes in the value of the dependent variable. behavior.

The Equivalence of Groups
The second essential characteristic of an experiment is that there are at least two groups
involved who are comparable at the outset of the experiment. In the simplest type of
experiment, one group of participants receives a treatment (for example, they are told
Social Psychology

there is open seating). The participants who receive the experimental treatment comprise
experimental group the experimental group. To know for sure that an experimental treatment (the indepen-
A group comprising dent variable) is causing a particular effect, you have to compare the behavior of partici-
participants who receive the pants in the experimental group with the behavior of participants who do not receive the
experimental treatment in an
treatment (they are told nothing about seating arrangements). The participants who do
not receive the experimental treatment comprise the control group. A simple example of
control group A group in this strategy is an experiment testing the effects of a drug on behavior. Participants in the
an experiment comprising
experimental group would receive a dose of an active drug (e.g., norepinephrine), whereas
participants who do not
participants in the control group would not receive the drug. The researcher then com-
receive the experimental
pares the behavior of the participants in the experimental and control groups. In essence,
the control group provides a baseline of behavior in the absence of the treatment against
which the behavior of the treated participants is compared.
In the real world of research, the distinction between the experimental and control
groups may not be this obvious. For example, in the Sturmer et al. (2005) experiment
on in-group versus out-group helping, there is no true control group in the true sense of
the concept. Instead, participants in both groups received a “treatment” (i.e., in-group or
out-group information). Most experiments you will encounter will follow this model.
In order to establish a clear cause-and-effect relationship between the independent
and dependent variables in an experiment, the participants in the groups must have the
same characteristics at the outset of the experiment. For example, in the experiment
on norepinephrine and aggression, you would not want to assign individuals with bad
tempers to the 15-mg group. If you did this and found that 15 mg produces the highest
levels of aggression, one could argue that the heightened aggression was due to the fact
that all the participants in that group were hotheads.
The best way to ensure that two or more groups will be comparable at the outset of
an experiment is random assignment of individuals to groups, which means that each
random assignment
participant has an equal chance of being assigned to the experimental or control group.
A method of assigning
Researchers can then be fairly certain that participants with similar characteristics or
participants to groups in
backgrounds are distributed among the groups. If the two or more groups in an experi-
an experiment that involves
each participant™s having ment are comparable at the outset, the experiment is said to have internal validity, and
an equal chance of being it can legitimately demonstrate a causal relationship.
in the experimental or
Researchers are also concerned about another kind of validity, known as external
control group.
validity, or generality. When researchers study how experimental treatments affect
groups of participants, they want to be able to generalize their results to larger popu-
lations. To do so, they have to be reasonably sure that the participants in their experi-
ments are representative (typical) of the population to which they wish to generalize
their results. For example, if the participants of a study were all male science majors at
a small religious college, the researchers could not legitimately generalize the results
to females or mixed populations, to younger or older people, or to music majors. If the
researchers have gotten a representative sample of their population of interest, then
they can legitimately generalize the results to that population, and the study is said to
have external validity.

Controlling Extraneous Variables
The goal of any experiment is to show a clear, unambiguous causal relationship between
the independent and dependent variables. In order to show such a relationship, the
extraneous variable
researcher must ensure that no other variables in¬‚uence the value of the dependent vari-
Any variable not controlled
able. The researcher must tightly control any extraneous variable that might in¬‚uence
by the researcher that could
affect the results of a study. the value of the dependent variable. An extraneous variable is any variable not con-
Chapter 1 Understanding Social Behavior 13

trolled by the researcher that could affect the results. For example, if the temperature in
the room where an experiment is run ¬‚uctuates widely, it could in¬‚uence participantsʼ
behavior. When it is hot, participants may get irritable and impatient. When it is cold,
participants may become sluggish and uninterested in the task at hand.
As just described, extraneous variables affect the outcome of an experiment by
adding a random in¬‚uence on behavior. In short, extraneous variables make it more dif-
¬cult to establish a causal connection between your independent and dependent variable.
In some cases, an extraneous variable can exert a systematic effect on the outcome of
an experiment. This happens when the extraneous variable varies systematically with
confounding variable
the independent variable. The result is that a confounding variable exists in the experi-
An extraneous variable in
ment. For example, letʼs say you are running an experiment on the relationship between
an experiment that varies
frustration and aggression. Participants in the experimental group perform a puzzle for
systematically with the
which there is no solution (frustration group), whereas participants in the control group
independent variable, making
do a puzzle that is solvable (no frustration group). As it happens, on the days when you it dif¬cult or impossible to
run the experimental group, the room you are using is hot and humid, whereas on the establish a causal connection
between the independent and
days when you run the control group, the temperature and humidity are normal. Letʼs
dependent variables.
say you ¬nd that participants in the experimental group show higher levels of aggres-
sion than those in the control group. You want to attribute the difference in aggression
between your two groups to the frustration levels. However, it may be that the higher
levels of aggression recorded in the experimental group are due to the high temperature
and humidity and not the frustrating task.
In the real world of research, confounding is seldom as obvious and blatant as in
our example. More often, confounding results because a researcher is careless when
designing an experiment. Confounding variables often creep into experiments because
independent variables are not clearly de¬ned or executed. The presence of confound-
ing variables in an experiment renders the results useless. The confounding variable
provides an alternative explanation for any results that emerge. Because of this, a clear
causal connection between the independent and dependent variables cannot be estab-
lished. Consequently, it is essential that a researcher identify potential sources of con-
founding and take steps to avoid them. The time to do this is during the design phase of
an experiment. Careful attention to detail when designing an experiment can go a long
way toward achieving an experiment that is free from confounding variables.

Factorial Experiments
An important aspect of real-world research is that experiments are usually more complex
than the simple experimental group/control group design we discussed previously. In
fact, a vast majority of research in social psychology has two or more independent vari-
ables. These are called factorial experiments. factorial experiment
As an example of a simple factorial experiment, consider one conducted by Patricia An experimental design
in which two or more
Oswald (2002) that investigated the effects of two independent variables on willingness
independent variables are
to help. Oswald had participants watch a videotape of a person presented as an older
manipulated, allowing for
adult (Michelle), who was discussing some of her thoughts and emotions about returning
the establishment of a causal
to college. The ¬rst independent variable was whether participants were instructed to connection between the
focus on Michelleʼs thoughts (cognitions) or emotions (affect) while watching her on the independent and dependent
videotape. The second independent variable was the type of affect (positive or negative) variables.
and cognitions (positive or negative) Michelle displayed on the videotape. Participants
¬lled out several measures after watching the videotape, including how much time they
would be willing to devote to helping the student shown on the tape. Before we get to
Oswaldʼs results, letʼs analyze the bene¬ts of doing a factorial experiment.
Social Psychology

The principal bene¬t of doing a factorial experiment as compared to separate one-factor
(i.e., one independent variable each) experiments is that you obtain more information from
the factorial experiment. For example, we can determine the independent effect of each
independent variable on the dependent variable. In Oswaldʼs experiment we determine the
effect of participant focus (the focus on either Michelleʼs affect or cognition) on willingness
to help. This is called a main effect of one independent variable on the dependent variable.
We could also determine, independently, the main effect of the second independent variable
(positive or negative cognition or affect) on the dependent variable.
The main advantage of the factorial experiment lies in the third piece of informa-
interaction When the tion you can determine: the interaction between independent variables. An interaction
effect of one independent exists if the effect of one independent variable (e.g., focus of attention) changes over
variable in a factorial levels of a second (e.g., type of affect displayed). The presence of an interaction indi-
experiment changes over
cates a complex relationship between independent variables. In other words, an interac-
levels of a second, indicating
tion shows that there is no simple effect of either independent variable on the dependent
a complex relationship
variable. For this reason, most social psychological experiments are designed to discover
between independent
variables. interactions between independent variables.
Letʼs go back to Oswaldʼs experiment to see what she found. First, Oswald found
a statistically signi¬cant main effect of focus of attention on willingness to help.
Participants who focused on Michelleʼs affect volunteered more time than those who
focused on Michelleʼs cognitions. If this were all that Oswald found, we would be content
with the conclusion that focus of attention determines helping. However, Oswald also
found a statistically signi¬cant interaction between focus of attention and the type of
affect (positive or negative) Michelle displayed. This interaction is shown in Figure 1.3.
As you can see, focus of attention had a signi¬cant effect when Michelle displayed posi-
tive emotion, but not when she displayed negative emotion. In the light of this interac-
tion, would you still be con¬dent in the broad conclusion that focus of attention affects
helping? Probably not, because whether focus of attention affects helping depends upon
the type of emotion displayed.

Evaluating Experiments
Most of the research studies described in this book are experimental studies. When
evaluating these experiments, ask yourself these questions:
• What was the independent variable, and how was it manipulated?
• What were the experimental and control groups?
• What was the dependent variable?
• What methods were employed to test the hypothesis, and were the methods
• Were there any confounding variables that could provide an alternative
explanation for the results?
• What was found? That is, what changes in the dependent variable were observed
as a function of manipulation of the independent variable?
• What was the nature of the sample used? Was the sample representative of the
general population, or was it limited with respect to demographics, such as age,
gender, culture, or some other set of characteristics?
Chapter 1 Understanding Social Behavior 15

Focus of Attention

Mean Time Volunteered




Figure 1.3
The interaction between
type of affect and focus of
Positive Negative attention.
Type of Affect Based on data from Oswald (2002).

Correlational Research
Although most research in social psychology is experimental, some research is corre-
lational. In correlational research, researchers do not manipulate an independent vari-
able. Instead, they measure two or more dependent variables and look for a relationship
between them. If changes in one variable are associated with changes in another, the
two variables are said to be correlated. When the values of two variables change in the
same direction, increasing or decreasing in value, there is a positive correlation between
them. For example, if you ¬nd that crime increases along with increases in tempera-
ture, a positive correlation exists. When the values change in opposite directions, one
increasing and the other decreasing, there is a negative correlation between the vari-
ables. For example, if you ¬nd that less help is given as the number of bystanders to an
emergency increases, a negative correlation exists. When one variable does not change
systematically with the other, they are uncorrelated.
Even if correlations are found, however, a causal relationship cannot be inferred.
For example, height and weight are correlated with each other”the greater one is,
the greater the other tends to be”but increases in one do not cause increases in the
other. Changes in both are caused by other factors, such as growth hormone and diet.
Correlational research indicates whether changes in one variable are related to changes
in another, but it does not indicate why the changes are related. Cause and effect can be
demonstrated only by experiments.
In correlational studies, researchers are interested in both the direction of the rela-
tionship between the variables (whether it is positive or negative) and the degree, or
strength, of the relationship. They measure these two factors with a special statistical
correlation coef¬cient
test known as the correlation coef¬cient (symbolized as r). The size of the correlation
A statistical technique used
coef¬cient, which can range from “1 through 0 to +1, shows the degree of the rela-
to determine the direction
tionship. A value of r that approaches “1 to +1 indicates a stronger relationship than a
and strength of a relationship
value closer to 0. between two variables.
Social Psychology

In Figure 1.4, the ¬ve graphs illustrate correlations of varying strengths and direc-
tions. Figure 1.4A shows a 0 correlation: Points are scattered at random within the
graph. Figures 1.4B and 1.4C show positive correlations of different strengths. As the
correlation gets stronger, the points start to line up with each other (Figure 1.4B). A
positive correlation positive correlation exists when the values of two variables increase or decrease in the
The direction of a correlation same direction. In a perfect positive correlation (r = +1), all the points line up along a
in which the values of straight line (Figure 1.4C). Notice that in a positive correlation, the points line up along
two variables increase
a line that slopes in an upward direction, beginning at the lower left of the graph and
or decrease in the same
ending at the upper right.
In a negative correlation (shown in Figures 1.4D and 1.4E), the same rules concern-
ing strength apply that held for the positive correlation. However, in a negative cor-
negative correlation
The direction of a correlation relation, as the value of one variable increases the value of a second decreases. Figure
in which the value of one 1.4E shows a perfect negative correlation (“1).


. 5
( 115 .)