Critical Thinking Research

90

2 Research Methods

AJ PHOTO/BSIP / Corbis Documentary / Getty Images

91

Core Questions

1. What are the elements of the scientific method?

2. How do social psychologists design studies?

3. How do social psychologists analyze their results?

4. How can research be analyzed in terms of quality?

92

Like all sciences, social psychology usually moves like snail: steady but slow. It is slow, in

part, because what social psychologists study is usually invisible—and therefore difficult to

measure. For example, prejudice, persuasion, altruism, and romantic love are all scientific

constructs, theoretical ideas that cannot be directly observed. Although the scientific

process is slow, social psychology is growing fast. It is growing fast because so many

students are attracted to Kurt Lewin’s vision of an applied science.

Perhaps social psychology’s popularity explains why so many passengers were carrying long

plastic or cardboard tubes on a recent plane ride. The plane was full of people presenting at

a conference sponsored by the Society for Personality and Social Psychology (SPSP), which

happens at the end of every winter. The tubes contained rolled-up posters summarizing the

most cutting-edge research in the field. This chapter describes how the professional

scientists, graduate students, and even a few undergraduates created those studies—and it

invites you to join us.

Ask the Experts: Keon West on Research Methods in Psychology

©SAGE Publications

Learning Objectives

2.1 Describe how the scientific method creates knowledge.

2.2 Compare the logic behind preexperiments, true experiments, quasi-experiments, and correlational

designs.

2.3 Summarize the most common ways to analyze and interpret data.

2.4 Describe reliability, validity, replication, and ethical standards for research in social psychology.

93

Basic researchers: Psychologists who increase our understanding of psychology by creating and improving

the theories that predict social behavior.

Applied researchers: Psychologists who translate the findings of basic researchers into social action and

apply psychological ideas to the real world.

What Are the Elements of the Scientific Method?

— LO 2.1: Describe how the scientific method creates knowledge.

Social psychologists tend to describe themselves as belonging to one of two groups. Basic

researchers increase our understanding by creating and improving the theories that predict

social behavior. Applied researchers translate those theories into social action. Applied

research is where theory confronts reality—but with the understanding that reality always

wins. If a theory does not describe reality, then the theory has to change. Basic research is

important because, as social psychology’s pioneer Kurt Lewin famously said, “There is

nothing so practical as a good theory” (Lewin, 1951, p. 169). Applied research has to have a

theoretical foundation—and a “good” theory has to describe and explain the real world.

How does this balance play out in actual scientific research?

94

The Cycle of Science: The Scientific Method

Most of the research stories on those conference posters followed the same, easy-to-follow

formula. The scientific method is a systematic way of creating knowledge by observing,

forming a hypothesis, testing a hypothesis, and interpreting the results. Your results lead

you to generate a new hypothesis that starts the process all over again. As you can see in

Figure 2.1, it cycles endlessly through the same steps (or phases) and demands a willingness

to change your mind if the data surprise you. Even though properly following the scientific

method is sometimes a challenge, the rewards of exploring the fascinating topics within

social psychology make it all worthwhile.

Figure 2.1 In the scientific method, scientists observe the world and notice patterns, then

formulate a hypothesis. Next, they create a procedure to scientifically test their hypothesis.

After interpreting the results, the process begins again as the hypotheses become more

refined or complicated or as they are applied to a wider variety of settings or people.

There are four phases in the scientific method:

Phase 1: Observe a pattern of behavior. Imagine that you are in a coffee shop quietly

observing other customers. You notice that men frequently interrupt people during

conversations—and that seems to be especially true when their conversation partner is a

woman. Welcome aboard; you’ve started the scientific journey. Your simple observation got

you started as your curiosity prompted you to ask, Is this a pattern? (By the way, this exact

95

observation was tested in coffee shops and drug stores back in 1975 by Zimmerman and

West.) Phase 1 of the scientific method occurs when we observe a reliable pattern of

behavior.

Phase 2: Generate a hypothesis. After you identify what looks like a pattern, you move on

to Phase 2 by generating a formal hypothesis, or a specific statement of what you believe

will happen in an experiment that tests your observation. Perhaps your hypothesis is that

men are more willing to interrupt others than women are, especially in cross-sex

interactions. In other words, you expect that (1) men interrupt more than women in

general, and (2) men interrupt women more than they interrupt other men. Hypotheses are

never stated as questions (such as, “Who interrupts more, men or women?”). As we

discussed in Chapter 1, hypotheses are always falsifiable statements that can be proved

wrong.

Consider the following hypotheses: (1) Every man on Earth has interrupted someone else at

least once, and (2) men have always been more likely to interrupt than women, in every

culture throughout history. Neither of these hypotheses stands up to the “falsifiability” rule

because they can’t be tested and proved wrong. The first hypothesis is untestable because it

simply can’t be done. Even a crack team of well-funded researchers couldn’t locate every

single man on the face of the planet and then observe him in an objective social setting.

The second hypothesis also can’t be falsified because there is no archive of historical records

that enables us to ask the question about every historical period for the past thousands of

years.

Phase 3: Test the hypothesis. Now that we have a hypothesis, we set up a specific

methodology or procedure to test it. For our current example, we might observe people in

public places (like the coffee shop) in a more structured way, such as by making a

spreadsheet of all the men and women and making tallies for each time someone interrupts.

We might also ask people to come to a classroom or lab on a college campus and set them

up in groups with a certain number of men and women and then observe who interrupts

whom.

Either way, we can gather real, measurable data that will support our hypothesis, allowing

us to modify it and move to the next step, or data that cause us to throw it out. We have to

be careful, of course, not to ruin our own experiment by staring as we eavesdrop on some

innocent couple drinking coffee (later we will discuss this issue in more depth, as well as

ethics in research).

Phase 4: Interpret the results and refine your hypothesis. Notice that the final stage of

interpreting results is not the end of the road. Once we have our data, we aren’t done with

science—in fact, we’ve only just begun! The scientific method cycles back so that we can

explore our topic in more complicated and refined ways. Perhaps we found support for the

basic idea that, overall, men are indeed more likely than women to interrupt someone.

96

Scientific method: A systematic way of creating knowledge by observing, forming a hypothesis, testing a

hypothesis, and interpreting the results. The scientific method helps psychologists conduct experiments and

formulate theories in a logical and objective manner.

Hypothesis: A specific statement made by a researcher before conducting a study about the expected

outcome of the study based on prior observation. Hypotheses are falsifiable statements that researchers

believe to be true (see falsification).

Constructs: Theoretical ideas that cannot be directly observed, such as attitudes, personality, attraction, or

how we think.

Operationalize: The process of specifying how a construct will be defined and measured.

However, this general pattern probably varies greatly based on the people involved and the

circumstances. In other words, our results have become a new hypothesis that requires us to

begin again. Consider the following possible new hypotheses as examples:

Women with more assertive personalities are more likely to interrupt others,

compared to women with less assertive personalities.

Men are less likely to interrupt women they find physically attractive, compared to

women they don’t find attractive.

Men interrupt others more in friendly or informal settings, compared to formal

settings such as at work.

Men from cultures with more traditional gender roles are more likely to interrupt

women than are men from more egalitarian cultures.

Can you see why the scientific approach is a constantly unfolding story? That story can

only move forward if we remain as objective as possible when forming hypotheses and

interpreting results. An individual research study, like the posters at a research conference,

is a very small piece of a very big puzzle. But every step we take brings us a tiny bit closer to

understanding the complicated world of social interaction.

97

Creating and Measuring Constructs

Many of the things social psychologists are interested in are abstract ideas or constructs,

theoretical ideas that cannot be directly observed; examples are attitudes, personality,

attraction, or how we think. Those measurement challenges—and the passion to conduct

meaningful social research—are what has made social psychologists so creative in designing

studies.

Because constructs are abstract and sometimes relatively broad ideas, the first step in using

them in research is to operationalize your variables by specifying how they will be defined

and measured. The process is called operationalizing because you must describe the specific

operations you will perform to measure each of the variables in your study. If a researcher

wanted to investigate the construct of “love,” for example, she could operationalize it in a

wide variety of ways such as (1) scores on a survey asking people to rate how much they

love someone on a scale of 1 to 10; (2) how long they have maintained a committed,

monogamous relationship; or even (3) how much their heart rate increases and their pupils

dilate when the other person comes into the room.

The myth of Sisyphus says that he’s constantly trying to roll a heavy boulder uphill. The

scientific method can sometimes feel like an uphill battle—but progress requires constant

small steps. The top of the scientific mountain may not even exist, but moving up remains

satisfying.

Sisyphus by Titian, 1549

Once we’ve operationalized the variables in our hypothesis, we have to decide how to

proceed. Here are common methodologies that you’ll see in several of the studies featured

in this book. It’s not a complete list of every possible study design, but it will give you a

98

good idea of how social psychologists do business.

99

Archival data: Stored information that was originally created for some other purpose not related to research

that can later be used by psychologists, such as census data.

Types of Research

Once you’ve noticed a pattern and generated a hypothesis, there are a lot of different ways

you can set up a scientific methodology or procedure to test that hypothesis. This book

isn’t about research methods, so we’ll just cover a few of your options here—these are the

most popular methods you’ll see throughout the rest of the book and in the field of social

psychology. Four options are (1) archival studies, (2) naturalistic observation, (3) surveys,

and (4) experiments. We’ll cover the first three methods here, and experiments will be

discussed in depth in the next section.

Archival Studies.

One of the sources of information available to social psychologists requires (almost) no

work because the data already exist. Archival data are stored information that was

originally created for some other purpose not related to research. Newspapers, census data,

Facebook posts, and even pop culture are all examples of archival data.

Archival data are being collected every day in every community, and it’s up to social

scientists to think about hypotheses that might be tested. For example, researchers

interested in patterns within domestic violence can look at police records to test hypotheses

about whether different types of people report this crime, whether couples who report once

are more likely to report again, whether demographic variables such as socioeconomic status

or certain neighborhoods have higher or lower rates of violence, and so on.

Naturalistic Observation.

Another approach is naturalistic observation, or scientific surveillance of people in their

natural environments. By “natural,” we don’t mean in a cornfield or a forest—we mean

people doing the behavior of interest where it normally would occur. Observing people in a

coffee shop is a good example. If we were interested in whether teachers are nicer to

physically attractive children, then we might go to an elementary school to observe classes.

If we were interested in leadership styles, we might go to a large corporate office and

observe how workers react to different types of managers. Either way, we’re simply

observing behavior in its natural setting.

You might be thinking, “If some scientist came to my workplace and followed me around,

writing down everything I do, then I probably wouldn’t react very naturally.” If that

thought occurred to you, then congratulations—you are thinking like a good scientist. The

presence of the researcher is one of the biggest challenges for naturalistic observations.

100

Naturalistic observation: A research design where scientists gather data by observing people in the

environment within which the behavior naturally occurs (for instance, observing leadership styles in a

corporate office).

When people change their behavior simply because they’re being observed, it’s called

reactivity. But social psychologists are clever people. How do you think they get around

this problem?

One creative solution is a technique called participant observation, in which scientists

disguise themselves as people who belong in that environment. It’s kind of like going

undercover. You pretend you’re not doing research at all and hope to fade into the

background—and still find some discreet way to record your observations. For example,

when observing schoolchildren, we might pretend to be substitute teachers. If we want to

observe people at work, we might pretend to be interns at the company. One set of

researchers wanted to photograph boys enjoying themselves at a summer camp (that was

secretly run by psychologists). So, one of the camp counselors played the role of a “shutter

bug”—someone who is taking pictures all the time. The boys quickly learned to ignore the

shutter bug, and the researchers came away with some beautiful, authentic photographs (see

the Sherif study described in Chapter 9).

Participant observation may create some ethical problems, so be careful. After all, you are

deceiving people about why you are there. And it may be an ethical violation to observe

people when they don’t know they are being observed. The advantage of this technique—

or any form of naturalistic observation—is that hopefully, we get to observe authentic social

behaviors.

In the movie Never Been Kissed (Isaac, Juvonen, & Gosnell, 1999), Drew Barrymore’s

character is a reporter who wants to write about the life of high schoolers. To get the “true

scoop,” she pretends to be a high school student herself. In Imperium (Taufique, Lee,

Ragussis, Walker, & Ragussis, 2016), Daniel Radcliffe works for the FBI and infiltrates a

White supremacist group, pretending to be racist. If either one of them had been social

psychologists in a real setting, then their technique would have been called participant

observation.

AF archive / Alamy Stock Photo & Atlaspix / Alamy Stock Photo

101

Survey: A research design where researchers collect data by asking participants to respond to questions or

statements.

Self-report scale: A type of survey item where participants give information about themselves by selecting

their own responses (see survey).

Social desirability bias: The tendency for participants to provide dishonest responses so that others have

positive impressions of them.

Surveys.

An alternative approach is simply to ask people to tell us about their own thoughts,

emotions, and behaviors in surveys. Psychological surveys typically ask people to react to

statements about themselves by choosing a number on a scale. It might range from 1

(strongly disagree) to 7 (strongly agree). These self-report scales ask people to give us

information about themselves in a straightforward, explicit manner (hence the name “self-

report”). There are self-report scales throughout this textbook, so you can see how you

score on a variety of social psychological concepts. Those scales will help you understand

just how interesting and complicated you are as you navigate your social world.

There are several considerable advantages to the survey method of research. One is that it is

relatively inexpensive and you can get hundreds of participants in your study relatively

quickly, especially if you put your survey online. This also allows for you to get a wider

diversity of participants as you can send your survey’s URL to people all over the world.

Self-report surveys also can ask people personal questions about their intimate lives that you

would never have access to (at least, not legally!) through naturalistic observation.

However, recall that one common problem with naturalistic observation is reactivity, or

people changing their behaviors because they know they are being observed. Self-report

surveys have their own concerns, and one of the big ones is dishonesty. For example, you

might not tell the truth if you were asked whether you’ve ever treated a romantic partner

badly, cheated on a test, stolen something, or had “bad thoughts” about another person.

The dishonesty problem is often attributed to the social desirability bias, the idea that

people shape their responses so that others will have positive impressions of them. (This

problem is also sometimes known as impression management.) For one creative way to get

around the social desirability bias in survey research, see the Applying Social Psychology to

Your Life feature.

Reactivity: When people change their behavior simply because they’re being observed (see social desirability

bias and good subject bias).

Participant observation: A technique used during naturalistic observation where scientists covertly disguise

themselves as people belonging in an environment in an effort to observe more authentic social behaviors.

102

Case Studies

In the first three common methods for testing hypotheses—archival studies, naturalistic

observation, and surveys—most studies will have several people who serve as the

participants. However, before we move on to discuss experiments, there’s one more term it

would be good for you to know.

Applying Social Psychology to Your Life: Measuring Social Desirability

Sometimes when people fill out self-report scales in research studies (or on job interviews, or anywhere else),

they aren’t completely honest. Instead, they answer in a way that they think makes them look good; this

tendency in people is called the social desirability bias. One creative way that social psychologists test for

this tendency in people is to give them the scale shown here, which is specifically designed to catch people

in small lies. Most people have done many of the bad behaviors listed here—so if research participants don’t

admit to them, they are probably showing the social desirability bias; they are changing their answers to

look good.

Instructions: Listed below are a number of statements concerning personal attitudes and traits.

Please read each item and decide whether the statement is true or false as it pertains to you personally.

Circle “T” for true statements and “F” for false statements.

T F 1. Before voting I thoroughly investigate the qualifications of all the candidates.

T F 2. I never hesitate to go out of my way to help someone in trouble.

T F 3. I sometimes feel resentful when I don’t get my way.

T F 4. I am always careful about my manner of dress.

T F 5. My table manners at home are as good as when I eat out in a restaurant.

T F 6. I like to gossip at times.

T F 7. I can remember “playing sick” to get out of something.

T F 8. There have been occasions when I took advantage of someone.

T F 9. I’m always willing to admit it when I make a mistake.

T F 10. There have been occasions when I felt like smashing things.

T F 11. I am always courteous, even to people who are disagreeable.

T F 12. At times I have really insisted on having things my own way.

Scoring: Give yourself 1 point if you said TRUE for Item 1, 2, 4, 5, 9, or 11. Then, give yourself 1 point if

you said FALSE for Item 3, 6, 7, 8, 10, or 12. The more points you have, the more you are trying to

manage your impression on others.

Source: Crowne and Marlowe (1960).

When a single example of an event or a single person is used to test a hypothesis or refine it

further, it’s called a case study. A case study represents a single example of the phenomenon

of interest. For example, one case study you’ll see later (in Chapter 11) describes the level of

violence in The Great Train Robbery, a 1903 movie that was the first film to tell a story

103

(Porter & Porter, 1903). The case study summarized the story in the film and computed

the ratio of violence per minute to analyze how much violence was shown in the movie.

Note that this case study used archival data. The movie itself was preexisting information,

sitting in YouTube and waiting to be analyzed. Case studies can be archival, but we might

also use naturalistic observation to record behaviors in a particular person over time, or we

might give one person a survey to complete—or we might even ask a single person to

engage in an experiment (see the next section). So, case studies can be used in any of the

forms of research we’ve covered here.

As we prepared to write this chapter, we became interested in whether case studies were

being more or less accepted by modern psychologists. We hypothesized that the rate of

referencing case studies had been declining over time, as the use of online survey software

has become more and more popular. To test our hypothesis, we used archival data that

existed in the PsycINFO database. PsycINFO is the most comprehensive database of

research books and journal articles across psychological subdisciplines. It’s like Google or

any other search engine, except your search collects data from books, chapters, or journal

articles published by psychologists around the world. When we searched PsycINFO for

publications that made use of case studies, would we see a decline over time?

Figure 2.2 The frequency and percentage of articles referring to case studies in PsycINFO.

Source: Created using data from the American Psychological Association.

Our hypothesis was not supported—in other words, we were wrong. Instead, we found a

104

Case study: A type of research where scientists conduct an in-depth study on a single example of an event or

a single person to test a hypothesis.

PsycINFO database: The most comprehensive database of research books and journal articles across

psychological subdisciplines.

The Main Ideas

The scientific method, which is used by social psychologists who conduct research, includes (1)

observing a pattern, (2) generating a hypothesis, (3) scientifically testing the hypothesis, and (4)

interpreting results so that the hypothesis can be refined and tested again.

Abstract ideas or variables are called constructs, and deciding how to define and measure constructs

is called operationalization.

Three ways to gather data are (1) using archival data, or sources originally gathered or created for a

different purpose; (2) naturalistic observation, or watching behavior where it would have occurred

anyway; and (3) surveys, or asking people directly to report their thoughts, emotions, or behaviors.

Any of these methods can use multiple participants or a single participant; single-participant studies

are called case studies.

trend of increased referencing of case studies (Figure 2.2). Again, these were archival data,

originally created for another purpose, waiting to be mined for insight. (Note: If you do

this study again, you probably will get slightly higher numbers, especially for the most

recent years, as new articles are admitted into the PsycINFO database.)

105

Critical Thinking Challenge

Think about the classrooms you’ve been inside recently. Consider the physical aspects of the room (such

as size, type of desks, color, art on the walls, and so on). Then consider how people choose to sit in the

room during classes (such as whether they prefer the front or back row, how much they spread out, what

kinds of people tend to sit together, and so on). Generate three hypotheses about how either the physical

environment or the social environment shapes learning.

Imagine that you want to do a study on how companies support leadership within their organizations.

First, describe how you might conduct the study using archival data; then, how you’d do it with

naturalistic observation. Finally, describe how you would conduct the study differently if you decided to

give people who work there a survey. What kinds of questions would you ask? How would you get people

to fill it out honestly?

Identify three different ways you could operationalize each of the following variables: (1) prejudice, (2)

high self-esteem, and (3) empathy toward other people.

106

How Do Social Psychologists Design Studies?

— LO 2.2: Compare the logic behind preexperiments, true experiments, quasi-

experiments, and correlational designs.

The working world is full of designers. We have fashion designers, graphic designers,

architectural designers, cookware designers, landscape designers, and game designers. To

become a clear-thinking social psychologist, you must become an experiment designer.

If the most famous book about experimental designs were going for a big audience in the

self-help market, then it might be called How to Think Clearly. However, the original book

by Donald Campbell and Julian Stanley (1966) had a less dramatic title: Experimental and

Quasi-Experimental Designs for Research.

Social Psychology in Action: Self-Deception Experiment

©SAGE Publications

Its original target market was the unruly world of education research. Education research

had been (and perhaps still is) like a pendulum clock hanging from a swinging rope,

lurching back and forth between “a wave of enthusiasm for experimentation [that] gave way

to apathy and rejection” (Campbell & Stanley, 1966, p. 2). Campbell and Stanley wanted

to calm things down and remind researchers that experimentation takes time, replications,

and multiple methods. They organized the world of research design into four categories

that we’ll cover here: preexperimental designs, true experiments, quasi-experiments, and

correlational designs.

107

Preexperimental Designs

The most basic methodology is called a preexperiment, in which a single group of people is

tested to see whether some kind of treatment has an effect.

One type of preexperimental design is the one-shot case study, which explores one event,

person, or group in great detail. This design can be either archival (data originally created

for some other reason) or a case study examined specifically for research purposes from the

outset. For example, interviewing a particular world leader in depth about how he or she

made leadership decisions might be a one-shot case study. In a classroom environment, one

teacher who uses a particular new technique—such as playing music in the background of

the classroom—and then measures how well the students do on a test is another example.

You might try a particular approach to studying one time, and your test score will be the

measurement of how well that technique worked.

The one-shot case study, then, involves just two elements: (1) identifying a particular case

of something or trying a technique once and (2) an observation or outcome.

Another type of preexperimental design is the one-group pretest-posttest design,

sometimes called a “before-after design.” Here, the expected outcome is measured both

before and after the treatment so that the researchers can assess how much change occurred.

The one-group pretest-posttest design is represented this way, with an initial observation

(pretest), the presentation of the treatment or technique, and a second observation

(posttest):

Notice that neither type of preexperiment counts as a true experiment. The major problem

with either type of preexperiment is that you can’t be sure that any outcome was really due

to the treatment or technique. For example, imagine that a classroom teacher tries a new

teaching technique and the students all do very badly. Can the teacher be sure this

disappointing outcome was really due to the technique? No, because there are other

possible explanations. Maybe the students all would have done badly regardless of how they

were taught because the material was particularly difficult. Maybe it was the time of year, or

the teacher was badly trained on the technique, or the teacher was sick that week, and so

on.

108

Preexperiment: A research design in which a single group of people is tested to see whether some kind of

treatment has an effect, such as a one-shot case study or a one-group pretest-posttest.

One-shot case study: A type of preexperimental research design that explores an event, person, or group in

great detail by identifying a particular case of something or trying a technique once, then observing the

outcome.

One-group pretest-posttest design: A type of preexperimental research design in which the expected

outcome is measured both before and after the treatment to assess change.

Confounding variables: Co-occurring influences that make it impossible to logically determine causality.

Confounding variables, such as bad weather or the inability to concentrate on a survey due to illness,

provide alternate explanations for the outcome of an experiment that make it impossible to know whether

the results are due to the independent variable (see internal validity).

Many studies attempt to establish cause-effect relationships, meaning the researchers want

to say, “X caused Y to happen.” Preexperiments can’t make causal claims because of these

alternate explanations (bad weather, someone was sick, etc.). Alternate explanations are

called confounding variables, which are co-occurring influences that make it impossible to

logically determine causality. Fortunately for science, this problem is solved by true

experiments.

109

Experiment: A research design where scientists randomly assign participants to groups and systematically

compare changes in behavior. Experiments allow scientists to control confounding variables and establish

cause-effect relationships.

Random assignment to experimental condition: A solution to the problem of confounding variables by

creating equivalent groups at the start of an experiment. Random assignment cancels out the influence of

confounds by distributing them equally across groups.

True Experiments

The “gold standard,” or best methodology, for most social psychological studies is an

experiment, which compares groups that have been created by the researchers on some

important outcome. A well-designed experiment extracts meaningful patterns from a

chaotic world. (Pretty amazing, if you stop to think about that one!) The main goal of a

true experiment is to be able to make claims about causes and effects. Establishing causality

is a rich philosophical question. But to the experiment designer, that means ruling out

alternative explanations by controlling confounding variables (the “enemy” of good

research).

Imagine you wanted to know whether being sleep-deprived makes people anxious. So, you

ask 10 people to take a nap and 10 people to not nap, and then you measure their anxiety.

If the freshly napped group is less anxious, can you be sure their relatively relaxed feelings

are due to the nap? How do you know that those 10 people weren’t already more relaxed

and happy-go-lucky than the other 10 people?

The best solution to the problem of confounding variables is random assignment to

experimental condition, which occurs at the beginning of your study. Each person in the

study has an equal chance of being put in the nap or no-nap group based on some

determination of chance, such as the flip of a coin. With random assignment, it’s

statistically likely that people who came into the study already high or low in anxiety will be

equally distributed across the two experimental groups, which in essence makes the groups

identical. If this is done, then any differences in the measurement at the end must be due to

experimental treatment; it’s the most plausible explanation. Let’s talk about the details of

how true experiments work.

Independent and Dependent Variables.

All true experiments have two types of variables, called “independent” and “dependent.”

Researchers use the independent variable to make the participant groups different from

each other at the start of the experiment. For example, researchers might have some people

in the study listen to classical music and have others listen to rock music. So, the

independent variable is type of music—it’s what makes the groups different from each

other from the very beginning of the study.

110

Independent variable: A variable that is manipulated at the beginning of an experiment to determine its

effect on the dependent variable.

Dependent variable: The measured outcome of an experiment that is affected by the independent variable.

Of course, we also have to measure some kind of outcome, which is the dependent

variable. Maybe while our two groups listen to either classical or rock music, everyone

reads a passage from a textbook. Then, they take a memory test. If the experiment is trying

to establish whether type of music causes better or worse memory, then performance on the

memory test at the end is the dependent variable. It’s called the “dependent” variable

because we are hypothesizing that memory scores are “dependent” on whether participants

heard classical or rock music. Does type of music cause people to do better or worse on a

memory test? In experiments, the independent variable is the cause being tested (here,

music) and the dependent variable is the effect or outcome (here, memory test scores).

The “studying while listening to different kinds of music” is a popular experiment among

new researchers, so we hope that you noticed some other problems. For example, what kind

of classical music was it? How well was each type of music performed? Were they played at

equal volume? Again, if some of those problems occurred to you, then you are starting to

think like a scientist. If any of these other differences by group occurred (such as the rock

music was loud and fast but the classical music was soft and slow), then both volume and

tempo are confounding variables. Possible criticisms such as these are important in the

scientific process, and they are one reason why so many studies need to be done to answer

what may have seemed like a simple question at the beginning of the process. Each study

helps us refine our hypothesis and begin again.

For several more examples of independent and dependent variables, see Table 2.1.

Types of True Experiment.

Just like we saw with preexperiments, there are several different kinds of true experiment.

One type of experimental design is called the pretest-posttest control group design. This

is basically a before-after design because the outcome (or dependent variable) is tested both

before and after the experimental manipulation or intervention occurs. For example,

memory could be measured in all participants—then they listen to either rock or classical

music—and then memory is measured again. The researchers can now measure how much

each type of music affects change in memory from the pretest to the posttest. The key

difference is that in a true experiment, we now have two groups instead of one, and

participants have been placed into one group or the other through random assignment. It

looks like this:

111

You need both groups, made equivalent by random assignment, to control for confounding

variables. However, sometimes the different levels of the independent variable are compared

to a “neutral” or baseline group. In our example about measuring whether memory scores

change after listening to music, imagine that memory scores went up in both conditions.

Maybe simply listening to any kind of music helps memory. To test this refined hypothesis,

we need another group that doesn’t listen to music at all. A group that serves as a neutral or

baseline group that receives no treatment at all is called a control group. That experimental

design would look like this:

Finally, you will learn about many studies in this book that are close to the design described

here, with one difference: They sometimes skip the pretest phase. When this happens, it’s

called a posttest-only control group design. Everything else is the same—you have two or

more groups, including a control group, and you measure outcomes for each group. You

just don’t have the pretest, which means you compare how the groups are different from

each other in only the posttests. This design is perfectly fine as long as you really did use

random assignment to put each participant in the different groups. Again, this should

logically mean that the groups started out on equal footing, so it’s okay to measure the

112

Pretest-posttest control group design: A type of true experiment where the dependent variable is tested

both before and after the experimental manipulation.

Control group: A group of participants in a true experiment that serves as a neutral or baseline group that

receives no treatment.

Posttest-only control group design: A type of true experiment where the dependent variable is measured

for two or more groups, including a control group, only after the experimental manipulation.

influence of the independent variable by simply looking at how the groups differ on the

dependent variable (posttest only).

Between-Participants Versus Within-Participants.

The different designs described so far are considered by many to be ideal choices for

research. However, sometimes real studies have challenges. For example, there might be

limitations to how many people you can get in your study or how you can create the

different conditions of your independent variable. So far when we’ve considered designs

with two or more groups, the assumption has been that each group is made up of different

people. This approach in general is called a between-participants design, meaning that

levels or conditions of the independent variable change between groups of different

participants. Maybe 20 people listened to rock music, and 20 different people listened to

classical music (and maybe 20 others listened to no music). This would require 60 different

people!

However, imagine you are a teacher and you want to test your hypothesis, but there are

only 20 students in your entire class. In addition, you need to teach all of them the

material, but you have only one set of speakers in your classroom, so everyone will have to

listen to the same music. With this kind of limitation, you can consider a within-

participants design, which means that the same group of participants all experience each

experimental condition. You put each person in every group, one group at a time.

For example, in Unit 1, the entire class listens to rock music; in Unit 2, the entire class

listens to classical music; and in Unit 3, there is no music. Ideally, different people would

go through all of the groups in different orders (for example, someone might listen to

classical music first, while someone else hears rock first; then they switch), but sometimes

this just isn’t possible. Just like we saw with archival research, naturalistic observation, and

surveys, there are advantages and disadvantages to each different design option within true

experiments. Within-participants designs have the advantage of needing fewer people (here,

it’s 20 people compared to 60). But a within-participants design comes with built-in

disadvantages. One of them is the possibility of order effects that influence the outcome

variable just because one condition of the independent variable happened to come first.

Choosing the right design is an important step in the research process.

The famous character Quasimodo from Victor Hugo’s (1831) novel The Hunchback of

113

Between-participants design: An experimental research design where the levels or conditions of the

independent variable are different for each group of participants; patterns are found by comparing the

responses between groups.

Within-participants design: An experimental research design where the same group of participants all

experience each experimental condition; patterns are found by comparing responses for each condition.

Order effects: Variations in participants’ responses due to the order in which materials or conditions are

presented to them.

Notre Dame was known for being “half-formed.” Quasi-experiments could be considered

“half-formed” versions of true experiments, because they compare groups but are missing

the key element of random assignment of people to those groups.

Photo 12 / Alamy Stock Photo

114

Quasi-experiment: A research design where outcomes are compared across different groups that have not

been formed through random assignment but instead occur naturally.

Quasi-Experimental Designs

As we’ve already indicated, there are many situations in which true experimental designs are

impossible, unethical, or impractical. In addition, sometimes the variables of interest aren’t

things that can be randomly assigned. For example, you can’t randomly assign some people

to experience a tornado. In these cases, quasi-experiments can help us reach reasonable but

still tentative conclusions. Quasi is the Latin word for “almost” or “partially.” Thus, a

quasi-experiment is almost like a true experiment.

Quasi-experiments compare outcomes across different groups, just like true experiments—

but the key difference is that the groups have not been formed through random

assignment. The groups in quasi-experiments formed on their own, naturally. For example,

you may want to compare men to women in terms of who interrupts more. Here, the

independent variable you’re testing is sex (men vs. women), and the dependent variable is

frequency of interruptions. However, these variables aren’t “true” independent and

dependent variables because they were not randomly assigned, and thus the experimental

setup hasn’t really eliminated any possible confounding variables.

Even if we find that one group interrupts more, we still don’t really know why. Is it because

of genetics? Hormones? How boys versus girls are raised? Because we are not able to

eliminate all confounding variables, we can’t truly make causal inferences or statements

with a quasi-experimental design. But, it’s impossible to randomly assign each participant

to sex—you’d find it odd to show up to an experiment only to be told, “Today you’re

going to be a woman.”

While participant sex is impossible to randomly assign, other variables may be simply very

difficult or unethical to randomly assign. If you want to study the effect of going through a

hurricane or being in a terrible car accident, you can’t randomly assign half of your

participants to experience these traumas. You might also want to study variables such as

religion or choice of career—and again, participants would probably protest if you

randomly assigned them to either of these important variables. In short, quasi-experiments

are needed to study several important variables within social psychology—but we need to

be more cautious when interpreting results, especially in regards to any claims about

causality.

115

Correlational design: A research design where scientists analyze two or more variables to determine their

relationship or association with each other.

Social Psychology in Popular Culture: Correlation and Causation in Harry Potter

Correlational Designs

Your head might be spinning with all of the different study designs we’ve gone through so

far—preexperiments, true experiments, quasi-experiments—so we’ve saved the most simple

design for last.

Correlational designs are relatively easy; they involve simply collecting or measuring two

pieces of information from each participant in the study, then seeing if there is a pattern.

For example, you can ask people (1) how many hours they study each week and (2) what

their grade point average (GPA) is. Most people would quickly see the hypothesis being

tested: More studying is associated with higher grades. (Note: We’ll talk more about how to

specifically test for this pattern in the next section.)

Correlational designs thus look for patterns in which two variables have a relationship or

association with each other. Height and shoe size are correlated; the general pattern is that

taller people have larger feet. Income and size of someone’s house are correlated; richer

people usually have bigger houses. It’s sometimes easy to think of correlations as patterns

that indicate a causal relationship: Studying for more hours each week causes a student’s

GPA to improve, or having more money causes someone to be able to buy a bigger house.

While it is possible that some correlations show causal relationships, in social psychology

(and in all forms of science), it is very important not to assume that correlations indicate

causal relationships. This word of caution is summarized in the popular motto,

“Correlation does not imply causation.” Why not? First, it might be the case that both

variables are actually caused by a third variable. In the case of a student who spends many

hours studying and has a very good GPA, both of these outcomes might have been caused

by the student’s (1) motivation to do well, (2) level of pressure from parents, or (3) amount

of enjoyment of class subjects.

116

United Archives GmbH / Alamy Stock Photo

Most people are familiar with the Harry Potter series of books and movies, in which a young boy discovers

he’s a wizard and attends a boarding school to learn spells. In the sixth book, Harry Potter and the Half-

Blood Prince (Rowling, 2005), Harry first gets the chance to learn the magical power of “apparition,” or the

ability to disappear and reappear in a different location. Harry’s first apparition lesson is an example of his

awareness that correlation does not imply causation.

Harry describes his apparition teacher’s appearance and how that appearance might be linked to the ability

to disappear:

He was oddly colorless, with transparent eyelashes, wispy hair, and an insubstantial air, as though a

single gust of wind might blow him away. Harry wondered whether constant disappearances and

reappearances had somehow diminished his substance, or whether his frail build was ideal for anyone

wishing to vanish. (p. 382)

Harry doesn’t know whether having a wispy appearance caused his teacher to have a greater apparition

ability—or whether his talent at apparition has caused him to appear wispy. While Harry suspects there’s a

correlation or association between the two things, he knows that without more information, he can’t know

which is the cause and which is the effect. It’s also possible that the association is merely a coincidence or

that both are caused by something else. Without logical research designs, the secrets of apparition may

remain unknown to Muggles (nonmagical folks) and wizards alike.

Consider another example: There is usually a correlation between amount of ice cream sold

per year and the number of people who drown that year. Is ice cream consumption causing

drowning? Probably not. It’s more likely that a third variable explains the correlation: heat.

Towns that have hotter temperatures (such as Miami, Florida, and Austin, Texas) sell more

ice cream. In addition, more people swim in these towns due to the heat (which

unfortunately sometimes leads to more drownings). So, while it might look like the two

variables of ice cream and drownings are related, both are actually driven by something else.

It’s also important to realize that with a simple correlation, even if there is a causal

relationship between the two variables, we don’t necessarily know which is the cause and

which is the effect. Think about the controversial idea that watching violence on TV causes

children to act more violently in real life. This hypothesis makes a lot of intuitive sense, and

there are studies showing that amount of violence viewed on TV is related to how violent

children are.

117

However, it’s also possible that the causal relationship goes the other way around. Maybe

children who are already inclined to have violent personalities and behaviors are more likely

to watch violent TV, because they find it interesting. Now, the cause is the violent

personality and the effect is watching violence on TV. Simple correlations aren’t enough to

assume anything about causation; we need more precise methodologies, like experiments,

to know the details.

It’s fun to be a scientific problem solver, addressing important social issues. But it is much

more than that. Research can be a satisfying, meaningful way to spend a life—words you

may never thought you would hear in a discussion about research designs.

118

The Main Ideas

A preexperiment is a method in which a single group of people is tested to see whether some kind of

treatment has an effect.

A true experiment compares outcomes on two or more groups that have been created by the

experimenter through random assignment to condition. True experiments are the only

methodology that can lead researchers to make claims about cause-effect relationships between

variables.

Cutting-Edge Combinations

Classic studies using naturalistic observation, archival data, surveys, and experiments have

now established many foundational theories and concepts in social psychology, and all of

these methods continue to provide new insights from research labs around the world.

However, social psychology’s research methodology is also able to grow and evolve as

technology advances. These advances result in new, cutting-edge procedures, many of

which will be described in the rest of this book.

One exciting way that social psychology research is becoming more advanced is by joining

forces with neuropsychologists to study how social interactions both influence and are

influenced by neurons in the brain. For example, one study (Cyders, Dzemidzic, Eiler, &

Kareken, 2016) used functional magnetic resonance imaging (fMRI) brain scans to show

that men have more neural reaction to sexual images than to nonsexual images and that

degree of neural reaction was positively correlated with the men’s survey responses on how

much they engage in high-risk sexual behaviors. Several other examples of how the brain is

connected to social psychology are highlighted in later chapters.

Other researchers examine physiological responses during studies, such as blood pressure

(e.g., Hill et al., 2017; Huntley & Goodfriend, 2012; Jennings, Pardini, & Matthews,

2017) or the release of stress hormones such as cortisol in blood or saliva (e.g., Simons,

Cillessen, & de Weerth, 2017; Tarullo, St. John, & Meyer, 2017) as ways of measuring

participants’ biological reactions, in addition to their psychological reactions. One

interesting example is pupil dilation, which has been correlated with sexual interest or

attraction to other people (see Lick, Cortland, & Johnson, 2016; Savin-Williams, Cash,

McCormack, & Rieger, 2017).

Finally, computer software has been specifically designed or updated by researchers in social

psychology labs both to create different independent variables and to measure dependent

variables; often, this takes the form of video games that participants play in which various

aspects of the game are manipulated, and how the participants play and respond serves as

the outcome being measured. Again, several examples will be seen in future chapters. As

technology advances, our ability to test hypotheses about social psychological phenomena

will also advance in exciting new ways.

119

A quasi-experiment tests for the effect of a treatment in groups that have formed naturally, such as

men versus women or people who prefer different kinds of music.

Correlational designs look for associations between two variables that are measured in a single

group. A correlation between two variables does not necessarily mean that one causes the other.

120

Critical Thinking Challenge

Earlier, we discussed challenges to naturalistic observation (reactivity) and to surveys (social desirability

bias). What are some challenges to experimental research designs? How could scientists overcome these

challenges?

Imagine that you have a hypothesis that people who drink a lot of caffeine will experience heightened

emotions over the course of a day. Explain three different ways you could design a study to test your

hypothesis, using three different methods from this section of the chapter.

Think of a correlation you’ve observed in the world where it’s hard to tell if the relationship is causal

because either of the two variables could be the cause or the effect. Which comes first and affects the

other? Then, think of a correlation that appears to show a relationship, but both variables are actually

being caused by a third variable that wasn’t directly measured.

121

How Do Social Psychologists Analyze Their Results?

— LO 2.3: Summarize the most common ways to analyze and interpret data.

“I don’t like statistics,” one of our students declared before class was starting, and then

asked his neighbor, “Do you know if the Pittsburgh Pirates won last night? They’re only

two games out of the lead.” He couldn’t see the obvious: He already loves numbers, what

they mean, and the stories they tell—but only when they are connected to something that

he cares about (in this case, baseball). We hope that in this next section, which reviews

some basic types of statistics used in social psychology, you can think about how each one

might apply to something that you care about.

People who are serious about understanding social psychology need an understanding of

statistics, or mathematical analyses that reveal patterns in data. They are important, both

on an individual level—the pleasure of thinking analytically is satisfying—and on a larger

level. Understanding patterns in data and how to interpret them correctly is absolutely

necessary for authentic understanding of how to create government policies, how to design

effective education, how to plan for a stable financial future, how to respond to global

climate change, and more. Critical, scientific, analytical thinking is needed for social

psychology to be able to influence the world for the better.

122

Statistics: Mathematical analyses that reveal patterns in data, such as correlations, t tests, and analyses of

variance.

Standard deviation: The amount of variability in a distribution. In other words, how widely dispersed the

data are.

Comparing Groups

Let’s think through the hypothetical results (in Figure 2.3) about the effects of different

types of music on memory. The horizontal x-axis indicates memory scores; higher memory

scores are shown to the right. The vertical y-axis indicates how many people got that score.

The average memory score for people listening to classical music is higher than the average

for people listening to rock music. These two averages provide the most important

statistical information.

But you also may have noticed something else in Figure 2.3. The shape of the memory

curve for people listening to rock music is a little flatter and wider. The flatter, wider

distribution of scores means that there was more variability in the memory scores for people

in that group (the highest score was further away from the lowest score, meaning a wider

range). The amount of variability in a distribution is called the standard deviation.

The two bell-shaped curves in Figure 2.3 also show overlap between the two groups. Some

people listening to rock music had higher memory scores than some of the people listening

to classical music. If the overlap were small and the average scores were far apart from one

another, then we can be more confident that this experiment found a real effect of type of

music on memory. But if the overlap is large and the average scores are close to one

another, then we could have very little confidence that this was a real difference.

Comparing Two Groups: The t Test Statistic.

The t test statistic uses both the mean and the standard deviation to test for differences

between two groups. William Sealy Gossett, a brewer at Dublin’s Guinness Brewing

Company, first published how to calculate it anonymously. He wasn’t supposed to publish

industrial secrets, and the t test was one of them. Gossett’s statistical invention used small

samples to make far-reaching judgments about the quality of Guinness’s many ingredients

(Mankiewicz, 2000). The t test solved many practical problems at Guinness.

One of Gossett’s jobs was to test the amount of stout in each batch of beer—think of it as a

taste test. It would have been impossible (although a fun challenge, at least for a while) for

him to sample from the thousands of casks produced every single day. Imagine that Gossett

tasted 10 random samples from one batch made on Monday morning and gave each a taste

rating. Then he did the same thing with a second batch made on Monday afternoon.

Gossett’s t test allowed him to compare the quality in thousands of casks even though he

123

t test: A statistical test that uses both the mean and the standard deviation to compare the difference

between two groups.

only sampled from a total of 20 batches.

There were two critical components to Gossett’s statistical invention: (1) the sampling had

to be random, and (2) the sample had to be big enough to be representative of all casks of

beer. It would not be good for Guinness’s business if the morning batch tasted different

from the afternoon batch. Gossett’s invention of the t test statistic for comparing two

groups (the morning vs. afternoon batches) ensured consistent quality in Guinness. So, the

next time you’re sipping a beer and wondering why you’re required to learn statistics, you

can lift a glass to W. S. Gossett for his liver-saving contributions to science. The secrets of

sampling are as important for making far-reaching judgments about human behavior as

they are for brewing a better beer.

Figure 2.3 One way social scientists look for patterns is by comparing average scores within

different groups of participants. When we compare two groups, as here, we use a t test.

When we compare three or more groups, we do an analysis of variance, or “ANOVA.”

Comparing Three or More Groups: Analysis of Variance.

You can use the t in the t test to remind you that it is used to compare two groups. But

think back to our example from earlier in the chapter, when we wanted to test the effects of

different kinds of music on memory. What if our study involved students listening to three

types of music, such as country music, hip hop, or jazz? What if we wanted to compare five

types of music? And what if Gossett wanted to compare taste samples from more than two

groups of beer—say, one from each day of the week?

124

Analysis of variance (ANOVA): A statistical test that uses both the mean and the standard deviation to

compare the differences between three or more groups.

We have Guinness to thank for the statistic known as the t test.

© istockphoto.com/Oktay Ortakcioglu

The principle for comparing multiple groups is the same as comparing two groups. For

each group, we calculate the average score and the standard deviation, just like before.

However, when we’re comparing three or more groups, the test is called an analysis of

variance, or “ANOVA” for short. For example, perhaps the classical music group in a

three-group experiment does better than the rock group—and maybe they are both better

than the control group with no music at all! ANOVA tests will tell you whether at least one

of the groups is different from the others. Different types of comparison groups require

different statistical tests. However, they all rely on just two basic ingredients: the average

and the standard deviation in each group.

125

Patterns in a Single Group: Correlations

While t tests and ANOVAs compare patterns of results in different groups, correlations

look for patterns of results in a single group. Specifically, as explained in the preceding

section, correlations test whether two different variables are systematically associated with

each other. Think about the following questions that all apply to the life of college

students:

Does a student’s high school grade point average (GPA) predict his or her college

GPA?

Does the distance between home and chosen college relate to how much

homesickness a typical student feels?

Are the number of hours spent studying related to test scores?

Are students with more individual interactions with professors going to be more

satisfied with their college experience?

Do hours of sleep in an average week predict levels of stress in that same week?

All of these questions would be tested with correlations. Correlations begin with two

variables of interest (for example, hours spent studying and GPA). To test for a pattern,

scores on each variable are gathered from as many people as possible and are then charted

on a graph called a scatterplot. One variable is on the (horizontal) x-axis, and the other is

on the (vertical) y-axis, and each dot on the scatterplot represents one person. Take a look

at Figure 2.4 for an example.

The pattern shown in Figure 2.4 indicates that, for most people, more hours studying each

week is associated with better grades. As you would expect, as studying increases, so does

GPA. Certainly, not everyone will fit the pattern—there might be brilliant people who can

get a high GPA without studying and people who study for hours and hours and still

struggle with grades. But, the general pattern is fairly stable. The line summarizes the story

told in detail by the scatterplot.

Figure 2.4 In this scatterplot, each dot represents one person. For each person, study hours

per week fall on the x-axis, and GPA falls on the y-axis. By looking at the general pattern,

we can determine whether the two variables are correlated.

126

When a correlation is calculated, the number you get is called a correlation coefficient. It

will always be a number between -1.00 and +1.00. How can you tell what the coefficient

means? It’s basically like a two-part code you can crack to understand what the pattern

looks like on a scatterplot. There are two parts to the code: (1) the sign or direction—

positive or negative—and (2) the number. Let’s talk about each part.

First, the sign will always be either a positive or a negative (unless the correlation is exactly

zero). A positive correlation (between +0.01 and +1.00) indicates that both variables move

in the same direction. In other words, if scores or values on one of the variables go up,

values on the other variable will also go up. If one goes down, the other will go down. The

example in Figure 2.4 shows a positive correlation: As a typical student spends more hours

studying each week, grades will go up. Hours and grades are both moving up, in the same

direction. Positive correlations are shown in scatterplots when the pattern or summary line

moves from the bottom left-hand corner to the upper right-hand corner.

In contrast, a negative correlation (between -0.01 and -1.00) indicates that the variables

move in opposite directions. As one variable goes up, the other goes down. Instead of hours

spent studying, think about hours spent partying. The more hours a student parties, the

lower his or her grade might be: Partying moves up, and grades move down. Negative

correlations will be shown in scatterplots with a pattern that goes from the upper left-hand

corner to the bottom right-hand corner.

The second part of a correlation coefficient is the number, which will always be between

zero and 1 (either positive or negative). The number tells you how clear the pattern is on

the scatterplot or how well the different dots (which represent people) fall along the

summary line. Basically, this number tells you how much variability there is in the data, or

whether some people don’t fit the pattern. In Figure 2.4, for example, you can see that

there is one student who studies an average of about 20 hours per week but only has a GPA

of about 2.75. If the dots all fall exactly on the line, meaning the pattern is perfect, the

number you get will be 1.00. As the number gets closer to zero, it means the pattern

127

Correlation: A type of statistical test that determines whether two or more variables are systematically

associated with each other by identifying linear patterns in data.

Scatterplot: A graph that demonstrates the relationship between two quantitative variables by displaying

plotted participant responses.

Correlation coefficient: A number that indicates the relationship or association between two quantitative

variables. It ranges from –1.00 to +1.00.

Positive correlation: A positive correlation occurs when the correlation coefficient is between +0.01 and

+1.00. In this case, as one variable increases, the other also increases.

Negative correlation: A negative correlation occurs when the correlation coefficient is between –0.01 and –

1.00. In this case, as one variable increases, the other decreases.

becomes slightly less clear.

Note that coefficients of +1.00 and -1.00 are equally strong—both indicate perfect patterns

—it’s just that in one case, the variables move together (+1.00), and in one case, they move

in opposite directions (-1.00). Figure 2.5 shows a summary of how to understand

correlations, showing a range of patterns that move from perfect and positive, through no

correlation at all, to perfect and negative. A zero correlation coefficient means that the two

variables have no relation to each other at all, such as GPA and someone’s height or love of

chocolate. These variables are not associated with each other at all, so the scatterplot looks

like a bunch of random dots.

Figure 2.5 Correlations always range from –1.00 to +1.00. The sign (positive or negative)

indicates whether the two variables move in the same direction or in opposite directions.

The number (from 0.0 to 1.0) tells you how well each data point fits onto a general

pattern. If a correlation is zero, it means there is no pattern or association between the two

variables.

128

Testing for Statistical Significance: p Values

So far, you’ve learned about t tests, ANOVA tests, and correlation tests. All of these

statistics look for patterns of data. The last statistical concept is to know what it means—

and what it does not mean—when your research results approach statistical significance.

In this case, the word significance has nothing at all to do with importance. A finding that is

statistically significant might be unimportant, very important, or just a trivial curiosity.

Statistical significance only means that a pattern of data identified by any statistical test is

strong enough that it probably wouldn’t have happened by chance. If it didn’t happen by

chance, then it probably would happen again if you did the study again. So, you can think

of statistical significance as the “happen again estimate.”

For example, consider once again the hypothesis that listening to different kinds of music

will have an effect on memory. Imagine that the scores on the memory test could range

from 0 to 100, and the average scores were 86 for the classical music group versus 85 for

the rock music group. Is that difference of 1 point enough for you to be confident that

classical music has a better effect on memory than rock music? Probably not.

However, you probably would be more confident if the difference were larger, such as

comparing 86 to 46. In addition, it might matter how many people were in the study. If

you found a difference between 86 and 46 but you had only a single participant in each

group, you probably still wouldn’t be very confident—maybe the person who listened to

classical music would have done better anyway, due to confounding variables such as

intelligence, interest in the material, and so on. And don’t forget that standard deviation

will matter as well.

When a researcher reports that a result is “statistically significant,” that only means that the

patterns of data found in t tests, ANOVAs, or correlations seems stable, that it is likely to

happen again. To determine whether the results of any statistical test reach the level of

“statistical significance,” a final calculation is usually done.

The last calculation is for a p value, which results in a number between zero and 1. The p

value number indicates the probability or likelihood that the pattern of data would have

been found by random chance, if the pattern doesn’t really exist. The p in p value stands for

probability. Let’s go back to our example of music and memory. If music has no effect on

memory, then we’d expect the groups we test to be equal in memory scores. The p value

tells us how likely it is that the results we actually find would exist if the groups were equal.

In other words, it tells us how confident we can be that the groups really are different from

each other. We also calculate p values for ANOVAs and correlations, and the logic is the

same; the number from the p value tells us whether the pattern would have been found by

chance if the groups are the same (for an ANOVA) or if the two variables we’re testing in a

correlation have nothing to do with each other.

129

Statistical significance: The likelihood that the results of an experiment are due to the independent

variable, not chance (see p value).

p value: A number that indicates the probability or likelihood that a pattern of data would have been found

by random chance. Commonly seen as a variation of “p < .05,” which, in this example, means there is a less

than 5% probability the patterns are due to chance.

The Main Ideas

Statistics are mathematical analyses that reveal patterns in data. Researchers use t tests to compare

two groups and analysis of variance (ANOVA) tests to compare three or more groups. Correlations

are used to see whether two variables within a single group have a consistent relationship with each

other.

Correlations can be positive (the variables move in the same direction) or negative (the variables

move in opposite directions), and they can range from zero (no association between the variables) to

a perfect score of 1.00 (all of the scores match the pattern exactly).

All three of these statistical tests (t tests, ANOVAs, and correlations) can have their strength tested

with a final test that calculates a p value. A statistical test’s p value indicates the probability that the

pattern of data found would have happened by random chance, if the pattern isn’t really there.

Most social psychologists want a p value of .05 or less, meaning a 5% or lower chance that the

pattern occurred due to chance.

For several decades, social psychologists (and other scientists) have generally agreed that to

be really confident in conclusions based on statistics, we want to say that we’re 95% sure

that the patterns we find aren’t the result of random chance. You will also hear the phrase

“p < .05” to refer to stable patterns of results. You should know, however, that there are big

changes coming to the traditional world of statistics. If you go to graduate school, then you

probably will hear much more about “effect sizes,” “confidence intervals,” “power analyses,”

“replications,” and “Bayesian statistics.” Many undergraduate textbooks on statistics are

already teaching these new approaches.

130

Critical Thinking Challenge

Scatterplots look for patterns in data within a single group. Experiments look for whether different

groups of people have different outcomes. How many people do you think need to be included in either

research design before we can be confident that the pattern would probably also apply to other people we

haven’t tested? If we chart the study hours and GPAs of five students at your school, do you think it

would be representative of every student? What about 20 students or 100? How do we explain people

who don’t fit into the “typical” pattern?

Most social psychologists have decided that a p value of .05 or less is the cutoff for “statistical

significance.” Do you think that a 5% chance of patterns being found by chance is a low enough

number? Or, do you think it should be higher—would 10% be good enough for you to feel confident?

131

How Can Research Be Analyzed in Terms of Quality?

— LO 2.4: Describe reliability, validity, replication, and ethical standards for research

in social psychology.

You now know quite a bit about how different studies are designed and even how results

are analyzed. Perhaps it is time to start forming and testing your own hypotheses. It may be

the best way to figure out how much you can trust the many studies you will read about in

this (and every other) psychology textbook. If you were going to assign a grade to each

study in this book, what criteria would you use to decide whether a study deserves an A+ or

something a bit lower? In this section, we will consider reliability and validity, the

importance of replication, and ethical considerations.

132

Reliability, Validity, and Replication

Imagine this: Your back hurts from lugging around heavy textbooks all day, so you become

curious about just how much those suckers weigh. You pile them on top of the scale at your

local gym and it gives you a result: 36.8 pounds. No wonder your back hurts! But then

your friend tells you that the scale is often off by several pounds, which she knows because

she sometimes compares what that scale tells her to the scale at her doctor’s office on the

same day. So, you try again, and this time the scale tells you that the exact same pile of

books weighs 33.2 pounds. Surprised, you take the books off and put them back on, and

now it says 40.6 pounds! Clearly, you can’t trust this scale much because it is unreliable.

If you use a scale like this one to weigh yourself, how confident are you that the result is

correct or that it wouldn’t change if you stepped off and back on?

© istockphoto.com/Image Source

This simple example illustrates the first criterion for assessing the quality of a particular

research study: reliability. Reliability is consistency of measurement, over time or over

multiple testing occasions. If you took an intelligence test that said you were smarter than

95% of people in the world but then tried the same test the following week and it said

you’re only smarter than 52% of people in the world, you’d probably be disappointed—

and you wouldn’t know what exactly to think about your intelligence. Reliability is

important because if measurements are not consistent, researchers won’t feel confident that

the tests are particularly useful.

In addition to reliability, scientists also want their studies to be high in validity. Two of the

most important types of validity are internal and external validity. Internal validity is the

level of confidence you have that the results you got from a study really mean what you

think they mean. In other words, internal validity refers to whether patterns of data are

really due to what you’re testing, as opposed to flaws in how the study was designed. For

example, if you tested the hypothesis about how listening to types of music affect memory

but you forgot to use random assignment, you might not have eliminated other

explanations (confounding variables). Random assignment of participants to different

conditions or levels of your independent variable is one of the best ways to increase the

133

Reliability: Consistency of measurement, over time or multiple testing occasions. A study is said to be

reliable if similar results are found when the study is repeated.

Internal validity: The level of confidence researchers have that patterns of data are due to what is being

tested, as opposed to flaws in how the experiment was designed.

External validity: The extent to which results of any single study could apply to other people or settings (see

internal validity of a study.

Good studies also want high levels of external validity, which is the extent to which results

of any single study could apply to other people or settings. Lack of external validity is one

of the most common concerns with single research studies. For example, many of the

studies you’ll read about in this book used participants who were college students. That’s

great for you in some ways, because it means that you can relate to the people who

participated and that the results might apply to you. However, in other ways, it means that

the results might be limited in their generalizability, or how much they apply to people “in

general.” Some people have even used the acronym WEIRD to refer to this problem by

saying that too many studies in psychology have “WEIRD” participants—meaning they are

Western, Educated, and from Industrialized, Rich, Democratic cultures.

If a study only tested a hypothesis on college students from schools in the United States, we

don’t know if the same results would be found among college students in other countries—

or among anyone not in college (children, older people, etc.). The best way to increase

generalizability of results—and thus also increase external validity—is to use a technique

called random sampling. Random sampling means that instead of asking for volunteers to

be in your study, you first identify what your “general” population of interest is, and then

you randomly choose people from that larger group to participate in the study. Statistically,

then, you’ll have a good chance of getting a wider variety of people who represent different

aspects of the larger group. Consider this each time you think about whether a study has

believable results—you can always ask, “Who was in the study? Do they represent everyone,

or just one particular type of person?”

Like all the statistical tools we discussed in the previous section, all of the criteria for a

good-quality study (reliability, internal validity, and external validity) can be strengthened

by replication, conducting the same experiment using the same procedures and the same

materials. Replication is one of the easiest ways to get started in research, yet some students

think that replication is a form of cheating. Replication is not cheating. You can replicate

someone else’s study by using their published procedures, scales, methodologies, and

analyses. You can even e-mail the researchers and ask them to send you their materials, if

needed. If you carefully replicate their study, then you should get pretty close to the same

results, right? Replication solves all sorts of problems for psychological scientists, because it

helps our confidence that our claims are backed up by a reliable pattern of results across

different people in different settings.

134

generalizability).

Generalizability: How much the results of a single study can apply to the general population (see external

validity).

Random sampling: A sampling technique used to increase a study’s generalizability and external validity

wherein a researcher randomly chooses people to participate from a larger population of interest.

Replication: The process of conducting the same experiment using the same procedures and the same

materials multiple times to strengthen reliability, internal validity, and external validity.

135

Ethical Considerations

There’s a certain level of trust that happens when anyone shows up to participate in a

psychology study. As researchers, we want to remember that we have a solemn

responsibility to treat people with respect. Even when we use unobtrusive methodologies

like naturalistic observation or archival studies, all people involved in the study of human

social behavior should be valued—otherwise, our research would be quite ironic.

Researchers across all the sciences provide ethical and legal guidance about what it means to

treat study participants with respect through institutional review boards (IRBs), which are

committees of people who consider the ethical implications of any study. Before any of us

begin formal research that might affect participants and be published, we are obliged to

submit our methods for review. Your local IRB committee is typically composed of

representatives from different departments in a college, university, research institute, or

corporation—and sometimes from a combination of such organizations. The committees

also often have a lawyer as a member, and sometimes they have a member with no

background in research at all, to represent the “average person’s” perspective.

In psychology, the specific guidelines we follow come from both our local IRB committees

and from the American Psychological Association (APA), an organization of professionals

in the field who have determined what the ethical standards of research should be. Note

that the standards currently in effect from the APA, which are described here for you, were

not always in place. Some of the studies you’ll read about in this book were done before

APA standards and IRB committees were standard—and thus, some of those studies have

been harshly criticized for being harmful to the people who were in them. Protecting and

respecting participants should be the highest priority for anyone conducting research. What

are the guidelines?

Ethical violations in the Tuskegee Syphilis Study helped establish the need for IRBs to

prevent future violations of trust. Participants in this study of syphilis were not (a) told the

purpose of the study, (b) given a chance to leave the study, or (c) treated with penicillin

that might have cured them.

136

The National Archives at Atlanta

Spotlight on Research Methods: Questions Your Institutional Review Board Might Ask

Institutional review boards (IRBs) are committees that review ethical considerations for any new study; you

need to get IRB approval before you can move forward. To know more about this process, consider the

questions below; they are the typical types of questions any researcher will need to answer before starting the

research process. Researchers are required to type their answers and submit them to the IRB committee for

review, along with copies of any materials such as surveys, videos, and so on.

What is the purpose of your study, and what are the hypotheses?

What kinds of participants do you plan to use? How many? How will you recruit them? Will they

be compensated for their participation through something like money or extra credit?

Are the participants from any kind of legally protected group, such as children, prisoners, people

with disabilities, and so on?

How will you get informed consent?

What are the specific questions you will ask or experimental procedures you will use?

Will there be any deception? If so, what is the justification for that deception?

How will debriefing occur?

What are the potential harms (physically, emotionally, or mentally) that participants might

experience, short term or long term?

Will you provide any resources if participants have questions or concerns?

Are there any potential benefits participants might experience as a result of being in your study?

How long will your study take? How will you ensure that the data collected will remain confidential

and/or anonymous?

Do you plan to present your results to the public, such as through a conference presentation or

publication? If so, will the participants have access to these results themselves if they are interested?

Hopefully, you can see that IRB committees take their jobs very seriously—as they should.

The American Psychological Association lists several “rights” that they say all participants

should have. Again, consider each of these as you read about studies in this book and

137

The Main Ideas

Reliability, validity, and replication are all criteria regarding the quality of any given research study.

Reliability is consistency of measurement. Internal validity is the extent to which results are

interpreted in an accurate way. External validity is the extent to which results could apply to other

people or settings. Replication means repeating a pattern of results over and over, with different

people and in different settings.

Internal validity is increased when participants are randomly assigned to experimental conditions.

External validity is increased when participants are selected from a larger population of interest

through random sampling, which means the results are more likely to be generalizable.

Ethical considerations are also very important when evaluating research studies. The American

Psychological Association lists several participant rights, such as informed consent and debriefing. In

addition, any study must be approved by an IRB before it can be conducted.

consider whether you think the study was ethical. In addition, if you have the chance to

participate in any studies yourself, it’s important for you to know what your rights are—so

that you can stop participating at any point if you are uncomfortable or feel like you don’t

want to continue for any reason. Some of the participant rights identified by the APA are as

follows:

Informed Consent: Participants should be told what they will be asked to do and

whether there are any potential dangers or risks involved in the study before it begins.

Deception: Participants should be told the truth about the purpose and nature of the

study as much as possible. Deception, or hiding the true nature of the study, is only

allowed when it is necessary because knowing the truth would change how the

participants responded.

Right to Withdraw: Participants have the right to stop being in the study at any

time, for any reason, or to skip questions on a survey if they are not comfortable

answering them.

Debriefing: After completing the study, all participants should be given additional

details about the hypotheses of the study, allowed the opportunity to ask questions,

and even see the results if they wish. This debriefing after the study is complete should

definitely include an explanation of any deception that was involved (if deception

occurred) so that participants have the right to withdraw their data if they are upset

about the deception.

If you want to design your own study, you should consider all of these criteria for the

quality of good research. In addition, you’ll have to get approval from your school’s IRB

committee as well before you begin. To learn more about the IRB approval process, see the

Spotlight on Research Methods feature.

138

Institutional review boards (IRBs): Committees of people who consider the ethical implications of any

study before giving the researcher approval to begin formal research.

American Psychological Association (APA): A large organization of professional psychologists who provide

those in the field with information in the form of scholarly publications, citation style guidelines, and ethical

standards for research.

Informed consent: Participants’ right to be told what they will be asked to do and whether there are any

potential dangers or risks involved before a study begins.

Deception: Hiding the true nature of an experiment from a participant to prevent a change in how the

participant would respond.

Right to withdraw: The right participants have to stop being in a study at any time, for any reason, or to

skip questions on a survey if they are not comfortable answering them.

Debriefing: Additional details given to participants after participation in an experiment, including

information about the hypotheses, an opportunity to ask questions, an opportunity to see the results, and an

explanation of any deception.

Chapter Summary

Critical Thinking Challenge

Imagine a study was done in 1930, before the APA enacted ethical guidelines and before IRB committees

were common. If that study were unethical but highly interesting in terms of the results, should textbooks

still talk about the study and what we learned from it? Does continuing to talk about the study disrespect

the participants, or does learning from it mean that at least we are attempting to get some good from the

bad that already occurred?

Different IRB committees have different levels of strictness regarding ethical thresholds. For example, one

committee might be fine with a study that causes participants to temporarily be angry, sad, or aggressive

—while another committee might consider the same study unethical. If you were to serve on an IRB

committee, how would you decide what the threshold of danger or harm should be? What’s the balance

between possible risk of harm versus what could be learned from the study?

139

What are the elements of the scientific method?

Basic researchers advance theories, while applied researchers translate those theories into real-world settings

or people. Both types of researchers use the scientific method, a systematic way of creating knowledge by (1)

observing patterns of behavior, (2) generating a hypothesis, (3) scientifically testing that hypothesis, and (4)

interpreting the results to form new, more refined hypotheses. In this way, the scientific method is a

circular, never-ending process. Most psychological hypotheses include constructs, which are abstract ideas or

concepts. One important step in research is operationalization of variables, in which researchers explicitly

define those constructs and decide how to measure them.

There are four common types of research in social psychology. The first is archival research, when data are

analyzed that were originally created or gathered for a different purpose (such as newspapers or police

records). The second is naturalistic observation, which is carefully watching people behave in their natural

environments. A common problem with this method is reactivity, which occurs when the people being

observed change their behavior because they are being watched. One specific technique within naturalistic

observation is called “participant observation” and occurs when researchers pretend to be part of the natural

environment they are observing. A third common method is surveys that include self-report scales, in which

participants are explicitly asked to report their thoughts, emotions, and behaviors. While surveys have

several advantages, one concern is the social desirability bias, which is when participants answer questions

dishonestly to make themselves look better. The fourth type of research is an experimental design, which is

described in the next section.

For all of these types of research, a study might have multiple participants or even a single participant.

When a study only examines one person or event, it’s called a case study. Simple archival data show that the

number of case studies being reported in psychology is increasing over time.

140

How do social psychologists design studies?

In addition to archival data, naturalistic observation, and surveys, researchers can design studies based on

experimental or correlational procedures. There are several types of experimental design. Preexperiments test

for the effects of a treatment or technique with a single group of people. This can be done with pretests and

posttests or with only posttests.

True experiments have more than one group of people who experience different conditions. The different

conditions are levels of the independent variable (such as some people who listen to classical music while

others listen to rock music). The outcome being measured at the end of the study is called the dependent

variable (such as scores on a memory test). Ideally, people have been put into different experimental

conditions based on random assignment, meaning everyone has an equal chance of being put in any of the

conditions. Random assignment eliminates confounding variables (other explanations for the results).

Sometimes control groups are included, which is a baseline or neutral group that receives no treatment at all

(for example, a group that listens to no music). When experiments have different people in each condition,

it’s called a between-participants design. When the same people go through all of the different conditions,

it’s called a within-participants design.

Quasi-experiments also compare groups, but the groups are not formed by random assignment. Instead,

they are formed naturally. Examples are men versus women, or people who already prefer classical music to

rock music. Finally, correlational designs simply measure two variables and then test to see if they seem to

have a pattern of association (such as when one goes up, the other also goes up). Height and shoe size are

correlated; taller people tend to have bigger feet. Again, it’s important to remember that just because two

variables are correlated, it doesn’t mean that one causes the other.

141

How do social psychologists analyze and interpret

research results?

To analyze the results of any study, researchers use statistics, which are mathematical tests that reveal

patterns. When two groups are being compared, a t test will compare the groups’ averages and standard

deviations to see if they appear different from each other. The same logic is used when comparing three or

more groups, but that test is called an analysis of variance (or ANOVA for short).

Correlations are the statistic used to detect patterns within a single group of people but between two

variables of interest. For example, a correlation test could tell you if there’s a reliable pattern between hours

spent studying each week and performance in school. Positive correlations mean the two variables move in

the same direction, while negative correlations mean that the two variables move in opposite directions.

Correlations also have a coefficient (or number) between zero and 1.00 (either +1.00 or -1.00). A number

closer to 1 means a stronger or more reliable association between the two variables. A number closer to zero

means that the two variables are weakly associated.

For any of these tests, a p value can tell researchers the probability that the pattern detected in the results

would have happened by random chance if the association isn’t really there. The p in p value stands for

probability, and researchers want it to be as close to zero as possible, meaning the pattern they found wasn’t

due to chance. Most researchers have agreed that to claim that a pattern of results is “statistically

significant,” the p value should be between zero and .05, meaning a less than 5% chance of finding the

pattern due to chance or luck.

142

How can research be analyzed in terms of quality?

Reliability, validity, and replication are all important things to consider when trying to evaluate the quality

of a research study. Reliability is whether the measures used are consistent. There are several forms of

validity, including internal and external validity. Internal validity is the extent to which results are

interpreted in an accurate way. For example, causal claims should only be made after true experiments and

even then should only be made if people were randomly assigned to the different experimental conditions.

External validity is the extent to which results could apply to other people or settings. If a certain pattern of

results can only be found within a specific group of people, then the results have low generalizability (they

are not true in general for the larger group of interest). The best way to increase generalizability (and thus

external validity) is to choose participants randomly from the larger group of interest; this technique is

called random sampling. Random sampling will hopefully result in a diverse variety of different kinds of

people in the study. Finally, we can only be confident with results if they are replicated (or repeatedly

found) across several studies with different people and different settings.

It is also essential to consider the ethical implications of any study. The American Psychological Association

has published a list of participant rights that should apply to any psychological study. In addition, local

organizations such as universities will have an institutional review board (IRB), a committee that reviews the

ethics of any study before it is carried out. Examples of ethical practices are that participants should have

informed consent (be told the nature of the study in advance), have the ability to quit the study or skip

parts that make them uncomfortable, and that deception (or deceiving the participants) should only be

done when necessary. Finally, a debriefing session should occur at the end, which is when participants are

given more details about the study and are given the opportunity to ask questions or express concerns.

143

Critical Thinking, Analysis, and Application

Which of the research methods described in this chapter seem the most appealing to you? Why is that

method appealing, and what issues or concerns do you have with the other methods?

Find a news report that makes a claim that one variable causes another (for example, “Drug X leads to

bad behavior,” or “Access to birth control leads to risky sex,” etc.). Is the causal relationship being

suggested one that seems valid? Why or why not? How could this causal claim be scientifically tested?

How many times does a study have to be replicated in order for researchers—or the general public—to be

confident in the results? Even if a study is never replicated, does that mean that the data are useless? What

other explanations could there be for results that seem to change (in other words, aren’t replicated)? Is it

possible that researchers simply haven’t identified the exact reason why the pattern happened the first

time, for example?

Ideally, studies have generalizable samples of people who participate. But what if you want to generalize

your findings to all of humanity? It’s clearly impossible to use random sampling across everyone in the

world, so most people just use participants who are nearby volunteers. What’s the balance between

convenience for the researcher and the need for a diverse, generalizable sample of participants? Are any

studies truly high in external validity if they don’t have true random sampling?

144

Visit edge.sagepub.com/heinzen to help you accomplish your coursework goals in an easy-to-use learning

environment.

Visual Key Concepts

Mobile-friendly eFlashcards

Mobile-friendly practice quizzes

Video and multimedia content

EXCLUSIVE! Access to full-text SAGE journal articles

LO2.1 Describe how the scientific method creates knowledge.

basic researchers 27

applied researchers 27

scientific method 27

hypothesis 28

constructs 29

operationalize 29

archival data 30

naturalistic observation 30

reactivity 30

participant observation 30

survey 31

self-report scale 31

social desirability bias 31

case study 32

Personal Reflection

Here’s the great and terrible thing about thinking like a scientist: It’s psychologically addicting. The track behind

my house is a weird shape because it circles both the baseball field and the football field. Do runners perform

better on this strange-shaped site? I could answer that question with the tools we described in this chapter. But I

won’t because I don’t have time; there are other, even more inviting questions that we can answer. Although it is

perfectly safe, the runway for the long jump at another high school in our region is situated so that athletes feel as

if they are almost jumping off a cliff. Apparently, my daughter was not alone in jumping her personal best at that

site—by almost 16 inches! But I don’t have the time for that one either. Right now, I want to know whether the

principles that make a well-designed game psychologically addictive can help at-risk students complete college.

We’re getting answers to that question, too, but every single answer prompts more questions. Research is fun,

frustrating, and addictive—you always want more—and more leads to wanting even more. But it’s not just a

healthy addiction—the scientific method is the engine of innovation across all the sciences. Social psychology,

public health, medicine . . . we all are making slow, steady progress by using the scientific method. [TH]

145

PsycINFO database 33

LO2.2 Compare the logic behind preexperiments, true experiments, quasi-experiments,

and correlational designs.

preexperiment 34

one-shot case study 34

one-group pretest-posttest design 35

confounding variables 35

experiment 35

random assignment to experimental condition 36

independent variable 36

dependent variable 36

pretest-posttest control group design 36

quasi-experiment 39

control group 37

posttest-only control group design 37

between-participants design 39

within-participants design 39

order effects 39

correlational design 39

LO2.3 Summarize the most common ways to analyze and interpret data.

146

statistics 42

standard deviation 42

t test 44

analysis of variance (ANOVA) 44

correlation 44

scatterplot 44

correlation coefficient 44

positive correlation 45

negative correlation 45

statistical significance 46

p value 46

LO2.4 Describe reliability, validity, replication, and ethical standards for research in

social psychology.

reliability 48

internal validity 48

external validity 48

generalizability 48

random sampling 48

replication 49

147

institutional review boards (IRBs) 49

American Psychological Association (APA) 49

informed consent 50

deception 51

right to withdraw 51

debriefing 51

Get a tour of All Visual Key Concepts and their definitions at edge.sagepub.com/heinzen