Neural Predictors of Message Effectiveness during Counterarguing in Antidrug Campaigns

Abstract

A substantial amount of research has focused on predicting the effectiveness of persuasive messages. However, characteristics of both the message itself and its receiver can impact theoretically predicted effects. For example, recent work published in this journal demonstrated that issue involvement modulates the relationship between message sensation value (MSV) and argument strength (AS). When exposed to anti-cannabis public service announcements (PSAs), high-drug-risk individuals rate these messages as having low effectiveness regardless of variation in MSV and AS. Accordingly, for high-risk individuals, MSV and AS lose their predictive power in message design; moreover, the all too common use of high MSV, high AS PSAs to dissuade drug use may be ineffective, as high-risk viewers are more likely to engage in counterarguing. In this paper, we use functional magnetic resonance imaging to investigate the neural correlates of counterarguing. Subsequently, we employ a brain-as-predictor approach using neural activation and self-report data to predict message effectiveness in two independent samples. We demonstrate that by adding two neural predictors within the middle frontal gyrus and superior temporal gyrus to self-report data, the prediction accuracy of message effectiveness in high-drug-risk individuals during counterarguing can reach, and even surpass, the prediction accuracy for low-drug-risk individuals.

Keywords:

• Counterarguing
• Persuasion
• Health Communication
• Antidrug Public Service Announcements
• Brain Imaging
• Functional Magnetic Resonance Imaging

Acknowledgments

We greatly thank Joseph N. Cappella (Annenberg School for Communication, University of Pennsylvania) for generously providing the original PSAs, three of the five data sources, and additional message effectiveness data for this study. We thank Scott Grafton and Michael Miller (Department of Psychological and Brain Sciences & Brain Imaging Center, University of California Santa Barbara) for generously providing the fMRI scanner time and also providing us with access to UCSB's brain-imaging facilities. We also thank Mario Mendoza, Grace Anderson, and Chadd Funk for their assistance with collecting the brain-imaging data.

Notes

[1] Unlike Falk et al. (2014), we derived signal change of PSA contrasts on the group (second) level and not on the subject (first) level. This procedure avoids using average PSA ranks per subject for a second ranking procedure in which PSA are ranked against each other. We found this procedure less constraining for the variance of ranks and thus more conservative. In other words, with the ranking procedure used in Falk et al., the stability and fit of our prediction models reported here would increase rather than decrease.

[2] Compared to the MPFC mask generated by Falk et al. (2014), our MFG activation is more superior, posterior, and lateral. We find MFG activation in the right hemisphere, whereas Falk and colleagues found more medial MPFC activation tending toward the left hemisphere. However, Chua's fROI is comparably close to Falk's MPFC.