Visual Exemplification and Skin Cancer: The Utility of Exemplars in Promoting Skin Self-Exams and Atypical Nevi Identification

Abstract

The present article reports an experiment investigating untested propositions of exemplification theory in the context of messages promoting early melanoma detection. The study tested visual exemplar presentation types, incorporating visual persuasion principles into the study of exemplification theory and strategic message design. Compared to a control condition, representative visual exemplification was more effective at increasing message effectiveness by eliciting a surprise response, which is consistent with predictions of exemplification theory. Furthermore, participant perception of congruency between the images and text interacted with the type of visual exemplification to explain variation in message effectiveness. Different messaging strategies influenced decision making as well, with the presentation of visual exemplars resulting in people judging the atypicality of moles more conservatively. Overall, results suggest that certain visual messaging strategies may result in unintended effects of presenting people information about skin cancer. Implications for practice are discussed.

Notes

¹ Representative exemplars offer people accurate referents from which to base judgments about larger event populations. This is counter to many exemplars—often nonrepresentative—that bias processing.

² Zillmann (2009, p. 81) suggested three ways to obtain a representative, nonbiased exemplar group. As an example, consider a phenomenon that varies on three attributes. Random exemplification would pull three examples from the total population of potential instances and, in theory, equally present the three attributes shared among all of the instances of a phenomenon within the entire population of instances (for further details, see Zillmann, 1999, p. 81, Model A). Another way to construct a representative exemplar is to divide some event population into subpopulations and then pull event subsamples that provide a representative exemplar (see Zillmann, 1999, p. 81, Model B and Model C).

³ Zillmann (1999) wrote, “It should be emphasized that the exemplification procedures … are presentational strategies. Which of these strategies will prove superior in terms of message reception, processing, retention, and retrieval is to be determined empirically” (p. 82).

⁴ The decision was made to include cover images on all pamphlets because results of a content analysis suggested that printed materials usually included images on the cover (see King, 2015). In addition, three organizations’ logos appeared on the cover to indicate sourcing of written messages (American Cancer Society, Centers for Disease Control and Prevention, National Cancer Institute). Two other panels were identical across conditions. The back panel featured steps on conducting self-exams at home, and an inside panel featured general information on skin cancer. See the Appendix for an example of the layout.

⁵ The control condition did not include any visual exemplars but included stock design patterns to avoid having unrealistic patches of white space in the brochure. The other conditions shared a layout but featured different arrays of moles. The exemplar condition featured one array of atypical moles, utilizing a random exemplification presentation (see Zillmann, 1999), in which the population exemplified was that of atypical moles. The juxtaposition condition featured the exemplar array of atypical moles, as well as an array of typical moles, using a stratified exemplification strategy (see Zillmann, 1999). The multiple comparison condition featured both arrays from the juxtaposition condition as well as an additional array of atypical moles using a biased stratified exemplification strategy. Visual exemplars of typical/atypical moles were taken from printed materials, images provided by a medical photography company, and an image database maintained by the National Cancer Institute (2014).

⁶ The mole images for this task were taken primarily from clinical photographs provided by a medical photography company. Moles in the risk identification task were unique to the task and not found in stimuli. Participants were instructed to circle moles that they believed demonstrated signs of atypicality. This circling behavior, although simple, requires the application of skills promoted in the pamphlet and is similar to clicking behavior in Internet research (e.g., Murphy, Hofacker, & Mizerski, 2006), as it requires participants to perform an action that they might do in the privacy of their own home (e.g., identify an atypical mole on their body). Participants were not allowed to consult the paper pamphlet when performing the visual identification task.

⁷ Rerunning the mediation analyses with procedures outlined by Hayes and Preacher (2014) in which dummy codes were assigned to each of the experimental conditions with the other conditions as referents (indicator coding) produced results consistent with the reported mediation for all conditions. One reason for this, even though reported Games-Howell post hoc comparisons did not find differences between juxtaposition/multiple comparison and control, is that Games-Howell is more conservative than indicator coding when making comparisons. Also of note is that using the same procedures specified by Hayes and Preacher—but engaging in contrast coding—produced results indicating a significant indirect effect of surprise for all exemplar conditions against control but not in comparing the exemplar conditions to one another. The main body text results are reported, as they are the most logically stated in describing and reporting exemplification theory tests.

⁸ This is eta-squared, calculated by hand, not partial eta-squared provided by the statistical package used (SPSS).