Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2024 Apr 1.
Published in final edited form as: Psychol Health. 2021 Aug 27;38(4):409–428. doi: 10.1080/08870446.2021.1969021

Expanding the Analysis of Mechanisms of Action in Behavioral Interventions: Cognitive Change Versus Cognitive Activation

Aya Avishai a,*, Noel T Brewer b,c, Jennifer R Mendel c, Paschal Sheeran a,c
PMCID: PMC8904159  NIHMSID: NIHMS1761143  PMID: 34445902

Abstract

Objective:

To understand the mechanisms of action underlying behavioral interventions, researchers typically examine whether the treatment changes cognitions and whether changes in cognition predict behavior (cognitive change). This current research explores an alternative mechanism whereby the intervention increases the impact of pre-existing cognitions on behavior (cognitive activation). We tested whether cognitive change or cognitive activation explains the impact of cigarette pack messages on smoking restraint.

Design:

The research comprised a validation experiment (N = 135) and a 4-week RCT (N = 719) with smokers.

Main Outcome Measures:

At both baseline and follow-up of the RCT, smokers self-reported threat appraisals, coping appraisals, and smoking restraint.

Results:

Intervention messages heightened the accessibility of threat appraisals compared to control messages (validation experiment). In the RCT, smoking restraint increased among intervention participants but not controls. Trial arm showed no corresponding change in threat or coping appraisals. However, trial arm interacted with baseline health cognitions such that synergies between threat appraisal components, and between threat appraisals and coping appraisals, predicted smoking restraint for intervention participants but not for controls.

Conclusion:

Our findings support a cognitive activation process whereby health messages on cigarette packs increase the impact of pre-existing threat appraisals on smoking restraint.

Keywords: mechanisms of action, butting out, risk appraisal, coping appraisal, risk communication

The present research analyzes mechanisms of action in a randomized clinical trial (RCT) of health messages on cigarette packs (Brewer et al., 2019a). The trial successfully increased smoking restraint (prematurely butting out cigarettes to smoke less), a measure that is positively related to quit attempts (Borland et al., 2010; Li et al., 2015; Partos et al., 2014). The standard strategy adopted in smoking trials and other health behavior interventions is to attempt to change cognitions about (a) a focal disease or illness, or (b) the recommended health behavior (Sheeran, Klein, & Rothman, 2017). The assumption is that cognitive change is the theoretically specified process (i.e., the mechanism of action) by which interventions promote behavior change (Sheeran et al., 2017; Michie et al., 2018). For instance, protection motivation theory (PMT; Rogers, 1975) and the extended parallel process model (EPPM; Witte, 1992) specify that interventions should seek to change threat appraisals (i.e., risk perceptions, fear, perceived severity) and coping appraisals (i.e., self-efficacy, response efficacy, response costs). Evidence indicates that changing threat appraisals and coping appraisals indeed leads to changes in health-related intentions and behavior (Sheeran, Harris, & Epton, 2014). An alternative approach, that may be especially relevant in the context of cigarette pack messages that smokers view every time they reach for a cigarette, is that the messages activate (i.e., increase the accessibility of) smokers’ pre-existing threat and coping appraisals. Our research offers the first tests of cognitive change versus cognitive activation as the mechanism underlying the impact of pack messages on smoking behavior.

Cognitive Change

Previous experimental research offers support for the role of cognitive change in health behavior change (Sheeran et al., 2014; but see Kok et al., 2018, and associated commentaries for discussion). In a meta-analysis of 217 intervention studies, increasing threat appraisals had positive effects on intentions (d+ = .31) and behavior (d+ = .23) (Sheeran et al., 2014). However, the main effects of threat appraisal variables were qualified by interactions both among the components of threat appraisal and between threat and coping appraisals. For instance, interventions that increased perceived severity and risk perceptions had larger effects on behavior than interventions that increased risk perception but did not increase perceived severity (d+ = .36 vs. .16). Similarly, increasing perceived severity amplified the impact of heightening fear on behavioral intentions (d+ = .39 vs. .21). Thus, the beneficial impact of changing threat appraisals accrues in part from two-way interactions among risk perception, fear, and perceived severity.

Heightening threat appraisal also had greater impact on intentions and behavior when coping appraisals were simultaneously increased (Sheeran et al., 2014; Tannenbaum et al., 2015). For instance, interventions that increased perceived severity led to stronger behavioral intentions if the intervention was successful in augmenting response efficacy than if it was not (d+ = .70 vs. .22). Relatedly, interventions that increased perceived risk and reduced response costs had a larger effect than interventions that increased perceived risk but had no influence on response costs (d+ = .88 vs. .35). These findings provide support for cognitive change as the process by which threat and coping appraisals lead to intention and behavior change. The findings also indicate, however, that direct effects of threat appraisal may be modest and that interactions among threat appraisal variables and between threat and coping appraisals warrant testing.

In the context of smoking, the cognitive change hypothesis has, at best, mixed support. Meta-analysis indicates that among interventions that successfully change risk perceptions, effects on smoking behavior are small (d+ = .11; Sheeran et al., 2014). Reviews by Noar and colleagues (Noar, Francis, Bridges, Sontag, Ribisl, & Brewer, 2016a; Noar, Hall, Francis, Ribisl, Pepper, & Brewer, 2016b) observed that pictorial warnings did not change risk perceptions concerning smoking, and a recent large-scale trial of pictorial warnings on cigarette packs found no effect on threat appraisals (Brewer, Parada, Hall, Boynton, Noar, & Ribisl, 2019b). These studies found changes in other health cognitions including attention to and thinking about the warnings. Thus, there is doubt concerning the role of cognitive change (increased threat and coping appraisal) in modifying smoking behavior.

Cognitive Activation

An alternative analysis of the impact of interventions on health behavior is that threat and coping cognitions are activated rather than changed. If individuals already appraise a disease as threatening and believe that they can cope effectively with that threat, then it should no longer be necessary to increase threat and coping appraisal. Rather, ensuring that pre-existing appraisals become salient at the critical juncture could serve to promote behavior change. Accessibility theory (Higgins, 2011) proposes that the frequency and recency with which concepts are activated determines their impact on decisions and behavior. Health messages on cigarette packs should heighten the accessibility of threat and coping appraisals as smokers encounter a message each time they reach for a cigarette.

One argument for why concept accessibility matters for behavioral performance comes from Fazio’s (Fazio, 1990b) attitude accessibility model. According to this model, how well attitudes predict behavior depends not only on the valence of an attitude (how favorable/unfavorable is the person’s evaluation of the behavior), but also on its accessibility (the activation level of the evaluation in working memory). When an attitude towards a behavior is both favorable and accessible, then behavioral performance is likely. On the other hand, when the attitude is unfavorable or accessibility is low, then performance is much less likely. Studies have shown that repeatedly expressing an attitude increases attitude accessibility and improves the consistency between attitudes and behavior (Fazio, Chen, McDonel, & Sherman, 1982; Glasman & Albarracín, 2006). Similarly, studies that measured accessibility using response latencies to attitude questions observed moderation of the attitude-behavior relationship such that greater accessibility led to stronger associations between attitudes and behavior (Fazio & Williams, 1986; Cooke & Sheeran, 2004 for meta-analysis). In the present research, we apply Fazio’s analysis of attitude accessibility to the concept of threat appraisal. The idea is that threat appraisals can vary in valence and accessibility in the same way as attitudes, and repeated exposure to messages on cigarette packs could increase the accessibility of threat appraisals.

The Present Research

We conducted two studies to test our hypotheses. In a validation experiment, we examined the effects of smoking-related messages on both the accessibility and valence of threat appraisals as assessed through self-report and reaction time. We anticipated that the messages would increase the accessibility of threat appraisals but have no effect on the valence of threat appraisals or item readability. Our second study was a 4-week randomized clinical trial (RCT) that explored whether cognitive change or cognitive activation explained the relationship between these pack messages and smoking restraint.

Validation Experiment

We conducted the validation experiment in response to feedback on our initial submission, and after we had conducted the RCT.

Method

Participants and Procedure

We recruited current smokers via Amazon Mechanical Turk and paid them more than the US minimum wage for participation. We randomized participants to either the control group (n = 66) or the intervention group (n = 69). Participants viewed and responded to two practice messages before beginning the main experiment. Participants randomized to the control group viewed three messages regarding not littering cigarette butts (e.g., ‘Please refrain from littering. Cigarette butts are the most littered item.’), and those randomized to the intervention group viewed three messages about the chemicals in smoke and resulting health harms (e.g., ‘Cigarette smoke contains uranium. This causes lung tumors and kidney damage.’). The order of presentation of messages within group was random. After each message, threat accessibility items or readability items appeared, one at a time in a randomized order. Next, participants completed valence of threat appraisal scales and then demographic and attention check items. Data analyses included only participants who answered at least one of the two attention check items correctly. To ensure that randomization was successful, we examined demographic variables (e.g., age, gender, race, education) and found that these characteristics did not significantly differ between the groups (all ps > .05). The University of North Carolina institutional review board approved both studies.

Measures

Smoking status.

Participants were included if they selected the response, “I currently smoke cigarettes.” We confirmed smoking status with items from the PhenX toolkit (Hamilton et al., 2011) (e.g., “Have you smoked at least 100 cigarettes in your lifetime?”, “Do you now smoke cigarettes every day, some days, or not at all?”).

Accessibility of threat appraisals and readability.

The accessibility of threat appraisal items were ‘This message brings the danger of smoking to mind’ and ‘This message makes me think of smoking-related health problems’ (Windschitl, 2003; Cooke & Sheeran, 2013) and had 5-point response scales ranging from Definitely no to Definitely yes (α = .96). Readability of each message was assessed with the item, ‘This message is clear and easy to read’ on a 5-point response scale ranging from Definitely no to Definitely yes (α = .80). Reaction times to each of the threat accessibility and readability items were measured in milliseconds from the presentation of the items to the time that participants responded, (see Fazio, 1990a). We averaged the reaction times to form a response latency index of accessibility of threat appraisals.

Valence of threat appraisals.

The survey assessed the valence of threat appraisals using a multidimensional model of risk perception, the TRIRISK model (Ferrer, Klein, Persoskie, Avishai-Yitshak, & Sheeran, 2016; Ferrer, Klein, Avishai, Jones, Villegas, & Sheeran, 2018). The TRIRISK model makes the traditional distinction between deliberative risk perceptions (likelihood judgments) and affective risk perceptions (fear or worry), but also establishes experiential risk perception (‘gut-level’ reactions to health risks) as a third component of risk perception. The survey had the 3 items for each risk component developed by Ferrer et al. (2016; 2018) accompanied by 7-point response scales. For affective risk perception, the survey included the items, ‘When you think about smoking-related health problems for a moment, to what extent do you feel _______?’ (1) ‘fearful’, (2) ‘worried’, and (3) ‘anxious’; all rated Not at all – Extremely. Deliberative risk perception items were: (1) ‘How likely is it that you will have smoking-related health problems at some point in the future?’ Unlikely – Likely, (2) ‘How do you think your chance of developing smoking-related health problems in the future compares to the average person of your gender and age?’ Much lower – Much higher, and (3) ‘The way I look after my health means that my odds of getting smoking-related health problems in the future are…?’ Very low – Very high. The experiential risk perception items were, (1) ‘How concerned are you about developing smoking-related health problems in your lifetime?’ Not at all – Extremely, (2) ‘How easy is it for you to imagine yourself developing smoking-related health problems in the future?’ Not at all easy – Extremely easy, and (3) ‘I feel very vulnerable to smoking-related health problems’ Strongly disagree – Strongly agree. Reliability values for the affective, deliberative, and experiential subscales were high (α = .93, .81, and .88, respectively)

Results

Self-reported accessibility of threat appraisals was higher in the intervention group (M = 3.60, SD = .52) than the control group (M = 2.13, SD = .72), t(117.4) = 13.51, p < .001, d = 2.33. Additionally, responses to these items were faster for the intervention group (M = 4323.47 ms, SD = 2174.77) than the control group (M = 5767.85 ms, SD = 2213.69), t(131) = 3.80, p < .001, d = −.66. Differences remained significant when outliers were removed.

As expected, readability ratings did not significantly differ between the intervention group (M = 3.72, SD = .41) and the control group (M = 3.66, SD = .45), t(132) = .79, p = .43. Additionally, speed of completing the readability items did not significantly differ between the intervention (M = 4035.18 ms, SD = 1972.46) and control groups (M = 4381.67 ms, SD = 1733.11), t(128) = 1.06, p = .29.

The valence of threat appraisals did not significantly differ between the intervention group and the control group (affective, deliberative, and experiential risk perceptions: Ms = 5.10, 5.38, and 5.38. respectively, for the intervention group; Ms = 4.78, 5.25, and 5.12, respectively, for the control group; SDs = 1.70, 1.03, 1.40, 1.51, 1.11, and 1.05, respectively), ts < 1.16, ps > .24.

Discussion

Findings supported the hypothesis that messages about chemicals in cigarette smoke and resultant health risks increase the accessibility of threat appraisals among smokers. These findings were observed for both the self-report and response latency measures of accessibility. The intervention messages did not alter the valence of threat appraisals. Thus, the messages lead to heightened activation of threat appraisals but do not alter beliefs about the threat that smoking poses to health. Importantly, the groups did not significantly differ on the control variable, readability, which helps to rule out social desirability bias or experimenter demand as explanations of the findings for accessibility. The validation experiment thus indicates that the messages used in the RCT influenced the accessibility of threat appraisals.

Randomized Clinical Trial

The research question addressed in the present RCT is, what mechanism of action explains how health messages on cigarette packs promote smoking restraint? The traditional approach to testing theoretically specified processes involves a mediation model wherein the health message generates cognitive change that, in turn, leads to behavior change (Judd & Kenny, 1981). Support for this cognitive change account will be obtained if intervention participants demonstrate higher scores on threat and coping appraisals at follow-up compared to controls, and these appraisals are related to behavior change. The alternative approach tested here is a moderation model wherein the health messages serve to increase the impact of participants’ pre-existing threat appraisals on behavior change. Because the RCT was conducted in a field setting, we were unable to obtain reaction time indices of the accessibility of threat appraisals. However, we anticipated that repeated exposure to the health messages in the intervention arm would increase the accessibility of threat appraisals which, in turn, should mean that threat appraisals become stronger predictors of smoking restraint in this arm compared to the control arm. Accordingly, and in line with the Testing a Process Hypothesis by an Interaction Strategy (TPIS; Jacoby & Sassenberg, 2011), we tested the accessibility hypothesis via the interaction between trial arm and threat appraisals in the RCT.

Support for the cognitive activation account will be obtained if study arm interacts with baseline threat and coping appraisals. Given the synergy observed among components of threat appraisals and between threat appraisals and coping appraisals in promoting health behaviors in previous reviews (Sheeran et al., 2014; Tannenbaum et al., 2015), we anticipated three-way interactions (i.e., arm × threat appraisal variable × threat appraisal variable, arm × threat appraisal variable × coping appraisal variable) for the prediction of smoking restraint. In particular, we predicted that smoking restraint would be greatest in the intervention arm when multiple components of threat appraisal were high and when threat appraisals and coping appraisals both were high.

Method

This manuscript analyzes data from a larger RCT; only relevant measures are reported here (see Brewer et al., 2019a, for an overview of the entire trial including inclusion/exclusion criteria and measures). The trial registration can be found at ClinicalTrials.gov identifier: NCT02785484.

Participants and Procedure

Participants (N = 719) were adults (21+ years of age) in California who were current cigarette smokers in 2016-2017. Randomization allocated 359 participants to the control arm of the trial and 360 to the intervention arm. To check whether randomization was successful, we confirmed that demographic characteristics did not significantly differ between trial arms. At baseline, participants in the control arm smoked an average of 9.97 cigarettes per day (SD = 12.2) and participants in the intervention arm smoked an average of 11.62 cigarettes per day (SD = 16.9). Mean participant age was 42.8 years (SD = 13.6) in the control arm and 42.1 years (SD = 13.2) in the intervention arm. Additional participant characteristics are available elsewhere (Brewer et al., 2019a).

The trial was comprised of 5 visits, each one week apart. Participants brought with them their own cigarette packs and study staff placed control or intervention messages on all participants’ cigarette packs for 3 weeks (placed during visits 2-4). Participants in the control arm of the trial received messages about not littering cigarette butts (e.g., ‘Please refrain from littering. Cigarette butts are the most littered item.’), whereas participants in the intervention arm received messages about the toxic chemicals in cigarettes and consequent health harms (e.g., ‘Cigarette smoke contains uranium. This causes lung tumors and kidney damage.’). Labels with the randomized messages were placed on the sides of participants’ cigarette packs (i.e., on the side opposite from the Surgeon General’s warning). Participants in both arms received a new label, in random order, at each visit for a total of three new messages during the trial. Participants completed surveys at each weekly visit and received up to $300 at the conclusion of the trial.

Measures

We refer to trial visit 1 as ‘baseline’ and trial visit 5 as ‘follow-up,’ as these are our trial visits of interest. The measures described below were obtained at both baseline and follow-up.

Threat appraisal variables.

The survey included the same items as the validation experiment to measure the TRIRISK variables. Affective, deliberative, and experiential subscale reliability values were high at baseline (α = .93, .86, and .83, respectively) and follow-up (α = .93, .88, and .85, respectively). Perceived severity of smoking behavior was measured by the following items: ‘How much would getting _______ because of smoking affect your life?’ (1) ‘lung tumors’, (2) ‘throat cancer’, (3) ‘kidney damage’, and (4) ‘heart damage’. These four health harms came from the intervention messages (e.g., ‘Cigarette smoke contains uranium. This causes lung tumors and kidney damage.’). The perceived severity items had a 5-point response scale ranging from Not at all to Very much at baseline (α = .95) and follow-up (α = .96).

Coping appraisal variables.

Response efficacy items (Brewer et al., 2016) reflected the same health risks listed in the perceived severity items: ‘How much would quitting smoking lower your chances of getting…’ (1) ‘lung tumors?’, (2) ‘throat cancer?’, (3) ‘kidney damage?’, and (4) ‘heart damage?’ rated on a 5-point response scale that ranged from Not at all to Very much at baseline (α = .93) and follow-up (α = .95). Five survey items assessed self-efficacy (Armitage, 2007): (1) ‘I believe I have the ability to quit smoking in the next 2 months if I wanted to’, (2) ‘I see myself as being capable of quitting smoking in the next 2 months if I wanted to’, (3) ‘I feel I have personal control over quitting smoking in the next 2 months if I wanted to’, (4) ‘If I wanted to quit smoking in the next 2 months, it would be difficult’ (reverse scored), and (5) ‘I am confident that I would be able to quit smoking in the next 2 months if I wanted to’. Self-efficacy was rated on a 5-point response scale that ranged from Strongly disagree to Strongly agree at baseline (α = .85) and follow-up (α = .87). The survey assessed response costs (McKee, O'Malley, Salovey, Krishnan-Sarin, & Mazure, 2005) using the following four items: ‘If you were to stop smoking, how likely would you be to…’ (1) ‘gain weight?’, (2) ‘be more irritable?’, (3) ‘miss the pleasure you get from cigarettes?’, and (4) ‘experience intense cravings for a cigarette?’ These response cost items were rated on a 5-point response scale that ranged from Not at all likely to Extremely likely at baseline (α = .77) and follow-up (α = .82).

Smoking restraint.

The surveys assessed smoking restraint with the item, ‘In the last week, how often have you butted out a cigarette before you finished it because you wanted to smoke less?’ Response options were Never, 1-2 times, 3-4 times, 5-9 times, and 10 or more times. We coded participants who did not respond to the item as ‘Never’, and this was done for 5 participants (0.7%) at baseline and 46 participants (6.4%) at follow-up.

Data Analysis Plan

We first determined whether smoking restraint increased during the trial after participants were exposed to cigarette pack messages using a two-way repeated measures ANOVA of arm (control vs. intervention) and time (smoking restraint at baseline vs. follow-up). If we found an interaction, we examined simple main effects.

Next, we tested whether the effect of intervention was due to cognitive change or cognitive activation. To test cognitive change, we again used two-way repeated measures ANOVAs to examine the effects of intervention on threat and coping appraisals. To test cognitive activation, we used moderated linear regression analyses to test 3-way interactions among arm, baseline threat appraisals, and baseline coping appraisals in predicting smoking restraint. We examined six 3-way interactions between arm and baseline threat appraisal variables (deliberative, affective, and experiential risk perceptions, and perceived severity), and examined twelve 3-way interactions between baseline threat appraisals and baseline coping appraisals (response efficacy, self-efficacy, and response costs). To probe interactions, we tested the simple slopes (M ± 1SD) for baseline threat and coping appraisal variables separately by arm. We report simple slopes as unstandardized regression coefficients (Bs; Aiken, West, & Reno, 1991). We supplemented this analysis by testing simple slopes for arm for relevant combinations of threat and coping variables. Analyses used two-tailed tests and a critical alpha of .05 in IBM SPSS Version 25.0 (IBM Corp., 2017).

Results

Smoking Restraint

The main effects of trial arm (control vs. intervention), F(1, 717) = 3.26, p = .07, d = .12, and time (baseline vs. follow-up), F(1, 717) = 0.16, p = .69, d = .02, on smoking restraint were qualified by the predicted interaction between trial arm and time, F(1, 717) = 6.78, p < .01, d = .19. The arms did not significantly differ on smoking restraint at baseline, F(1, 717) = 0.04, p = .84, d = .01. By follow-up, smoking restraint was higher in the intervention arm (M = 3.36, SD = 3.33) than the control arm (M = 2.67, SD = 2.96), F(1, 717) = 8.66, p < .01, d = .22. Smoking restraint increased over time in the intervention arm, p = .03, d = .11, but did not change in the control arm, p = .12, d = −.09 (see Figure 1 and Figure S1 in the Supplemental Materials). These findings indicate that health messages on cigarette packs increased smoking restraint.

Figure 1.

Figure 1.

Open in a new tab

Mean (Standard Error) Smoking Restraint by Arm and Time in the RCT.

The Role of Cognitive Change

Trial arm did not interact with time, all Fs < 3.20, ps > .07, ds < .15 (see Table 1), indicating that there were no changes in threat or coping appraisal as function of the health messages. Thus, the increase in smoking restraint observed over the course of the trial cannot be explained in terms of cognitive change.

Table 1.

Mean (Standard Deviation) Threat and Coping Appraisals in the RCT

Arm: Control
 Intervention
 Arm x Time
Time: Baseline Follow-Up Baseline Follow-Up
Threat appraisal
Affective risk perception 4.38 (1.63) 4.45 (1.80) 4.31 (1.66) 4.52 (1.68) F(1, 700) = 1.58,
p = .21
Deliberative risk perception 4.73 (1.50) 4.53 (1.60) 4.75 (1.41) 4.72 (1.49) F(1, 699) = 2.89,
p = .09
Experiential risk perception 4.60 (1.57) 4.49 (1.65) 4.51 (1.54) 4.59 (1.59) F(1, 699) = 3.19,
p = .07
Perceived severity perception 4.33 (0.95) 4.26 (1.05) 4.25 (0.99) 4.24 (0.97) F(1, 701) = 0.58,
p = .45
Coping appraisal
Response efficacy 3.99 (1.00) 4.03 (1.03) 4.04 (0.94) 4.02 (1.00) F(1, 700) = 0.90,
p = .34
Self-efficacy 3.42 (0.98) 3.56 (1.07) 3.42 (0.92) 3.57 (0.90) F(1, 701) = 0.00,
p = .97
Response costs 3.39 (0.97) 3.21 (1.06) 3.39 (0.93) 3.23 (1.05) F(1, 702) = 0.03,
p = .87
Open in a new tab

Note. Means did not significantly differ between control and intervention arms at baseline or follow-up.

The Role of Cognitive Activation

We observed three interactions of trial arm and baseline threat and coping appraisals that significantly predicted smoking restraint at follow-up (Table 2). First, we found an interaction between arm, affective risk perception, and perceived severity, p = .04 (Figure 2, top panel). Affective risk perception was strongly associated with smoking restraint at follow-up when participants had both high perceived severity and were exposed to health messages on their cigarette packs (B = 0.61, p < .001). Simple slopes for intervention participants with low perceived severity (B = 0.32, p = .01) and for control participants with high perceived severity (B = 0.10, p = .45) or low perceived severity (B = 0.35, p = .02) indicated that affective risk perceptions were weaker predictors, or were not predictive, of smoking restraint for these groups.

Table 2.

Three-Way Interactions Predicting Smoking Restraint at Follow-Up in the RCT

IV: Affective RP Deliberative RP Experiential RP Deliberative RP Response Efficacy
Moderator: Severity Severity Severity Response Costs Severity
B SE B SE B SE B SE B SE
Intercept 2.73 .17 *** 2.78 .18 *** 2.71 .18 *** 2.72 .17 *** 2.70 .18 ***
IV .23 .11 * .03 .12 .12 .11 .01 .12 .12 .18
Moderator −.05 .19 −.08 .24 .00 .22 .16 .18 .05 .21
Arm .60 .24 * .52 .25 * .56 .24 * .55 .24 * .55 .25 *
IV x Moderator −.13 .10 −.16 .11 −.07 .11 −.11 .11 −.08 .16
IV x Arm .24 .15 .13 .18 .22 .16 .28 .17 .28 .26
Moderator x Arm .26 .26 .47 .31 .40 .29 −.36 .26 .29 .27
IV x Moderator x Arm .28 .14 * .33 .15 * .38 .15 ** .32 .16 * .47 .21 *
Model Statistics F(7, 710) = 6.04,
p < .001,
R2 = .06		 F(7, 710) = 2.77,
p < .01,
R2 = .03		 F(7, 710) = 4.83,
p < .001,
R2 = .05		 F(7, 710) = 2.61,
p = .01,
R2 = .03		 F(7, 711) = 3.42,
p = .001,
R2 = .03
Open in a new tab

Note. RP = risk perception. B = unstandardized regression coefficient. SE = standard error of the estimate. Control arm was coded as 0, and experimental arm was coded as 1. Nine other 3-way interactions that were not statistically significant appear in Tables S1 and S2.

*

p < .05

**

p < .01

***

p < .001.

Figure 2.

Figure 2.

Open in a new tab

Predicted Simple Slopes for Baseline TRIRISK Components × Baseline Perceived Severity × Trial Arm Predicting Smoking Restraint at Follow-up in the RCT.

Second, trial arm, deliberative risk perceptions, and perceived severity interacted, p = .03 (Figure 2, middle panel), following the same pattern as affective risk perceptions. Deliberative risk perceptions predicted smoking restraint when participants were in the intervention arm and had high perceived severity scores (B = 0.33, p = .05). All other simple slopes were not different from zero (intervention/low perceived severity: B ~ 0, p = .99; control/high perceived severity: B = −0.12, p = .43; control/low perceived severity: B = 0.19, p = .26).

Third, we found a three-way interaction of arm, experiential risk perception, and perceived severity, p = .01 (Figure 2, bottom panel). Baseline experiential risk perceptions were associated with smoking restraint when perceived severity was high and participants were in the intervention arm (B = 0.64, p < .001). Experiential risk perceptions did not predict smoking restraint for intervention participants with low perceived severity (B = 0.03, p = .84), control participants with high perceived severity (B = 0.05, p = .74), or control participants with low perceived severity (B = 0.18, p = .25). (Findings for interactions that were not statistically significant appear in Tables S1 and S2 in the Supplemental Materials.)

Next, we examined the 3-way interactions between arm, baseline threat appraisal variables, and baseline coping appraisal variables. Two interactions were statistically significant (Table 2): arm × deliberative risk perception × response costs, p = .04 (Figure 3, top panel), and arm × perceived severity × response efficacy, p = .03 (Figure 3, bottom panel). Deliberative risk perceptions predicted smoking restraint for participants in the intervention arm when response costs were high (B = 0.48, p = .007) but not when response costs were low (B = 0.09, p = .57). For control participants, deliberative risk perceptions did not predict smoking restraint whether response costs were high (B = −0.10, p = .54) or low (B = 0.11, p = .43).

Figure 3.

Figure 3.

Open in a new tab

Predicted Simple Slopes for Baseline Threat Appraisals × Baseline Coping Appraisals × Trial Arm Predicting Smoking Restraint at Follow-up in the RCT.

Perceived severity predicted smoking restraint when response efficacy was high and participants were exposed to health messages on cigarette packs (B = 0.78, p = .004). The other simple slopes did not differ from zero (intervention/low perceived severity: B = 0.03, p = .87; control/high perceived severity: B = 0.04, p = .86; control/low perceived severity: B = 0.20, p = .37). These findings are generally consistent with the cognitive activation analyses of the impact of pack health messages on smoking restraint: Threat appraisals better predicted restraint when participants were repeatedly exposed to messages on their cigarette packs and participants held high perceptions of severity, high response costs, and high response efficacy.

To double-check this conclusion, we also decomposed the 3-way interactions treating arm (control vs. intervention) as the predictor and threat and coping appraisals as moderator variables. Simple slopes indicated that arm predicted smoking restraint when (a) affective risk perceptions and perceived severity both were high (B = 1.69, p < .001), (b) deliberative risk perceptions and perceived severity both were high (B = 1.63, p < .001) (c) experiential risk perceptions and perceived severity both were high (B = 1.89, p < .001), (d) deliberative risk perceptions and response costs were both high (B = 1.06, p < .05), and (e) perceived severity and response efficacy were both high (B = 1.55, p < .001). Thus, health messages on cigarette packs engendered smoking restraint predominantly when participants had high threat and coping appraisals at the outset of the trial.

Concerns About Multiple Testing

Readers might be concerned that the significant three-way interactions observed here capitalize on chance. We take this concern seriously and attempted to resolve the issue empirically. Messages about chemical constituents and health harms of cigarettes clearly target threat appraisals so analyses focused on interactions between study arm and both threat appraisal × threat appraisal, and threat appraisal × coping appraisals. A total of 18 three-way interactions were tested, of which 5 proved statistically significant (28%). A binomial test indicated that this significance ratio is well above chance (z = 3.89, p = .0014, 2-tailed) and obviates concerns about capitalization on chance.

General Discussion

We tested two hypotheses about the mechanisms underlying the impact of cigarette pack health messages on smoking restraint. The cognitive change hypothesis postulates that messages change health cognitions (i.e., increase threat and coping appraisals), thereby influencing smoking restraint. This is the traditional approach used to analyze mechanisms of action in behavioral interventions (Judd & Kenny, 1981). The second hypothesis, cognitive activation, was tested here for the first time. According to this analysis, messages on cigarette packs increase the accessibility of threat appraisals and so change the relationship between ratings of threat appraisal (valence) and smoking restraint; repeatedly viewing cigarette pack health messages activates cognitions and renders threat appraisals more powerful predictors of restraint.

In the validation experiment, we obtained evidence that messages about chemicals in cigarette smoke increase the accessibility of threat appraisals, according to both self-report and response latency indices of accessibility. The same messages were used in the RCT and findings showed that smoking restraint increased in the intervention arm over the course of the trial, but not in the control arm. Health messages on packs did not affect threat and coping appraisal scores, indicating that change in these appraisals could not explain increased smoking restraint in the intervention arm – such cognitive change did not occur. Instead, we observed support for the cognitive activation hypothesis. We found five 3-way interactions involving arm, threat appraisal variables, and coping appraisal variables. Exposure to health messages on cigarette packs improved the predictive validity of (a) deliberative, affective, and experiential risk perceptions when perceived severity was high, (b) deliberative risk perceptions when response costs were high, and (c) perceived severity when response efficacy was high. Looked at differently, pack message exposure generally led to greater smoking restraint when participants’ deliberative, affective, and experiential risk perceptions, perceived severity, response costs, and response efficacy each were high at the outset of the trial. Activating threat and coping appraisals via health messages on cigarette packs could promote smoking restraint if participants already had high appraisals of threat and coping to begin with.

Health messages on cigarette packs failed to change threat appraisals in the present study. One potential explanation is that simple declarative statements about the chemicals in cigarette smoke and resultant health harms are not sufficiently compelling to change perceptions of risk and severity. Evidence indicates that it is difficult to change threat appraisals. For instance, a recent meta-analysis located 208 studies that obtained significant intervention effects on threat appraisals but identified a further 133 studies wherein threat appraisals were targeted, but were not changed, by the intervention (Sheeran et al., 2014). Thus, the null effects of trial arm on risk perceptions or perceived severity observed here are not unusual given that ~2 in 5 of (often highly intensive) interventions fail to increase threat appraisals. Furthermore, meta-analyses of the impact of cigarette pack warnings shows that they do not change risk perceptions (Noar et al., 2016a; Noar et al., 2016b). The implication is that a concerted program of research is needed to identify the content, format, and mode of delivery of messages that can increase risk perceptions, fear, and perceived severity among smokers with low risk appraisals.

The present findings are consistent with a new theoretical framework for understanding the impact of cigarette pack warnings, the Tobacco Warnings Model (TWM; Brewer et al., 2019a). According to the TWM, tobacco warnings do not exert their influence by changing threat appraisals. Rather, the model posits that attention paid to warnings elicits negative affect and social interactions (i.e., conversations about the warnings) that in turn lead people to think about the warnings (i.e., cognitive elaboration of the messages). Thinking about the warnings is construed as the proximal determinant of quit intentions and behavior change (smoking restraint, quit attempts, and smoking cessation).

An innovation of the present research may be to offer insights into how cognitive elaboration of health messages on cigarette packs gets transformed into smoking restraint. In particular, thinking about warnings could serve to activate smoker’s pre-existing risk perceptions or perceptions of the severity of tobacco use. We found no evidence of cognitive change (increased threat appraisals) in the present trial but observed that repeated exposure to pack messages increased the impact of threat appraisals on behavior. Two notable patterns are present in the data. First, consistent with the TRIRISK analysis of risk perceptions (Ferrer et al., 2016; Ferrer et al., 2018) all three postulated risk components – deliberative, affective, and experiential – interacted with trial arm and other health cognitions to predict smoking restraint. Second, consistent with meta-analytic findings (Sheeran et al., 2014), we found interactions among the components of threat appraisal and between threat appraisals and coping appraisals for each of the significant interactions with trial arm. We did not find any two-way interactions between arm and threat appraisals. Instead, risk perceptions predicted smoking restraint only when participants were exposed to health messages on cigarette packs and had high perceived severity or high response costs, and perceived severity predicted restraint only when participants were exposed to messages and had high response efficacy. Thus, the pattern of cognitive activation observed in the present trial pertained to all three types of risk perception and to particular combinations of (a) threat appraisals, and (b) threat and coping appraisals.

It is notable that deliberative risk perceptions predicted smoking restraint when response costs were high. PMT and the EPPM both predict that risk perceptions should predominantly predict outcomes when response costs are low, not high. Additional research is needed to corroborate the findings obtained here. However, it is intriguing to note that when participants were repeatedly exposed to health messages on cigarette packs, deliberative risk perceptions became powerful predictors of smoking restraint in the face of challenges such as irritability, weight gain, and intense cravings, and not when response costs were low (see Drach-Zahavy & Erez, 2002 for discussion).

The present findings also have implications for efforts to understand the mechanisms underlying successful health behavior change. The traditional approach to analyzing mechanisms of action posits a mediation model wherein the intervention changes cognitions, and cognitive change leads to behavior change (Judd & Kenny, 1981). The idea is that interventions ‘change people’s minds’ and thereby alter behavior. The innovation of the present research was to propose and test an alternative route to health behavior change – cognitive activation. This route derives from decades of research in social cognition (e.g., Higgins, 2011 and Fazio, 1990b) demonstrating that the behavioral impact of cognitions depends not only on valence (e.g., how high or low are appraisals of threat and coping) but also on accessibility (how activated are threat and coping appraisals; the extent to which such appraisals are at the ‘top of one’s mind’). We obtained support for a moderation model wherein exposure to health messages on cigarette packs promoted smoking restraint when participants had high threat and coping appraisals even before the trial began. The value of these findings resides in offering researchers another way to think about and analyze trial data. If a trial proves successful and cognitive change is not observed, it may be worthwhile to test whether cognitive activation could explain the findings. Cognitive activation and cognitive change may work hand-in-hand of course; an intervention could both change people’s minds and increase the accessibility of respective cognitions. In this instance, mediation and moderation might both be observed: the treatment changes health cognitions that, in turn, relate to behavior change and health cognitions have stronger associations with behavior change in the intervention as compared to the control arm. Considering cognitive activation as a route to behavior change should prove generative for intervention research as it opens a new avenue for understanding why interventions prove, or do not prove, effective.

The present study has several limitations that we should acknowledge. First, findings for smoking restraint come from a single trial, conducted in a single location in the USA over a relatively brief period of time (4 weeks). This is a first test and replication of the present findings would be valuable, for instance, to test whether habituation or maladaptive coping responses (e.g., denial) are observed with prolonged deployment of pack messages. Second, the trial used an outcome (self-reported smoking restraint) that was appropriate for testing the impact of health messages on cigarette packs. Smoking restraint is associated with quit attempts (Borland et al., 2010; Li et al., 2015; Partos et al., 2014). However, additional longer-term studies are needed to evaluate intervention effects on biochemically verified smoking cessation. Third, it was not possible in the RCT to obtain direct measures of the accessibility of threat appraisals though our validation experiment indicated that the RCT messages indeed increased the accessibility of threat appraisals. Further research using cognitive and neuroimaging paradigms (Cunningham & Zelazo, 2007; Fazio, 1990a) would be valuable to corroborate our findings and to rule out potential alternative explanations of our findings.

Notwithstanding these limitations, the present research offers a new analysis of the mechanisms underlying efficacious behavioral interventions and new insights into how health messages on cigarette packs promote behavior change. This is only the first comparison of cognitive change versus cognitive activation in explaining intervention effects. Further tests are warranted. The approach to analyzing mechanisms of action advocated here opens up several questions for future research: How much message exposure is needed to meaningfully increase the accessibility of threat appraisals? Do variations in message content lead to improved cognitive activation compared to repeated and consistent use of the same messages? Which behavior change techniques are liable to maximize both cognitive change and cognitive activation? These research questions can and should be addressed in future studies.

Supplementary Material

Supp 1

Acknowledgements

This work was supported by the National Cancer Institute and FDA Center for Tobacco Products (CTP) under grant number P50CA180907. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH or the Food and Drug Administration. The funding institutions had no role in the design and conduct of the study; collection, management, analysis and interpretation of the data; preparation, review or approval of the manuscript; and decision to submit the manuscript for publication.

Footnotes

Ethical Adherence

The authors confirm that the research was conducted in accordance with APA ethics guidelines. The University of North Carolina at Chapel Hill IRB approved the study.

Declaration of Interest Statement

NTB has served as a paid expert consultant in litigation against tobacco companies. The other authors declare no conflicts of interest.

Data Availability Statement

Due to our university’s requirements on grant-funded research, we can share the study data with a signed data use agreement. Investigators wishing to access the data may contact NTB.

References

  1. Aiken LS, West SG, & Reno RR (1991). Multiple regression: Testing and interpreting interactions. Sage. [Google Scholar]
  2. Armitage CJ (2007). Efficacy of a brief worksite intervention to reduce smoking: The roles of behavioral and implementation intentions. Journal of Occupational Health Psychology, 12(4), 376–390. [DOI] [PubMed] [Google Scholar]
  3. Borland R, Yong HH, Balmford J, Cooper J, Cummings KM, O'Connor RJ, … & Fong GT (2010). Motivational factors predict quit attempts but not maintenance of smoking cessation: Findings from the International Tobacco Control Four country project. Nicotine & Tobacco Research, 12(Suppl. 1), S4–S11. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Brewer NT, Hall MG, Noar SM, Parada H, Stein-Seroussi A, Bach LE, … & Ribisl KM (2016). Effect of pictorial cigarette pack warnings on changes in smoking behavior: A randomized clinical trial. JAMA Internal Medicine, 176(7), 905–912. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Brewer NT, Jeong M, Mendel JR, Hall MG, Zhang D, Parada H Jr, … & Ribisl KM (2019a). Cigarette pack messages about toxic chemicals: A randomised clinical trial. Tobacco Control, 28(1), 74–80. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Brewer NT, Parada H Jr, Hall MG, Boynton MH, Noar SM, & Ribisl KM (2019b). Understanding why pictorial cigarette pack warnings increase quit attempts. Annals of Behavioral Medicine, 53(3), 232–243. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Cooke R, & Sheeran P (2004). Moderation of cognition-intention and cognition-behaviour relations: A meta-analysis of properties of variables from the theory of planned behaviour. British Journal of Social Psychology, 43(2), 159–186. [DOI] [PubMed] [Google Scholar]
  8. Cooke R, & Sheeran P (2013). Properties of intention: Component structure and consequences for behavior, information processing, and resistance. Journal of Applied Social Psychology, 43(4), 749–760. [Google Scholar]
  9. Cunningham WA, & Zelazo PD (2007). Attitudes and evaluations: A social cognitive neuroscience perspective. Trends in Cognitive Sciences, 11(3), 97–104. [DOI] [PubMed] [Google Scholar]
  10. Drach-Zahavy A, & Erez M (2002). Challenge versus threat effects on the goal–performance relationship. Organizational Behavior and Human Decision Processes, 88(2), 667–682. [Google Scholar]
  11. Fazio RH (1990a). A practical guide to the use of response latency in social psychological research. Research Methods in Personality and Social Psychology, 11, 74–97. [Google Scholar]
  12. Fazio RH (1990b). Multiple processes by which attitudes guide behavior: The MODE model as an integrative framework. Advances in Experimental Social Psychology, 23, 75–109. [Google Scholar]
  13. Fazio RH, & Williams CJ (1986). Attitude accessibility as a moderator of the attitude–perception and attitude–behavior relations: An investigation of the 1984 presidential election. Journal of Personality and Social Psychology, 51(3), 505–514. [DOI] [PubMed] [Google Scholar]
  14. Fazio RH, Chen JM, McDonel EC, & Sherman SJ (1982). Attitude accessibility, attitude-behavior consistency, and the strength of the object-evaluation association. Journal of Experimental Social Psychology, 18(4), 339–357. [Google Scholar]
  15. Ferrer RA, Klein WM, Avishai A, Jones K, Villegas M, & Sheeran P (2018). When does risk perception predict protection motivation for health threats? A person-by-situation analysis. PloS ONE, 13(3), e0191994. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Ferrer RA, Klein WM, Persoskie A, Avishai-Yitshak A, & Sheeran P (2016). The tripartite model of risk perception (TRIRISK): Distinguishing deliberative, affective, and experiential components of perceived risk. Annals of Behavioral Medicine, 50(5), 653–663. [DOI] [PubMed] [Google Scholar]
  17. Glasman LR, & Albarracín D (2006). Forming attitudes that predict future behavior: A meta-analysis of the attitude-behavior relation. Psychological Bulletin, 132(5), 778–822. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Hamilton CM, Strader LC, Pratt JG, Maiese D, Hendershot T, Kwok RK, … & Nettles DS (2011). The PhenX Toolkit: Get the most from your measures. American Journal of Epidemiology, 174(3), 253–260. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Higgins ET (2011). Accessibility theory. In Van Lange PAM, Kruglanski AW, & Higgins ET (Eds.), Handbook of theories of social psychology (pp. 75–109). Sage. [Google Scholar]
  20. IBM Corp. (2017). IBM SPSS Statistics for Windows, Version 25.0 Armonk, NY: IBM Corp. [Google Scholar]
  21. Jacoby J, & Sassenberg K (2011). Interactions do not only tell us when, but can also tell us how: Testing process hypotheses by interaction. European Journal of Social Psychology, 41(2), 180–190. [Google Scholar]
  22. Judd CM, & Kenny DA (1981). Process analysis: Estimating mediation in treatment evaluations. Evaluation Review, 5(5), 602–619. [Google Scholar]
  23. Kok G, Peters GJY, Kessels LT, Ten Hoor GA, & Ruiter RA (2018). Ignoring theory and misinterpreting evidence: The false belief in fear appeals. Health Psychology Review, 12(2), 111–125. [DOI] [PubMed] [Google Scholar]
  24. Li L, Borland R, Fong GT, Jiang Y, Yang Y, Wang L, … & Thrasher JF (2015). Smoking-related thoughts and microbehaviours, and their predictive power for quitting. Tobacco Control, 24(4), 354–361. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. McKee SA, O'Malley SS, Salovey P, Krishnan-Sarin S, & Mazure CM (2005). Perceived risks and benefits of smoking cessation: Gender-specific predictors of motivation and treatment outcome. Addictive Behaviors, 30(3), 423–435. [DOI] [PubMed] [Google Scholar]
  26. Michie S, Carey RN, Johnston M, Rothman AJ, De Bruin M, Kelly MP, & Connell LE (2018). From theory-inspired to theory-based interventions: A protocol for developing and testing a methodology for linking behaviour change techniques to theoretical mechanisms of action. Annals of Behavioral Medicine, 52(6), 501–512. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Noar SM, Francis DB, Bridges C, Sontag JM, Ribisl KM, & Brewer NT (2016a). The impact of strengthening cigarette pack warnings: Systematic review of longitudinal observational studies. Social Science & Medicine, 164, 118–129. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Noar SM, Hall MG, Francis DB, Ribisl KM, Pepper JK, & Brewer NT (2016b). Pictorial cigarette pack warnings: A meta-analysis of experimental studies. Tobacco Control, 25(3), 341–354. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Partos TR, Borland R, Thrasher JF, Li L, Yong HH, O'Connor RJ, & Siahpush M (2014). The predictive utility of micro indicators of concern about smoking: findings from the International Tobacco Control Four Country study. Addictive Behaviors, 39(8), 1235–1242. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Rogers RW (1975). A protection motivation theory of fear appeals and attitude change. The Journal of Psychology, 91(1), 93–114. [DOI] [PubMed] [Google Scholar]
  31. Sheeran P, Harris PR, & Epton T (2014). Does heightening risk appraisals change people’s intentions and behavior? A meta-analysis of experimental studies. Psychological Bulletin, 140(2), 511. [DOI] [PubMed] [Google Scholar]
  32. Sheeran P, Klein WM, & Rothman AJ (2017). Health behavior change: Moving from observation to intervention. Annual Review of Psychology, 68, 573–600. [DOI] [PubMed] [Google Scholar]
  33. Tannenbaum MB, Hepler J, Zimmerman RS, Saul L, Jacobs S, Wilson K, & Albarracín D (2015). Appealing to fear: A meta-analysis of fear appeal effectiveness and theories. Psychological Bulletin, 141(6), 1178. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Windschitl PD (2003, February). Measuring and conceptualizing perceptions of vulnerability/likelihood. Paper presented at the Conceptualizing and Measuring Risk Perceptions Workshop, Washington, DC. Retrieved from: https://cancercontrol.cancer.gov/sites/default/files/2020-06/windschitl.pdf [Google Scholar]
  35. Witte K (1992). Putting the fear back into fear appeals: The extended parallel process model. Communications Monographs, 59(4), 329–349. [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supp 1
NIHMS1761143-supplement-Supp_1.docx (77.9KB, docx)

Data Availability Statement

Due to our university’s requirements on grant-funded research, we can share the study data with a signed data use agreement. Investigators wishing to access the data may contact NTB.

RESOURCES