4
Persuasion Research: How Do
We Know What We Know?
Wavebreakmedia Ltd/Wavebreak Media/Thinkstock
Learning Objectives
After reading this chapter, you should be able to:
• Appraise the value of research in understanding how we know what we know about persuasion.
• Understand the three criteria for establishing causation.
• Explain the methodology involved in conducting experiments, as well as the limitations of experimental
research.
• Evaluate the utility of verbal reports, behavioral measures, and physiological measures in assessing
experimental outcomes.
• Discuss the different ways in which verbal reports, behavioral measures, and physiological measures can
be conducted.
mag81516_04_c04_073-096.indd 73
8/15/14 2:46 PM
Persuasion Research: How Do We Know What We Know?
Look at this video of a persuasion experiment: http://www.youtube.com/watch?feature=player
_detailpage&v=osUwukXSd0k. The researchers wanted to know what it would take to get kids
to clean their hands before eating. They divide 80 kids into four experiment conditions, and
each condition includes one or more factors that might encourage kids to clean their hands.
You can think of each of these four elements as a cause and hand cleaning as the effect, with
three different forms of influence—1) personal motivation (i.e., fear of getting sick from germs),
2) changing the environment (i.e., moving the hand sanitizer and adding a sign), and 3) increasing personal ability (i.e., practice using the hand sanitizer).
They concluded that having four sources of influence is sufficient to get people to change their
behavior. They called it “peer pressure,” but really the influence was a verbal reminder from
a peer to the kids to wash their hands. Note that in the condition in which the kids practiced
using the hand sanitizer that some kids (i.e., three kids) indeed used the hand sanitizer, but
that was not enough to lead the other nine children to use it.
Notice, however, that the data did not support the conclusion. An alternate explanation for
kids washing their hands is that the “social influence” factor alone might have influenced their
behavior. We don’t know if the verbal reminder by itself would be sufficient to get the kids
to wash their hands. That possibility was not tested. Further, if you ask kids, or anyone else,
after the fact to provide an explanation for why they did something, they will readily come up
with a plausible explanation. Drawing conclusions that are not supported by the evidence is
called “arguing beyond the data,” and we should be careful not to draw conclusions that the
data do not support. The book that is promoted at the end of the video might provide ample
evidence for the conclusion, but the experiment portrayed in the video itself does not provide
compelling evidence.
Such ambiguity in reaching definitive conclusions even in controlled environments prompts
questions as to how we ever reach definitive conclusions generally in real life. How, in effect,
do we know what we know? For example, if you’re trying to persuade someone to engage in
a diet and exercise plan, should you use a testimonial that illustrates a typical case, or should
you use statistics that summarize a lot of cases? How do you know which one is more effective? Or, if you are a salesperson, how does your appearance affect your performance? Would
wearing a white shirt instead of a blue shirt help you close more sales? What about your
hairstyle and hair color? Would wearing eyeglasses make you more credible? Are these factors in the sales process? If you’re a marketing executive, should you consider using product
placements in movies to promote your brand? How would you be able to know if that would
increase sales? In short, how do you know if a persuasion technique is effective? The way you
can know—the only way we can really know anything for certain—is through research.
In this chapter, you will learn about the experimental method, which is an important source
of generating evidence for the science of persuasion and social influence. In doing so, you will
also become a better “consumer” of experimental research so that you will be better able to
evaluate the strengths of the evidence. Just as we were able to assess the experiment concerning children washing their hands, you will also be able to evaluate the merits of the research
that is used to test many theories of persuasion and social influence.
mag81516_04_c04_073-096.indd 74
8/15/14 2:46 PM
How to Show Causation
Section 4.2
4.1 The Value of Research
Epistemology is the study of knowledge. Research is a way in which we acquire knowledge.
Much of what we know today is due to systematic research endeavors. Often, though, we take
our own personal experience and extend it, or generalize it, to a larger group. The problem
with this is that our own experience may be different from the larger population. How can you
know if what you experience is common to most other people? What is true in our own experience may not hold true for many other people. So, we conduct research to find out how we
can know what we know, or what we think we know. Simply put, research provides evidence.
Empirical research can employ many methods, such as focus groups, surveys, in-depth interviews, field observation, or experiments. What all of these methods have in common, though,
is their purpose to help us discover an answer to a question. They are intended to help us
acquire knowledge. Each method has strengths and weaknesses, and, often, the weaknesses
of one method are the strengths of another method. The choice of best method to use depends
on the question the researcher is asking. If the researcher wishes to explore a phenomenon,
focus groups can be a good choice. If the researcher wants an in-depth understanding of why
certain people do something, then in-depth interviews might be a good choice.
Experiments are best suited for testing causation. Most of the theories in the science of persuasion deal with causal relationships; therefore, most of these theories are tested through
experiments. If you are asking causal questions, then you must use experiments to test your
answers, which is why this chapter focuses heavily on the experimental method. This is important because each of the theories that we will study and learn to apply must provide evidence
that supports them. By understanding the experimental method, you can better evaluate the
evidence that was used to test a theory. Before we describe experiments, let’s look at the three
elements that are necessary to show evidence of a causal relationship.
4.2 How to Show Causation
Philosopher John Stuart Mill established that three conditions are necessary to conclude that
a causal relationship exists. To show that A, an independent variable, causes B, a dependent
variable, you must show that A occurred before B. You must also show that there is a relationship between A and B. Thirdly, you must rule out alternate explanations for the relationship
between A and B. The experiment is the best method for showing causation because it can be
used to meet all three of these conditions. Through measures of the independent variable and
the dependent variable, you can show a relationship between the variables. By manipulating
the way you introduce the independent variable, you can establish time order. The experiment allows you to control for other types of factors and rule out alternate explanations. Let’s
look at each of the three conditions.
mag81516_04_c04_073-096.indd 75
8/15/14 2:46 PM
How to Show Causation
Section 4.2
Time Order
To demonstrate a causal relationship between A
and B, you must show that A happened before B.
In laboratory experiments, it is rather easy to control the order in which the procedures take place.
For example, Yang and Roskos-Ewoldsen (2007)
tested the effect of type of product placement in
movies. They showed people movie clips that featured product placements in which the protagonist used the product or the product just appeared
in the background. Then they measured people’s
brand attitudes and recorded their product choice.
They found that having a protagonist use a product
(compared with the product merely appearing in a
scene) led to higher positive attitudes toward that
brand. They also found, though, that people were
more likely to choose a product if it appeared in
the movie, regardless of whether it was used by a
character or it merely appeared in the background.
Because the researchers were able to show the clips
before people reported their attitudes and made a
choice, the researchers were able to demonstrate
time order.
TGPRN Kraft Foods Inc./PR Newswire/
Associated Press
Besides the use of a celebrity athlete,
what else might a social scientist study
in this food advertisement?
In other types of studies, particularly those that
occur in the field, time order can be more difficult to
ascertain. Some relationships are cyclical, and it can be hard to determine which factor causes
which. For example, if you wanted to test the effect of Mountain Dew® product placements in
movies on a person’s attitude toward the brand in a natural setting, you would have a harder
time demonstrating time order. Because people have preexisting attitudes toward Mountain
Dew, it can be hard to tell if seeing the Mountain Dew in a television show would increase
a favorable brand attitude, or if people who have a favorable brand attitude would be more
likely to notice the product placement in the first place.
Existing Relationship
It is not enough to show that A occurred before B. You must also show that there is a relationship between A and B. If there is a direct relationship, then when A moves higher, B also
moves higher. If there is an inverse relationship, then when A moves higher, B moves lower, or
vice versa. The relationship does not need to be directly proportional. That is, it is possible,
for example, that when A moves two steps higher, B moves only one step higher. Ultimately,
the experiment is trying to determine if there is a relationship between two variables, but
this relationship can vary in degrees of strength; some relationships are stronger than others.
mag81516_04_c04_073-096.indd 76
8/15/14 2:46 PM
How to Show Causation
Section 4.2
For example, there is a strong relationship between smoking and lung cancer. The more one
smokes, the higher the incidence of lung cancer. There is also a link between smoking and skin
cancer, but this relationship is not as strong. In the product placement study (Yang & RoskosEwoldsen, 2007), the researchers were able to manipulate the type of product placement (i.e.,
product used by the character versus only appearing in the background) and then measure
attitudes toward the different brands to show that brand attitudes were higher when a protagonist used the product than when the product only appeared in the background. Thus, they
were able to show a relationship between the type of product placement and brand attitudes.
Rule Out Alternate Explanations
To conclude that A caused B, it is not enough to demonstrate that A occurred before B and
that there is an association between A and B. The third condition for demonstrating a causal
relationship between A and B is to show that no other factor could account for the relationship between A and B. This third condition is the most difficult to establish because it involves
accounting for many other potential variables—but often there are too many variables for
research to adequately account for. This means no experiment will allow a researcher to claim
causation between two variables with 100% certainty, but it also makes researchers strive to
do everything they can to obtain the most convincing results. A variable that can influence a
study’s findings is called a confound and diminishes the confidence we have in the study’s
results because it provides an alternate explanation for the study’s findings. In other words, a
confound is some variable outside the relationship of interest that can influence the behavior
of the dependent variable.
Researchers must begin by trying to account for other factors that are likely explanations
for the apparent relationship between two variables. The relationship must be probable or
plausible, not merely possible. Determining the extent of confidence in the research findings requires the refutation of as many alternative explanations as possible. If a legitimate
alternative explanation of a study’s results can be presented, then the study’s conclusions are
suspect. Importantly, an outcome such as this also prevents possibly passing on erroneous
information to the rest of the scientific community. If, on the other hand, no alternate explanation is found, then that increases the confidence in the relationship found between the two
variables. For example, Yang and Roskos-Ewoldsen (2007) had to consider several alternate
explanations for why someone might choose a product that had appeared in the movie over a
similar product that did not appear in the movie. It is likely that some people had previously
seen one of the movies, or that they already were familiar with the brand, or that they already
had a favorable attitude toward that brand. To account for these potential alternate explanations, the researchers recorded this information and then included it in their statistical
tests. They were able to show that there was a relationship between product placements and
people’s choice even after they took into account (or “controlled for”) these other variables.
Example
Imagine that Anna runs a clothing boutique and spends money on advertising only in November of each year. She also records a surge in sales in December compared with the other
months of the year when she does not spend money on advertising. Can she conclude that the
mag81516_04_c04_073-096.indd 77
8/15/14 2:46 PM
Experiments: Ways to Test Causation
Section 4.3
independent variable of advertising caused the dependent variable, namely, the increase in
sales? Let’s examine the three criteria to establish causation.
Anna can show time order because the advertising expenditure occurred before the increase
in sales. She can also use the measures of advertising expenditures and sales to demonstrate
an association. That leaves us with the third criterion of accounting for alternate explanations. Are there any other ways to explain this increase in sales? In the United States, retail
sales surge in December because of holiday-season gift giving. The variable of holiday gift
giving, or more generally, seasonality, can explain the increase in sales in December. So, Anna
cannot be fully confident that her advertising is the cause of the increase in her sales. Now,
let’s describe experiments.
4.3 Experiments: Ways to Test Causation
Most of the theories in persuasion science concern cause-effect relationships. For example,
a theory might propose that demonstrating to someone the personal relevance of the topic
of a persuasion appeal will increase that person’s motivation to pay attention to the appeal.
Another theory might propose that using vivid imagery in an entertaining story will increase
the likelihood that a reader or viewer will adopt beliefs that are embedded in the story. These
types of theorized relationships are causal in nature.
Among the many research methods available to investigators, the experiment is the best way
to test a causal relationship (or lack thereof) between two variables because it allows for
some element of control not available in most real-life settings. In simplest terms, an experiment is a research study whereby a researcher manipulates one variable in order to see the
consequences of that manipulation.
Experiment Basics
In experiments, an independent variable can be thought of a cause that has some sort of effect
on a dependent variable, or outcome variable. In the video of the hand- cleaning experiment,
the four different elements—the sign, the verbal reminder, and so on—were the independent
variables. Cleaning hands was the dependent variable, measured as the number of kids who
used the hand sanitizer, not the number of kids who wiped their hands on their pants. Experimental researchers often refer to the independent variable as a factor. An independent variable is sometimes called a manipulated variable because the researcher changes the variable
(e.g., adding a sign) to see what effect it might have on the dependent variable (e.g., using hand
sanitizer before eating a cupcake). The independent variable is also referred to as a treatment
variable because one group, the treatment condition, will receive the experimental treatment, while another group, the control condition, will not receive the treatment. A control
condition serves as a comparison group or baseline against which the researcher can compare
the treatment’s effects. With these two conditions, the researcher can then compare the effect
of the manipulation on the dependent variable.
mag81516_04_c04_073-096.indd 78
8/15/14 2:46 PM
Experiments: Ways to Test Causation
Section 4.3
A true experiment includes random
assignment, which means that the
study’s participants were assigned
to the treatment and control groups
in a random way (Keppel & Wickens,
2004). Doing this lets the researcher
assume that the groups are equal in
virtually every way. So, if the treatment and control groups differ after a
manipulation, then the only plausible
explanation is the manipulation itself.
The children in the hand-cleaning
study likely were assigned to the four
groups in a random manner. If so, then
we can assume, for example, that even
©Getty Images/Jupiter Images/Goodshoot/Thinkstock
though some of the kids’ parents might
When performing a field experiment, a researcher
observes relationships between variables in natural have stressed cleanliness, these kids
were randomly distributed across all
settings, such as an elementary school.
four groups, that they didn’t somehow
all wind up in the fourth group. Sometimes, however, the researcher cannot realistically assign people to the treatment and control
groups in a random fashion. In this case, when random assignment is absent, the experiment is
called a quasi-experiment. For example, let’s say the researchers had wanted to test whether
boys, more than girls, would respond more favorably to the experiment’s treatments. The
researchers can’t randomly assign some kids to be boys and others to be girls. In this case, the
study would be a quasi-experiment.
Usually, a researcher wants to remove, or control for, all the possible sources of extraneous
influence on the independent and dependent variables so that she can focus on the relationship between the independent and dependent variables. To do so, she conducts the experiment in a controlled setting such as a laboratory. An experiment that occurs in this type of
setting is called a laboratory experiment. In the hand-cleaning experiment, for example,
the researchers wanted to have control over the placement of signs and hand sanitizing gel.
Sometimes, however, a researcher will want to test the relationship between the independent
and dependent variables in a natural setting to see if the relationship will hold up in the field.
An experiment that occurs in this type of environment is called a field experiment (Shadish,
Cook, & Campbell, 2002). For example, you will read (in Chapter 12) about an experiment
involving a donation box that took place at the entrance to an art gallery, not in a controlled
laboratory setting. As you can imagine, most lab experiments are true experiments (if they
include random assignment), and most field experiments are quasi-experiments (because
random assignment is difficult to achieve).
Because lab experiments involve tight control, researchers are concerned with external
validity, which is the degree to which the phenomena manipulated and measured in the
experiment can be extended beyond the lab (Shadish, Cook, & Campbell, 2002). A lab experiment that has high levels of external validity tends to have independent variable manipulations that reflect what occurs in the “real world.” Field experiments tend to be strong in
mag81516_04_c04_073-096.indd 79
8/15/14 2:46 PM
Experiments: Ways to Test Causation
Section 4.3
external validity if the relationship between the independent variable manipulations and
dependent variable observations hold up in a natural setting.
Random Assignment
Random assignment is critical to the validity of a research study because it makes sure the
treatment and control groups are equal before the study begins, and it also helps eliminate
potential researcher bias. If you implement random assignment correctly, then you can
assume that each of the groups in the experiment is essentially equal to the other (Shadish,
Cook, & Campbell, 2002). In this way, random assignment is an important feature that helps
rule out potential confounds.
If you assign your participants to the experiment’s treatment and control conditions using a
method of randomization, then you can assume that any unique or idiosyncratic characteristic that might exist among the participants is distributed across both conditions. If, for example, you think that a person’s birth order might make a difference in how they respond to an
advertisement, you can randomly assign people to the treatment and control conditions and
assume that the factor of birth order is distributed across both conditions. Likewise, if you
think that ethnicity might influence responses to a persuasive appeal, then random assignment will help ensure equal distribution among experimental groups.
Many times random assignment is not possible for practical or ethical reasons. As mentioned earlier, quasi-experiments, or field experiments, occur outside of a laboratory setting.
Because quasi-experiments are conducted outside of a lab, researcher control is diminished
and random assignment is not possible. Some variables are naturally occurring in society,
and you cannot really assign people to one of those variables. For example, you cannot assign
someone to be a particular age, to be of a particular race, or to live in a particular region. In
these cases, you cannot have random assignment. This type of quasi-experiment can be a useful test of causal relationships, but you must take extra care to demonstrate that anything that
might have led the study participants to be in either the treatment or control condition did
not also affect the way they responded during the study (Shadish, Cook, & Campbell, 2002).
To assert a causal relationship between an independent and dependent variable, it is rather
easy to demonstrate time order and the existence of a relationship between the variables.
Ruling out alternate explanations, however, is the hardest part of drawing conclusions about
a causal relationship. That is why random assignment is so important—because it makes it
easier for you to conclude that a difference between the treatment and control conditions is
due to your experimental manipulation and not something else.
Interaction Effect
As we noted earlier, the independent variable that is manipulated in an experiment is often
referred to as a factor. When you have two factors that are crossed with each other, that is,
mag81516_04_c04_073-096.indd 80
8/15/14 2:46 PM
Experiments: Ways to Test Causation
Section 4.3
when you have two independent variables that are manipulated in conjunction with each
other, you have what is called a factorial design (Keppel & Wickens, 2004). This type of
experimental design lets you see if one independent variable interacts with another independent variable. Let’s say you wanted to examine ways to persuade children to take baths, and
let’s say that seeing an animated character sing a song in the bathtub leads to an increase in
children’s bathing intentions compared with a video in which the character does not sing a
song. Another technique that leads to an increase in children’s bathing intentions is to present a rubber duck compared to an absence of a rubber duck. You might then wonder if the
combination of seeing a rubber duck and having a character sing a song might interact with
each other to have a joint effect. Now we have two factors, the presence or absence of a yellow
rubber duck, and a character who either sings a song or does not sing a song. This is referred
to as a 2 by 2 (or 2 3 2) factorial design because each of the factors has two levels: the yellow
rubber duck (present versus absent) and the character (singing versus not singing).
So, to conduct the experiment, we would prepare four videos. The first video would feature
a yellow rubber duck present and a character singing a song. In the second video, the yellow
rubber duck would be absent, and the character would sing a song. The third video would
feature a yellow rubber duck, with the character not singing a song. In the fourth video, the
yellow rubber duck would be absent and the character would not sing a song. Then, we could
measure children’s bathing intentions in each of the four conditions. If bathing intentions are
highest in the condition in which the yellow rubber duck is present and the character sings
a song, then we would have evidence of an interaction. That is, the presence of the yellow
rubber duck and the character singing a song work together to increase in bathing intentions more than they would on their own. To express it more precisely, the effect of one factor
depends in part on the other factor. These effects of such a study are illustrated in Table 4.1.
Table 4.1: Bathing intentions (on a scale of 1–5)
Yellow Rubber Duck
Character
Yes
Singing
No
Present
Absent
4.7
3.5
3.8
2.2
In practice, a single experiment rarely goes beyond the manipulation of three factors simultaneously because the statistical tests become too complex and difficult to interpret. These
experiments can also become harder to implement. If scientists wish to examine several
factors, they tend to do so with a series of simple focused experiments, and not a single
complex experiment.
mag81516_04_c04_073-096.indd 81
8/15/14 2:46 PM
Experiments: Ways to Test Causation
Section 4.3
Persuasion in Focus: Aztec Queso Dip and Scarcity Appeals
Ever heard of the Aztec Queso Dip from Taco Bell? Don’t feel left out if you haven’t. It was
a fictitious product made up for an experiment measuring the persuasiveness of ad copy.
Aguirre-Rodriguez (2013) designed a set of experiments to assess whether people would
be more likely to be persuaded by an ad if it made claims about scarcity, in terms of supply
versus demand. She was interested in discovering if supply-related scarcity (SRS) was
more effective to use than demand-related scarcity (DRS). Specifically, the supply was the
number of coupons for the dip and the demand was interest in getting the coupons.
To test her initial hypotheses, she conducted experiments. Here we will focus on her
second experiment, which tested whether specificity in supply or demand would influence
people’s reactions to the ads.
First, she developed four ads. All ads started with the following language:
“New! The Aztec Queso Dip. Try it for free!”
The supplemental text for each of the four ads follows, based on supply-related appeals
and also demand-related appeals:
Supply-Related Scarcity Appeals
a) Specific—”Only 500 households per zip code can claim a coupon. Due to
restricted supply, this offer’s availability is limited!”
b) Nonspecific: “A maximum number of households per zip code can claim a coupon.
Due to restricted supply, this offer’s availability is limited!”
Demand-Related Scarcity Appeals
c) Specific: “Over 500,000 have already responded to this offer. Due to popular
demand, this offer’s availability is limited!”
d) Nonspecific: “Thousands have already responded to this offer. Due to popular
demand, this offer’s availability is limited!”
This was a 2 × 2 factorial experiment with type of scarcity (supply versus demand)
as one independent variable, and specificity (specific versus nonspecific) as the other
independent variable. First, Aguirre-Rodriguez created four ads, and each of these ads
represented one of the combinations of the independent variables. Then, she randomly
assigned each of the study’s participants to view one of the four ads.
Both types of appeals proved to have degrees of persuasiveness. But what was the most
persuasive strategy to gain participant interest? It was b)—Using nonspecific numbers
and appealing to the scarcity of supply. With that strategy, audiences could logically
assume the company could control how many coupons are released.
The supply-related scarcity appeals were more persuasive because the were less likely
to generate skepticism. In fact, Aguirre-Rodriguez explained that these types of scarcitybased appeals were perceived as being informative instead of being persuasive. Demandbased appeals were more likely to generate skepticism in the audience, as they could
easily identify a persuasive element in the ad (i.e., the ads sounded like it was trying
to persuade). With demand-based appeals, participants were more likely to question
the ability to even measure demand, and, therefore, they questioned the whole ad, a
phenomenon described in more detail in Chapter 9.
—Cheri Ketchum, Ph.D.
(continued)
mag81516_04_c04_073-096.indd 82
8/15/14 2:46 PM
Experiments: Ways to Test Causation
Section 4.3
Persuasion in Focus: Aztec Queso Dip and Scarcity Appeals
(continued)
Critical Thinking Questions
1.
2.
Look closely at the language of both supply and demand appeals. What language
do you think set supply apart from demand appeals? That is, what about the
supply appeals in option b proved most effective?
Which of these appeals would you use as the marketer for this product? Choose
one, or create a new appeal entirely that you feel would work better.
Reference
Aguirre-Rodriguez, A. (2013). The effect of consumer persuasion knowledge on scarcity appeal
persuasiveness. Journal of Advertising, 42, 371–379.
Strengths and Weaknesses of Experiments
Experiments are the “gold standard” for demonstrating causation. This is because experiments provide the control that researchers need to show causation. Experiments allow the
researcher to show time order, namely, that the independent variable occurred before the
dependent variable. They also allow the measurement that is needed to show a relationship
between the independent and dependent variables. Finally, they allow the researcher to take
into account and rule out alternate explanations.
Keep in mind, however, that experiments are not intended to address all of the factors that
might influence the outcome of a persuasion appeal. Experiments are intended to answer a
specific question that focuses on one factor or on one process. In the real world, you know
that a multiplicity of factors can interact to influence the outcome of a particular persuasion
episode. Most empirical studies are not designed to test all of the factors in a single setting.
Experiments provide control, in the sense that they allow the researcher either to eliminate
or to control for the influence of other factors so that the researcher may concentrate on a
particular phenomenon of interest. This focus and tight control is an advantage of the experimental method. It is also a weakness because it can come at the expense of external validity.
A scholar may demonstrate that a particular phenomenon occurs in a laboratory, but once
he moves to the real world, this phenomenon also would interact with any other host of factors that tend to accompany a persuasion episode. Experiments rest on the ceteris paribus
assumption that “all else is equal.” In the real world, though, “all else” is never equal.
College Students as Participants
Some people criticize experiments for using college students as participants. However, using
college students in a study is not a weakness in itself. The appropriateness of any experiment’s
sample depends on the question that the researcher is asking (Mook, 1983). To criticize a study
for using college students, you first must explain how college students are different from some
other population on an important variable that matters to the study. Then, you must explain
mag81516_04_c04_073-096.indd 83
8/15/14 2:46 PM
Measuring Experiment Outcomes
Section 4.4
how the study’s results would have been different had the experiment used a different type of
participant sample. Using a sample of college students actually can make for a stronger study
because it is harder to detect a difference between treatment and control conditions when the
conditions are very similar in every other way (Lynch, 1982; Mook, 1983).
Number of Participants
Some people criticize experiments for the small number of participants that are involved.
However, a large sample size is not necessarily better in an experiment. It is not true that a
greater number of participants leads to a more generalizable study, or greater external validity
(Mook, 1983). The purpose of an experiment is to test a causal relationship, and the processes
through which an effect occurs. If you can detect a causal relationship with a small sample,
then the effect must be quite strong. If, for example, you can show that having a speaker wear
glasses increases the speaker’s persuasiveness, and you can do so with a sample of 50 participants, you should expect the same result if you had 100 participants. As long as the research
methodology was sound, you would not expect the pattern of results to change in a different
direction merely because you added more participants to the experiment’s conditions.
Overall, experiments remain the best method for testing causal relationships. By manipulating one or more independent variables, a persuasion researcher can see precisely what type
of effect it might have on a person’s attitudes.
4.4 Measuring Experiment Outcomes
Attitudes are the primary dependent variable in persuasion research. We defined attitude
(in Chapter 1) as a tendency toward a particular object that has some sort of valence. That
valence can be either favorable or unfavorable. It should occur to you now, however, that we
cannot really see attitudes. Attitudes are mental states, so how can we measure something
that we cannot see?
Scientists have developed many ways to assess a person’s attitudes, but they can be grouped
into three categories. The first way involves verbal reports—in other words, simply asking
people to express how they feel about something. The second way involves measuring behavior, which reflects their attitudes. The third way involves taking measures of psychophysiological responses. That is, the body’s responses can be used to draw conclusions about a
person’s attitudes.
Verbal Reports: What People Say
Obtaining verbal report data involves asking people questions, such as asking someone in
person via an interview or via a survey that requires the person to answer a series of
mag81516_04_c04_073-096.indd 84
8/15/14 2:46 PM
Measuring Experiment Outcomes
Section 4.4
questions. The most common type of verbal reports consists of self-report data. Experiments
often involve survey-like questionnaires to assess a person’s attitudes.
Verbal reports are the most common form of dependent variable in persuasion research because they
are efficient and cost effective. If you want to know
how your friend Heidi feels about your new website,
you can just ask her. This is a type of self-report data.
Usually people will tell you the truth. Sometimes,
though, they do not want to hurt your feelings, and
they might not tell you what they really feel. Another
problem is that a person may not have an attitude
one way or the other about an issue, but she will
make one up for you just to please you. Researchers
need to be aware of these potential pitfalls on the
part of their research participants, but they also need
to be mindful not to bias responses or elicit information from participants that might skew results in
inauthentic ways. Consequently, researchers must
be mindful of the methodology they employ in creating and administering questionnaires. The following
sections look more closely at some of the elements
that go into developing sound verbal report tools.
Question Order
©Mohd Khairi Ibrahim/iStock/Thinkstock
Many researchers utilize self-report
questionnaires to measure attitudes.
Many studies employ questionnaires because they
are an efficient way of gathering information from
many people. However, you need to be aware of things that might influence the way a person responds to the questions. For example, the order in which questions are asked can
influence how a person responds to questionnaire items. In normal conversation, once we
have provided an answer to a question, we assume that the next question takes the previous information into account. In one study, married people were asked two questions, one
about how satisfied they were with their life as a whole, and another about how satisfied
they were with their marriage (Schwarz, 1999). When they were asked first about their
life in general, the two answers correlated at r = 0.32. (A value of r = 0 means there is no
association; a value of r = 1.0 means a perfect 1-to-1 association.) When respondents were
asked about their marriage first, the two answers correlated at r = 0.67, a much stronger relationship (see Figure 4.1). This is because asking about their marriage first made
marriage-related information more accessible, and this information then influenced how
they responded about their life in general. For a more in-depth and easy-to-read look at
why this happens, see Schwarz (1999).
mag81516_04_c04_073-096.indd 85
8/15/14 2:46 PM
Measuring Experiment Outcomes
Section 4.4
Figure 4.1: Effect of question order Response Alternatives
0.8
0.7
Correlation
0.6
0.5
0.4
0.3
0.2
0.1
0
Asked about
life in general first
Asked about
marriage first
Source: Schwarz, N. (1999). Self-reports: How questions shape the
answers. American Psychologist, 54, 93–105. Published by the
American Psychological Association. Adapted with permission.
An open-ended question provides a blank
space for a person to provide an answer.
However, after collecting the responses,
the researcher must read and code every
response, which can be time consuming.
A closed-ended question asks a person
to choose a response from a list of likely
or anticipated responses. However, the
list might lead the person to provide an
answer that he had not thought of before.
For example, people were asked what they
thought was “the most important thing for
children to prepare them for life.” When
a closed-ended scale of alternatives was
presented, more than 60% of the respondents selected “To think for themselves.”
However, when the question was posed in
an open-ended format, less than 5% of the
respondents said something equivalent
(Schwarz, 1999).
Scale Values
Scales come in many forms, but you are probably familiar with Likert-type scales, the most
common type of scale. A Likert scale consists of five ordered response alternatives that include
1 = strongly disagree, 2 = disagree, 3 = neither agree nor disagree, 4 = agree, and 5 = strongly
agree. If the scale includes any number other than five responses, researchers call it a Likerttype scale. A semantic differential scale consists of presenting a person a target object, such
as an organization followed by pairs of opposite adjectives (e.g., bad–good; boring–interesting,
etc.) with seven response alternatives between them. The person then selects the option that
reflects how he feels about that organization.
The values on a response scale can influence a person’s responses. In one study, respondents
were asked, “How successful would you say you have been in life?” and presented an 11-point
scale (Schwarz, 1999). Some respondents saw a scale with the points labeled 0 through 10,
while other respondents saw the scale’s points labeled –5 to +5. In the first group of respondents, about 34% marked an answer on the scale’s lower half. In the second group of respondents, only 13% chose a value on the scale’s lower half. What do you think accounted for
the difference? When people saw the scale labeled 0 to 10, they interpreted 0 to mean the
absence of successes. People who saw the scale labeled –5 to +5 interpreted the label –5 to
mean the presence of failures.
The number of points on a scale can be important, too. In one study, Hispanic respondents tended
to give more extreme responses on a 5-point scale than on a 10-point scale, while the extremity
of Anglo-Saxon respondents did not vary on the different scale types (Hui & Triandis, 1989).
When researchers want to keep people from selecting a middle, neutral, option, they will
use a forced-choice scale. This type of scale typically has four or six options, with no middle
mag81516_04_c04_073-096.indd 86
8/15/14 2:46 PM
Measuring Experiment Outcomes
Section 4.4
point. For example, the General Social Survey (Davis, Smith, & Marsden, 2007), a national U.S.
survey, asked respondents the degree to which they agreed with this statement: “Morality is a
personal matter and society should not force everyone to follow one standard.” The responses
were the following: “A. Agree strongly; B. Agree somewhat; C. Disagree somewhat; D. Disagree
strongly.” This format encourages people to indicate a preference, however so slight, toward
one side or the other. This format can be used with topics for which people may wish to take
a “safe” neutral ground. Of course, a person is always free not to respond. Also, some people
can become upset over being unable to express a neutral stance if they believe they truly are
neutral toward that topic. This type of scale should be used with caution.
Strengths and Weaknesses of Verbal Data
Verbal reports have several strengths. Verbal reports, usually questionnaires, are efficient and
cost-effective ways of gathering lots of information. Most people have experience completing
questionnaires, so they are easily understood. Likewise, most people will make an earnest
effort to tell a researcher what they think, so questionnaires can be a direct way of assessing
a person’s attitudes.
Keep in mind, though, that verbal reports are not always the best measure of a person’s attitude. Question type and question order can influence people’s responses so that you do not
obtain as accurate an assessment as you want. Sometimes, the problem is not with the scale.
People will just simply lie. Sometimes they do this merely to protect someone’s feelings.
Sometimes they even lie to themselves (“I’m a moral person,” or “I don’t stereotype people”).
Sometimes, people are actually afraid to tell the truth. At other times, they really have no attitude, but they will make one up on the spot, just to “help” the researcher.
Also, by asking people questions, you risk sensitizing them to the purpose of the experiment, and they might adjust their answers or behavior because they are aware that they are
part of a study. So, because self-report data may not always be truthful or accurate, we need
to find other ways of measuring people’s attitudes. One of those ways involves measuring a
person’s behavior.
Behavioral Measures:
What People Do
whitetag/iStock/Thinkstock
Many retail outlets use behavioral data to learn
about their customers.
mag81516_04_c04_073-096.indd 87
The number of ways to measure a
person’s behavior is almost limitless.
Instead of asking a person how they
feel about an attitude object, researchers can observe behavior and infer
an attitude from the behavior. Many
times, a person’s behavior can be a
more authentic indicator than verbal
data of their attitude toward an object.
In other words, to obtain an accurate assessment of people’s attitudes,
researchers can watch what people do
and not just listen to what they say.
8/15/14 2:46 PM
Measuring Experiment Outcomes
Section 4.4
Choice Behavior
A person’s choice behavior is an important dependent variable in experiments that marketers
conduct. Many studies of persuasion offer a person a choice among two or more items, with
the choice sometimes framed as a gift for participating in the study (e.g., Ferraro, Shiv, & Bettman, 2005; Yang & Roskos-Ewoldsen, 2007). Coupon redemption can be another behavioral
measure in persuasion research. For example, supermarket shoppers were invited to sample
jams from an assortment of either 6 or 24 flavors, after which they were offered a coupon that
they could redeem for a jam within 1 week (Iyengar & Lepper, 2000). When 24 flavors were set
out for sampling, 60% of shoppers stopped at the tasting booth. When 6 flavors were set out,
only 40% of shoppers stopped for a taste. However, among the shoppers who encountered a
wide array of choices, only 3% of them redeemed the coupon for a jam. In contrast, 30% of the
shoppers who sampled jam from a limited variety redeemed the coupon. In this field experiment, stopping at a taste booth and redeeming a coupon were two behavioral measures that
were used to see if the manipulation of assortment size would have an effect on shoppers.
Response Latency
The speed, or latency, with which someone answers a question can be an important outcome
measure. Response latencies are measures of speed of activation, which are reflections of
attitude and belief accessibility. If information is accessible to you, then you should be able to
report that information more quickly, without having to “search” for or formulate an answer.
For example, scientists can use keyboards that are accurate to the millisecond to record a baseline for how much time it takes for Xavier to respond to a set of questions. Then, they show
him a product advertisement that includes either an appeal to pro-environment features or a
focus on features that are unrelated to the environment. The pro-environment appeal should
cause Xavier’s environmental values to become more accessible. After viewing the advertisement, Xavier completes a measure of his environmental values. If the values are truly important to Xavier, then the time it takes him to respond to the environmental scale items should
be slightly faster after he views the pro-environment appeal. So, measuring the difference in
response latencies between the treatment and control conditions can yield insight into the
mental processes that are involved in persuasion.
A specific test of attitudes that relies on response latencies is the Implicit Association Test
(IAT). Remember, an attitude can be defined as an action tendency toward an attitude object
(Chapter 1), and an implicit attitude (Chapter 2) is an attitude evaluation that we are not
aware of and that we cannot control.
The IAT assesses the relative strength of automatic associations between an attitude object
(e.g., Barack Obama versus Hillary Clinton) and a valence (e.g., good versus bad). This test
can also be used to measure the strength of an object (e.g., Black people versus White people)
with a stereotype (e.g., athletic). The IAT shows a person an alternating series of objects and
valence words and records the speed with which the person links the valence words with the
object. If an object (e.g., Barack Obama) and a valence word (e.g., honest) are strongly associated in a person’s mind, then that individual will respond more quickly when seeing the word
honest. So, if Krista feels that Barack Obama is more honest than Hillary Clinton, then she will
associate the word honest with Barack Obama more quickly than she will with Hillary Clinton.
If Krista does not associate honest with Barack Obama, then it will take her longer to link the
two terms. The IAT is particularly useful for assessing biases concerning race, gender, and
mag81516_04_c04_073-096.indd 88
8/15/14 2:46 PM
Measuring Experiment Outcomes
Section 4.4
other topics for which people are unable or unwilling to be truthful. For a more complete
description of the IAT and its many uses, see Banaji and Greenwald (2013).
Web Analytics
Many marketers are interested in people’s behavior in online settings. Web analytics involves
measuring a person’s interaction with online information (Kaushik, 2010). Many corporations collect and analyze this type of data because it is unobtrusive and occurs in a natural
setting. This includes webpage navigation behavior such as which links a person clicks, how
many pages a person browses, how much time is spent on a page, and so on. At a basic level,
a report of a clickstream can provide a description of how people are navigating a website.
However, if you have a good theory to help make sense of the data and discern relationships, a
more advanced analysis can yield useful insights. Some marketers will try a promotion in one
market (the treatment condition) but not in another (the control condition) to measure the
promotion’s effectiveness using web analytics before trying it nationwide (Kaushik, 2010).
Eye Tracking
Scanning behavior can be measured
by using eye-tracking equipment that
follows a person’s focused gaze on a
screen (Milosavljevic & Cerf, 2008).
This equipment can be used to see
what elements on a webpage, say, banner ads at the top or box ads on the
side, first attract a person’s attention.
It can also be used to see what type
of content interests people most on a
news site (see Bucher & Schumacher,
2006). In a study of magazine advertisements, consumers viewed magaReprinted by permission of LookTracker. www.looktracker.com
zine advertisements to test which
Eye-tracking
devices measure what a person looks
size factors (the brand, a picture, or
at
and
how
long
he looks at it.
ad copy) might influence visual attention (Pieters & Wedel, 2004). Eyetracking data revealed that the size of the text or copy (but not the brand or picture) predicted
whether a consumer paid attention to the advertisement. Also, the larger the text, the more
time consumers spent viewing the advertisement. However, the amount of time spent viewing ad copy was unrelated to the time spent viewing the picture. Instead of asking people to
tell them about which ads interested them, the researchers were able to measure actual scanning behavior.
Strengths and Weaknesses of Behavioral Measures
Behavioral measures can be good indicators of a person’s attitudes because people sometimes cannot or will not report their attitudes accurately. Further, instead of assuming a relationship between attitudes and behavior, researchers can measure the actual behavior that
they wish to study. Sometimes behavior can be measured unobtrusively in a field experiment
mag81516_04_c04_073-096.indd 89
8/15/14 2:46 PM
Measuring Experiment Outcomes
Section 4.4
to maximize an experiment’s external validity. In this way, people are not sensitized to the fact
that researchers are conducting a study and their behavior is more likely to be natural.
With behavioral data, though, researchers have to infer the underlying attitude that is driving
the behavior. Further, much of a person’s behavior can be the result of a complex interaction
of several factors beyond any single attitude. A person’s shopping behavior can be influenced
by the person they are with, if they are not shopping alone. Or, sometimes people’s motivation
in a retail environment can vary. That is, some people go shopping as a pastime, while at other
times they are seeking a specific item, such as a birthday gift. It can be difficult to an infer
differences in attitude when the motives for behavior also vary, especially given that attitude
functions can vary, too.
Psychophysiological Measures: How the Body Responds
Recall from Chapter 2 that instead of thinking of attitudes as things, you can also think of an
attitude as mental states. Recall also that our definition of attitude includes the idea of an action
tendency, or a readiness to respond, toward something, either favorably or unfavorably. If you
have a mental state in which you are ready to respond favorably or unfavorably, or respond to
something that is either favorable or unfavorable, then your body should be prepared to take
action. Therefore, psychophysiological measures can also be used to assess attitudes, defined
as a mental state. The prefix “psycho-” refers to the mind, and the root “physiology” refers to
the body. So, scientists take physiological measures to infer psychological processes and states.
Galvanic Skin Response
Galvanic skin response (GSR) is like the lie detector test that you might be familiar with.
When people have a favorable attitude toward something, their skin has slightly higher electric conductivity. You can hook a probe to someone’s ankle or hand and measure changes in
GSR when they’re shown an advertisement or a product or person.
Heart Rate
You can also measure someone’s heart rate. When someone shows interest in something,
their heart rate tends to drop. So, a decrease in someone’s heart rate indicates an interest in
whatever they are looking at. This type of measure does not indicate a valence, that is, your
attention can be drawn to something positive or something negative, and your heart rate
would respond the same way toward both of them. It is merely an orienting response.
Facial EMG
You can also measure the electrical impulses of a person’s muscles. This is known as facial
electromyography, or facial EMG. You can put probes around the corrugator muscles, which
lie between the eyebrows on the forehead. People tend to contract those muscles, furrowing
their brow, when they see something they do not like. You can also place probes on the zygomatic muscles, which lie above the corners of the mouth. We contract these smiling muscles
when we see something we like. A person does not actually have to smile or furrow the brow.
The electrical impulses appear even when a person tries not to smile. All scholars do is measure the increase in electrical activity, whether or not those muscles are fully activated.
mag81516_04_c04_073-096.indd 90
8/15/14 2:46 PM
Measuring Experiment Outcomes
Section 4.4
Functional Magnetic Resonance Imaging
Functional magnetic resonance imaging (fMRI) is used to see which parts of the brain
become active after a person has been exposed to a particular stimulus. For example, the amygdala, which processes emotion and motivation, is activated after people see fearful faces or
other stimuli that tend to create a fear or arouse fear in the person (Ito & Cacioppo, 2007). There
is a modest amount of amygdala activation when we see positive stimuli. The prefrontal cortex
of the brain, the part right behind the forehead, is involved in planning and control. This is the
part of the brain that we use when we engage in hard thinking. The amygdala is activated when
we see something negative that might instill fear or require caution. When we see something
threatening, the amygdala is activated. However, we can control our fight-or-flight tendencies
through our prefrontal cortex.
Of course, marketers are particularly excited about
using fMRI to explore how consumers react to a
brand or a persuasive appeal (Penenberg, 2011).
For example, some film studios use fMRI to gauge
people’s reactions to movie trailers to create more
appealing trailers, which, of course, are used to
draw audiences for a film (Randall, 2013).
One well-known study illustrated the usefulness
of fMRI for understanding people’s attachment to
brands (McClure, Li, Tomlin, Cypert, Montague, &
Montague, 2004). Some individuals were served
Pepsi® in two cups, one labeled Pepsi and one unlabeled, and were asked which one tasted better.
Similarly, other individuals were given Coca-Cola®
in two cups, one labeled Coca-Cola and the other
unlabeled. The people who tasted Pepsi were
evenly divided between the labeled and unlabeled
cups of Pepsi. Among the people who tasted CocaCola, however, 85% of them chose the labeled cup
of Coca-Cola over the unlabeled cup of the same
Kul Bhatia/Science Source
thing. In a follow-up study, the researchers used
By using fMRIs, researchers can tell
fMRI to see what parts of the brain were activated.
how a person responds to ads. In this
Before participants drank soda through a straw,
fMRI scan, the visual cortex is active
they were exposed briefly to an image of either
as the person views an object (in this
a Coca-Cola or Pepsi label. When people were
case, a red hibiscus flower).
exposed to the Coca-Cola, but not the Pepsi, label,
they showed significant activity in the regions of
the brain that are associated with memory, behavioral control, and emotion. These areas
are not directly related with taste sensations.
However, most of our attitude processes are rather complex, and researchers need to be careful not to attribute any particular attitude to a particular place that’s activated in the brain.
For example, when people see ambivalent or conflicting information such as both positive
and negative information about a target, different parts of the brain that tend to be relatively
distinct are activated at the same time (Ito & Cacioppo, 2007). Of course, the activation of different parts of the brain depends on how strong the stimulus is, as well. There is also a time
delay of a few seconds between when a person sees something and when the change in blood
mag81516_04_c04_073-096.indd 91
8/15/14 2:46 PM
Measuring Experiment Outcomes
Section 4.4
flow is seen in the brain. The change in blood flow begins about 2 seconds after the person
sees the stimulus, and the change in blood flow peaks about 4 to 6 seconds afterward and
returns to its baseline measure about 10 to 15 seconds later (Ito & Cacioppo, 2007). So, it is
harder to measure more complex evaluations using fMRI. Nevertheless, the method is useful
for measuring cognitive processes involved in attitude formation, resistance, and change.
Event-Related Brain Potentials
Another type of measure is event-related brain potentials, known as ERP. Probes are
placed on different points of the scalp, and changes in brain waves can indicate the degree
to which someone likes or dislikes something. A person is presented with a series of positive items, and a measure of certain brain waves is taken. Then one negative item is placed
in the series, and researchers measure the distance and change in the brain waves between
the positive and negative items. ERPs are useful because they can also measure the degree
of change or difference between positive and negative targets.
Strengths and Weaknesses of Psychophysiological Measures
Psychophysiological measures are believed to provide a “direct” link to attitudes that bypass
some of the mental correction processes that people use to control what they say and do.
This is particularly useful when researchers need to evaluate a phenomenon, such as racial
stereotypes in advertisements, for which people try to control how they respond. Further,
psychophysiological measures are one of the most effective ways of testing mind-body connections in persuasion. In some cases, they can be the only means of testing some of the mental processes that are believed to explain relationships between independent and dependent
variables in some persuasion theories.
However, psychophysiological measures are more cumbersome, more expensive, and more
intrusive. You have to have specialized equipment to take the measures. Generally, you can
only administer the experiment to one person at a time, unlike questionnaires, which can be
administered to large groups of people at one time. And psychophysiological measures usually are insufficient, by themselves, to yield compelling information about attitudes, which
is why they tend to be used in conjunction with verbal or behavioral measures. Media measurement giant Nielsen uses both psychophysiological measures and eye tracking to test ads,
product packaging, the in-store experience, and so on (Nielsen, n.d.). Naturally, given the
number of methods and research firms that offer neuroscience-based methods, advertisers,
broadcasters, and other potential clients have begun to seek ways to measure their effectiveness (Advertising Research Foundation, n.d.).
The Value of Multiple Measures
How can you measure something you cannot see? As you have learned, different types of measures have their strengths and weaknesses. Any one of these measures likely is an imperfect
assessment of a person’s attitude, but several types of measures, taken together, can produce
a better assessment of a person’s attitude and the processes through which the independent variable affects the dependent variable (Fabrigar, Krosnick, & MacDougall, 2005). For
example, psychophysiological measures by themselves are not necessarily good indicators
of an attitude, but if you take psychophysiological measures and couple them with verbal
measures such as questionnaires, you can get a better measure of someone’s attitude. As we
mag81516_04_c04_073-096.indd 92
8/15/14 2:46 PM
Measuring Experiment Outcomes
Section 4.4
discussed earlier, sometimes people do not want to tell the truth. So, in addition to recording
verbal reports, scientists take behavioral or psychophysiological measures because it is much
harder for someone to avoid telling the truth with these types of measures.
When measuring an attitude, scholars often try to get several indicators of that attitude
because several joint indicators should be a better assessment of an attitude than any single
indicator. Any single measure might not be a good indicator of what we think we are measuring (see Figure 4.2). Let’s say that Elaine has a boyfriend, Ed, and she believes that Ed
loves her. How does Elaine know that Ed loves her? Maybe Elaine knows that Ed loves her
because he smiles whenever he sees her. Maybe she believes that Ed loves her because he
brings her flowers. Maybe she believes that Ed loves her because he washes the dishes. Maybe
she believes he loves her because every now and then he prepares a meal for her.
However, maybe Ed smiles when he sees her because he thinks that she dresses funny. Maybe
Ed brings her flowers whenever he feels guilty about something he has done. Maybe he does
the dishes because he has some sort of cleanliness compulsion, not because he loves her.
Maybe he cooks meals for her because she really can’t cook, and he figures that if he wants to
eat a good meal he might as well prepare it himself.
Any one of these actions, by itself, might not be a good indicator of whether Ed loves Elaine.
But, taken together, they can be pretty good indicators that he loves Elaine. If he buys her
flowers, and smiles when he sees her, and prepares meals for her, and does the dishes, maybe
he really does love her. Or, maybe he just feels sorry for her (we can never be certain). The
point is that multiple indicators of a phenomenon are better than a single indicator.
Figure 4.2: Multiple measures of an attitude
Attitude
A single measure can be
an incomplete assessment
of an attitude...
mag81516_04_c04_073-096.indd 93
Attitude
...but multiple measures can
provide a better assessment
of the attitude.
8/15/14 2:46 PM
Summary and Resources
Summary and Resources
Summary
•
•
•
•
•
•
Most of the theories in the science of persuasion deal with causal relationships.
Therefore, most of these theories are tested through experiments.
There are three ways to show causation between two variables: create a time order,
establish a relationship, and rule out alternate explanations.
If you are asking causal questions, then you must use experiments to test your
answers. Experiments are the ideal method for meeting all three of the conditions that are necessary for showing a causal relationship. An experiment gives the
researcher control over the order in which the stimulus materials are shown and
how the dependent variable measures or observations are gathered, so it provides
evidence for time order.
Through the judicious use of measures, the researcher can also demonstrate an
association between the independent variables and the dependent variables. An
experiment allows the researcher to control the influence of extraneous factors to
have a tight and focused test on the relationship that the researcher wants to study.
The design and control of an experiment also allows the researcher to take into
account the potential role of confounds. This type of control helps to rule out alternate explanations.
Attitudes are the typical effects of experimental manipulations in persuasion
research. Ways to assess attitudes can be grouped into three types: verbal, behavioral, and psychophysiological. Each method has its own strengths and weaknesses,
but by combining several types of measures, scientists can have a more accurate
measurement of attitudes.
Questions for Reflection and Application
1. According to John Stuart Mill, what are the three elements needed to show a causal
relationship?
2. Do you think that using psychophysiological measures is an improvement over traditional and more cost-effective self-report measures of attitudes? Why or why not?
3. What are the ethical implications of gathering behavioral data in a field experiment
if you do not inform people that you are conducting a study?
4. A political consultant for a politician wants to test television ads that focus either on
the politician’s character or on the politician’s policy position. What potential confounds should the consultant consider?
5. The Center for Science in the Public Interest has expressed concern over the number
of celebrities who endorse sugar-laden beverages. (See http://cspinet.org/new
/pdf/power-of-celebrity-soda-endorsements.pdf.) Design an experiment to test the
hypothesis that a celebrity athlete’s endorsement of sugary beverages will increase
the likelihood that children would try the beverages.
mag81516_04_c04_073-096.indd 94
8/15/14 2:46 PM
Summary and Resources
Key Terms
behavioral measures The measurement of
what people do, such as choices they make,
speed of response, and eye movements,
rather than what people self-report.
causation A relationship between two variables in which one variable causes the other
variable to behave in a certain way.
control condition The group in an experiment that does not receive the treatment or
change.
event-related brain potentials (ERP)
The measurement of changes in brain waves
to indicate the degree to which someone
likes or dislikes something.
experiment A research method whereby
a researcher manipulates one variable to
see the consequences of that manipulation;
the best way to test a causal relationship
between two variables.
external validity The degree to which the
phenomena manipulated and measured in
the experiment can be extended beyond the
lab to the “real world.”
eye tracking A type of behavioral measure
that uses specialized equipment to follow
a person’s focused gaze on the screen; can
be used to see what elements on a webpage
first attract a person’s attention and what
type of content interests people most.
facial electromyography (EMG)
The measurement of the electrical impulses of
a person’s muscles, such as those between the
eyebrows on the forehead and those above
the corners of the mouth; can be measured
whether or not the muscles are fully activated.
factor Another word for the manipulated
independent variable, which has an effect on
the dependent variable.
mag81516_04_c04_073-096.indd 95
factorial design An experimental design in
which two independent variables are manipulated in conjunction with each other.
field experiment An experiment that takes
place in a natural setting.
functional magnetic resonance imaging
(fMRI) A scan used to see which parts of
the brain experience more blood flow after
a person has been exposed to a particular
stimulus.
galvanic skin response (GSR) The tendency of a person’s skin to have a higher
electric conductivity when they have been
exposed to a stimulus; measured with a
probe on the ankle or hand.
heart rate Tends to drop when someone
shows interest in something.
Implicit Association Test (IAT) A specific
test of attitudes that relies on relative differences in response latencies; assesses the
relative strength of automatic associations
between an attitude object and a valence.
laboratory experiment An experiment
that takes place in the controlled setting of
a laboratory.
manipulated variable Another word for
the independent variable, so used because
the researcher changes the variable to see
what effect it might have on the dependent
variable.
psychophysiology The measurement of the
body’s responses; physiological measures
are used to infer psychological processes
and states.
quasi-experiment An experiment in which
random assignment is absent, typical of field
experiments.
8/15/14 2:46 PM
Summary and Resources
random assignment A condition in a true
experiment in which the study’s participants were assigned to the treatment and
control groups in a random way; allows the
researcher to assume that the groups are
equal in virtually every way.
response latency A measure of speed
of activation; how quickly attitudes are
accessed in a given situation.
treatment condition The group in an
experiment that receives the treatment or
change; the independent variable.
mag81516_04_c04_073-096.indd 96
treatment variable Another word for the
independent variable, so used because this
is the group receiving a treatment or change,
whereas the control group does not receive
the treatment.
verbal reports Self-report data based on
what a person says; the most common form of
dependent variable in persuasion research.
web analytics Involves measuring a person’s interaction with online information,
such as which links a person clicks, how
many pages a person browses, and how
much time is spent on a page.
8/15/14 2:46 PM
Purchase answer to see full
attachment