When Cigarette Package Warnings Contain Imagery, Does Textual Content Matter? Field Experiments to Test the Impact of Testimonial and Scientific Narratives

Theoretical Background and research questions/hypothesis: Little evidence exists on the most effective content for use in pictorial warning labels on cigarette packages.  Previous experimental research in Mexico indicated which images were most effective.  The present study aimed to determine which textual content was most effective in combination with these images.

Methods:  Mock cigarette packs were constructed to integrate pictorial and textual  elements into health warning labels.  For each of six different topics (i.e., lung cancer, throat cancer, stroke, heart disease, premature birth, addiction), three packs had the same image, with two packs containing short testimonials and one pack containing factual text adapted from available educational material.  For each of two other topic areas (i.e., gangrene, emphysema), two mock packs were constructed that included the same testimonial content, but one warning used a gruesome image of damaged organs and one image showed someone suffering from the disease.  Field experiments were conducted with 529 adult smokers and 530 young adults recruited from public venues in two major Mexican cities.  Participants evaluated all packs for four different topics, and packs were presented in random order.  Participants rated each pack in terms of message acceptance (i.e., attention, credibility, relevance) and impact (e.g., negative emotional arousal, perceived impact on others).  After all warnings for a topic were rated, participants selected which warning for that topic motivated them most want to not smoke.  T-tests were used to compare mean ratings for packs, and linear mixed effects models were estimated to assess which textual strategy was rated as having a higher impact, while controlling for topic and participant sociodemographics.  Bi-variate and multivariate logistic regression models were estimated to determine factors associated with selecting testimonial text as more motivating than the factual text. 

Results:  T-tests indicated that warnings with factual content had significantly higher (i.e., stroke, lung cancer, premature birth) or equal (i.e., throat cancer, addiction, heart attack) message acceptance and impact than warnings with testimonial content.  Mixed effects models that combined data across topics confirmed that warnings with factual content were rated more highly than testimonial content.  However, a statistical interaction was found between textual content and both educational attainment and sensation seeking.  Stratification of the models indicated that factual content was consistently rated most highly among participants with higher educational attainment and higher sensation seeking.  Participants with lower educational attainment or low sensation seeking rated testimonials as equivalent or higher than factual content.  These results were consistent with those from bi-variate and multivariate logistic models regressing the selection of testimonials over factual content as most motivating to not smoke.  When the testimonial content was held constant and ratings of warnings with gruesome imagery were compared with warnings with imagery of human suffering, the gruesome imagery was consistently rated higher.

Conclusions:  The results from this study indicate that pictorial warning labels that contain factual text generally have equal or greater impact than testimonial text. 

Implications for research and/or practice:  If regulators must select one type of text, then it should be factual.  However, some warnings could include testeimonial text that targets disadvantaged groups.