How do psychologists study prejudice and discrimination? - Chapter 2

A chapter about research is important, because research informs us of what stereotyping and prejudice are. It gives us information with which we can make theories and try to find explanations for certain things. Research can also help testing theories and predicting behaviour and to see whether/ how well certain interventions work.

How do you formulate hypotheses?

In the behavioural sciences the goal of research is to find out why people behave differently from each other and which factors limit or push certain behaviours.

Researchers use observations of everyday life to compose their research questions out of theories. Theories have certain links between variables. A variable is a characteristic that varies across people. Sex is a variable: some people are male and other people are female. Variables don’t just differ from person to person, but they can also differ across time and situation. One person might be very prejudiced against Black people, but then years later this person might be less prejudiced against them because he had more encounters/ experiences. Also, people might be more prejudiced when a lot of things are going on in their head or when they’re distracted than when they have the time to think carefully about a certain person. Proposed links among variables are called postulates. These postulates can be based on research, observations, experiences, speculations or a combination of these things. Theories can change because postulates may or may not be correct. Hypotheses are theoretical postulates that can/will be tested in research. Hypotheses are usually based on abstract concepts like prejudice. If you want to test certain abstract concepts, you must use operational definitions. These are observable, concrete representations of hypothetical constructs. Questionnaires that contain questions about self-esteem are directly observable indexes of someone’s self-esteem. The scores on the questions give a good view of someone’s self-esteem. There are different ways to measure self-esteem or other concepts and the researcher needs to find which one suites him/her best.

Sometimes researchers don’t measure variables, but manipulate them. One research showed that challenging someone’s worldview makes him/her more prejudiced. Worldview is linked to self-esteem and if you challenge someone’s worldview he or she will get anxious. This person will be more prejudiced, because having negative views about others gives a person more self-esteem. The way researchers made people more anxious was by having people think about their own death. They manipulated anxiety that way.

When a researcher has chosen operational definitions, hypotheses become predictions.

How do you measure stereotypes, prejudice and discrimination?

Measuring certain things can be difficult, because some things can only be measured indirectly. Researchers therefore need to know for sure that their methods are accurate. They can do this by looking at reliability and validity.

Reliability means that every time a measure is used with the same person, the same results are provided. This is a certain stability across time. Although characteristics can change over time, they do so slowly. So a person’s racial attitude now and in a month will be pretty much the same.

There are two types of reliability that are often used. The first one is test-retest reliability. This means that researchers let a group of people do the same test two times. The time between the tests could be weeks, a month or a couple of months. They then look at the correlation between these tests. The higher the correlation, the more reliable the measure will be. The second type is internal consistency. This means that every item that measures a certain characteristic will be answered in the same way. If people have negative attitudes about something, they should answer every item about this thing negatively. With the help of statistics researchers measure the consistency of the response. The two types of reliability are usually correlated. So a measure that is internally consistent is also likely to be consistent over time.

A measure might be consistent, but it does not mean that it measures characteristic that it’s supposed to measure. Validity means that a measure assesses the characteristics that it’s supposed to assess assesses different aspects of this characteristic and only measures this characteristic. If researchers want to measure racial attitude, they should find certain questions that only measure racial attitudes and not a person’s general positive or negative attitude towards people in general.

There are also two types of validity that will be discussed here:

1. Convergent validity means that the scores on a measure correlate with the scores of a measure of the same or related characteristic. These scores also relate with behaviours that are related to the characteristic.
2. Discriminant validity discriminates between characteristics. It tells researchers to what extent a measure does not access characteristics that it is not supposed to assess. For example, this type of validity is used to see if people are not really prejudiced or if they just want to give social desirable answers.

What are the different measures used?

There are different types of measures used to measure a characteristic.

What are self-report measures?

The method that is used the most to assess stereotypes is self-report. People are asked stuff about their opinions and attitudes. There are different questionnaires researchers use to assess prejudice.

The Katz and Braly checklist is used widely and it has questions about certain characteristics and the respondent needs to write if he/she thinks that a certain group of people has these characteristics. When a researcher uses a checklist consistently, he/she can see how stereotypes change over time. A downside of using a checklist is that it needs to be up-to-date. Stereotypes change over time, so checklists need to change over time too. Also, researchers need to remember that there are social stereotypes, also knows as culturally shared beliefs, and personal beliefs, which is what individuals personally believe. One person might know that the social belief about a certain group is negative, but his or her own belief might be positive. Stereotypes can also be assessed by asking people how likely or unlikely they think that certain groups/group members have certain characteristics. A third way to assess stereotypes is by free response measures. Respondents are asked to write down a couple of characteristics about a group and say if those characteristics are positive or negative.

How to assess prejudice and behavior?

Prejudice can be assessed by questionnaires. Emotional response can be assessed by just asking how people feel when they interact with those groups. Behaviour towards certain groups can also be assessed by questionnaires. In those questionnaires there are questions about how many times the respondent has performed a certain behaviour against members of a certain group or what they would do against members of a certain group.

What are advantages and limitations?

Self-report measures are efficient, a lot of people can fill them in at the same time and these questions are easy to administer. They also can cover a lot of topics. The negative side about self-report measures is that people can lie about their actual opinions and behaviour. The best thing to do is to secure anonymity. People are more willing to tell the truth when they can stay anonymous.

What are unobtrusive measures?

These measures give the impression that they don’t have to do something with prejudice. The measures used are behaviour and judgement. In these types of measures people really look what someone does and not says about his behaviour. They usually do this with helping or sitting distance. People who are prejudiced against Black people are less willing to help them and will sit further from them.

What are physiological measures?

These types of measures assess changes in the body’s responses, like heart rate, to a stimulus. This can be used to see if somebody has a positive or negative attitude towards a certain person and to see how intense this is. A big advantage of these types of measures is that they are not really controllable so people can’t fake them. It is also usually a valid method. The downside is that it’s expensive.

What are implicit cognition measures?

One could implicitly assess a certain characteristic. When you prime somebody, he or she will activate concepts associated with the category you primed. When something is primed it is more accessible in the mind and it is easier to recognize for the person. These measures are good, because people don’t know that they have been primed and thus can’t hide their actual feelings. The downside is that such studies have to be done in an environment where there’s not much distraction. So they are usually limited to lab settings.

There are various examples:

1. The Affective Priming Paradigm uses exposure to a member of a category to activate associated concepts. If a person, for instance, associates old with forgetful, forgetful becomes activates by seeing a picture of a grandparent. 
2. The Implicit Association Test assesses the extent to which unassociated concepts make responding more difficult. Response competition between habitual and opposing response is used. The stronger the habitual reponse, the longer it takes to get the opposing one, because the individual has to suppress the habitual response. 
3. The Affect Misattribution Procedure uses priming. It examines the extent to which the affect associated with a certain prime is transferred to a neutral stimulus. 

What is the difference between self-report measures and physiological and implicit cognition measures?

There are usually low correlations between scores on self-report measures and physiological, behavioural and implicit measures. How can this be, given the fact that measures of same constructs should be related (if they’re valid)? The point is, when people can, they will usually control a behaviour that makes them look bad. This will usually happen on self-report measures. Implicit measures usually find the correct attitude, because respondents can’t hide behaviours they are not aware of.

Why do researchers use multiple measures?

A good way to study prejudice is by using multiple measures. If the results from these different measures point in the same direction, we can be quite sure that our results are valid. Also, by using different measures the strength of one measure can compensate for the limitations of another. Another reason why multiple measures should be used for assessing prejudice is that prejudice has different aspects and by using different measures, one could assess all these aspects. It is also important to use different measures because it is not only important to look at the uncontrollable expressions, but also to the controllable expressions. One might want to know under what circumstances people try to control their expressions or prejudice.

Which research strategies can be used?

The next part will give an overview of research strategies that psychologists use.

What are correlational studies?

Researchers measure two or more variables and look for certain relationships among them.

What is a survey research?

One way to do this is by survey research. In this type of research, people are asked questions about their attitudes, beliefs, opinions, personality and behaviour. Researchers need to find people to use for their study. This is called sampling. There are two ways of sampling. The first one is probability sampling. First, researchers need to think of what their research population is. To whom do they want to apply their results? From this population a sample is drawn. In this sample, all of the characteristics of the population (age, ethnicity) are represented. The sample accurately reflects the population and is reliable. Probability sampling also has a downside. It is very expansive to contact all those people. Also, probability sampling usually uses telephone interviews and therefore not every question could be asked.

A lot of researchers also use convenience sampling. The sample exists of people from whom the researchers can easily collect data. Data can be gathered quickly and easily and a lot of questions can be asked. On the other side, with convenience sampling it is hard to know how well the sample represents any given population. Researchers must be cautious about drawing conclusions.

What is the correlation coefficient?

The relationship between two variables is often described by a statistic that is known as the correlation coefficient, r. This consists of a number and a sign. The number can be a value between 0 and 1 and the sign can be a – or a +. The sign indicates the direction of the relationship. A plus indicates a positive relationship and a minus a negative one. Positive means that when the value of one variable increases, the other also increases. A negative relationship means that when the value of one variable increases, the value of the other decreases.

The number indicates the strength of the relationship. Zero indicates no relationship at all and one indicates a perfect relationship. Correlations between 0.1 and 0.3 are small, between 0.3 and 0.5 are moderate and greater than 0.5 are large.

What is the difference between correlation and causality?

Finding a relationship between two variables doesn’t mean that one variable is causing the other. For causality, three criteria have to be met. The first one is covariation. This means that the causal variable is related to the effect variable. If the cause is present, the effect should also be present. The second criterion is the time precedence. The cause needs to come before the effect. It is sometimes hard to figure this out in correlational research, because all variables are measured together. The third criterion is the absence of alternative variables for the effect. When the number of ice creams consumed increases, the number of burglaries also increases. Does this means that ice creams are responsible for the increase in burglaries? No, there is a third variable. Ice creams are usually eaten when it’s hot outside. When it’s hot, people go outside or on vacation and that’s when thieves usually strike. To figure out which variable is responsible for the effect, researchers must eliminate (control) other variables. However, it is hard to control for every single variable.

Correlational research can’t establish time precedence of a cause or eliminate alternative variables of explanation and thus correlational research is no causal research.

What are experiments?

To establish causality, researchers conduct experiments. In experiments, three criteria for causality are met. The proposed cause is called the independent variable and the proposed effect is called the dependent variable. Researchers need to manipulate the independent variable, they do this by creating two or more conditions. All these conditions represent a different aspect of the independent variable.

Experiments can be conducted in different settings:

1. Laboratory experiments: In this type of experiment, researchers have a high degree of control. Researchers can meet almost every criterion for causality. The down-side to a laboratory experiment is that it loses the natural setting. Because of the artificial setting, researchers have to wonder whether the found effects will also have been found in naturalistic settings.
2. Field experiments: If you want more naturalistic settings, you have to conduct a field experiment. Researchers manipulate an independent variable in a natural setting. This way they try to keep as much control as possible over the situation and try to keep it as natural as they can. Field experiments, however, are difficult to conduct. One can’t have a very high degree of naturalism and control together. You have to give some from the one up to the other.
3. Experiments within surveys: both laboratory and field experiments use convenience samples. It is much too expensive to do a probability sample in the laboratory and not a lot of people would want to participate. Researchers, however, can do a probability sample by conducting experiments as part of surveys. There have to be different parts of the survey, to represent the different conditions of the independent variable. This type of experiment gives the researchers more security about generalizing the results to the population as a whole. A down-side to this type of research is that only a couple of independent variables can be used. These are only the variables that can be manipulated by changing the question. Also, naturalism is low in this type of experiment.
4. Individual difference variables within experiments: A study does not only have manipulated experimental variables, it can also have non-manipulated individual difference variables, like personality traits.

What are ethnographic studies?

Ethnographic research tries to understand behaviour by observing behaviour, conducting interviews and studying behaviour in a context. This type of research concentrates on naturalism and not really on control. This type of research finds a participant’s point of view much more important than manipulating experiments.

What is content analysis?

Researchers from this type of research area don’t study people, but objects people create. Examples of these are documents, artworks and photographs. Researchers can study websites to figure the ideals of certain groups out.

What is a meta-analysis?

As you have probably noticed, every type of research has it pros and cons. Researchers like to have data connected from different methods. If the data from these methods points to the same conclusion, a researcher can be quite sure that his/her conclusion is the right one. A meta-analysis is a research method that statistically combines the results of multiple studies to determine the average between the variables used in the studies. 

How do you draw conclusions?

After the data has been collected, researchers have to draw conclusions. They want to see if there’s support for their hypotheses. This can be answered very easily if the data are quantitative. If a relationship is found, there are two explanations for this. The first one is that the relationship exists and the second one is that an error occurred in research. With the help of statistics, one can figure out if an error occurred or whether there is a real relationship. Quantitative data are analysed by looking for patterns of responses of behaviour.

What does the data mean?

When the data has been analysed, researchers need to figure out what it means. If researchers have found that men score higher on measures of prejudice than women, what does this mean? Because there can be different possibilities. Some researchers might say that testosterone has an effect on prejudice. Others say that social norms teach males to be more prejudiced than females. So it all depends on the different theoretical orientation and the psychologist’s background. So how do we know which explanation is the correct one? This is quite difficult, because some explanations can be tested easily (like the testosterone explanation) and others can’t. Also, there could be more than one explanation.

Why do you have to verify the results?

To ensure the accuracy of the results, one could verify the results. One way to do this, is to redo the study using the same measures and to see if the same results will occur. This is exact replication. The other way to redo the study is with different measures or participants with different characteristics. This is called conceptual replication. Generalizability is an important issue in psychology and if a relationship between self-esteem and prejudice is found, it should be found every time and with every questionnaire/ way you measure self-esteem and prejudice.

When can you apply the theory?

When everything has been examined and researchers are confident about their findings, they have to see if it fits their initial hypothesis. If it doesn’t fit, the hypothesis should be revised. When everything is finished, researchers can start using their hypotheses and see how well the theory works in certain settings. If a certain application did not work, researchers should put certain questions to the theory.