Adolescent Attention to Disgust Visuals in Cigarette Graphic Warning Labels

Abstract

Purpose:

Adolescents are often a target audience for disgust-eliciting anti-smoking messages, including graphic warning labels (GWLs) on cigarette packages. Yet, few studies have examined how adolescents attend and respond to disgust imagery frequently depicted in these messages.

Method:

A within-subjects eye-tracking experiment with middle school youth (n = 436) examined attention for GWLs that feature disgust versus non-disgust images. Hypotheses were based on emotion theory and previous findings with adult participants. This study also tested whether living with a smoker moderated effects of attention on negative emotions and risk beliefs.

Results:

Participants paid similar levels of attention to warnings with disgust visuals as they did warnings with non-disgust visuals, accounting for other differences in the warnings. The presence of a disgust visual drew greater attention to the warning image and reduced attention for the warning text. These viewing patterns were similar for youth who live with a smoker and those who do not. Attention to disgust imagery was the only attentional factor to predict negative emotional reactions, and this relationship was driven by results observed among youth who live with a smoker. Attention to neither image nor text predicted risk beliefs.

Conclusions:

GWLs with disgust imagery do not trigger more or less attention to the overall warnings but do influence allocation of attention to images over text. Future work should confirm whether attention to disgust imagery itself is important for triggering negative emotional responses, particularly with youth for whom the message is more personally relevant.

Nearly 90% of smokers in the United States began smoking before age 18¹—making youth a primary target of anti-smoking messages. Prior research suggests messages featuring disgust may make smoking less appealing to adolescents.^2,3 Indeed, cigarette graphic warning labels (GWLs) often depict disgust-elicitors like death, decay, deformity, and disease⁴ to emphasize smoking risks.⁵ In 2009, the U.S. Food and Drug Administration proposed nine GWLs to appear on cigarette packages—four labels fe ature a corpse (death), diseased lungs (disease), lip cancer (decay), and a tracheotomy or hole in the throat (deformity). Such stimuli qualify as disgust-elicitors because they reliably evoke feelings of disgust.⁴

Although prior research suggests disgust-eliciting messages are an effective anti-smoking strategy with adolescents, studies on attention to disgust-eliciting anti-smoking messages have mainly looked at adult smokers, and nonsmokers, reactions to disgust imagery in televised anti-smoking advertisements rather than GWLs appearing on cigarette packages alongside tobacco branding.^6–8 The current study examines how disgust imagery in the FDA-proposed GWLs influence adolescent attention and how attention is related to emotional reactions and smoking risk beliefs. This study also examines differences in these outcomes for youth who live with a smoker—a risk indicator that may influence reaction s to GWLs.

Rationale

Adolescents are particularly attracted to the sophisticated, colorful designs of cigarette packaging; their exposure to tobacco branding correlates with favorable smoking attitudes and intentions.^9,10 Therefore, one goal of GWLs should be to detract attention from tobacco branding.^11,12 Attention is a primary component of many communication effects models; the literature on effective warnings suggests GWLs must attract and hold attention in order to aid risk information processing^13–15. Imagery motivating people to look away would be less effective at holding attention and, therefore, limit processing.

Negative emotional stimuli are often regarded as attracting attention, but disgust has also been characterized as repelling attention. According to discrete emotion theories, emotion experiences are distinguished according to distinct feelings, thoughts and behaviors. Disgust stimuli trigger repulsive feelings and a desire to distance oneself from the offensive object leading to an involuntary rejection response. Disgust elicitors—like gory depictions of smoking consequences—are predicted to motivate low-attentional behavior.^16,17

In contrast, dimensional emotion theories characterize emotions with similar valence and arousal as functioning in similar ways. Disgust, like fear, attracts attention because it signals a threat in one’s environment requiring further processing.¹⁸ Although there is significant evidence for the dimensional proposition in anti-smoking research, most studies define attention as allocating resources to message processing measured via cardiac response rather than visual attention. One study, however, found adults paid greater visual attention to aversive than non-aversive GWLs.¹⁹

Consistent with evidence supporting the dimensional perspective, we propose disgust GWLs attract attention more quickly (H1a) and hold attention longer (H1b) than non-disgust GWLs among adolescent participants. Conversely, we predict disgust GWLs detract attention from brands, such that participants attend to branding more slowly (H1c) and for less time (H1d) when accompanied by disgust GWLs.

Once participants direct attention toward the warning, it may be more difficult to disengage attention from disgust stimuli.²⁰ This attentional bias may increase attention to disgust images and reduce attention to text. Although dimensional theories would predict increased attention to text to understand the threat, studies measuring visual attention have shown aversive images attract greater attention while reducing attention for text.^19,21 Thus, we predict participants will spend more time looking at images (H2a) and less time looking at text (H2b) for disgust compared to non-disgust GWLs.

Personal Relevance by Living with a Smoker

When health messages are personally relevant, individuals may be motivated to attend to them; alternatively, individuals may engage defensive processes, such as avoiding the message, or its more threatening parts, to minimize risk perception.^22–24 In this study, relevance is conceptualized as living with someone who smokes. Anti-smoking messages have greater relevance for youth who live with smokers—portrayin g risk for important people in their lives. Household smoking is also a significant indicator of future smoking.²⁵ We are not aware of previous research examining the connection between household smoking and attention; therefore, we ask whether participants living with a smoker attend to disgust GWLs less quickly or for less time (RQ1) and whether they exhibit greater (weaker) bias toward disgust imagery (RQ2)?

The Influence of Attention on Emotions and Risk Beliefs

Eliciting negative emotional reactions is considered a key mechanism driving GWL effectiveness.^26,27 Emotions provide information about an issue’s importance²⁸ and motivate judgements and behavior accordingly.²⁹ Disgust-eliciting imagery have been found to evoke stronger reactions than fear-evoking content^6,30, and may conveniently increase emotional impact.³¹

Research on emotion regulation show increased attention to emotional cues predicts changes in emotional state.³² Although studies have demonstrated a positive relationship between exposure to GWLs and negative emotional reactions, the emotional outcomes of increased attention to more aversive GWLs have not yet been explored for adolescents. Based on previous support for a relationship between attention to aversive content and negative emotional responses for adults, we expect greater attention to disgust imagery positively predicts negative emotional reactions (H3). We also examine whether household smoking interacts with attention to increase negative emotions (RQ3).

GWLs may also convey risk information to individuals who are not motivated to process text-only warnings.³³ Research has both supported and refuted this outcome.^2,19 In light of initial supporting evidence, we test whether adolescents’ risk perceptions are positively associated with attention for disgust imagery (H4). Again, we consider whether household smoking interacts with attention to increase risk perceptions (RQ4).

In summary, we predict disgust GWLs attract attention faster and longer, delaying and reducing attention to branding (H1), and shifting greater attention to images over text (H2). We predict greater attention to disgust images positively predicts negative emotional reactions (H3) and risk beliefs (H4); and ask whether household smoking predicts differences in attention to disgust warnings and images (RQs 1 & 2) and moderates the relationship between attention to disgust images and both emotions and risk beliefs (RQs 3 & 4).

Method

Study Design

A within-subjects experiment examines adolescent responses to disgust versus non-disgust GWLs. The data come from three larger experiments examining between-subject effects for variations of the nine FDA-proposed GWLs.^34,35 All studies were approved according to University IRB procedures for conducting research with children.

Materials

This study analyzes responses from participants randomly assigned to view full-color GWLs appearing on the top 50% of cigarette packages. Four GWLs contain disgust images: a corpse, lip cancer, blackened lungs, and a man with a tracheotomy (Figure 1). Non-disgust images include a baby in an incubator, a child exposed to second-hand smoke, a man wearing a breathing mask, a man wearing a t-shirt with the words “I Quit”, and a woman crying (Figure 2).^a We removed the quitline number and made typeface and font size consistent. We identified areas of interest (AOIs) that captured attention within brand and warning regions, including separate AOIs for images and text. AOI markings were not discernable to participants.

Figure 1. GWLs with disgust visuals.

(left to right): death, decay, disease, and deformity

Figure 2.

GWLs with non-disgust visuals.

Participants

Participants (n = 436) attended middle schools in the Northeast where 40–100% of students receive free- or reduced-price lunch—indic ating participants are primarily from low-income households. Table 1 provides participant demographics.

Table 1.

Participant Demographics N = 436

	Means (SD) or n’s (valid %)
Age	M = 12.91 (0.90)
Gender
Male	219 (50.50%)
Female	204 (47.00%)
Prefer not to say	11 (2.50%)
Hispanic	50 (11.50%)
Race
Black	177 (40.60%)
White	213 (48.90%)
Other	88 (20.20%)
Smoking variables
Household smoking	205 (47.00%)
Tried a cigarette	39 (8.90%)
Colorblind	28 (6.50%)

School administrators sent opt-out forms to parents/guardians in advance explaining the study. Interested participants, whose parents did not opt out, completed assent forms. Students were informed participation was voluntary and they could discontinue at any time. To accommodate district policies, we offered schools compensation choices: a $10 gift card for each participant, $10 to the school per participant toward student initiatives, or a 50/50 split.

Procedures

A mobile laboratory, with five stations, parked on-site at participating schools. Participants were told they would view cigarette packages and answer questions about them. Participants viewed pack images measuring 312 pixels x 482 pixels on Tobii T60XL 24-inch LCD monitors at 1920 × 1200 resolution. TobiiStudio 3.4.4 recorded their gaze. We asked participants to keep their eyes on the screen during the experiment. These instructions appeared on screen and were also read to participants.

Images appeared in random order for 10 seconds each. One of three popular cigarette brand logos appeared in the bottom half of each image. Brand and GWL pairings were rotated across three sub-conditions to ensure the same brand and GWL did not always appear together. Fixation crosses appeared in between images to reset gaze. A random number generator determined where in one of eight positions around the outer screen each cross appeared (i.e. bottom center); placement varied for each sub-condition. After viewing all images, participants used tablet computers to complete a post-exposure questionnaire. Most participants completed the questionnaire independently. Two participants needed clarification of questions; the survey was read to one of these participants. Finally, participants were debriefed and received permitted compensation.

Measures

The independent variable is manipulation of disgust versus non-disgust imagery. All other variables are measured.

Attention.

We measured attention with two eye-tracking statistics. Time to first fixation (TFF) measures the time participants take to initially fixate on AOIs. Fixation duration measures the amount of time participants look at AOIs. We calculated proportional fixation duration (PFD) for attention to each AOI as a fraction of overall attention for warning and brand regions to compare relative attention for disgust versus non-disgust GWLs. Average fixation duration (AFD) accounts for the unequal number of disgust and non-disgust stimuli in models where attention predicts emotions and beliefs (see Table 2).

Table 2.

Descriptive Statistics for Time to First Fixation (TFF) in Seconds, Average Fixation Duration (AFD) in Seconds and Proportion Fixation Duration (PFD) on Warning, Image, Text and Brand.

	Time to First Fixation & Average Fixation Duration		Proportion Fixation Duration
	Mean (SD)	Range	Mean (SD)	Range
TFF Disgust Warning	0.59 (0.88)	0.00 – 9.38	-	-
TFF Disgust Brand	1.86 (2.12)	0.00 – 9.99	-	-
TFF non-Disgust Warning	0.55 (0.86)	0.00 – 9.97	-	-
TFF non-Disgust Brand	2.51 (2.52)	0.00 – 9.99	-	-
Disgust Warning	4.52 (1.62)	0.00 – 10.61	0.67 (0.21)	0.04 – 1.00
Disgust Image	3.14 (1.35)	0.00 – 9.74	0.47 (0.20)	0.01 – 0.97
Disgust Text	1.21 (0.78)	0.00 – 6.22	0.19 (0.15)	0.01 – 0.93
Disgust Brand	2.09 (1.15)	0.00 – 7.91	0.33 (0.21)	0.00 – 0.96
Non-Disgust Warning	4.99 (1.52)	0.41 – 8.33	0.73 (0.19)	0.05 – 0.99
Non-Disgust Image	2.31 (0.94)	0.13 – 5.84	0.35 (0.18)	0.01 – 0.99
Non-Disgust Text	2.52 (1.04)	0.07 – 6.28	0.37 (0.19)	0.01 – 0.94
Non-Disgust Brand	1.67 (0.91)	0.00 – 6.62	0.27 (0.19)	0.01 – 0.95

Negative Emotion.

Participants indicated how afraid, angry, disturbed, grossed out, sad, and scared they felt after viewing stimuli on 5-point scales ranging from “Not at all” to “Extremely” (adapted from PANAS ³⁶). These items were averaged to create a negative emotion scale (M = 2.70, SD = 1.05; Cronbach’s α = 0.83).^b

Risk Beliefs.

Participants responded to items measuring perception that cigarette smoking is related to heart disease, lung disease, cancer, stroke, a hole in the throat, asthma and problems in babies whose moms smoke using a 4-point scale ranging from “Definitely not ” to “Definitely yes”. These items were averaged to crea te a risk beliefs index (M = 3.73, SD = 0.43; Cronbach’s α = 0.91).

Household Smoking.

Participants answered whether “…anyone who lives with you smoke cigarettes?” Nearly half reported living with a smoker (N = 205).

Covariates included age, gender, race and Hispanic ethnicity, prior smoking (even a puff), and being diagnosed as colorblind. Three items assessed sensation seeking on a 4-point scale from “strongly disagree” to “strongly agree” (M = 2.02, SD = 0.73; Cronbach’s α = 0.73).

We included word count as a control variable to account for the impact of differences in message complexity on attention.

Results

To test differences in TFF and PFD between disgust and non-disgust GWLs, we ran mixed-effects regression models with condition as the fixed effect. We used contrast coding to identify the conditions (.5=disgust; −.5=non-disgust) so model intercepts represent the mean of all stimuli. Word count varied across warnings. On average non-disgust warnings had more words; therefore, we included word count as a covariate to control for the influence of message complexity on attention. Demographic variables were also included as covariates. We included the intercept and slope across participants and the intercept across stimuli as random effects to account for variation in participant responses and responses elicited by individual warnings. Covariances between the random effects for participants were estimated with an unstructured covariance matrix. The model for TFF to brand did not converge—the simpler compound symmetry covariance structure was imposed instead. Covariate tests for the random effects show significant variation in the intercept and condition slopes across participants (p’s ≤ 0.05) for each hypothesized model below. Variances across stimuli were marginally significant (p’s > 0.07) for all models except TFF for the warning (p = 0.37).Analyses were run using SPSS version 25.

We predicted participants would orient more quickly (H1a) and attend disgust GWLs longer (H1b) than non-disgust GWLs, and examined whether attention differs based on household smoking (RQ1). Results show participants did not look at disgust warnings any faster or longer than non-disgust warnings (p’s > 0.12), rejecting H1a and H1b. The effect of household smoking on TFF to the GWLs was marginally significant, suggesting a trend whereby those who live with a smoker orient more slowly toward any warning (b = −0.08, t_397.98 = −1.81, p ≤ 0.07). However, the interaction of household smoking and exposure to disgust GWLs was not significant for TFF or PFD (RQ1; p’s > 0.85). Message complexity was not a predictor of either attention measure (p’s > 0.14).

Next, we tested whether disgust GWLs delayed TFF and reduced PFD for brands. Contrary to H1c and H1d, there were no significant differences in TFF or PFD for disgust compared to non-disgust GWLs, (p’s > 0.57). Message complexity did not predict TFF or PFD for branding (p’s > 0.12).

We predicted disgust GWLs increase PFD for images (H2a) and decrease PFD for text (H2b). PFD on images was significantly longer for disgust GWLs (b = 0.11, t_10.17 = 2.93, p ≤ 0.02), supporting H2a. PFD for text was significantly reduced for disgust compared to non-disgust GWLs (b = −0.15, t_9.71 = −3.12, p ≤ 0.01), supporting H2b. Again, there were no significant differences in PFD for images or text based on household smoking (RQ2; p’s > 0.16). Message complexity did not predict attention to images or text (p’s > 0.23).

Finally, to test whether AFD for disgust images positively predicted negative emotional responses (H3) and risk beliefs (H4)^c we conducted moderated OLS regression analyses using Hayes (2018) Process Model 1 with 5000 bias-corrected bootstrapping samples at 95% CI. Household smoking was included as the moderator for both emotions and beliefs (RQ3 & 4). AFD for disgust text and brands, and non-disgust images, text and brands were added alongside demographic covariates (Table 3). The model with AFD predicting negative emotions was significant (F(17, 404) = 2.27, p ≤ 0.01, R² = 0.09). For participants who live with a smoker, negative emotions increased 0.14 points for every second of attention to disgust images (b = 0.14, t(404) = 2.37, p ≤ 0.02; R²-change = 0.01). The interaction slope was not significant for participants who do not live with a smoker. The overall model predicting risk beliefs was not significant (F(17, 411) = 0.59, p = 0.90, R² = 0.02).

Table 3.

Regression Coefficients for Relationships between Average Fixation Duration (AFD) in Seconds and Interaction with Household Smoking Status on Negative Emotions and Risk Beliefs

	Negative Emotion		Risk Beliefs
	b (SE)	t	b (SE)	t
Constant	3.79 (0.70)***	5.39	3.71 (0.34)***	10.86
AFD Disgust Image	−0.01 (0.06)	−0.20	−0.01 (0.03)	−0.51
AFD*Live w/Smoker	0.16 (0.07)*	2.33	<−0.01 (0.03)	−0.02
AFD non-Disgust Image	0.01 (0.06)	0.09	0.06 (0.03)	1.76
AFD Disgust Text	0.04 (0.07)	0.61	<0.01 (0.04)	0.06
AFD non-Disgust Text	−0.04 (0.06)	−0.79	−0.01 (0.03)	−0.18
AFD Disgust Brand	−0.08 (0.06)	−1.52	−0.02 (0.03)	−0.58
AFD non-Disgust Brand	−0.09 (0.07)	−1.30	<0.01 (0.03)	0.05
Covariates
Age	−0.05 (0.05)	−1.02	<0.01 (0.02)	0.06
Male (vs. female)	−0.23 (0.09)**	−2.47	−0.01 (0.04)	−0.32
Other gender	0.08 (0.28)	0.28	0.03 (0.14)	0.25
Hispanic	−0.17 (0.14)	−1.15	0.02 (0.07)	0.24
Black	0.13 (0.12)	1.07	<−0.01 (0.06)	−0.01
White	0.15 (0.12)	1.22	0.07 (0.06)	1.21
Previous smoking	−0.08 (0.17)	−0.49	−0.06 (0.08)	−0.78
Sensation seeking	−0.14 (0.06)*	−2.16	−0.02 (0.03)	−0.80
Colorblind	0.11 (0.18)	0.58	0.01 (0.09)	0.16
Live w/Smoker	−0.43 (0.23)	−1.85	−0.04 (0.11)	−0.36

^*p ≤ .05,

^**p ≤ .01,

^***p ≤ .001

Results suggest gender and sensation seeking account for changes in negative emotions. We conducted post-hoc analyses examining the interaction of attention with these variables. Although both models yielded similar significant results (F(16, 405) = 2.04, p ≤ 0.01, R² = 0.07), neither variable moderated attention to increase negative emotions (p’s > 0.47).

Discussion

Disgust-evoking content is frequently used to attract attention to messages communicating smoking risks with adolescents. We examined whether disgust-evoking GWLs increased attention for warnings and decreased attention to brands. Results show no significant differences in attention to warnings or brands between disgust and non-disgust GWLs. Although this finding does not provide support for increased attention to disgust GWLs, it also suggests adolescents do not avoid disgust content as theories of discrete emotion might predict.

In fact, consistent with dimensional emotion theories results show significant increases in attention to disgust imagery, a finding similar to previous research with adults.¹⁹ We also found evidence of an attentional bias for disgust images over text. After controlling for word count, the presence of disgust remained a significant predictor of reduced attention to text, suggesting reading time for shorter messages on disgust GWLs does not account for this difference alone. This finding appears inconsistent with dimensional theories which predict greater attention to sources of information about a potential threat such as GWL text.¹⁸ Potential explanations for this effect include avoidance of threatening information following exposure to highly aversive content, distraction from processing text due to increased processing of aversive imagery (the inability to look away), and efficiency of processing (whereby the aversive images provide sufficient information regarding risk circumventing the need for further processing).²¹

One rationale for including images in warnings is to enhance processing among those with low motivation to process text-only warnings.⁴⁰ Disgust-evoking images may be well suited for this goal. We found a significant relationship between attention to disgust images and increased negative emotions regarding smoking; this relationship was driven by those who reported living with a smoker, a subpopulation at greater risk for smoking initiation.²⁵ Additionally, youth who live with a smoker are more likely to see GWLs on cigarette packs belonging to smokers in their homes.

Negative emotional reactions have been identified as an important factor influencing processing of anti-smoking messages. Inducing negative emotions may help to stimulate negative attitudes toward smoking at a time when youth are beginning to form more favorable smoking perceptions.²⁵ However, the observed association between attention to disgust imagery and negative emotions was small. Future research with longer exposure duration and/or stronger disgust-elicitation would be valuable to assess whether these results have sufficient magnitude for practical application.

Ultimately, emotions are expected to influence cognitions and motivate behavior²⁶, such as avoiding smoking or cues encouraging smoking. The current study suggests disgust GWLs did not reduce attention for cigarette brands over non-disgust GWLs. Additionally, we found no evidence of a connection between increased attention for disgust imagery and increased risk perceptions—consistent with previous results among adults.^19,21 However, we found no evidence of a relationship between attention to text (for disgust and non-disgust GWLs) and risk beliefs either. It should be noted that the risk belief index mean was fairly high with restricted variance, suggesting a ceiling effect. Evoking negative emotions without mechanisms that allow emotions to influence cognitions and behaviors may be counterproductive. Future research should consider disgust’s influence on additional cognitions such as increasing recall or reducing curiosity about smoking.

Limitations and Future Research

These findings should be interpreted in light of several limitations. Due to the secondary nature of these analyses, the stimuli were not designed to exclude potential confounds such as word count variation between disgust and non-disgust GWLs. Although our analyses control for this confound, they do not completely rule out the influence of message complexity on attention. Future research should use tightly controlled stimuli holding message complexity constant or manipulating text saliency to examine its influence on responses to disgust imagery versus text. The current study cannot determine whether defensive processing, processing efficiency, or some other factor accounts for shifts in attention to image versus text for disgust GWLs. Future studies should consider the motivation driving this bias and its implication for message outcomes.

In this study, viewing time was fixed, which may have influenced how participants allocated time to images, text and branding. Attentional patterns may differ when participants control exposure duration. Time to first fixation, which theoretically indicates prioritizing an object for further processing, did not appear to be a meaningful attention measure. Future work should consider additional contexts where faster or slower orienting has practical implications. Our design does not capture changes in attention as participants view more stimuli; future research should consider the potential for habituation to disgust warnings and how it impacts self-directed attention, cognitions and behavior. A between-subjects design would also address limitations of interpreting causation from the attention-emotion correlations in this study.

Previous research identified a relationship between disgust exposure and risk beliefs among adolescents when messages featured younger individuals.² Such findings may also be due to using more vivid imagery than those in the FDA-proposed warnings. Future research should examine adolescent attention to highly vivid messages portraying youth. Given the significant differences in responses across participants, additional individual characteristics should be considered as moderators. For instance, only a few participants reported prior smoking behavior. Research involving more young smokers is important.¹ Further, current findings relate to early adolescence and may not generalize to older youth. Older youth often have more positive smoking perceptions and are more likely to be influenced by their peers smoking behavior in addition to smoking behavior modeled by relatives.²⁵ Older and younger adolescents differ in reasoning and emotion regulation skills that may influence reactions to disgust-eliciting anti-smoking messages³⁹.

Conclusion

Prior research has found a connection between exposure to disgust and adolescents’ perceptions of smoking.² The present study goes beyond exposure, providing evidence that adolescents attend to disgust imagery, which in turn influences their emotional responses to GWLs. The findings also raise additional questions with implications for GWL policy considerations in the United States. For instance, evidence that cigarette brands draw attention regardless of the type of GWL used may point to policies, like plain packaging, that have proven effective in other countries.¹² Future studies and policy decisions should consider the impact of message complexity, processing motivation, and individual differences such as developmental stage on additional outcomes related to attention for disgust-evoking GWLs.

Implications and Contributions:

The current study provides evidence that adolescents attend to disgust imagery, which in turn may influence emotional responses to warning labels. These findings add evidence that youth process graphic warning labels in ways consistent with dimensional theories of emotion but in contrast to discrete emotion theories.

Abbreviations

AFD

Average Fixation Duration

PFD

Proportion Fixation Durationa

TFF

Time to First Fixation

AOI

Area of Interest

GWL

Graphic Warning Label