PARTI FOUNDATIONS

46 minimizes
assignment the
degree to likelihood
whether,
andthe participants
in one condition
to assess that
researcher cause) manipu- fromthose in another differ
on the
(the possible
in condition
which, a variable to a change
on
leads (e.g. average
three variables intelligence,
the experimenter There are
lated by how personality
extraneou
(the effect). that might influence
another
variable inde they
relevant to
experiments:
measurement of the dependentrespondfactor
to
kinds ofvariables
and extraneous. second form control is
of

pendent, dependent,
variable.
that is actively experimental The
variable is one which involves ensuring that
An independent
bythe experimenter. contro)
(i.e., changed) different experimental conditions
participants in
manipulated like an indepen- are th
a house is
A thermostat
dial in
to the same environmental
stimuli expOsed to
in that it
can be manipulated (e.g.,
dent variable experimenter, i instructions) with
furnace generates. the roOm,
regulate how much heat the varia the independent variable. Thus, exception
at least two levels
(i.e., with of
Experiments involve imental control, there will be good
variable, for example, no exper-
tions) of an independent stimuli
warm setting or environnmental (e.g., an uncontrolled
the thermostat at eithera warm and
setting
who are introduced
who is
friendly vs. oneexperimenter
a cool setting. Participants whois
vari and distant)that might account for gruff
to a level of an independent
(i.e., exposed)
behavior across the experimental differences
able are said to be ina particular experimental in
conditions.
warm setting condition).
condition (e.g.,

often involve more than one inde- Example of a true experiment:

Experiments Effect of
example, you be video
pendent variable. For might cial
games on helpful
proso-
behavior. Gentile
interested in the effect of warm and cool tem and his colleagues (2009) conducted
peratures when the furnace fan is on or off. This the effects of prosocial video
games
research on
on pros0cial
study would be referred to as a "two by two behavior (i.e., "behavior
intended to he
design. If vou were reading about it in a journal ers"). Although considerable research ba
article, would be represented as a 2 (cool vs. ined the behavioral
effects of violent video
it

games
warm) x 2 (fan on vs. fan off) design. little research
has been conducted on the
behay.
A dependent variable is one that is measured ioral effects of prosocialvideo games. The
by the experimenter to
determine if it changes in significance of
social
this line of research is under
response to the manipulation of the
independent scored by the fact that almost all U.S. teens plav
variable. Thus, house temperature (dependent
video games; one survey indicated 99% of bovs
variable) would be measured to see if it
changed and 94% of girls (Lenhart et al., 2008).
with (ie., "depended" on) adjustments of An exper
the iment was set up to test the
thermostat dial (independent hypothesis (predic
variable). An tion) that playing prosocial
extraneous variable is one that the video games would
experimenter increase helpful, prosocial
wishes to hold constant across behavior, whereas
levels of the inde-
playing violent video games
pendent variable so as to rule it out would increase
as a possible unhelpful, aggressive behavior.
reason why the dependent
variable changes in
The independent
variable was the type of video
accordance with the manipulation game, which had
of the inde- three levels: prosocial
pendent variable. In the
home heating example, game, violent game, and
neutral game. There
extraneous variables might include other heat were two dependent vart
ables: prosocial behavior
sources (e.g., open windows,
lit fireplace)
and aggressivebehavior.
that
could affect house Each type of video game was represented
temperature and make it by
dif two
ficult to determine the actual games as follows:
exact influence of prosocial game
the (Super Mario
thermostat setting on Sunshine and Chibi-Robo!), Voe
house temperature.
Experimenters rule out game (Ty the Tasmanian Tiger 2 and Crash
the possible biasing
effects of
extraneous variables by Twinsanity), and neutral game (Pure Pinball and
types ofcontrol., using two
One
form of control is basic Super Monkey Ball Deluxe). (Duringthe experi-
assignmentofparticipants random
to ment, depending on the condition (e.g., violent
ditions (i.e., levels
experimental con
of
the game), one of the two games was randomly
This involves independent variable).
using a procedure (e.g.,
tothe participant to play.) To ensure that
assigned
able of coin toss,
random numbers) that the games selected reflected the intended charac
ensures that all
have the same game type (e.g., the two
"articipants teristics of the violent
probability of being
5signed to each of games were in fact violent in content), a plot
the conditions.
Random
study was carried out in which 27 college students
 Chapter3 Research Methodsin Applied Social Psychology Rragg
played the games and rated each game (using a The participant heard the
7-point scale) on a series of dimensionsthat per experimenter go
through the same cover story (i.e., stated objec
tained to how much their character helped/did tives of the study) and instructions and then
ask
nice things harmed/shot or killed other
for or the partner (confederate) to go to a second
characters. The ratings were combined to yield a cubicle. At no point during
the experiment did
helpful score and a harmful score. The mean the participant see or interact with the
confeder
scores are shown in Table 3.2 and confirm that the ate. Then the participant
played for 20 minutes
independentvariable was effectively constructed either a prosocial, a neutral, or a
violent video
in that the prosocial games and violent games
game, depending on the condition. Immediately
were experienced (rated) by game players to con afterward, the participant selected 11 puzzles
tain relatively high levels of helpful content and for his or her partner to try to solve from a list
harmful content, respectively, and the neutral of 30 puzzles were classified
that by difficulty
games were experienced to contain low levels of level: 10 easy, 10 neutral, and 10 hard. The exper
both helpful and harmful content. imenter explained that if the partner solved 10 of
Participants were 161 American college stu the 11 puzzles in 10minutes, the partner would

dents (66 men, 95 women) who were randomly win a S10 gift certificate. Thus, the participant
assigned to one of the three experimental condi- could help the partner by assigning easy puzzles
tions (levels of the independent variable). When or hurt the partner by assigning hard puzzles.

a participant arrived, the experimenter explained The participant was asked to completea ques
the main purposeof the study was toinvestigate tionnaire while waiting to receive 11 puzzles
how different types of video games influence from the Once the questionnaire was
partner.
performance on puzzles. The participant was completed, the experimenterexplained that the
told that he or she would have a partner, and studywas over and debriefed the participant.
after each had separately played a video game, There were two dependent variables: helpful
they wouldassign each other 1lpuzzles to try to behavior and hurtful behavior. Helpful behavior
solve. The experimenter explained further that was operationally defined as the number of easy

part of the study was to see if motivation to win puzzles participants assigned to their partners,
a reward affects performance.The experimenter whereas hurtful behavior was operationally
added that on a strictly random basis the partner defined as the number of hard puzzles assigned.
had been chosen to be eligible to win a gift cer The results are presented in Figure 3.1. The
tificate i the partner could later solve 10 of
heights of the three columns on the left side of the
ll
puzzles, whereas the participant would not have figure reflect the amount of helpful behavior (i.e.,
such an opportunity. number of easy puzzles assigned to the partner)
After receiving the experimentalinstructions, in the experimental conditions. Consistent with
the participant was asked to wait in a nearby the experimental hypothesis, students who had
cubicle. Soon afterward, the participant over played a prosocial game were more helpful
video
heard the arrival of another person of the same to their partners than were those who had played

sex who actually was an experimental confeder a violent video game (also a neutral game). Also

ate pretending to be the participant's partner. in support of the hypothesis, the three columns

Table3.2 Pilot Test Mean Ratings of Content of Video Games Used in Experiment

Prosocial games Neutral games Violent games

How 2.5
often one does nice things to or for other 5.6 1.1

characters

How 6.2
often one shoots at or kills other characters 1.8 1.1

SOURCE:Gentile et al. (2009).

48 PART I FOUNDATIONS OF APPLIED SOCIAL PSYCHOLOGY

show that those who cause-and-effect relationships. Although tre

on the right side of Figure 3.1

experiments may allow us to draw causal concl

had played a violent game displayed more hurtful
sions, steps must be taken in conductingan
exper
behavior than those who played a prosocial video
iment so that our confidencethat a cause-and-effert
game (also a neutral game). Gentile and his col-
that playing video games relationship has been demonstrated remains high.
leagues (2009) concluded
had caused the players to The steps involve using extreme care in Carrvine
with prosocial content
out the two key types of control: random
become more helpful afterward. They pointed out assign
relationship had not ment and experimentalcontrol. Actually, when we
that this cause-and-effect
can be confident that an experiment has demon
been demonstrated previously in the research lit
researchers also con strated causal relationship, the experimentis said
erature on video games. The

a
cluded that the finding of a cause-and-effect to have typeof validity known as internal validity
a

between video games and

violent Internat validity refers to theextent to which on
relationship
which had been can infer that an independent variable has inlu
aggressive behaviorreplicated that
enced the dependent variable, that is, has caused
well established in the research literature
on video
Bushman, 2001). the difference in behaviors between the partici
games (Anderson 8&

pants in the different experimental conditions.

Falure to achieve internal validity seriously under.
Level of confidencethat a cause-and-effect rela-

demonstrated. important to mines the value of an experiment because conclu

tionship has been It is

and his coresearchers (2009) con sions about causality cannot be made with
note that Gentile
validity directly to
cluded that the above investigation allowed them
Internal is related
confidence.

to draw causal conclusions about the relationship how experimenterhas nullified the infu
well the

video games and players' sub ence of possible extraneous variables through
between playing
social behavior. Recall that Table 3.1 experimentalcontrol and random assignment.
sequent
indicates that experimentaldesigns surpass the Ifwe discern a possible failure to adequately
controlfor an extraneous variable in the Gentile
other basic designs in the ability to demonstrate

Figure 3.1 Helpful and Hurtful Behavior as a Function of Type of Video Game

4
Choices
3
Puzzle
of
Number
Mean

Helpful Choices Hurtful Choices

Type of Behavior

Prosocial Neutral Violent

1though Chapter 3 Research Methods in Applied
causal Social Psychology 49
true
ting
anondi would constitute a threat time and
et al. (2009) study, this among groups with
exper
to internal
thesemptionthat
Fortunately,the experiment
validity. these groups may not be equileht io each 6ther
Jse-and-effe
in
remains
high
seems to have been conducted with proper con (for a thorough review of
thes esigns, see Cook
trols, and we can be confident that
causal rela && Campbell, 1979).
domcarrying
ly, assign
tionships were demonstrated. First, participants fCENTRAL
were randomly assigned to the three experimen Pretest-posttest Drunkdrivers
design.
has
whenwe that LIBRARhaor
tal conditions, thus reducing the
likelihood
concern in many communit. Dunk drivipg has,
demon the differences between the conditions in help been linked to serious
mentis accider ahd traffio fata
mal said fulness and harmfulness could be due to differ
ties. As a response problen, peliee
to this shoni
among the participants in factors (e.g., trait toring programs have been launched in
which
validityy ences many
places in the United States and Canada, and
has one of empathy; personal history of playing violent penal

as influ video games) relevant to the dependent variables. ties for drunk driving have become
more severe.
caused Second, there is no evidence of a breakdown in Do these strategies have any on the number
e effect

partic experimental control in that participants in the of people who engage in this behavior? To address
nditions. three experimental conditions seem to have been this question,we could employ a pretest-posttest
y exposed to the same environmental stimuli (e.g., design (Reichardt & Mark, 1998). Briefly, initial
under
conclu same physical setting, same oral instructions) observations are taken (the pretest), the treatment
le except for the type of video game they played. (e.g., an intervention) is implemented, and obser
with
ated to Had there been a difference between the experi vations are taken again (the posttest) so that a

e influ mental conditions in a potentially relevant comparison can be made between what happened
experimental variable, confidence in the internal before the treatment and what happened after the
arough
validity of the study would have been under treatment. This design is depicted as follows:
nt.
mined, and we rightly would question the con
uately
clusions related to causality. Observations Before Treatment
entile
Observations After

Quasi-Experiments Does implementing more severe penalties for

As we have seen, true experiments permit drunk driving the number of
actually reduce
causal conclusions because they entailrandomly drunkdriving incidents? Suppose that a researcher
assigning participants to the conditions and used a pretest-posttest design to evaluate the

exerting experimental control, thereby eliminat effect of legislative reforms on alcohol-related

ing a host of alternative explanations. However, fatal and injury accident rates. By determining
many events worthy of study are outside of the that rates before the reforms were higher than
researcher's control (e.g., people's exposure to a
those after the reforms, the researcher concluded

natural disaster). In addition,people are already that the reforms had been effective.

members of particular groups or categories (e.g., Now the question is, based on the research

gender, religion, socioeconomic level, right- or left- design used in the preceding hypothetical study,
harsher
handedness), and so random assignment to those can we conclude with confidence that

groups is either unethical orimpossible. Finally, in penalties reduced alcohol-related injuries and
many cases where the researcher may implementa deaths? Not necessarily. A number of other extra

treatment (i.e., independent variable), he or she neous factors may have led to the reduction, pos

the internal of the

might not be able to assign participants randomly ing serious threats to validity

to the different levels of that treatment because study. For instance, an event in the communty
of the
participant assignment not under the research
is may have coincided with the introduction
was a particularly
er's control (e.g., using hunmorous examples in new penalties. Perhaps there
to drunk driv
teaching one section of a course but not a second serious traffic accident attributed
section where students themselves choose their ing, and in the aftermath
many people may have
Sections when registering). So, how do we address become more reluctant todrive while intoxicated.
process of time could
yield
questions involving differential effects on the basis In addition, a natural
That is, even
of preassigned groups? The researcher may employ such an effect among drivers.

before the introduction of severe penalties, the

any one of a collection of research designs known have already