Abstract
Vaidis et al. (2024) provided results from a multilab conceptual replication of an experiment reported in Croyle and Cooper’s (1983, Experiment 1) seminal article on the role of physiological arousal in cognitive dissonance. The original experiment relied on an induced-compliance manipulation of cognitive dissonance in which participants selected for holding negative attitudes toward a campus drinking ban were randomly assigned to write arguments against the ban (high-choice consonant condition) or assigned to write arguments in favor of the ban under conditions designed to encourage either low or high perceived freedom in writing the arguments (low-choice/counterattitudinal condition, high-choice/counterattitudinal condition). Croyle and Cooper found evidence of attitude change toward more favorable evaluation of the drinking ban that was higher among participants in the high-choice/counterattitudinal condition compared with participants in the low-choice/counterattitudinal and high-choice/consonant conditions combined, d = 2.40, t(27) = 6.23, p < .001.
Vaidis et al. (2024) conducted a multilab replication of this experiment with several notable procedural changes. One change was use of an increase in university tuition as the issue of evaluation. Another change was substitution of a neutral-attitude condition (describe arguments for other university issues) in place of a congruent-attitude condition. In contrast to Croyle and Cooper (1983), Vaidis et al. found a vastly smaller comparative effect of d = 0.14, t(2,721) = 3.66, p < .001 (all analyses in this paragraph based on descriptive statistics reported in Vaidis et al., 2024, Table 4). Not only is this effect very small, but it is mostly due to smaller attitude change among participants in the neutral-attitude condition with essentially no difference in effect between the low-choice and high-choice counterattitudinal conditions, d = 0.02, t(1,995) = 0.38, p = .70. Vaidis et al. did find that participants in both counterattitudinal conditions reported greater change in attitude than did participants in the neutral-attitude condition. The authors proposed this finding may suggest that cognitive inconsistency produced by counterattitudinal behavior generates cognitive dissonance regardless of the extent to which one perceives freedom to engage in the behavior.
No (Low) Choice or Moderate Choice?
These findings appear to signal a failure to replicate Croyle and Cooper (1983) and appear inconsistent with the theoretical claim that freely choosing to engage in counterattitudinal behavior is one source of cognitive dissonance. However, as in the case of any conceptual-replication failure, it is important to consider plausible theoretical explanations for apparent discrepancies in results. A notable difference between the results of Croyle and Cooper and those of Vaidis et al. (2024) is an apparent mismatch in the degree of perceived freedom to engage in the counterattitudinal behavior reported by their respective participants. Rescaling the 31-point perceived-choice manipulation check used by Croyle and Cooper to a 9-point scale like the one used by Vaidis et al. reveals participants in Croyle and Cooper’s low-choice/counterattitudinal condition reported a mean perceived freedom rating of 2.80 (in the bottom third of the scale), whereas those in the high-choice/counterattitudinal condition reported a mean perceived freedom rating of 6.10 (in the top third of the scale). In the Vaidis et al. studies, the respective means were 4.44 (middle third of scale) and 6.50 (top third of scale). This discrepancy suggests that Croyle and Cooper’s manipulation of choice produced low-choice and high-choice conceptual conditions, whereas Vaidis et al.’s manipulation of choice produced moderate-choice and high-choice conceptual conditions. The typical participant in Vaidis et al.’s “low-choice” condition may not have perceived sufficient external justification for their behavior and so experienced cognitive dissonance that motivated attitude change such as participants assigned to the high-choice/counterattitudinal condition.
Also notable is an internal analysis based on the data provided by Vaidis et al. (https://osf.io/9xsmj), which revealed a significant correlation between ratings of perceived freedom and attitude change among participants within and across both counterattitudinal conditions (across both conditions: r = .12, p < .001, N = 1,997; low-choice condition: r = .17, p < .001, N = 937; high-choice condition: r = .09, p < .003, N = 1,060). Although these effects are small and do not permit strong causal inference, when considered in conjunction with the potential manipulation failure to sample low conceptual values of perceived choice, it seems more caution may be warranted before questioning the validity of the induced-compliance paradigm.
Cognitive Inconsistency or Self-Persuasion?
What about the difference in attitude change between the counterattitudinal conditions and the neutral condition? Is this evidence that counterattitudinal behavior generates cognitive inconsistency more generally regardless of one’s perceived freedom to engage in the behavior? Possibly, but there is an alternative explanation that proves far simpler: Writing a counterattitudinal essay generates reasonable propositional arguments for changing one’s negative attitude in a more favorable direction. Vaidis et al. (2024) did find participants reported feeling more discomfort and conflict in the counterattitudinal conditions than in the neutral-attitude condition (seemingly consistent with a general-inconsistency explanation), but these ratings did not appear to statistically mediate the corresponding effect on attitude change (seemingly inconsistent with a general-inconsistency explanation). Alternatively, these ratings may reflect increased attitude ambivalence (e.g., Conner & Armitage, 2008) due to changing one’s negative attitude in a more favorable (positive) direction.
Conceptual-Replication Failures Remain a Challenge for the Field
Taken as a whole, conceptual questions regarding interpretation of Vaidis et al.’s (2024) results leave the field in a theoretical quandary. After 37 replication studies and 2,723 participants, do researchers conclude in favor of a freedom-of choice-explanation, a general-inconsistency explanation, or a self-persuasion explanation for their results? In fairness, the major problem this international team faces is the same problem the entire field faces: How should one interpret failed conceptual replications? Their endeavor is a valuable contribution because it offers insight into the range of variability one can expect from replication studies conducted by different labs and by different cultural groups. Their endeavor is a valuable contribution because it clearly identifies interpretation issues pertaining to induced-compliance paradigms that require theoretical resolution and are not readily explainable in terms of sampling error and measurement error (e.g., Stanley & Spence, 2014). Moreover, the alternative theoretical interpretations presented here in no way diminish these contributions. These interpretations are not easily anticipated a priori and are facilitated greatly by the benefit of hindsight and the considerations of empirical results. Nonetheless, it seems unlikely that multilab replication endeavors will be useful tools for providing insight into whether published studies or their theoretical implications are faulty. Instead, they may be more useful as tools for facilitating theoretical coherence by more clearly identifying empirical inconsistencies in need of theoretical resolution.
