CHAPTER 2 Research Methods
Total Page:16
File Type:pdf, Size:1020Kb
CHAPTER 2 Research Methods social psychologist, like any other sci- TYPES OF STUDIES A entist, utilizes a specific set of proce- dures to examine the phenomena of interest Descriptive to him. This is otherwise known as the sci- entific method. The scientific method is One of the first questionsdistribute researchers a way of understanding the natural world. need to ask is what type of data they are seek- It is a set of assumptions and procedures ing. Do we merely want to ask questions of that guide the way we gather data, and they the subjects? Do weor merely want to describe specify the conditions under which we can people, or a person, or an interaction? This trust the data that are obtained. For exam- basic type of research is termed descrip- ple, one principle of the scientific method tive research and encompasses very simple is standardization. It stipulates that, in an methods such as surveys, or case studies, or experiment, all aspects of the experiment naturalistic observation (whereby a subject is are the same for all experimental condi- observedpost, but is unaware of being observed; tions except the levels of the independent for example, people-watching in a mall). It is variables. Doing this helps the researcher important to note that these studies do not make the assertion that, if she obtains dif- manipulate variables, and they don’t allow us ferences on the dependent variable between to make any cause–effect statements about experimental conditions, then those must behavior. But these types of studies are good have been caused by the differencescopy, in the places to get ideas for experiments. independent variable between the condi- tions, because each group was exposed to Correlations the same procedures and stimuli in all other respects. In thisnot chapter, we discuss fur- Another source for experiment ideas is ther how social psychologists conduct their correlational research. Remember that a research. Of course, this is by no means correlation is the naturally occurring rela- meantDo to be an exhaustive review, and those tionship between any two variables. For interested in a more complete discussion example, there is a strong positive correla- are advised to consult other detailed sources tion between physical violence and ice cream (e.g., Aronson, Brewer, & Carlsmith, 1985; sales. Huh? Does that mean that eating a lot Aronson, Ellsworth, Carlsmith, & Gonzales, of ice cream makes you more likely to hit 1990; Reis & Judd, 2000). someone? Or is it the other way around? Copyright ©2019 by SAGE Publications, Inc. 37 This work may not be reproduced or distributed in any form or by any means without express written permission of the publisher. 38 Classic and Contemporary Studies in Social Psychology: A Text-Reader Hitting someone makes you crave a nice on the dependent variable(s) (DV; the vari- bowl of rocky road ice cream? Both of those ables the experimenter measures). seem unlikely. Perhaps a third explanation In the smallest, simplest type of experi- would make more sense. What do the two ment design, a 2 × 2, there are two inde- have in common? They both increase in the pendent variables, with two levels of each summer. So, perhaps it is the case that this variable. This forms four experimental third variable, summer, causes an increase conditions. Subjects are randomly selected in both, and in so doing, causes an artificial (everyone in the population of interest has correlation between ice cream sales and rates an equal chance of being selected for the of physical violence. This illustrates one of experiment) and then randomly assigned the problems with correlational research: (each subject has an equal chance of being the so-called third variable problem whereby assigned to any of the experiment condi- two variables are correlated only because tions) to one of the four experimental con- they are linked to a third variable that is the ditions. Remember, an IV is something that reason for elevations in those two variables the experimenter manipulates. We cannot (Brewer, 2000). Correlations tell us that manipulate aspects of the subject, such as two variables are related but not how they their gender, religion, socioeconomic status. are related. You may have heard the axiom Those are termed subject variables. Suppose “Correlation does not equal causation.” That our variables distributeare mood and comfort. For is unfortunate, because more often than not, mood, we will assign people to either expe- social psychologists are seeking to discover rienceor happiness or anger. For comfort, we the causes of behavior in a given situation. will assign people to either a high comfort or If the goal of science in general is to explain low comfort condition. We write this design and predict observed phenomena, then it like this: it is a 2 (mood: happy vs. anger) × would be best to be able to know what tends 2 (comfort: high vs. low) between-subjects to reliably cause a particular behavior (or design. The between-subjects designation thought, or feeling) to occur. So, post,just like says that each subject is exposed to one level descriptive research, correlational studies of each of the variables. In a within-subjects are good points for researchers to get ideas design, all the subjects are exposed to all for experiments that will help reveal causation. levels of all variables. So, if we hold all vari- ables constant, and only vary those we want Experiments to vary (the IVs), then to the extent that we copy, find differences between the experimental The great majority of research in social groups on their answers on the dependent psychology comes in the form of experiments. variable (in this example, let’s say the DV is There is a good reason for that. Experiments expectations for future success on a task), arenot the only way that we can establish cause we can say with some confidence that the and effect relationships between variables. different levels of the IV are the cause of the Experiments allow researchers to examine differences on the DV. the influence of one variable (or several) on Do the individual, and by holding other vari- ables constant, we can establish the causal THEORY AND HYPOTHESES influence (if any) of that variable on behav- ior. We are seeking to discover if there is a Good research must be based on a theory. causal effect of the independent variable(s) The goal of the research is to develop data (IV; the variables the experimenter manipulates) that speak to the theory either in terms of Copyright ©2019 by SAGE Publications, Inc. This work may not be reproduced or distributed in any form or by any means without express written permission of the publisher. Chapter 2 Research Methods 39 support or no support. A theory is an inte- DOING EXPERIMENTAL RESEARCH: grated set of principles that explains and FURTHER CONSIDERATIONS predicts some phenomenon. In psychology, we can’t say that our data “prove” a theory Suppose you are ready to start recruiting (as one can in some other sciences), because your subjects for your experiment. First, how we are dealing with humans and humans are many subjects do you need? Researchers messy. There will always be outliers in any can get a fairly exact number (N) for their experiment, testing any theory. We never study from a computer statistical program. get to 100% perfect results all the time, A rough rule of thumb I used to use was to but we can get very close. We set very high strive for about 20 per experiment condi- standards for accepting that there is a true tion for between-subjects designs. Too few cause and effect relationship between two subjects, and one will not have the statistical variables. This is the statistical likelihood power to find an effect if it is there (Type II that the results are due to chance. We say error—incorrectly believing there is no effect that we will only consider a finding a reli- when there is one). On the other hand, if able and causal factor in changing the DV you have too many subjects, tiny effects may when there is less than a 5% chance that that look significant erroneously (Type I error— change is due to random error. The hypoth- incorrectly believing there is a significant eses we derive from the theory are testable distribute finding when there is none in reality). predictions about what we believe will hap- Second, how and who do we recruit? pen in a given situation, with certain factors Ideally we would likeor to have our results gen- present. For example, given the theory of eralize to the entire world, but that isn’t pos- cognitive dissonance (Festinger, 1957), we sible because of cross-cultural differences. might make a prediction about what a per- So we (researchers in the United States) son will do when they say one thing that is will restrict our population to people in the incompatible with a behavior that they just United States.1 How do we find our sample performed. If we find evidence that the pre- (thepost, group of people from the population diction was not supported, and assuming we who do the experiment)? We randomly didn’t have any major flaws in our design, select them from the population. This means method, or statistical power, we would say that everyone in the population has an equal that the theory is thus not supported. If the chance to be selected for the experiment. theory continues to bring no supportive This, as you might guess, is an ideal that no data, it will be abandoned in favor of a differ- copy, one ever reaches.