Research methods are the ways in which researchers measure variables and design studies to test hypotheses. For example, if a researcher wants to study whether people in a happy mood are more likely to offer help to a stranger than are people who are not happy, the researcher might measure or manipulate how research Page 748  |  Top of Articleparticipants feel and then measure how likely people are to offer help.

Researchers can choose among many different ways to measure variables. They can directly observe people's behaviors, directly ask people for their perceptions, or infer people's perceptions on the basis of behaviors or responses that only indirectly relate to the variables of interest. In most areas of social psychology, researchers want to learn what causes the phenomenon of interest (in the example, whether differences in mood causes differences in helping). Thus, whenever possible, researchers seek to manipulate variables of interest (e.g., mood) in an effort to make confident claims about causes (e.g., happy mood causing larger amounts of helping). Of course, for some variables or in some settings, the researcher cannot or chooses not to manipulate variables but instead looks at the relations between presumed cause and effect variables (such as mood and helping, respectively, in the example).

Measurement techniques will be discussed first and then research designs. Social psychologists commonly use a variety of measurement techniques including self-report, behavioral observation, response latency (time to answer), and physiological measures. Each type of measure has its strengths and limitations, but the extent to which one can draw conclusions from measured data is also a function of the type of research design employed. Social psychological research designs can be broadly classified into experimental and non-experimental research methods. Nonexperimental approaches are well-suited for identifying associations among variables; however, these approaches are less well-suited to determining cause-and-effect relations. However, experimental designs can demonstrate causal relations because of random assignment to conditions and greater control over variables that may covary (go along) with the cause variables under study.

Whenever possible, researchers try to collect supporting evidence using more than one type of measure. By doing this, the strengths of some measures can offset the weaknesses of other measures. Researchers often have greater confidence in the research conclusions when a particular theory can be supported by more than one type of measure.

Self-reports are perhaps the most widely used measurement technique in social psychology. Self-report measures ask people to directly report their feelings, behaviors, or thoughts. In some cases, self-report questions may require open-ended responses (e.g., "What is your current mood state?"). Other types of self-reports may require people to respond according to a provided scale (e.g., "Please rate your current mood state." 1 = negative mood to 7 = positive mood). Using a set of items that all tap into the variable of interest (e.g., asking people to rate mood on measures of how negative/positive, bad/good, and unpleasant/ pleasant their mood is) generally provides a better measure than using only a single item (e.g., only the negative/ positive question). The primary advantage of self-report measures is that variables of interest can be directly measured from the source of those experiences.

However, at times, people may not be able or willing to provide accurate reports. When this is true, data collected from self-reports may be inaccurate or misleading. In some cases, for example, the validity of self-reports may depend on respondents' verbal abilities. Self-report data from children or those who have cognitive deficits may be inaccurate because of an inability to understand the questions or express responses. Even when ability to accurately report is not in question, people may not be willing to express their true feelings, thoughts, and behaviors. For example, when asked about socially undesirable opinions or behaviors, people may be inclined to respond in ways that make them look good to themselves and others (i.e., social desirability bias). Thus, when there are concerns that research participants might lie on self-reports (e.g., when addressing socially sensitive issues such as stereotyping, prejudice, or aggression), other forms of measurement may provide especially useful information.

Behavioral measurement consists of observing and recording people's actions. Social psychologists typically measure a particular behavior(s) (e.g., smiling) because the behavior directly relates to a variable of interest (e.g., mood). Unlike self-reports, one advantage Page 749  |  Top of Articleof behavioral measures is that assessment can occur without participants realizing that the measurement is taking place. Thus, researchers might be able to assess reactions that research participants would not willingly share.

However, one limitation of behavioral measurement is that researchers must infer the reasons for the behavior. For example, imagine a study of opinions toward consumer products where participants are asked to choose one product (from several) that they can take from the study. If a participant selects one product rather than another, this could indicate that he or she has a more positive opinion of the chosen product, but this choice could have also been made for other reasons unrelated to his or her opinion (e.g., taking it to give to a friend). Another possible limitation of behavioral measures (and of some self-reports) is that behaviors are often situation-specific. That is, the behavior may occur in one situation, but not in another similar situation. In most social psychological studies, people's behaviors are assessed in only a single situation. Therefore, the behavior-based assessment of the variable might not reflect a general perception that would work across circumstances; it may reflect a more limited tendency to act a certain way in a certain circumstance.

Researchers may often be able to make inferences about psychological variables based on how quickly or slowly people make responses. More often than not, computers are used to present words or pictures on screen, and the computer records how quickly people respond to the word or picture (e.g., pronouncing the word, naming the pictured object, or evaluating the object). When speed of response is important, responses often take the form of hitting one of two computer keys as quickly but accurately as possible. One common use of response time is to index whether people have recently been thinking about a concept related to the word or picture on the computer screen. For example, imagine that a researcher believes people are likely to spontaneously think about the concept of race in a certain circumstance. If so, reaction time to label a pictured person as belonging to one racial category or another might be faster in that condition than would the same reactions to the same picture in a condition where previous consideration of race is unlikely.

Like behavioral measures, reaction times may be less susceptible than self-reports are to participant control over socially undesirable responding. Because response times often differ across conditions in very small amounts (fractions of seconds), participants may not even realize that they are responding more quickly to some stimuli than to others, and they may be even less likely to identify any such differences as reflecting the conceptual relations between certain conditions of the study and the critical responses to the specific words or pictures. Like behavioral measures, however, the researcher must infer the meaning of relatively fast or slow responses. Research participants can take time to engage in many different psychological processes. If a number of different processes would make people take longer to respond, then long response times alone may not help the researcher to distinguish among those potential thought processes. This may be especially true when time measures are taken for engaging in an activity such as reading information, rather than responding to a question. People can take a long rather than short time to read a passage for many different reasons. Thus, measures of time taken to read (or, in some cases, respond) may require additional measures or manipulations to help characterize why the additional time was taken.

Social psychologists (and perhaps especially social neuroscientists) may use a variety of measures that assess physiological responses to social stimuli. These measures include those that assess brain or muscle activity, activation of the autonomic nervous system, and others. For example, even if observers cannot see that a participant is smiling, electrical activity may be greater in the smiling muscles when the person is listening to information with which they agree rather than disagree.

Physiological responses are often involuntary or not under people's conscious control. Therefore, concerns about people's inability or unwillingness to respond in a certain way are minimal. Thus, like behavioral and response latency measures, physiological measures can be useful when measuring reactions to socially sensitive material. Disadvantages include the time and expense involved in taking physiological measures. Even relatively minimal physiological recording equipment is expensive, and more advanced physiological measures (especially scanning techniques) Page 750  |  Top of Articleinvolve very expensive equipment. Most physiological measures are also especially sensitive to participant movements during the study and to the environment in which the measurements are taken. Thus, relatively long periods are taken to acquaint participants with the recording equipment and to get baseline measures (to control for individual differences in baseline activation of the systems under study). In addition, limitations in movement for many physiological measures restrict the kinds of interactions in which research participants can engage while physiological recording occurs. Finally, much remains to be learned about how various patterns of physiological reaction relate to particular psychological processes and variables. Many physiological systems become active during more than one type of psychological process. Therefore, in many circumstances, there may not be a one-to-one mapping of activation of a particular brain area or a particular system with one particular psychological process or outcome. This can make inferences based on physiological measures quite complex when compared with other types of measurement.

Experiments are perhaps the most prominent research approach used in social psychology. Experiments offer many advantages over nonexperimental approaches. In particular, because experiments control extraneous variables through random assignment to conditions, they allow researchers to confidently determine cause-and-effect relations. Random assignment is the procedure of assigning research participants to different experimental groups such that each participant has an equal chance of being assigned to any experimental condition. This is important because researchers can be assured that the background characteristics of the participants in each group are equivalent before a manipulation is applied. For example, let's return to our example of a study of mood and helping. Some people may simply be more likely to offer help to a stranger. Yet, if each person has an equal chance of being assigned to a happy mood group or a neutral mood group, then personal tendencies to offer help should be equal across the groups before any mood manipulation takes place. Later, if there are differences in helping across mood groups, this difference cannot be attributed to differences across groups in the background tendencies of the people in each group; the differences must have been created by the manipulation.

Even though experimental designs offer many advantages, they do have limitations. For instance, experimentation cannot be performed when variables such as gender, personality traits, or ethnicity are under study because these variables cannot be manipulated. Also, many possible manipulations of variables such as ethnic prejudice, marital status, and physical aggression would not be undertaken because of ethical concerns. Therefore, research addressing influences of variables such as these must be conducted nonexperimentally.

Another potential issue with experimental designs concerns to the extent to which findings can generalize to real life. Increases in experimental control can result in increased artificiality of the experimental setting. This is less of an issue when the goal of the study is to test psychological theory rather than to produce results that are relevant to a particular applied setting. For instance, a researcher may believe that ethnic categories are activated when people encounter group members as they walk down a hallway. Yet, it may be much more straightforward to show such activation in a laboratory showing pictures or video on a computer screen. However, all else being equal, researchers would often prefer that their research findings (or at least the psychological processes that produced the findings) would translate to real world settings. Researchers can increase the likelihood of their results translating to real world contexts by using experimental activities that closely reflect similar activities in everyday life, by showing that the research findings are the same across different kinds of manipulations and measures, or by conducting field research that shows parallel effects without the same level of experimental control over extraneous variables.

Although experimentation is the primary way to determine causal relations among variables, a nonexperimental design may be more appropriate for some research questions. Some research questions do not involve cause and effect. For example, when a researcher is developing a multi-item measure of a Page 751  |  Top of Articleparticular concept or idea, the researcher may only be interested in identifying the presence of relations among those items, not causal relations. Even when researchers are interested in cause and effect, some variables cannot be experimentally manipulated (e.g., gender, personality traits) or manipulation would not be ethical (e.g., marital status, physical aggression). When this is the case, nonexperimental research is the best that researchers can do. In addressing cause and effect relations, however, nonexperimental approaches face a number of challenges.

Consider nonexperimental investigation of the question of whether people are more likely to offer help while in happy moods. For example, participants could be asked to complete diaries in which they report their moods and their major activities each day for a month or more. It could be, in such a study, that people who report being generally happy also more often report helping others. One of the problems with non-experimental designs is that causes might often occur in either direction (i.e., happy mood could increase helping, or increased helping could create happy mood). Even when this is not as likely (e.g., if mood were measured before a specific opportunity to help, so the helping opportunity itself cannot be the source of the mood), a measurement of the independent variable (mood) might identify people who are also disposed to help for reasons other than their mood per se. For example, the people who report being happy at a particular point in time may be happy because of positive events in their lives (e.g., getting a raise at work), and those positive events themselves may make helping more likely separate from mood (e.g., by making people feel like they have an excess of resources, so they can afford to share). At times, the researcher can measure potential alternative reasons for the effects or can include measurements over time that make a stronger case for the preferred explanation. However, these solutions are often less compelling than running an experiment in which random assignment to conditions equate the conditions on variables not influenced by the manipulation of interest.