The data collected from naturalistic observations, case studies, and surveys provide valuable descriptions of behavior and mental processes, but they can do more than that. These data can also be examined to see what they reveal about the relation-ships, or correlations, between one research variable and another. For example, fear surveys show that most people have fears, but correlational analysis of those sur-veys also shows that the number of fears is related to age. Specifically, adults have fewer fears than children (e.g., Kleinknecht, 1991). examine relationships between variables in order to describe research data more fully, to test predictions, to evaluate theories, and to suggest new hypotheses about why people think and act as they do.
Consider the question of how aggression develops. One theory suggests that people learn to be aggressive by seeing aggressiveness in others. Psychologists have tested this theory through correlational studies that focus on the relationship between children’s aggressiveness and the amount of aggression they see on tele-vision. Just as the theory predicts, those who watch a lot of televised violence do tend to be more aggressive than other children. Another theory asserts that sexual aggressiveness in adults can be triggered by viewing pornography. And in fact, correlational analyses of case studies and surveys show that sex criminals often view pornographic material just prior to committing their offenses. And correla-tional studies of observacorrela-tional data indicate that children in day care for more than thirty hours a week are more aggressive than those who stay at home with their mothers.
Do violent television, pornography, and separation from parents actually cause the various forms of aggressiveness with which they have been associated? They might, but psychologists must be careful about jumping to such conclusions. The most obvious explanation for the relationship found in a correlational study may not always be the correct one (see Table 2.2). Perhaps the correlation between aggres-sion and violent televiaggres-sion appears because children who were the most aggressive in the first place are also the ones who choose to watch the most violent television.
Perhaps sex offenders exaggerate the role of pornography in their crimes because they hope to avoid taking responsibility for those crimes. And perhaps the aggres-siveness seen among some children in day care might have something to do with the children themselves or with what happens to them in day care, not just with separa-tion from their parents.
쐌correlational study A research method that examines relationships between variables in order to analyze trends in data, to test predictions, to evaluate theories, and to suggest new hypotheses.
Correlational studies
38 ●
CHAPTER 2 Research in PsychologyOne way psychologists evaluate hypotheses such as these is to conduct further correlational studies in which they look for trends in observational, case study, and survey data that support or conflict with those hypotheses. Further analysis of day-care research, for example, shows that the aggressiveness seen in preschoolers who spend a lot of time in day care is the exception, not the rule. Most children don’t show any behavior problems, no matter how much time they have spent in day care.
This more general trend suggests that whatever effects separation has, it may be dif-ferent for difdif-ferent children in difdif-ferent settings, causing some to express aggres-siveness, others to display fear, and still others to find enjoyment (see the chapter on human development). To explore this possibility, psychologists will have to conduct further studies to examine correlations between children’s personality traits, quali-ties of different day-care programs, and reactions to day care (NICHD Early Child Care Research Network, 2007). Throughout this book you will see many more examples of how correlational studies help to shed light on a wide range of topics in psychology.
Research Methods in Psychology
● 39
Correlation Possible Explanation
A survey found that the more sexual content that It might have been some teens’ greater interest in sex that led them to U.S. teenagers reported watching on television, watch more sexually oriented shows and also to become sexually the more likely they were to begin having sex active.
themselves during the following year (Collins et al., 2004).
The number of drownings in the United States This relationship probably reflects a third variable—time of year—that rises and falls during the year, along with the affects both ice cream consumption and the likelihood of swimming amount of ice cream sold each month. and boating.
In places where beer prices are raised, the If price increases cause less beer consumption, people might stay number of new cases of sexually transmitted sober enough to remember to use condoms during sexual encounters.
disease falls among young people living in those The relationship could also reflect coincidence, because prices do not
places. always affect alcohol use. More research is required to understand this
correlation.
A study found that the more antibiotics a woman Long-term antibiotic use might have impaired the women’s immune has taken, and the longer she has taken them, systems, but the cancer risk might also have been increased by the the greater is her risk of breast cancer (Velicer diseases that were being treated with antibiotic drugs, not the drugs
et al., 2004). themselves. Obviously, much more research would be required before
condemning the use of antibiotics.
Individuals and teams wearing red are more Does red convey a signal of dominance that intimidates opponents?
likely to win fights and other athletic contests Possibly, but other research has found color-related outcome patterns than those wearing other colors (Hill & Barton, when neither contestant wears red (Rowe, Harris, & Roberts, 2005), so
2005). other factors, including coincidence, might be at work.
The U.S. stock market rises during years in which The so-called Super Bowl Effect has occurred 30 times in 37 years;
a team from the National Football Conference striking as this might seem, coincidence seems to be the most likely wins the Super Bowl and falls during years in explanation.
which an American Conference team wins.
Look at the relationships described in the left-hand column, then ask yourself why the two variables in each case are correlated. Could one variable be causing an effect on the other? If so, which variable is the cause, and how
might it exert its effect? Could the relationship between the two variables be caused by a third one? If so, what might that third variable be? We suggest some possible explanations in the right-hand column. Can you think of others?
Correlation and Causation TA B L E
2.2
THISTRY
Experiments: Exploring Cause and Effect
Still, the surest way to test hypotheses and confirm cause-effect relationships between variables is to exert some control over those variables. This kind of research usually takes the form of an experiment. are situations in which the researcher manipulates one variable and then observes the effect of that manipulation on another variable, while holding all other variables constant.
Consider the experiment Francine Shapiro conducted in an attempt to better understand the effects of EMDR. As illustrated in Figure 2.1, she first identified twenty-two people who were suffering the ill effects of traumas such as rape or mil-itary combat. These were her research participants. She then assigned each of the participants to one of two groups. The first group received a single fifty-minute ses-sion of EMDR treatment; the second group focused on their unpleasant memories for eight minutes, but without moving their eyes back and forth (Shapiro, 1989b).
The group that receives an experimental treatment such as EMDR is called, nat-urally enough, the . The group that receives no treatment or some other treatment is called the . Control groups provide baselines against which to compare the performance of other groups. In Shapiro’s experiment, having a control group allowed her to measure how much change in anxiety could be expected from exposure to bad memories without EMDR treatment. If everything about the two groups were exactly the same before the experiment, then any differ-ence in anxiety between the groups afterward would have something to do with the EMDR treatment rather than with mere exposure to unpleasant memories.
Notice that Shapiro controlled one variable in her experiment, namely which kind of treatment her participants received. In an experiment, the variable controlled by the experimenter is called the . It is called independent because the experimenter is free to adjust it at will, offering one, two, or three kinds of treatment, for example, or perhaps setting the length of treatment at one, five, or ten sessions.
Notice, too, that Shapiro looked for the effects of treatment by measuring a different variable, namely her clients’ anxiety level. This second variable is called the because it is affected by, or depends on, the independent variable.
So in Shapiro’s experiment, the presence or absence of treatment was the independent variable, because she manipulated it. Her participants’ anxiety level was the depend-ent variable, because she measured it to see how it was affected by treatmdepend-ent. (Table 2.3 describes the independent and dependent variables in other experiments.)
The results of Shapiro’s (1989b) experiment showed that participants who received EMDR treatment experienced a complete and nearly immediate reduction in anxiety related to their traumatic memories, whereas those in the control group showed no change. This difference suggests that EMDR caused the improvement.
dependent variable
independent variable 쐌experiment A situation in which the
researcher manipulates one variable and then observes the effect of that manipulation on another variable, while holding all other variables constant.
쐌experimental group In an
experiment, the group that receives the experimental treatment.
쐌control group In an experiment, the group that receives no treatment or provides some other baseline against which to compare the performance or response of the experimental group.
쐌independent variable The variable manipulated by the researcher in an experiment.
쐌dependent variable In an experiment, the factor affected by the independent variable.
control group experimental group
Experiments
Ideally, the only difference between the experimental and control groups in experiments such as this one is whether the participants receive the treatment the experimenter wishes to evaluate. Under such ideal circumstances, any difference in the two groups’ reported levels of anxiety at the end of the experiment would be due only to whether or not they received treatment.
F I G U R E
2.1
A Simple Two-Group Experiment
1. Preliminary screening of participants
3. Treatment phase 4. Post-treatment phase
40 ●
CHAPTER 2 Research in PsychologyBut look again at the structure, or design, of the experiment. The EMDR group’s session lasted about fifty minutes, but the control group focused on their memories for only eight minutes. Would the people in the control group have improved, too, if they had spent fifty minutes focusing on their memories? We don’t know, because the experiment did not compare methods of equal duration.
Anyone who conducts or relies on research must be on guard for such flaws in experimental design. So before drawing conclusions from research, experimenters must consider factors that might confound, or confuse, the interpretation of results.
Any factor, such as differences in the length of treatment, that might have affected the dependent variable along with or instead of the independent variable can become a . When confounding variables are present, the exper-imenter cannot know whether the independent variable or the confounding variable produced the results. Let’s examine three sources of confounding: random variables, participants’ expectations, and experimenter bias.
●
Random Variables
In an ideal research world, everything about the exper-imental and control groups would be the same except for their exposure to the inde-pendent variable (such as whether or not they received treatment). In reality, however, there are always other differences between the groups that reflect random variables. are uncontrolled, sometimes uncontrollable, factors such as the time of year when research takes place and differences in the participants’backgrounds, personalities, life experiences, and vulnerability to stress, for example.
In fact, there are so many ways in which participants might vary from each other that it is usually impossible to form groups that are matched on all of them. Instead, experimenters simply flip a coin or use some other random process to assign each research participant to experimental or control groups. These procedures—called
—are presumed to distribute the impact of uncontrolled variables randomly (and probably about equally) across groups, thus minimizing the chance that these variables will distort the results of the experiment (Shadish, Cook, &
Campbell, 2002).
●
Participants’ Expectations
After eight minutes of focusing on unpleasant memories, participants in the control group in Shapiro’s (1989b) experiment were instructed to begin moving their eyes. At that point they, too, said they began to experience a reduction in anxiety. Was this improvement caused by the eye move-ments themselves, or could it be that the instructions made the participants feel random assignment쐌confounding variable In an
experiment, any factor that affects the dependent variable, along with or instead of the independent variable.
쐌random variable In an experiment, a confounding variable in which uncontrolled or uncontrollable factors affect the dependent variable, along with or instead of the independent variable.
쐌random assignment The procedure by which random variables are evenly distributed in an experiment by putting participants into various groups through a random process.
Random variables confounding variable
Research Methods in Psychology
● 41
1. Children’s reading skill is measured after taking either a The independent variable is . special reading class or a standard reading class. The dependent variable is . 2. College students’ memory for German vocabulary words is The independent variable is .
tested after a normal night’s sleep or a night of no sleep. The dependent variable is . 3. Experiment title: “The effect of a daily walking program on The independent variable is .
elderly people’s lung capacity.” The dependent variable is .
4. People’s ability to avoid “accidents” in a driving simulator is The independent variable is . tested before, during, and after talking on a cell phone. The dependent variable is .
Fill in the names of the independent and dependent variables in each of these experiments (the answers are listed at the bottom of page 42). Remember that the
independent variable is manipulated by the experimenter. The dependent variable is measured to determine the effect of the independent variable. How did you do on this task?
Independent and Dependent Variables TA B L E
2.3
THISTRY
more confident that they were now getting “real” treatment? This question illus-trates a second source of confounding: differences in what people think about the experimental situation. If participants who receive an impressive treatment expect that it will help them, they may try harder to improve than those in a control group who receive no treatment or a less impressive treatment. When improvement is cre-ated by a participant’s knowledge and expectations, it is called the placebo effect. A (pronounced “pla-SEE-boe”) is a treatment that contains nothing known to be helpful but that nevertheless produces benefits because a person believes it will be beneficial.
How can researchers measure the extent to which a result is caused by the independent variable or by a placebo effect? Usually, they include a special control group that receives only a placebo treatment. Then they compare results for the experimental group, the placebo group, and a no-treatment group. In one quit-smoking study, for example, participants in a placebo group took sugar pills described by the experimenter as “fast-acting tranquilizers” that would help them learn to endure the stress of giving up cigarettes (Bernstein, 1970). These people did far better at quitting than those who got no treatment; in fact, they did as well as participants in the experimental group, who received extensive treatment. These results suggested that the success of the experimental group may have been due largely to the participants’ expectations, not to the treatment methods. Placebo effects may not be as strong as experimenters once assumed (Hrobjartsson &
Gotzsche, 2001), but some people do improve after receiving medical or psycho-logical treatment, not because of the treatment itself but because they believe that it will help them (e.g., Wager et al., 2004).
Research on EMDR treatment suggests that the eye movements themselves may not be responsible for improvement, inasmuch as staring, finger tapping, feeling vibrations or listening to rapid clicks or tones while focusing on traumatic memo-ries has also produced benefits (e.g., Carrigan & Levis, 1999; Cusack & Spates,
쐌placebo A physical or psychological treatment that contains no active ingredient but produces an effect because the person receiving it believes it will.
Answer key to Table 2.3: The independent variable (IDV) in experiment 1 is the type of reading class; the dependent variable (DV) is reading skill. In experiment 2, the IDV is the amount of sleep; the DV is the score on a memory test. In experiment 3, the IDV is amount of exercise;
the DV is lung capacity. In experiment 4, the IDV is using or not using a cell phone; the DV is performance on a simulated driving task.
Ever Since I Started Wearing These Magnets . . . Placebo-controlled experiments are vital for establishing cause-effect relationships between treatment and outcome with human participants. For example, many people swear that magnets held against their joints relieve the pain of sports injuries and even arthritis. Some research supports this view (Harlow et al., 2004), but most experiments have found magnets to be no more effective than placebo treatment with an identical, but nonmagnetic, metal object (e.g., Collacott et al., 2000; Feingold &
Flamm, 2006; Winemiller et al., 2003).
Something other than magnets—wishful thinking, perhaps—appears to be causing the reported benefits.
They Are All the Same Scientists have succeeded in cloning mice, thus creating a population of genetically identical animals.
These animals can be assigned to various experimental and control groups with no worries about the confounding effects that individual differences might have on the dependent variable. Laws and research ethics rule out creating a pool of cloned people, so the process of random assignment will remain a vital component of psychological research with human beings.
placebo
42 ●
CHAPTER 2 Research in Psychology1999; Rosen, 1999; Seidler & Wagner, 2006; Servan-Schreiber et al., 2006). In fact, although EMDR appears to benefit some clients, it often fails to outperform impres-sive placebo treatments or other established anxiety treatment methods (e.g., Rothbaum, Astin, & Marsteller, 2005; Seidler & Wagner, 2006; Taylor, 2003).
These results have led many researchers to conclude that EMDR should not be a first-choice treatment for anxiety-related disorders (Davison & Parker, 2001; Lohr et al., 2003; Taylor, 2004; Taylor et al., 2003).
●
Experimenter Bias
Another potential confounding variable comes from , the unintentional effect that experimenters may exert on their results. Robert Rosenthal (1966) was one of the first to demonstrate one kind of experimenter bias, called experimenter expectancies. His research participants were laboratory assistants whose job was to place rats in a maze. Rosenthal told some of the assistants that their rats were “maze-bright”; he told the others that●