Psychologists try to avoid the pitfalls of intuitive thinking by using the scientific method. They observe events, form theories, and then refine their theories in the light of new observations.

Chatting with friends and family, we often use theory to mean “mere hunch.” In science, a theory explains behaviors or events by offering ideas that organize what we have observed. By organizing isolated facts, a theory simplifies. There are too many facts about behavior to remember them all. By linking facts to underlying principles, a theory offers a useful summary. It connects many small dots so that a clear picture emerges.

13

A theory about the effects of sleep on memory, for example, helps us organize countless sleep-related observations into a short list of principles. Imagine that we observe over and over that people with good sleep habits tend to answer questions correctly in class, and they do well at test time. We might therefore theorize that sleep improves memory. So far so good: Our principle neatly summarizes a list of observations about the effects of a good night’s sleep.

Yet no matter how reasonable a theory may sound—and it does seem reasonable to suggest that sleep could improve memory—we must put it to the test. A good theory produces testable predictions, called hypotheses. Such predictions specify what results (what behaviors or events) would support the theory and what results would cast doubt on the theory. To test our theory about the effects of sleep on memory, our hypothesis might be that when sleep deprived, people will remember less from the day before. To test that hypothesis, we might assess how well people remember course materials they studied before a good night’s sleep, or before a shortened night’s sleep (FIGURE 1.1). The results will either support our theory or lead us to revise or reject it.

Our theories can bias our observations. The urge to see what we expect to see is always present, both inside and outside the laboratory. Having theorized that better memory springs from more sleep, we may see what we expect: We may perceive sleepy people’s comments as less insightful.

As a check on their biases, psychologists use operational definitions when they report their studies. “Sleep deprived,” for example, may be defined as “2 or more hours less” than the person’s natural sleep. These exact descriptions will allow anyone to replicate (repeat) the research. Other people can then re-create the study with different participants and in different situations. If they get similar results, we can be more confident that the findings are reliable.

Replication is an essential part of good science. When 270 psychologists recently worked together to redo 100 psychological studies, the results made news: Only 36 percent of the results were replicated (Open Science Collaboration, 2015). (None of the nonreproducible findings appears in this text.) But then another team of scientists found most of the failed replications flawed and “the reproducibility of psychological science” to be “quite high” (Gilbert et al., 2016).Other fields, including medicine, also have seeming issues with nonreplicated findings (Collins & Tabak, 2014). Especially when based on a small sample, a single failure to replicate can itself need replication (Maxwell et al., 2015). In all scientific fields, replication either confirms findings, or enables us to correct or refine our knowledge.

Let’s summarize. A good theory:

14

We can test our hypotheses and refine our theories in several ways.

To think critically about popular psychology claims, we need to understand the strengths and weaknesses of these methods. (For more information about some of the statistical methods that psychological scientists use in their work, see Appendix A, Statistical Reasoning in Everyday Life.)

In daily life, we all observe and describe other people, trying to understand why they think, feel, and act as they do. Professional psychologists do much the same, though more objectively and systematically, using

The Case Study

A case study examines one individual or group in depth, in the hope of revealing things true of us all. Some examples: Medical case studies of people who lost specific abilities after damage to certain brain regions gave us much of our early knowledge about the brain. Jean Piaget, the pioneer researcher on children’s thinking, carefully watched and questioned just a few children. Studies of only a few chimpanzees jarred our beliefs about what other species can understand and communicate.

Intensive case studies are sometimes very revealing. They often suggest directions for further study, and they show us what can happen. But individual cases may also mislead us. The individual being studied may be atypical (not like those in the larger population). Viewing such cases as general truths can lead to false conclusions. Indeed, anytime a researcher mentions a finding (Smokers die younger: 95 percent of men over 85 are nonsmokers), someone is sure to offer an exception (Well, I have an uncle who smoked two packs a day and lived to be 89). These vivid stories, dramatic tales, and personal experiences command attention and are easily remembered. Stories move us, but stories—even when they are psychological case examples—can mislead. A psychologist’s single case of someone who reportedly changed from gay to straight is not evidence that sexual orientation is a choice. As psychologist Gordon Allport (1954, p. 9) said, “Given a thimbleful of [dramatic] facts we rush to make generalizations as large as a tub.”

The point to remember: Individual cases can suggest fruitful ideas. What is true of all of us can be seen in any one of us. But just because something is true of one of us (the atypical uncle), we should not assume it is true of all of us (most long-term smokers do suffer ill health and early deaths). To uncover general truths, we must look to methods beyond the case study.

Naturalistic Observation

A second descriptive method records behavior in a natural environment. These naturalistic observations may describe parenting practices in different cultures, students’ self-seating patterns in American lunchrooms, or chimpanzee family structures in the wild.

The scope of naturalistic observations is expanding. Until recently, naturalistic observation was mostly “small science”—possible with pen and paper rather than fancy equipment and a big budget (Provine, 2012). But new technologies have expanded the scope of naturalistic observations. The billions of people entering personal information on sites such as Facebook, Twitter, and Google have created a huge new opportunity for “big data” observations. To track the ups and downs of human moods, one study counted positive and negative words in 504 million Twitter messages from 84 countries (Golder & Macy, 2011). When were people happiest? As FIGURE 1.2 shows, spirits seemed to rise on weekends, shortly after waking, and in the evenings. (Are late Saturday evenings often a happy time for you, too?) Another study found that the proportion of negative emotion words (especially anger-related words) in 148 million tweets from 1347 U.S. counties predicted the counties’ heart disease rates (Eichstaedt et al., 2015). How well did it predict heart disease rates? Better than other traditional predictors, such as smoking and obesity.

15

Smart-phone apps and body-worn sensors are also offering new ways to collect data. Using such tools, researchers can track willing volunteers’ locations, activities, and opinions—without interfering with the person’s activity. In one study, 79 introductory psychology students donned electronically activated recording devices (EARs) (Mehl et al., 2010). For up to four days, the EARs captured 30-second snippets of the students’ waking hours, turning on every 12.5 minutes. By the end of the study, researchers had eavesdropped on more than 23,000 half-minute life slices. Was happiness related to having simple talks or deeply involved conversations? The happiest participants avoided small talk and embraced meaningful conversations. Happy people would rather talk than tweet. Does that surprise you?

Like the case study method, naturalistic observation does not explain behavior. It describes it. Nevertheless, descriptions can be revealing: The starting point of any science is description.

The Survey

A survey looks at many cases in less depth, asking people to report their own behavior or opinions. Questions about everything from sexual practices to political opinions are put to the public. In recent surveys,

But asking questions is tricky, and the answers often depend on the way you word your questions and on who answers them.

WORDING EFFECTS Even subtle changes in the wording of questions can have major effects. Should violence be allowed to appear in children’s television programs? People are much more likely to approve “not allowing” such things than “forbidding” or “censoring” them. In one national survey, only 27 percent of Americans approved of “government censorship” of media sex and violence, though 66 percent approved of “more restrictions on what is shown on television” (Lacayo, 1995). People are much more approving of “aid to the needy” than of “welfare,” and of “revenue enhancers” than of “taxes.” Wording is a delicate matter, and some words can trigger positive or negative reactions. Critical thinkers will reflect on how a question’s phrasing might affect the opinions people express.

16

RANDOM SAMPLING For an accurate picture of a group’s experiences and attitudes, there’s only one game in town. In a representative sample, a smaller group can accurately reflect the larger population you want to study and describe.

So how do you obtain a representative sample? Say you want to survey the total student population at your school to get their reaction to an upcoming tuition increase. To be sure your sample represents the whole student population, you will want to choose a random sample, in which every person in the entire population has an equal chance of being picked. You would not want to ask for volunteers, because those extra-nice students who step forward to help would not necessarily be a random sample of all the students. But you could assign each student a number, and then use a random number generator to select a sample.

Time and money will affect the size of your sample, but you would try to involve as many people as possible. Why? Because large representative samples are better than small ones. (But a smaller representative sample of 100 is better than a larger unrepresentative sample of 500.)

Political pollsters sample voters in national election surveys just this way. Using only 1500 randomly sampled people, drawn from all areas of a country, they can provide a remarkably accurate snapshot of the nation’s opinions. Without random sampling, large samples—including call-in phone samples and TV or website polls—often give misleading results.

The point to remember: Before accepting survey findings, think critically. Consider the wording of the questions and the sample. The best basis for generalizing is from a random sample of a population.

Describing behavior is a first step toward predicting it. Naturalistic observations and surveys often show us that one trait or behavior relates to another. In such cases, we say the two correlate. A statistical measure (the correlation coefficient) helps us figure how closely two things vary together, and thus how well either one predicts the other. Knowing how much aptitude tests correlate with school success tells us how well the scores predict school success.

The point to remember: A correlation coefficient helps us see the world more clearly by revealing the extent to which two things relate.

Correlation and Causation

Consider some recent headlines:

What shall we make of these correlations? Do they indicate that students would achieve more if their parents supported them less? That stopping smoking would improve mental health? That more sleep would produce better mental health?

17

No, because such correlations do not come with built-in cause-effect arrows. In two recent studies, sexual hook-ups correlated with college women’s experiencing depression; delaying sexual intimacy correlated with positive outcomes such as greater relationship satisfaction and stability (Fielder et al., 2013; Willoughby et al., 2014). Do these findings mean that sexual restraint causes better outcomes? It might. But in this case, as in many others, causation might work the other way around (more depressed people are more likely to hook up). Alternatively, some third factor, such as lower impulsivity, might underlie both sexual restraint and psychological well-being.

But correlations do help us predict. Here’s an example: Self-esteem correlates negatively with (and therefore predicts) depression. (The lower people’s self-esteem, the more they are at risk for depression.) But does that mean low self-esteem causes depression? If you think the answer is Yes, you are not alone. We all find it hard to resist thinking that associations prove causation. But no matter how strong the relationship, they do not!

How else might we explain the negative correlation between self-esteem and depression? As FIGURE 1.3 suggests, we’d get the same correlation between low self-esteem and depression if depression caused people to be down on themselves. And we’d also get that correlation if something else—a third factor such as heredity or some awful event—caused both low self-esteem and depression.

This point is so important—so basic to thinking smarter with psychology—that it merits one more example, this one from a survey of over 12,000 adolescents. The more those teens felt loved by their parents, the less likely they were to behave in unhealthy ways—having early sex, smoking, abusing alcohol and drugs, behaving violently (Resnick et al., 1997). “Adults have a powerful effect on their children’s behavior right through the high school years,” gushed an Associated Press (AP) news report on the study. But correlations don’t prove causation. Thus, the AP could as well have said, “Children’s behavior has a powerful effect on their parents’ feelings toward them.”

The point to remember (turn up the volume here): Correlation indicates the possibility of a cause-effect relationship, but it does not prove causation. Knowing that two events are associated does not tell us anything about which causes the other. Remember this principle and you will be wiser as you read and hear news of scientific studies.

18

Descriptions, even with big data, don’t prove causation. Correlations don’t prove causation. To isolate cause and effect, psychologists have to simplify the world. In our everyday lives, many things affect our actions and influence our thoughts. Psychologists sort out this complexity by using experiments. With experiments, researchers can focus on the possible effects of one or more factors by

Let’s consider a few experiments to see how this works.

Random Assignment: Minimizing Differences

Researchers have compared infants who are breast-fed with those who are bottle-fed with formula. Some studies, but not others, show that children’s intelligence test scores are a tad higher if they were breast-fed (von Stumm & Plomin, 2015; Walfisch et al., 2014). The longer they were breast-fed, the higher their later scores (Jedrychowski et al., 2012; Victora et al., 2015). So we can say that mother’s milk may correlate modestly but positively with later intelligence. But what does this mean? Do smarter mothers (who in modern countries more often breast-feed) have smarter children? Or do the nutrients in mother’s milk contribute to brain development? Even big data from a million or a billion mothers and their offspring wouldn’t tell us.

To find the answer, we would have to isolate the effects of mother’s milk from the effects of other factors, such as mother’s age, education, and intelligence. How might we do that? By experimenting. With parental permission, one British research team directly experimented with breast milk. They randomly assigned 424 hospitalized premature infants either to formula feedings or to breast-milk feedings (Lucas et al., 1992). By doing this, they created two otherwise similar groups:

Random assignment (whether by means of a random-number generator or by the flip of a coin) minimizes any preexisting differences between the two groups. If one-third of the volunteers for an experiment can wiggle their ears, then about one-third of the people in each group will be ear wigglers. So, too, with age, intelligence, attitudes, and other characteristics, which will be similar in the experimental and control groups. When groups are formed by random assignment, and they differ at the experiment’s end, we can assume the treatment had an effect.

Is breast best? The British experiment found that, at least for premature infants, breast milk is indeed best for developing intelligence. On intelligence tests taken at age 8, those nourished with breast milk scored significantly higher than those who had been formula-fed.

The point to remember: Unlike correlational studies, which uncover naturally occurring relationships, an experiment manipulates (varies) a factor to determine its effect.

The Double-Blind Procedure: Eliminating Bias

In the breast-milk experiment, babies didn’t have expectations that could affect the experiment’s outcome. Adults do have expectations.

Consider: Three days into a cold, many of us start taking vitamin C tablets. If we find our cold symptoms lessening, we may credit the pills. But after a few days, most colds are naturally on their way out. Was the vitamin C cure truly effective? To find out, we could experiment.

And that is precisely what investigators do to judge whether new drug treatments and new methods of psychotherapy are effective (Chapter 14). They use random assignment to form the groups. An experimental group receives the treatment, such as a medication. A control group receives a placebo (an inactive substance—perhaps a look-alike pill with no drug in it). Often, the people who take part in these studies are blind (uninformed) about which treatment, if any, they are receiving.

Many studies use a double-blind procedure—neither those taking part in the study nor those collecting the data know which group is receiving the treatment. In such studies, researchers can check a treatment’s actual effects apart from the participants’ belief in its healing powers and the staff’s enthusiasm for its potential. Just thinking you are getting a treatment can boost your spirits, relax your body, and relieve your symptoms. This placebo effect is well documented in reducing pain, depression, and anxiety (Kirsch, 2010). Athletes have run faster when given a fake performance-enhancing drug (McClung & Collins, 2007). Decaf-coffee drinkers have reported increased vigor and alertness—when they thought their brew had caffeine in it (Dawkins et al., 2011). People have felt better after receiving a phony mood-enhancing drug (Michael et al., 2012). And the more expensive the placebo, the more “real” it seems—a fake pill that cost $2.50 worked better than one costing 10 cents (Waber et al., 2008). So what do you think happens when people have repeatedly received a placebo and have repeatedly reported decreased pain? Even when they learn they are taking a placebo, they continue to report reduced pain (Schafer et al., 2015). To know how effective a therapy really is, researchers must control for a possible placebo effect.

19

Independent and Dependent Variables

Here is an even more potent example: The drug Viagra was approved for use after 21 clinical trials. One trial was an experiment in which researchers randomly assigned 329 men with erectile disorder to either an experimental group (Viagra takers) or a control group (placebo takers). The pills looked identical, and the procedure was double-blind—neither the men taking the pills nor the people giving them knew who received the placebo. The result: Viagra worked. At peak doses, 69 percent of Viagra-assisted attempts at intercourse were successful, compared with 22 percent for men receiving the placebo (Goldstein et al., 1998).

This simple experiment manipulated just one factor—the drug (Viagra versus no Viagra). We call the manipulated factor an independent variable: We can vary it independently of other factors, such as the men’s age, weight, and personality. These other factors, which could influence a study’s results, are called confounding variables. Thanks to random assignment, the confounding variables should be roughly equal in both groups.

Experiments examine the effect of one or more independent variables on some behavior or mental process that can be measured. We call this kind of affected behavior the dependent variable because it can vary depending on what takes place during the experiment. Experimenters give both variables precise operational definitions. They specify exactly how the

Operational definitions answer the “What do you mean?” question with a level of precision that enables others to replicate (repeat) the study.

Let’s see how this works with the British breast-milk experiment (FIGURE 1.4). A variable is anything that can vary (infant nutrition, intelligence). Experiments aim to manipulate an independent variable (type of milk), measure a dependent variable (later intelligence test score), and control confounding variables. An experiment has at least two different groups: an experimental group (infants who received breast milk) and a comparison or control group (infants who did not receive breast milk). Random assignment works to control all other (confounding) variables by equating the groups before any manipulation begins. In this way, an experiment tests the effect of at least one independent variable (what we manipulate) on at least one dependent variable (the outcome we measure).

20

In another experiment, psychologists tested whether landlords’ perceptions of an applicant’s ethnicity would influence the availability of rental housing. The researchers sent identically worded e-mails to 1115 Los Angeles–area landlords (Carpusor & Loges, 2006). They varied the sender’s name to imply different ethnic groups: “Patrick McDougall,” “Said Al-Rahman,” and “Tyrell Jackson.” Then they tracked the percentage of landlords’ positive replies. How many e-mails triggered invitations to view the apartment? For McDougall, 89 percent; for Al-Rahman, 66 percent; and for Jackson, 56 percent.

Throughout this book, you will read about amazing psychological science discoveries. TABLE 1.2 compares the features of psychology’s main research methods. In later chapters, you will read about other research designs, including twin studies (Chapter 3) and cross-sectional and longitudinal research (Appendix A). But how do we know fact from fiction? How do psychological scientists choose research methods and design their studies in ways that provide meaningful results? Understanding how research is done—how testable questions are developed and studied—is key to appreciating all of psychology.

In psychological research, no questions are off limits, except untestable ones. Does free will exist? Are people born evil? Is there an afterlife? Psychologists can’t test those questions, but they can test whether free will beliefs, aggressive personalities, and a belief in life after death influence how people think, feel, and act (Dechesne et al., 2003; Shariff et al., 2014; Webster et al., 2014).

Having chosen their question, psychologists then select the most appropriate research design—experimental, correlational, case study, naturalistic observation, twin study, longitudinal, or cross-sectional—and determine how to set it up most effectively. They consider how much money and time are available, ethical issues, and other limitations. For example, it wouldn’t be ethical for a researcher studying child development to use the experimental method and randomly assign children to loving versus punishing homes.

21

Next, psychological scientists decide how to measure the behavior or mental process being studied. For example, researchers could measure aggressive behavior by measuring participants’ willingness to blast a stranger with intense noise.

Researchers want to have confidence in their findings, so they carefully consider confounding variables—factors other than those being studied that may affect their interpretation of results.

Psychological research is a fun and creative adventure. Researchers design each study, measure target behaviors, interpret results, and learn more about the fascinating world of behavior and mental processes along the way.

When you see or hear about psychology research, do you ever wonder whether people’s behavior in a research laboratory will predict their behavior in real life? Does detecting the blink of a faint red light in a dark room say anything useful about flying a plane at night? Or, suppose an experiment shows that a man aroused by viewing a violent, sexually explicit film will then be more willing to push buttons that he thinks will electrically shock a woman. Does that really say anything about whether violent pornography makes men more likely to abuse women?

Before you answer, consider this. The experimenter intends to simplify reality—to create a mini-environment that imitates and controls important features of everyday life. Just as a wind tunnel lets airplane designers re-create airflow forces under controlled conditions, a laboratory experiment lets psychologists re-create psychological forces under controlled conditions.

An experiment’s purpose is not to re-create the exact behaviors of everyday life, but to test theoretical principles (Mook, 1983). In aggression studies, deciding whether to push a button that delivers a shock may not be the same as slapping someone in the face, but the principle is the same. It is the resulting principles—not the specific findings—that help explain everyday behaviors. Many investigations have shown that principles derived in the laboratory do typically generalize to the everyday world (Anderson et al., 1999).

The point to remember: Psychological science focuses less on specific behaviors than on revealing general principles that help explain many behaviors.