A Comparison of Three Approaches for Measuring Negative Cognitions for Psychological Reactance

ABSTRACT

Psychological reactance, routinely operationalized as a latent construct comprised of anger and negative cognitions, represents an aversive state following a threatened or eliminated freedom. The current study sought to extend the literature by comparing three distinct measures of negative cognitions: (a) trained coder (thought-listing), (b) participant coding (thought-listing), and (c) Likert scale. Participants (N = 540) were randomly assigned to view messages in a 2 (language: forceful and non-forceful) x 3 (topic: exercise, fruit and vegetable consumption, and sleep) between subjects factorial design. Exposure to forceful (vs. non-forceful) language resulted in higher negative cognitions across all three measures. Moreover, all three measures of negative cognitions were negatively associated with behavioral intention. Three competing structural models were examined, each using a different measure of negative cognitions. Results demonstrated the three models performed similarly based on comparisons of model fit and variance explained. Despite similar performance among the three measures of negative cognitions, there was a slight, but consistent advantage for the Likert scale measure model in terms of fit, variance explained, and factor loadings. Validation of these three measures of negative cognitions provides communication researchers with the flexibility to choose the measure most appropriate for their needs.

Acknowledgments

The authors thank Rahat Rizwan, Erin Wildman, and Alyssa Rivera for their coding of the thought-listing data.

Disclosure Statement

No potential conflict of interest was reported by the authors.

Notes

1. Measurement variance could not be tested in the models that included the observed negative cognition variables (i.e., trained coder and participant coding).

2. A structural model with reactance, anger, and negative cognitions (Likert scale) all modeled as latent variables (i.e., a 2nd order latent factor) also produced good model fit, χ² (86, 540) = 133.33, RMSEA =.03 (90% CI = 0.021 to 0.042), CFI = .99, SRMR = .03, AIC = 19242.51, BIC = 19444.21, and similar results (See Table A1).

3. The standardized factor loadings for anger were λ = 1.00, SE = 0.00 (trained coders); λ = .99, SE = 0.05 (participant coding); and λ = .89, SE = 0.03 (Likert scale).

Additional information

Funding

This work was supported by the Campus Research Board at the University of Illinois at Urbana-Champaign under award [RB18129].