Trust in science moderates the effects of high/low threat communication on psychological reactance to COVID-19-related public health messages

ABSTRACT

Background

As illustrated by the COVID-19 pandemic, communicating evidence-based health recommendations represents a tremendous challenge; among some recipients, public health messages can cause anger and negative cognitions, also known as psychological reactance, and consequently lead to negative attitudes and low intentions to perform the promoted behavior. The present study investigated the role of message characteristics (i.e. high vs. low freedom-threat messages), individuals’ trust in science (i.e. high vs. low trust in science), and their interaction in determining responses to public health messages.

Methods

We conducted an experimental study, in which participants (N = 228) with high or low trust in science were exposed to high or low freedom-threat messages promoting mask-wearing to reduce the spread of COVID-19 and regular physical activity.

Results

We found support for the notion that messages imposing high threat to freedom lead to higher state psychological reactance, and more negative attitudes and behavioral intentions. Moreover, our results showed that trust in science has a main and interaction effect (together with message characteristics) on state reactance, behavioral intentions, and – to a lesser degree – attitudes, in the case of COVID-19, but not physical activity messages. The findings remained the same regardless of controlling for other relevant variables.

Conclusions

While our study has some limitations, such as a rather homogeneous sample, a limited number of experimental stimuli, and a relatively artificial experimental environment, it offers some insight into the important role of health communication recipients’ trust in science and provides advice on how to communicate health recommendations to skeptics.

KEYWORDS:

• Trust in science
• psychological reactance
• health communication
• COVID-19
• physical activity

Introduction

Developing persuasive health messages requires walking a fine line between concisely and confidently communicating the evidence-based recommendations and ensuring that they are communicated in a way that does not interfere with the individual’s sense of freedom. Despite extensive literature describing different channels of communication and message features that can be leveraged to communicate health-related advice (e.g. [1–4]), effective health communication is still one of the key challenges faced by public health institutes. While the problem is not new, it has gained considerable attention during the recent COVID-19 pandemic, when public health messages promoting the preventive measures, such as mask-wearing, faced additional challenges [5] and were often met with significant backlash, leading to reduced compliance with health guidelines [6]. In Slovenia, where the present study took place, about one in five participants believed that the protective guidelines recommended by the government and health institutions were unnecessary [7], and a similar percentage found the use of protective masks in enclosed public spaces unimportant [8]. However, such negative responses were predominantly observed in certain subgroups, such as among less conscientious [8], less empathic, and more politically conservative individuals (e.g. [9]), pointing to the vital role of individual differences in responding to persuasive health messages.

While the responses to public health messages are generally determined by an interplay of various factors, including proneness to reactance [10, 11], the present study builds on previous research by investigating the role of a poorly-understood factor – trust in science. Previous research has established that trust in science is linked to health-related attitudes and behavior, including compliance with COVID-19-related guidelines and aspects of health-promoting lifestyle, such as physical activity [12–15]. However, research is still in its early stages, and one major issue is an insufficient understanding of the role of trust in science in health communication.

In the present study, we investigate the possibility that the relationship between trust in science and health behavior may be partly attributed to distrustful individuals being more sensitive to how health-related advice is conveyed. We will test this notion with an experimental study, in which participants with varying levels of trust in science will be exposed to two messages, a high and low threat one, promoting two health behaviors, specifically wearing face masks in healthcare facilities and being active at least 150 min per week. Therefore, this study could contribute to research on trust in science and psychological reactance by explaining why distrustful individuals exhibit lower levels of evidence-based health behavior. Second, the study may expand the literature on the interindividual differences that influence psychological reactance with a relatively malleable factor that could be addressed with future interventions. Third, the results may answer how to effectively communicate health-related advice to distrustful individuals and increase their likelihood of compliance with communicated guidelines. For example, while previous research (e.g. [16]) shows benefits of adapting health communication to individual’s ideology and other characteristics, this is not yet possible in the case of trust in science due to the lack of studies. We argue that by addressing these issues, we could take a critical step toward bringing science closer to those who may need it the most and building a resilient and healthy society.

Literature review

Psychological reactance and the intertwined process model

The theory of psychological reactance was constructed to describe a motivational state that drives individuals toward regaining a sense of freedom after being exposed to something subjectively threatening it [17, 18]. The theory may be particularly beneficial in the context of health persuasion, where it helps explain why persuasive messages sometimes do not yield the desired results or even lead to results that are opposite to what the message intended to achieve [19]. It is hence not surprising that one of the most elaborated upgrades of the theory emerged specifically in the context of persuasive health communication. The so-called Intertwined Process Model, developed by Dillard and Shen [20], suggests that when persuasion poses a threat to individuals’ free behavior, a reaction in terms of negative cognitions and anger will follow, which will, in turn, lead to more negative attitudes towards the persuasive message and, finally, decrease the likelihood of the promoted behavior. Contrary to previous conceptualizations, negative cognitions and anger are treated as two intertwined indicators of state reactance. The model is backed by empirical evidence [20, 21].

High versus low threat messages and their influence on psychological reactance

It is well-established that message characteristics can affect whether psychological reactance will occur and to what degree. The present study varies the level of threat imposed by public health messages via experimentally manipulating the level of freedom-threatening language (FTL), presence of restoration postscripts, and gain versus loss message framing.

FTL refers to language that explicitly limits the autonomy of the message recipient by using controlling and directive phrases such as ‘you must’ as opposed to using more implicit terminology such as ‘you may consider’ [3, 21]. Previous research consistently shows that high FTL, compared to low FTL, increases freedom threat and reactance [22, 23]. For example, a recent study by Ma and Miller [24] found that high FTL led to greater freedom-threat perception, reactance, and less positive attitudes towards COVID-19 vaccination messages compared to low FTL.

As opposed to FTL, choice-enhancing language has the potential to reduce reactance arousal [3, 18]. A well-known example are restoration postscripts – brief statements at the end of a message that emphasize that the decision to comply with the message recommendations is the recipient’s choice. An example of such a statement is ‘The choice is yours’ [3, 25]. A recent study by Richards and colleagues [26], for instance, compared two strategies, including restoration postscripts, and found that both equivalently reduced the degree of reactance in response to health appeals. While the existing body of literature is not completely consistent, several other studies have also concluded that adding restoration postscripts can reduce freedom threat and reactance [27, 28].

Lastly, another message feature that may affect state reactance is gain-loss message framing. Whereas gain-framed messages emphasize the benefits of adopting the recommended action (e.g. ‘Drinking enough water daily reduces the risk of headaches’), loss-framed messages emphasize the negative consequences of failing to adopt the recommended action (e.g. ‘Not drinking enough water daily increases the risk of headaches’). While findings are somewhat mixed, multiple studies have found that loss-framed messages elicit a greater threat to freedom and reactance [29, 30]. These effects have also been observed in a recent study [31], which tested how gain-loss framing influences the effectiveness of COVID-19 vaccination promotion messages. The study found a significant main effect of gain-loss frames on perceived threat to freedom and aspects of state reactance, with participants exposed to loss-framed messages experiencing greater threat to freedom and anger.

H1: Regardless of the topic, high threat public health messages will lead to a) higher state reactance, b) more negative attitudes, and c) lower behavioral intentions compared to low threat public health messages.

The role of reactance proneness and trust in science

Message characteristics are not the only determinant of state reactance; as different individuals respond differently to the same health persuasion appeals, individual characteristics may be equally important. This idea is not new; Brehm and Brehm [32] discussed the role of the need for autonomy and self-determination, followed by other authors who hypothesized the importance of general resistance to rules and guidelines and degree of defensiveness [33, 34]. These ideas later consolidated in the form of a construct called reactance proneness, a personality trait that describes individual’s propensity to react negatively in response to perceived attempts to influence their behavior [10, 11]. According to previous research, individuals with high reactance proneness experience a higher perceived threat to freedom after being exposed to persuasive messages and are hence more resistant to persuasive appeals compared to those with low reactance proneness [23, 35].

While the concept of reactance proneness summarizes the general characteristics of message recipients, it does not consider who the source of the message is and the recipient’s perception of this specific source. Moreover, reactance proneness is a relatively stable individual characteristic [36]. Accompanying reactance proneness with variables such as perceived source credibility and trust, defined as the intention to accept vulnerability based on positive expectations about others, may thus help improve our understanding of state reactance [37–39]. This assumption is supported by the research of Song and colleagues [40], who found that individuals’ perceptions of the source are an important determinant of state reactance. Specifically, the findings showed that the more the participants perceived the source as trustworthy, the less likely it was for the perceived threat to freedom and reactance to occur. Shifting the focus to trust would also shed light on more malleable determinants, which could be addressed with future interventions.

While different forms of trust may be important in different contexts, we argue that trust in science and scientists (i.e. individual’s belief in scientists’ honesty and their capacity as providers of information; [41, 42]) may be particularly relevant in the context of communicating evidence-based health recommendations since scientists represent the ultimate source of these recommendations. Studies investigating this specific form of trust show that it varies significantly in the population and is determined by individual factors such as political ideology [43], religiousness, conspiracy ideation, openness, intellectual humility [44], and news consumption habits [45]. Studies additionally show that it is associated with adherence to COVID-related preventive measures and other health-related decisions [12, 14, 15, 46], with previous authors suggesting that it is especially important in contexts that are difficult to understand, highly emotional and personally relevant [47]. Moreover, recent studies support the notion that trust in science can be intervened upon and increased on the individual [48] or societal level [49]. However, its role in the process of psychological reactance has yet to be investigated. For example, while there is some theoretical support for the notion that trust in science influences state reactance and related outcomes (due to previously observed associations with health behavior), much less is known about if and how trust in science modifies the effects of message characteristics. Moreover, it is unclear if trust in science explains variance above and beyond reactance proneness.

H2 and H3: After controlling for reactance proneness, individuals with lower trust in science will experience a) higher state reactance, b) more negative attitudes, and c) lower behavioral intentions than individuals with higher trust in science after exposure to (H2) COVID-19 – and (H3) physical activity-related public health messages.

RQ1 and RQ2: After controlling for reactance proneness, does trust in science interact with message characteristics in determining a) state reactance, b) attitudes, and c) behavioral intentions after exposure to (RQ1) COVID-19- and (RQ2) physical-activity-related public health messages?

Methods

Participants

The institutional ethics committee approved the study procedure on August 31st, 2022 (ID: 038-33-123/2022/8/FFUM), whereas the data collection began on September 22^nd and concluded on October 16^th, 2022.

A total of 1026 participants clicked on a sponsored post on Facebook, targeted towards users living in Slovenia and aged 18–65, but only 291 (28.4%) began filling out the study. After excluding those that did not complete the whole study (n = 52; 17.2%) and those that failed more than one attention check (n = 11; 4.6%), the final sample consisted of 228 participants. Of those, 79.4% were female, 18.0% were male, and the remaining 2.6% identified as ‘other’ or did not want to answer the question regarding their gender. The average age of participants was 38.63 years (SD = 10.74), and they were relatively highly educated (82.1% with some form of tertiary education). In general, the majority of participants described themselves as liberal (76.3% in terms of social issues and 46.9% in terms of economic issues), followed by politically neutral (14.0% in terms of social issues and 32.5% in terms of economic issues) and conservative individuals (9.7% in terms of social issues and 20.6% in terms of economic issues). The socioeconomic status of participants was rather diverse, with the largest share of participants stating that their status is slightly higher than the average¹ (26.3%) or average (22.4%).

While participants did not receive any compensation, they were given a chance to participate in a raffle to win one of the three 20 EUR gift cards for one of the major Slovenian supermarkets.

Stimulus material

Prior to the study, we prepared four public health messages – two addressing COVID-19 (wearing face masks in healthcare facilities) and two addressing physical activity (being active 150 min per week). Both stimuli within a given topic share the same design and contain identical scientific insights. However, the stimuli differ in how these findings are communicated. In the case of COVID-19, one message employs low FTL (e.g. ‘Please wear a face mask’) and includes a restoration postscript (‘Your decision matters’), while the other employs high FTL (e.g. ‘Mask is MANDATORY!’) and no restoration postscript. Similarly, in the case of physical activity, one message employs low FTL (e.g. ‘It is worth considering how to incorporate enough exercise in your life’) and gain-framing (‘Physical activity is good for your health’), while the other employs high FTL (e.g. ‘Ensure enough minutes of exercise as soon as possible’) and loss-framing (‘If you are not physically active, you are at risk’). The experimental manipulation was validated in a pilot study described in Supplementary materials. The translated materials are depicted in Figure 1.

Figure 1. Stimulus material.

Experimental design and procedure

To test our hypotheses, we conducted an experiment, in which participants were assigned two of four stimuli in a mixed repeated measures design (two topics: COVID-19 and physical activity x two messages: high and low threat). After providing consent, participants either received 1) a low threat COVID-19 message and high threat physical activity message (n = 117; 51.3%) or 2) a high threat COVID-19 message and low threat physical activity message (n = 111; 48.7%). Specifically, participants first saw one of the two public health messages. During their exposure to the message, they were asked to write down any thoughts that popped up in their mind. Next, they filled out the scales that measured perceived threat to freedom and anger pertaining to the first message. The same procedure (measuring negative cognitions, perceived threat to freedom, and anger) was then repeated with the second public health message. Afterward, participants additionally filled out the attitude and behavioral intentions instruments pertaining to both messages, measures tackling individual characteristics, demographic questions, and control questions. The study procedure took approximately 10 min to complete.

Measures

Perceived threat to freedom

The perceived threat to freedom was measured with the scale developed by Dillard and Shen [20]. It consists of four items (e.g. ‘The message tried to make a decision for me’), answered with a 5-point Likert scale (from ‘Strongly disagree’ to ‘Strongly agree’). The alpha reliability of the scale was .93 (COVID-19) and .89 (physical activity).

Anger

Participants rated the anger they experienced after exposure to public health messages with a scale [50] that consists of four items (e.g. ‘irritated’), answered with a 5-point response scale (from ‘None of this feeling’ to ‘A great deal of this feeling’). The alpha reliability was .97 (COVID-19) and .95 (physical activity).

Negative cognitions

To measure negative cognitions, we asked the participants to write out all thoughts that appeared in their minds while reading the message, regardless of their relevance. Similar prompts have previously been used in other studies investigating psychological reactance [20, 51]. The resulting data were independently coded by four postgraduate students working in pairs (i.e. each answer was coded by two students). The raters were asked to 1) segment the data into psychological thought units, 2) mark whether the thought is relevant to the message, and 3) assess whether the thought is positive, negative, or neutral (the magnitude of positive/negative thoughts was not assessed). They were blinded to the study hypotheses and the experimental condition to which a given participant was allocated. For relevant negative thoughts, which form the negative cognitions variable, the agreement among coders - calculated using the intraclass correlation coefficient (ICC) with a one-way random-effects model, single measure [52] - was good (COVID-19: ICC = .80, physical activity: ICC = .82). All inconsistencies between the coders were resolved by a third independent rater.

Attitudes

Six items assessed the participant’s attitudes towards wearing face masks (e.g. ‘I generally believe that wearing face masks in healthcare facilities is valuable’), and six items assessed the participant’s attitudes towards physical activity (e.g. ‘I would describe physical activity as enjoyable’). Both scales include items that refer to both the instrumental and experiential component, and were self-constructed using Ajzen’s [53] guidelines. The 7-point response scale ranged from ‘Strongly disagree’ to ‘Strongly agree’. The alpha reliabilities were .94 (mask-wearing) and .76 (physical activity). Both self-constructed measures can be found in Supplementary materials.

Behavioral intentions

Three items assessed the participant’s behavioral intentions regarding mask-wearing (e.g. ‘I plan to wear a face mask on my next visit of healthcare facilities’), and three items assessed the participant’s behavioral intentions regarding physical activity (e.g. ‘In the next few weeks, I intend to be regularly physically active’). Both scales were self-constructed using Ajzen’s [53] guidelines and employed a 7-point scale (from ‘Strongly disagree’ to ‘Strongly agree’). The alpha reliabilities were .97 (mask-wearing) and .96 (physical activity). Both measures of behavioral intentions can be found in Supplementary materials.

Reactance proneness

Reactance proneness was measured with the short version of the Hong Psychological Reactance Scale [10, 11]. The scale consists of 11 items (e.g. ‘Regulations trigger a sense of resistance in me’), answered using a 5-point agreement scale (from ‘Strongly disagree’ to ‘Strongly agree’). The alpha reliability was acceptable (.87).

Trust in science

We also used a shorter version of the Trust in Science and Scientists Inventory (TSSI; [46, 47]). This version of the scale consists of 14 items (e.g. ‘Scientific theories are trustworthy’), answered with a 5-point Likert scale (from ‘Strongly disagree’ to ‘Strongly agree’). As in previous studies, the scale exhibited great internal consistency (α = .96).

Attention checks, demographic questions, and control questions

Participants were asked to respond to three attention checks spread throughout the study; two in the form of directed questions (e.g. ‘Please choose ‘Somewhat agree’ and continue with the next question’; [54] and one open-ended question asking participants ‘Which topic was addressed by the message on the previous page?’.

Participants were also asked to answer basic demographic questions and two control questions asking participants about their past behavior related to mask-wearing and physical activity.

Statistical analysis

Statistical analyses were carried out using IBM SPSS Statistics 26. First, we prepared the database for analysis by calculating all factor scores. As in previous research (e.g. [51]), the composite reactance score was calculated as the sum of standardized Z scores of anger and negative cognitions (separately for each topic). Based on the response format of the TSSI, individuals were divided into two groups; individuals with average scores ranging from 1.00–3.00 were assigned to the low trust group (n = 50), while individuals with average scores ranging from 3.01–5.00 were assigned to the high trust group (n = 178). Second, we calculated basic descriptive statistics, performed preliminary analyses (such as calculating Pearson correlation coefficients between focal variables), and tested the assumptions of the analyses used for hypotheses testing.

We employed two-way MANCOVA to test the main omnibus effects of message manipulation (low threat/high threat) and trust in science (low trust/high trust), as well as their interaction effect on the combined dependent variables (state reactance, attitudes, and behavioral intentions) while controlling for reactance proneness. To test the robustness of our results, we also performed analyses without the covariate and analyses with additional sociodemographic covariates (age, education level, political orientation, and socioeconomic status). Significant omnibus tests were followed up by additional post-hoc tests and simple effects analyses. All p-values are accompanied by effect sizes (Cohen’s d and η_p²).

Results

Randomization and manipulation check

Concerning randomization, the analyses showed that the two independent groups did not differ significantly in any demographic variable, past mask-wearing or physical activity behavior, and other individual characteristics (i.e. reactance proneness and trust in science). Moreover, to perform a manipulation check, we compared the independent groups according to the perceived threat to freedom experienced after exposure to a COVID-19 message and physical activity message. In the case of both topics, the results showed that high threat messages elicited a higher threat to freedom compared to low threat messages. The detailed results pertaining to randomization and manipulation checks can be found in Supplementary materials.

Descriptive statistics

Table 1 reports the means and standard deviations for the two individual characteristics and all dependent variables, separately for each of the two experimental conditions. Descriptively, we observed higher state reactance, more negative attitudes, and lower behavioral intentions among individuals exposed to high threat messages, regardless of the topic.

Table 1. Descriptive statistics.

		M	SD
1. Reactance proneness		2.45	0.71
2. Trust in science		3.77	0.95
		COVID-19		Physical activity
		M	SD	M	SD
3. State reactance	Low threat	−0.30	1.45	−0.33	1.20
	High threat	0.32	1.84	0.31	1.87
4. Attitudes	Low threat	4.96	1.79	6.00	0.85
	High threat	4.66	1.99	5.94	0.87
5. Beh. intentions	Low threat	5.77	1.81	5.80	1.57
	High threat	5.17	2.28	5.67	1.45

We also calculated bivariate associations between the main variables, which are presented in Table 2. We found that both reactance proneness and, in particular, trust in science are strongly associated with state reactance (reactance proneness: r = .49, p < .001; trust in science: r = -.58, p < .001), attitudes (reactance proneness: r = -.55, p < .001; trust in science: r = .76, p < .001), and behavioral intentions (reactance proneness: r = -.50, p < .001; trust in science: r = .67, p < .001) reported after exposure to a COVID-19-related message. On the other hand, both are only negligibly to weakly associated with state reactance (reactance proneness: r = .14, p = .029; trust in science: r = -.01, p = .842), attitudes (reactance proneness: r = -.02, p = .739; trust in science: r = -.06, p = .334), and behavioral intentions (reactance proneness: r = -.15, p = .024; trust in science: r = .07, p = .310) reported after exposure to a physical activity-related message.

Table 2. Bivariate associations.

	1.	2.	3.	4.	5.
1. Reactance proneness	-	-.56***	.49***	-.55***	-.50***
2. Trust in science	-.56***	-	-.58***	.76***	.67***
3. State reactance	.14*	-.01	-	-.67***	-.64***
4. Attitudes	-.02	-.06	-.27***	-	.81***
5. Beh. Intentions	-.15*	.07	-.19**	.46***	-

Notes. Results related to the COVID-19 messages are displayed above the diagonal, while results related to the physical activity messages are displayed below the diagonal. * p < .050, ** p < .010, *** p < .001

Hypotheses testing

In the next step, we analyzed the omnibus effects of message characteristics, trust in science, and their interaction on three correlated outcomes – state reactance, attitudes, and behavioral intentions. In the case of COVID-19, we found that there is a significant main effect of experimental manipulation (F(3, 221) = 8.93, p < .001, Wilks’ Λ = .892, η_p²= .108) as well as a significant main effect of trust in science on the combined dependent variables (F(3, 221) = 49.87, p < .001, Wilks’ Λ = .596, η_p²= .404). Additionally, we found a statistically significant interaction effect between message characteristics and trust in science, F(3, 221) = 3.72, p = .012; Wilks’ Λ = .952, η_p²= .048. In the case of physical activity, on the other hand, we found a significant main effect of experimental manipulation (F(3, 221) = 4.32, p = .006, Wilks’ Λ = .945, η_p²= .055), while the main effect of trust in science (F(3, 221) = 1.77, p = .154, Wilks’ Λ = .977, η_p²= .023) and the interaction effect on the combined dependent variables (F(3, 221) = 1.20, p = .311, Wilks’ Λ = .984, η_p²= .016) were not statistically significant. The results of follow-up univariate tests to the two-way MANCOVA related to COVID-19 messages are presented in Table 3. Two-way MAN(C)OVA results without a covariate and with additional control variables, which led to the same conclusions, can be found in Supplementary materials.

Table 3. Follow-up tests: COVID-19.

	F	df1, df2	p	ηp^2
State reactance
Threat	17.84	1, 223	< .001***	.074
Trust in science	46.40	1, 223	< .001***	.172
Threat*Trust	4.73	1, 223	.031*	.021
Attitudes
Threat	5.93	1, 223	.016*	.026
Trust in science	140.12	1, 223	< .001***	.386
Threat*Trust	3.30	1, 223	.071	.015
Behavioral intentions
Threat	17.77	1, 223	< .001***	.074
Trust in science	84.35	1, 223	< .001***	.274
Threat*Trust	9.79	1, 223	.002**	.042

Notes. * p < .050, ** p < .010, *** p < .001

As can be seen in Table 3, there was a significant effect of message characteristics and trust in science on all three outcomes. Moreover, we observed a significant interaction effect on state reactance and behavioral intentions, but not attitudes. All three interactions are visualized in Figure 2, while means and standard deviations can be found in Supplementary materials. Simple slopes analyses using the Bonferroni adjustment for multiple comparisons revealed that individuals low (p < .001) and individuals high in trust in science (p = .030) experienced higher state reactance after exposure to a high threat message compared to exposure to a low threat message. Furthermore, individuals low (attitudes: p = .017, intentions: p < .001), but not individuals high in trust in science (attitudes: p = .509, intentions: p = .247), reported more negative attitudes and lower behavioral intentions after exposure to a high threat message compared to exposure to a low threat message.

Figure 2. Interaction effects of message characteristics and trust in science on state reactance, attitudes, and behavioral intentions.

Discussion

In the present study, we aimed to investigate the message characteristics and individual variables associated with psychological reactance experienced after exposure to public health messages, and the resulting attitudes and behavioral intentions.

Our results showed that state psychological reactance, negative attitudes, and low behavioral intentions are more likely to arise after exposure to high threat messages, compared to low threat messages, thereby supporting H1. This finding is in line with other studies reporting that low freedom-threatening language, restoration postscripts, and gain framing can decrease state reactance and be beneficial in terms of changing attitudes and behavioral intentions in the desired way (e.g. [24, 26]). On the other hand, we found that the role of trust in science may be more nuanced; while our results suggest that distrustful individuals experienced more intensive state reactance, more negative attitudes, and lower behavioral intentions after exposure to COVID-19 messages, compared to more trustful individuals, supporting H2, we did not observe a significant main effect of trust in science in the case of physical activity, thereby rejecting H3. While these findings were not completely anticipated, as trust in science has previously been linked to COVID-19-related health behavior as well as other health-promoting behaviors [15], they may be explained by the notion that trust-based decision-making occurs especially in the case of complex topics that are poorly understood by the general public, as well as topics that are highly emotional and highly personally relevant [47]. As COVID-19, and mask-wearing specifically, are contested phenomena characterized by a high amount of false claims [55, 56] and strong emotional responses [57], individuals’ responses may be more reliant on trust compared to individuals’ responses to physical activity messages. Such an explanation is also supported by rare empirical data; while the study by Plohl and Musil [15] showed significant associations between trust in science and compliance with COVID-19-related guidelines as well as physical activity, the association with COVID-related behavior was much stronger.

Lastly, we found that trust in science interacts with message characteristics in determining message-related outcomes in the case of COVID-19 ads, supporting RQ1, but, again, not in the case of physical activity ads, thus rejecting RQ2. The explanation for this finding is similar as for the main effect of trust in science – COVID-19 is a more contested, complex, and emotionally-charged topic than physical activity. Furthermore, in regards to RQ1, the post-hoc tests revealed a significant interaction effect on state reactance (RQ1a) and behavioral intentions (RQ1c), but not attitudes (RQ1b), although the general pattern was similar in this case as well and the p-value approached significance. The fact that our study was unable to convincingly demonstrate that trust in science and message characteristics interact to determine individuals’ attitudes towards mask-wearing could be explained in at least two ways. First, while state reactance is directly linked to the message at hand, attitudes and behavioral intentions are a conglomerate of different perceptions and experiences, making them less susceptible to the effects of a single message. Second, attitudes were measured with items capturing both the instrumental and experiential component, while our experimental stimuli were much more targeted towards changing instrumental attitudes (i.e. explaining how valuable face masks are) as opposed to experiential attitudes (i.e. how pleasant face mask-wearing can be). In fact, the additional analyses we performed to explore the non-significant finding showed a significant interactive effect between trust in science and message characteristics on the instrumental aspect of mask-wearing attitudes but not on the experiential aspect (see Supplementary materials).

Limitations and implications

The present study has some limitations. First, our sample is not representative of the broader population; for example, the sample consists mainly of women, highly-educated individuals, and participants who trust science to a relatively high degree. Future research should hence consider employing more diverse samples. Second, participants were only exposed to a single experimental stimulus per topic, and the number of topics was limited to only two. They also filled in the outcome measures almost immediately after exposure. As such, future studies should replicate our findings on different materials and with research designs that allow for delayed measurement of outcomes. Third, the study took place in an artificial environment, with participants knowing that they are participating in an academic study, which could, in turn, skew their responses. Therefore, future studies should focus on finding ways to increase the external validity of findings. Lastly, since we wanted to keep participants’ effort low, we did not ask participants to self-code the relevance and valence of their thoughts, as done in some previous studies (e.g. [51]). Therefore, it is possible that some thoughts were misinterpreted by external coders.

Despite certain limitations, the study’s findings highlight the vital role of message recipients’ trust in science that goes beyond the role of a well-known factor of state reactance, reactance proneness (as well as sociodemographic variables). Moreover, they point to specific and tangible ways of effectively communicating evidence-based health recommendations to a high-risk group - individuals who are distrustful of science. In particular, the results suggest that, especially in the case of sensitive and disputed topics, such as COVID-19, individuals with low trust in science need to be addressed with public health messages that avoid FTL and use restoration postscripts. Despite employing a more implicit way of communicating, such messages seem to not only cause less reactance but also lead to higher behavioral intentions among distrustful individuals without bearing any negative outcomes for highly trustful individuals.

Author contribution

NP was involved in conceptualization of the study, methodology, formal analysis, investigation, writing of the original draft, and reviewing and editing of the manuscript. BM was involved in conceptualization of the study, methodology, and reviewing and editing of the manuscript.

Ethics

The University of Maribor’s Faculty of Arts institutional ethics committee approved this study on August 31st, 2022 (ID: 038-33-123/2022/8/FFUM). All participants provided informed consent.

Acknowledgments

The authors would like to thank Ivona Balent, Tanja Goltnik, Tina Goznik, and Tinkara Koračin for their help with coding participants’ responses to the negative cognitions prompts.

Data availability

The data that support the findings of this study are available from the corresponding author, NP, upon reasonable request.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Notes on contributors

Nejc Plohl

Nejc Plohl is a PhD student as well as a research and teaching assistant at the Department of Psychology, University of Maribor, Slovenia. His research focuses on social and health psychology, with a particular emphasis on attitudes towards science and technology.

Bojan Musil

Bojan Musil is an associate professor and researcher at the Department of Psychology, University of Maribor, Slovenia. His primary research interests are cross-cultural studies of values and value orientations, personality concepts, health, educational studies, and computer-mediated communication.

Notes

1 Participants were given the average Slovenian net salary (which was 1.236 EUR at the time of the study) for comparison.