Bonus Article for Researchdesign explained
You may want to assign thefollowing classic article:
Festinger,L., & Carlsmith, J. M. (1959). Cognitive consequences of forced compliance.Journal of Abnormal and Social Psychology,58, 203-210.
The article is available from
The authors use a three-group, between-subjects experimentto help launch one of the most powerful theories in psychology: cognitivedissonance theory. The article is short and relatively easy for students toread. To make the article even easier to read, give students Table 1 and one ofthe following links to sites that provide summaries of the article:
Guide to Understanding the Article
Tips, Comments, and Problem Areas
Forced compliance: the participant is asked to do something that the participant may not want to do. Although the participant could turn down the request, the researcher exerts enough social pressure so that virtually all participants go along with the request.
This article does not have an Abstract. Instead, like many older articles, there is a “Summary” section at the end of the article.
Note that this paper was written before APA had a policy on gender-neutral (nonsexist) language.
The table below describes Janis and King’s (1956) results:
Note that, when Kelman’s (1953) study was conducted, the behaviorism viewpoint dominated psychology. Consequently, most psychologists would believe that the bigger rewards would lead to bigger effects.
Note that today we would use the word “participants” instead of the word “subjects.”
One problem in interpreting Kelman’s results is that the results do not clearly contradict the behavioristic prediction. For example, as behaviorism would predict, participants offered bigger rewards were more likely to contradict their own private opinions whereas participants offered smaller rewards were less likely to contradict their own private opinions. Another problem deals with self-selection: participants selected themselves into the “contradiction” group by choosing whether to contradict themselves. In the little money condition, few participants chose to contradict themselves; in the more money condition, many participants chose to contradict themselves. It could be that the participants who chose to contradict themselves for little money were less committed to their original position than the participants who chose to contradict themselves for big rewards (for more about the problem of self-selection, see Chapter 8 of Research design explained).
Unequivocal: unquestionably correct; clear-cut; not open to other interpretations
derivations: predictions that logically follow from something (the theory). In this case, you can think of derivations as hypotheses.
cognition: knowledge; thought
dissonant: inconsistent with; conflicting beliefs that need to be resolved
Consonant: consistent with; A participant who says “not x” will feel that making such a statement is “consonant” (consistent with) the pressure to say “not X.”
Amount of dissonance = inconsistencies
inconsistencies + consistencies
Imagine we have a participant named Joe. Usually, it is going to take some pressure to get Joe to act against his own beliefs. Without pressure, he will not act in a way that is inconsistent with his own beliefs. Consequently, he will not feel dissonance:
Amount of dissonance =
_______________________________ = 0 = 0
0 (inconsistencies) + 0 (consistencies) 0
With a little pressure, Joe will feel some dissonance
_______________________________ = 1
1 (inconsistencies) + 1 (consistencies) 2
However, note that with more pressure, Joe will feel less dissonance:
_______________________________ = 1
1 (inconsistencies) + 3 (consistencies) 4
In short, according to the formula, getting people to feel the most dissonance requires a delicate touch. You need just enough pressure to get Joe to act inconsistently with his beliefs, but not so much pressure that Joe thinks he acted consistently with the pressure exerted on him.
If you use too little pressure, Joe will not act inconsistently with his beliefs, and so he will feel no dissonance. If you use too much pressure, Joe will act inconsistently with his beliefs, but he will feel little dissonance because he will feel justified in thinking, “I did it because of the pressure.” Thus, beyond a certain point, more pressure leads to less dissonance.
Note that rules have changed so that today
Note that “mariner” is a typo; it should be “manner.”
How Enjoyable the Tasks Were
Today, the traditional analysis would have been analysis of variance (ANOVA) rather than a t test (to see why, see page 307-308 of Research design explained). Current researchers would probably do an ANOVA and, if that were significant, current researchers would probably use post hoc tests to look for differences between groups (to understand why, see the last paragraph of page 313 and all of page 314 of Research design explained).
Desire to Participate in a Similar Experiment
Today, many psychologists would say that any result that is not significant at the .05 level is not statistically significant.
Statistical analyses are more sophisticated today than they were back then. As we suggested, most reviewers would not have allowed the authors to do only t tests. At the very least, most reviewers would have demanded ANOVAs followed up by post hoc t tests. However, many reviewers would have demanded analyses that are even more sophisticated than ANOVA because the authors wanted to do more than test whether the manipulation had an effect on the four different measures. Instead, the authors wanted to test the assumption that the manipulation had different effects on the different measures. The authors tested that assumption by doing separate analyses on each measure and discussing differences between the outcomes of those different analyses. Today, many reviewers would insist that the authors determine whether the manipulation had a different effect on some of the four measures by doing a statistical test that used group (control, 1 dollar, or 20 dollars) and measure (enjoyability, learning, importance, desire to participate) as predictors. That test would usually be a 3 (group)X 4(measure) mixed analysis of variance. A statistically significant interaction between group and measure would indicate that the treatment manipulation had a different effect on at least one measure than it had on other measures. For more about mixed analysis of variance, see Chapter 12 of Research design explained.
The authors do a good job of considering—and then eliminating—an alternative explanation for their results. Today, reviewers would probably ask the authors to add a section that discusses how future research should follow up on these findings.
If the article had been written today, this summary would have appeared at the beginning of the article and been labeled “Abstract.”