Experimental and Nonexperimental Designs in Social Psychology

Abraham S. Ross

• Publisher: Routledge
• Year: 2019
• ISBN:
  9780429710971
• Edition: 1
• Topic: Psicologia
• Subtopic: Storia e teoria della psicologia

This book considers experimental designs, alternatives to experimental designs, survey methods, and how systematic collection of information can minimize alternative explanations in social psychology. It discusses meta-analysis for interpreting the results of many social psychology experiments.

Section I

Causality

Chapter 1
Inferences About Causality

The Search for Alternative Explanations

Evidence That X Causes Y

Evidence That X and Y Occur Together

Evidence That X Precedes Y

Evidence That Alternative Explanations Are Unlikely

Internal Validity

External Validity

Ethical Problems

Stress

Invasion of Privacy

Deception

Ethical Safeguards

Infonned Consent

Anonymity and Confidentiality

Debriefing of Subjects

Review by an Independent Ethics Conimittee

The Search for Alternative Explanations

All humans are scientists. We observe nature and the behavior of others. We seek consistencies and the causal rules that help explain these consistencies. Once we understand the rules, we can make predictions. These predictions, if they are accurate, make it easier for us to deal with nature and with the people around us.

Let us consider an example. We see that Michael helps Richard with his assignments, that he lends money to Morris, and that he takes a lost child to the police station. Looking for consistency in his behavior, most observers would agree that his actions are “helpful.” However, they might not agree about why Michael has taken these actions. Some might believe that the cause lies outside Michael; for example, Richard may have put pressure on Michael to help him. Others might argue that the cause of Michael’s actions is “inside” him; Michael may be a “helpful person.”

Some social psychologists study why people choose these alternative explanations for others’ behaviors. This area of research is known as attribution theory.

Many times people attribute causes for events or for other people’s behavior without realizing there are alternative explanations. Someone who believes Michael is a helpful person might not consider the alternative that Richard put pressure on him to help. We may be more likely to overlook alternative explanations when our first explanation fits with our beliefs about the other person.

There are almost always alternative ways of explaining even the most obvious events and behavior. Considering alternative explanations will often prevent hasty and erroneous conclusions. We share the view of Lord, Lepper, and Preston (1984) that many of the judgmental biases to which people are prone can be corrected or lessened by a conscious effort to “consider the opposite.”

Like other people, scientists and social psychologists observe, seek consistencies, and infer causal rules to explain them. As you will see, researchers too sometimes overlook alternate explanations for their observations. Much of what we discuss in this book will focus on the search for alternative explanations.

If psychologists and armchair scientists both search for consistencies and causal explanations of behavior, what are the essential differences between them? Two of the most important have to do with how systematically they collect the information and how effectively they can eliminate alternative explanations.

There are several methods of gathering information about thoughts, emotions, behavior, and the variables that may influence them. Some of these methods are more systematic than others. Suppose you are interested in finding out how people will vote in the next election. A less systematic approach might involve asking your friends about their choice of candidates. A more systematic approach might involve administering a questionnaire with highly specific questions to a random sample of a population.

Researchers try to collect information that will tell them whether particular explanations are true or false. Ideally, the information will support one explanation and disconfirm all others. In practice, however, there are usually alternative explanations for any results. Less systematic research methods leave many alternative explanations; more systematic methods leave few. We can classify research designs by their efficiency in eliminating alternative explanations—that is, in terms of how rigorous they are.

As you search for alternative explanations and try to evaluate the evidence for them you should realize that these processes are influenced by your own view of the world. This is true of scientists as well. Mahoney (1977) was interested in the possibility that scientists may allow their own theoretical orientations to bias their evaluations of research reports submitted by other researchers for publication. Typically an article that is submitted for publication is sent by the editor of the journal to three or four well-known researchers in the area for evaluation. The opinions of these reviewers largely determine whether or not the article is accepted for publication.

Mahoney (1977) decided to use as subjects in his experiment the 75 referees listed by the Journal of Applied Behavior Analysis for the year 1974. This journal publishes articles “advocating the refinement and expansion of applied behavioristic psychology” (Mahoney, 1977, p. 164) and the theoretical perspective of the reviewers was assumed by Mahoney to be clearly in this direction.

To each of the reviewers Mahoney sent a manuscript which he told them was being considered for inclusion in a volume titled “Current Issues in Behavior Modification.” In addition to responding to some open-ended questions, each reviewer was asked to evaluate the manuscript in terms of the relevance of its topic, its methodology, the presentation and discussion of the data, and its scientific contribution. Finally, each reviewer was asked to make a recommendation concerning the manuscript: accept, accept with minor revisions, accept with major revisions, or reject.

The manuscript that Mahoney fabricated for the purposes of his experiment dealt with the effects of reinforcement on intrinsic motivation. The traditional behavioristic position on this question is that when people are reinforced for working on a task, their motivation increases. Other psychologists have challenged this idea and have argued that extrinsic reinforcement can sometimes result in diminished intrinsic motivation. The fictitious manuscript which Mahoney devised described an experiment with children which attempted to test the validity of these two competing points of view.

Mahoney created five versions of the manuscript and each reviewer received one of these versions to evaluate. In one version of the manuscript, the results described were positive in the sense that they supported the behavioristic prediction of increased motivation following reinforcement. (Remember that this is the theoretical orientation presumably held by the reviewers.)

A second version of the manuscript described negative results which refuted the behavioristic prediction. This was done in part simply by reversing the identification labels on the lines of a graph in the first version. A third and fourth version of the manuscript contained mixed results which could be interpreted as either supporting or refuting the behavioristic prediction. In the third version, the mixed results were followed by a “positive” discussion in which the author concluded that the results supported the behavioristic point of view. In the fourth version, the conclusion in the discussion was that the results refuted the behavioristic point of view.

Apart from these variations all versions of the manuscript were identical in terms of the procedural details, bibliography, etc. Reviewers rated the study with negative results as having the least relevance, the weakest method, the weakest data, and as making a relatively minor contribution. The studies with negative results or mixed results were most likely to be rejected.

The reviewers were more likely to reject the study with results that didn’t agree with their theories. Even though the methods were the same in all versions, the reviewers rated the method in the rejected study as weaker. Mahoney’s results indicate a considerable degree of bias on the part of the scientist-reviewers in the direction of their own theoretical perspective.

Students and other nonscientists often hold the view that science is a purely objective, value-free enterprise. According to this view, the design of scientific experiments is dictated by the research hypothesis and the conclusions that are drawn are dictated by the results. Moreover, scientists may be thought to be less susceptible than are other people to preconceptions and biases. The results of the Mahoney experiment that we have just described clearly challenge this view. Scientists, like other people, are vulnerable to bias and hasty judgments and, again like other people, they can benefit from being reminded of it. An awareness of the possibility of alternative interpretations can be an enormous benefit to researchers both in the design and interpretation phases of their work.

Evidence That X Causes Y

In everyday conversations we often use the concept of causality. We ask, “What caused you to do that?” Or we may ask a friend, “Why did Sarah tell Charles that I liked him?” In this case “why” means, what caused Sarah to tell Charles. We search for causes or explanations of events and behavior occurring in the world around us, and in searching for an explanation, we may think of several possible causes. Perhaps Sarah behaved in a certain way because she is angry with me. Maybe she did it because she thinks it would help our relationship. When scientists search for such possible causes they think of them as causal hypotheses.

A causal hypothesis often takes the form, “If X then Y.” “If I comment on John’s weight, then he will become angry.” “If I help Jane with her term paper, then she will help me with mine.” People find it comparatively easy to generate these kinds of causal hypotheses. Such hypotheses may stem from a need to explain what has already happened and predict what is about to happen.

If you read mysteries you know that detectives often have hypotheses, or hunches about “who done it.” Their problems occur when they try to find evidence to support their hypotheses. For the scientist, like the detective, it may be difficult to collect evidence that a particular hypothesis is correct. The fictional detective has one advantage over a scientist. When she uncovers all of the evidence she can confront the villain and know that she was right.

As scientists we can never prove, logically, that any hypothesis is correct. No matter how many instances of an event we see, we cannot be sure that it will recur. It is logically impossible to prove that the sun will rise tomorrow. All scientists can do is gather evidence to support or disconfirm an hypothesis. Let us consider the kinds of evidence that are necessary.

There are, in general terms, three kinds of evidence that are useful for supporting or disconfirming an hypothesis:

1. evidence that X and Y occur together
2. evidence that X precedes Y and,
3. evidence that alternative explanations for our results are unlikely.

Evidence That X and Y Occur Together

The first evidence we need is that X and Y occur close together in time. In general, the closer X and Y are in time, the greater the likelihood there is a causal connection between them. When X and Y can occur together in varying degrees, we look for evidence that changes in X occur together with changes in Y. For instance, as we step down on the gas pedal in a car, the car goes faster. In some cases, the association may be of a negative kind; increases in X are associated with decreases in Y. The more novocaine the dentist injects, the less pain the patient feels. We call the variation of X and Y together concomitant variation.

A variety of statistical tests are available for detecting concomitant variation between variables and assessing the strength of the association. In fact, of the three kinds of evidence, this is the only one we can assess statistically. We can establish that X precedes Y and eliminate alternative explanations only with systematic data collection and good experimental designs.

Evidence That X Precedes Y

Evidence of concomitant variation alone does not warrant the conclusion that X is causing Y. We need to be sure that X precedes Y. Although this is an obvious point, it is overlooked surprisingly often. Consider the following example. Gerbner et al. (1976) reported that people who watch a lot of television, compared to those who watch less, are more likely to fear that they will be involved in some kind of violence in the near future. Although Gerbner et al. (1976) considered many alternative explanations for their finding, they eventually concluded that watching a lot of television causes people to fear violent crime.

Gerbner and colleagues (1976) did not consider the possibility that heavy television viewing was a consequence rather than a cause of people’s fear of crime. Citizens who fear crime may be less likely to venture out of their homes, particularly at night. Since they do not go out, they may spend more time watching television.

Establishing the correct time order will sometimes involve considering additional variables and gathering new data. Doob and Macdonald (1979), for example, discovered that people who lived in areas of the city with a high crime rate, compared to those in areas with a low crime rate, were more likely to fear crime and more likely to be frequent viewers of television. In neither the high nor the low crime areas, however, did the researchers find any relationship between fear of crime and television viewing. Researchers need to be alert to the possibility that different causal sequences may underlie the data they are attempting to explain.

Evidence That Alternative Explanations Are Unlikely

Once we have determined there is concomitant variation and that X indeed precedes Y, we must try to reduce the number of alternative explanations for the X-Y relationship. Usually, alternative explanations will take the following fo...

Table of contents

• Cover
• Half Title
• Title
• Copyright
• CONTENTS
• Preface
• SECTION 1: CAUSALITY
• SECTION II: EXPERIMENTAL RESEARCH
• SECTION III: ALTERNATIVES TO THE EXPERIMENTAL DESIGN
• SECTION IV: SURVEYS—MEASURING ATTITUDES OR OPINIONS OF A POPULATION
• SECTION V: ANYTHING NEW?
• Appendix A—Codes of Ethics
• Appendix B—How to Read Research Articles
• Glossary
• References
• Index