Ambivalence Toward the Implementation of Preventive Measures in (Un-)Vaccinated German Citizens
Abstract
Abstract:Background: The COVID-19 pandemic has changed life around the world. To fight the pandemic, preventive measures were implemented. Despite being accepted by a majority of citizens in Germany, these measures elicited fierce protest from others. It seems that people either like or dislike them. Given the immense complexity of the topic, it is also likely that people hold ambivalent attitudes (i.e., simultaneously positive and negative). Aims: The purpose of this study is to explore ambivalence toward the (non-)implementation of preventive measures in vaccinated and unvaccinated German citizens. Method: Vaccinated (N = 136) and unvaccinated (N = 170) participants indicated their positivity, negativity and experienced ambivalence toward the (non-)implementation of preventive measures (e.g., mask mandatory). Results: The structure of positive and negative evaluations (i.e., objective ambivalence) indicates that unvaccinated people are more univalent (negative) and vaccinated people are neutral toward the preventive measures. Interestingly, results indicate a dissociation between objective ambivalence and experienced ambivalence. Limitation: The results are limited by the measurement choice, data collection time, and sample. Conclusion: The findings indicate that the experienced ambivalence increases with increasing personal costs. Implications for behavior change interventions and health psychology are discussed.
The Corona pandemic is one of the major health crises in the world. But it is also a social and political crisis. Never have vaccines been developed so quickly, social behavior changed so rapidly and laws passed in such a short period of time. At the same time, the question of compliance, that is, whether people are adhering to the guidelines for protecting their health, has come into focus like never before.
Simultaneously with the restrictions, a protest movement was formed exhibiting strong anti-vaccination attitudes and rebelling against the restrictions in Germany. It seems that the issue of COVID-19 measures has divided society into two parts: those who approve of the measures and those who strictly reject them. This polarization even appeared to increase as the pandemic and its restrictions continued: on the one hand, more people participated in anti-vax protests (Grande et al., 2021) on the other hand, there were also counterprotests defending them (Hellmeier, 2022). This may have led to the impression that people are either strictly in favor and approve, or against and disapprove of the preventive measures.
However, research shows that attitudes are often more complex, that is, they are ambivalent (e.g., Thompson et al., 1995; Schneider et al., 2015; van Harreveld et al., 2015). Ambivalence is defined as the simultaneous existence of positive and negative associations toward one attitude object (van Harreveld et al., 2015). For example, the preventive measures albeit having clear positive effects like the reduced spread of the coronavirus and thereby preventing deaths, simultaneously had negative effects, like the reduction in freedom of movement. This simultaneous existence of positive and negative effects could enhance attitudinal ambivalence.
Ambivalence and Health Behavior
There is ample evidence that self-relevance enhances the probability of attitudinal ambivalence. This is evident, especially in the health context. Researchers obtained ambivalence toward unhealthy foods (Norris et al., 2019), physical exercise (Sparks et al., 2004), receiving a heart transplant (Kuhn et al., 1989), organ donation (Contiero & Wilson, 2019), water pipe use (Mays et al., 2020), drug use (Hohman et al., 2014), or condom use (Kane, 1990) to mention only a few health-related topics.
Investigating ambivalence is particularly relevant because of its relationship to behavior. According to the theory of planned behavior (Ajzen, 1991), intentions are an important antecedence of behavior. Intentions in turn are predicted by subjective norms, perceived behavioral control, and attitude toward the behavior. However, research shows that ambivalence also influences the subjective norm-intention relationship. For example, subjective norms were stronger predictors of intention to use marijuana for people high in ambivalence compared to people low in ambivalence (Hohman et al., 2014). Additionally, people who experienced ambivalence used norms to reduce their ambivalence and guide their intentions (Hohman et al., 2016). Besides the influence on the subjective norm-intention relationship Cooke and Sheeran (2004) found in a meta-analysis that ambivalence moderates the attitude-intention and attitude-behavior relationship. That is, people with lower ambivalence showed more consistency between their attitudes and intentions as well as between their attitudes and behavior than people with higher ambivalence. For health behavior, for example, Armitage and Conner (2000) addressed adherence to a low-fat diet and found that attitudes were more predictive of intentions and behavior when ambivalence was low.
Ambivalence also seems to influence behavior directly. Regarding COVID-19, Schneider and colleagues (2021) found ambivalence regarding the pandemic as well as ambivalence regarding the behavior recommendations to prevent the spread of COVID-19. Their findings showed that ambivalence toward preventive behavior was negatively related to self-reported adherence to preventive behavior whereas ambivalence regarding the pandemic was positively related to self-reported adherence to preventive behavior.
Hence, if behavior (change) is the goal, investigating ambivalence is mandatory. Regarding behavior change, researchers found that people who reported high ambivalence also reported a higher desire to quit the ambivalent behavior (e.g., for water pipe use; Lipkus & Noonan, 2017). Thus, ambivalence might be an indicator of willingness to change the behavior and might therefore be a good starting point for behavior interventions. Interventions that help people resolve their ambivalence should lead to attitude-consistent behavior.
Consistent with this idea, ambivalence is central in motivational interviewing because this counseling technique aims to facilitate behavior “change by exploring and resolving ambivalence” (Miller & Rollnick, 2002, p. 25). Motivational interviewing, which originated in substance abuse treatment, has proven helpful in other (health) behavior domains such as dieting and exercising (see Martins & McNeil, 2009 for a review). Even though ambivalence can be a catalyst for change, unresolved ambivalence can also be a barrier. If unresolved ambivalence leads to tension, this tension often produces resistance. Resistance can be, for example, noncompliance with homework (Westra & Norouzian, 2018) and increases when the therapist meets client resistance with confrontation, arguments, or persuasion (Moyers & Rollnick, 2002). Therefore, motivational interviewing describes specific counseling strategies for exploring and resolving ambivalence as well as dealing with resistance.
Whereas the previously described processes of ambivalence eliciting tension which is met with resistance are placed in client-therapist interaction, the Model of Ambivalence Induced Discomfort (MAID model; van Harreveld et al., 2009) describes similar processes more generally. According to this model, the structure of positive and negative associations (i.e., objective ambivalence; OA) leads to the experience of ambivalence (i.e., subjective ambivalence; SA) when a decision about a potentially ambivalent object must be made. If the consequences of this decision are uncertain and people anticipate regret, they experience discomfort and need to cope. Coping strategies could be procrastination, denial of responsibility, or (un)biased systematic processing. Thus, as in motivational interviewing, people need to cope with the discomfort. This coping can be constructive and resolve discomfort and OA (e.g., systematic processing) or non-constructive and reduce discomfort but maintain OA (e.g., procrastination, noncompliance with homework). The authors of the MAID model argued that the ambivalence-discomfort relation is amplified if “a behavioral choice related to the attitude object has direct consequences for the self” (van Harreveld et al., 2009, p. 57). As health-related behavior has immediate consequences for the self, OA should lead to SA and there should be a necessity to cope. Considering COVID-19 prevention, for example, people might have positive associations (e.g., lower risk of infection) and simultaneously negative associations to the COVID-19 vaccine (e.g., possible side effects). This OA elicits SA if people ponder the decision to get vaccinated or not. However, because vaccination is not mandatory, people might procrastinate on the decision to reduce SA.
Measuring Ambivalence
Consistent with the MAID model (van Harreveld et al., 2009), researchers differentiate between OA and SA when measuring ambivalence. OA is measured by capturing positivity and negativity on two separate unipolar scales and combining them into an index (Thompson et al., 1995). The most commonly used index is the similarity intensity model index (SIM-Index; e.g., Thompson et al., 1995) and is calculated with the following formula where P = positivity and N = negativity:
Thus, it is conceptualized as the intensity of the components (i.e., (P + N)/2) corrected by the polarization (i.e., |P − N|). With this measuring method, it is possible to differentiate between rather neutral, rather univalent, and rather ambivalent attitudes (Buttlar et al., 2021, 2023; Schneider et al., 2021; Schneider & Mattes, 2021). Ambivalence is characterized by high intensity and low polarization – in other words, high ratings on the positivity and negativity scale. Neutrality is characterized by low intensity and low polarization (i.e., low ratings on both scales). Univalence is characterized by high intensity and high polarization (i.e., high ratings on one of the two scales). However, the SIM-Index does not indicate the direction of the univalence (i.e., univalent positive or negative). Thus, if the SIM-Index indicated univalence additional analyses (e.g., difference scores) are needed to determine the direction of the univalence. SA is assessed by inquiring about the experience of mixed feelings and thoughts toward an attitude object (van Harreveld et al., 2015).
In this paper, we assume that OA is a prerequisite of SA and that vaccinated and unvaccinated people differ in ambivalence. Whereas vaccinated people might be rather univalent because they resolved their ambivalence toward COVID-19 by getting vaccinated, unvaccinated people might experience more ambivalence regarding all COVID-19-related issues and therefore postpone the decision to get vaccinated to reduce this ambivalence.
The Present Research
In the present study, we tested this hypothesis. Specifically, we investigated OA and SA toward the implementation of preventive measures (IPM) and non-implementation of preventive measures (NIPM) in vaccinated and unvaccinated participants. We were interested in these groups because vaccinated and unvaccinated people were differently affected by the preventive COVID-19 regulations implemented by the government in Germany. Whereas vaccinated people enjoyed at least some freedom, the unvaccinated were much more restricted in their freedom of movement, for example, for some periods, entry to restaurants and theaters was only allowed to vaccinated citizens. Thus, unvaccinated people were more affected in their freedom of movement (i.e., negative association) by the preventive measures, although the restrictions simultaneously protected their health (i.e., positive association). In contrast, if the preventive measures were not implemented, they could move freely (i.e., positive association), however, no restrictions also posed a risk to their health (i.e., negative association). Considering these simultaneous positive and negative associations (i.e., high OA), we expected high SA toward the IPM and NIPM for unvaccinated participants. In contrast, we expected vaccinated people to have univalent positive attitudes toward the IPM as the preventive measures protected their health. They also were expected to have univalent negative attitudes toward the NIPM as it decreased the protection of their health. In sum, our hypotheses are:
Unvaccinated people show higher OA toward the IPM and NIPM compared to vaccinated people.
Vaccinated people evaluated the IPM positive and the NIPM negative.
Unvaccinated people show higher SA toward the IPM and NIPM compared to vaccinated people.
Method
Design
The online study was programmed and run in Enterprise Feedback Suite Survey from QuestBack and distributed via snowball system to online social networks (e.g., Facebook) as well as in Telegram groups of German skeptics of the COVID-19 preventive measures. The study was advertised with a short text including information about the study procedure and duration. Because the sampling method did not allow a prediction of the distribution of participants to the subgroups (i.e., vaccinated and unvaccinated), we strived to collect as large a sample as possible. Data were collected between January 6, 2022 and February 28, 2022.
Stimuli
Due to the sampling method via the snowball system, it seemed likely that people would participate on mobile devices. Therefore, we chose preventive measures that could be represented easily as pictograms (see Figure 1). Additionally, we chose multiple stimuli to increase reliability. This resulted in the following stimuli: face mask mandatory, COVID-19 vaccination, 3G, 2G, and 2G+. The last three represented levels of entry restrictions for non-essential businesses (e.g., restaurants) in the fourth wave of the COVID-19 pandemic in Germany. 3G means recovered from COVID-19 (German: genesen), vaccinated (German: geimpft), or being tested (German: getestet), 2G means being recovered from COVID-19 or being vaccinated, and 2G+ means being recovered from COVID-19 plus tested negative, being vaccinated plus tested negative or being vaccinated three times (see Figure 1A). For these stimuli, an inverted version was also presented representing the NIPM. This was achieved by crossing the pictograms with a red cross (see Figure 1B). In addition to these stimuli, a pictogram of the Coronavirus was also presented to the participants as well as a happy and a sad smiley as positive and negative control stimuli. Before the self-reported measures all stimuli were presented with a brief description (Figure 1 gray text) to ensure that all participants interpreted the stimuli in the same way.
Participants
A total of N = 380 participants completed the questionnaire. The exclusion of missing values led to a sample of N = 364. Based on answers to the open-ended questions indicating that participants were confused about the study or that they did not participate honestly (e.g., writing insulting comments) 25 participants were excluded. Additionally, participants rating a positive [negative] control stimulus in the lowest 25% of the positive [negative] scale or the highest 25% of the negative [positive] scale, were excluded (n = 23). To achieve the extreme groups, we excluded participants, who were considering or preparing to be vaccinated, or those with one vaccination only. This led to a final sample of N = 306 participants.
Self-Reported Ambivalence Measure
For all self-reported measures, all stimuli were presented in randomized order. Only one stimulus (i.e., pictogram) was presented per page. The stimulus was presented in the center of the screen with the question and slider below it.
Before indicating the positivity and negativity of the stimuli, participants were informed that objects might be evaluated positively regardless of their negative aspects and negatively regardless of their positive aspects. The instruction also stated that they should move the slider to the position that best described their current assessment of the stimulus. Positivity and negativity were measured consecutively. The question for positivity (negativity) was “How positively (negatively) do you evaluate the image, regardless of the negative (positive) aspects?”. Answers were indicated by moving a slider on a scale (0–100) with only the endpoints labeled not at all positive (negative) on the left and very positive (negative) on the right. McDonald’s ω indicated good reliability for the positivity scale (implemented: ω = .95, 95% CI [.94, .95]; non-implemented: ω = .90, 95% CI [.89, .93]) and the negativity scale (implemented: ω = .93, 95% CI [.91, .94]; non-implemented: ω = .88, 95% CI [.86, .9]). Therefore, the average of all stimuli depicting the IPM and the average of all stimuli depicting the NIPM was used. The SIM-Index was calculated and used for analysis of the OA (Thompson et al., 1995). Thus, 100 would indicate maximum ambivalence, −50 maximum univalence, and values around zero neutrality.
SA toward the stimuli was indicated by the extent to which participants experience conflicting thoughts or feelings. Again, they indicated their response on a slider from 0 to 100 with the endpoints labeled not at all (left) and maximally (right). Because McDonald’s ω indicated good reliability for the implemented (ω = .93, 95% CI [.92, .95]) and the non-implemented stimuli (ω = .89, 95% CI [.87, .91]), the average of all stimuli representing the IPM and the average of all stimuli representing the NIPM were used in the following analysis.
Vaccination Status
Participants had to select a statement which they most agreed with. The options were (1) I have not been vaccinated against COVID-19 and I do not plan to be vaccinated, (2) I have not been vaccinated against COVID-19, but I am considering getting vaccinated, (3) I have not yet been vaccinated against COVID-19, but I have made an appointment (e.g., vaccination bus, vaccination center), (4) I have received at least one vaccination against COVID-19 or my second vaccination was not long ago (less than 2 weeks), and (5) I have already been fully vaccinated against COVID-19 and my second vaccination was more than 2 weeks ago (cf. Prochaska et al., 1994). Responses were used to build extreme groups. Thus, only participants selecting the first (i.e., unvaccinated) or the last (i.e., vaccinated) statement were included in the analyses.
Procedure
Before the data was collected, the study was approved by the ethics committee of Trier University (EK Nr. 46/2021). In the informed consent form, participants were informed about the topic of the study, procedure, and duration. Then, they gave informed consent to data processing and data storage. When participants did not consent to the study procedure or data processing, they were thanked for their interest and informed that they could only participate after giving consent. Next, they indicated whether they used a computer or notebook for the survey and whether they used a computer mouse. Participants using a computer mouse completed a different survey collecting mouse-tracking data. However, due to a technical error, the data was not recorded correctly, so this data cannot be used. All other participants read a short description of the stimuli before they completed the self-reported ambivalence measures. Afterward, they indicated their vaccination status and answered COVID-19 specific and demographic questions (see Codebook in the supplementary materials). At the end of the survey, they had the chance to leave a comment about what elements are disregarded in the discussion surrounding the COVID-19 vaccination and the questionnaire in general. The COVID-19 specific questions and the comments about what elements are disregarded in the discussion surrounding COVID-19 vaccination were not analyzed in the current study. On the final page, participants were thanked for their participation and debriefed.
Data Analysis
The data analyses were conducted after the collection of the data. We used R (Version 4.2.3; R Core Team, 2017) for all data analysis. To investigate the hypotheses, three 2 (status: vaccianted vs. unvaccinated) × 2 (preventive measures: implemented [IPM] vs. non-implemented [NIPM]) repeated measure analysis of variance (ANOVA), with repeated measure on the last factor were calculated. When analyzing post hoc pairwise comparison in interactions Bonferroni correction was used.
To investigate the hypothesis that unvaccinated participants show greater OA compared to vaccinated participants the SIM-Index was used as dependent variable. In the SIM-Index high positive values indicated ambivalence, negative values indicated univalence, and values around zero indicated a rather neutral attitude. Hence, we expected unvaccinated participants to exhibit higher scores for the IPM and the NIPM. Whereas vaccinated participants should exhibit negative scores as we expected their attitudes to be univalent.
Because the SIM-Index only differentiates ambivalence, neutrality, and univalence it does not indicate the direction of the univalence. Thus, we further explore the direction of the evaluation by using difference scores (i.e., subtracting the ratings of the negativity scale from the positivity scale) as the dependent variable. Negative values indicate a rather negative evaluation and positive values a rather positive evaluation. Hence, we would expect negative values for vaccinated participants for the NIPM and positive values for the IPM. For separate analyses of positivity and negativity and absolute difference scores, see the supplemental material (Table S1 for means and standard deviations).
To investigate the experience of ambivalence, SA was used as dependent variable. We expect unvaccinated participants to experience greater SA compared to vaccinated participants. Based on the suggestion of a reviewer, we use multilevel modeling to further explore the relationship between OA and SA. Supplementary materials include stimuli, codebook, data, and analysis script and are available on Open Science Framework (OSF) https://osf.io/7emh8/.
Results
The final sample consisted of N = 306 (Mage = 31.7 years, SD = 12.68, range = 18–73, 96 males, 208 female, and 2 non-binary) with n = 170 unvaccinated (Mage = 31.06 years, SD = 11.54, range = 18–73, 57 males, 111 female and 2 non-binary) and n = 136 vaccinated participants (Mage = 32.5 years, SD = 13.97, range = 18–68, 39 males, 97 female).
Objective Ambivalence
See Table 1 for the means and standard deviations of the different measures. In contrast to our prediction, unvaccinated participants had rather univalent whereas vaccinated participants had rather neutral attitudes, F(1, 304) = 30.38, p < .001, 
= .07. Additionally, results indicated more univalence for the IPM and rather neutral attitudes for the NIPM, F(1, 304) = 37.31, p < .001, 
= .03. These main effects were qualified by the significant interaction of vaccination status and preventive measure, F(1, 304) = 51.06, p < .001, 
= .04. Contrary to our hypothesis that unvaccinated people show more ambivalence regarding the IPM and the NIPM, unvaccinated participants showed more univalent attitudes to the IPM than to the NIPM (p < .001). Furthermore, instead of univalent attitudes toward the IPM and the NIPM, vaccinated participants were rather neutral regardless of whether the preventive measures were implemented or not implemented (p > .05). Inconsistent with our hypothesis, unvaccinated participants showed more univalence regarding the IPM compared to vaccinated participants (p < .001). Regarding the NIPM, there was no difference between the groups (p > .05; see Figure 2A). Thus, both groups rated the NIPM rather neutrally. In sum, results from the OA measurement provided unexpected results. Whereas vaccinated people were quite neutral regarding either the IPM or NIPM, unvaccinated people were also quite neutral regarding the NIPM but held strong attitudes regarding the IPM.
Difference Score
Unvaccinated participants showed more negative attitudes compared to vaccinated participants, F(1, 304) = 45.21, p < .001, 
= .037. Additionally, results indicated that the IPM was more negative compared to the NIPM, F(1, 304) = 28.15, p < .001, 
= .064. However, these effects were qualified by the interaction of vaccination status and preventive measure, F(1, 304) = 503.93, p < .001, 
= .552. This disordinal interaction indicated that unvaccinated people held strong negative attitudes toward the IPM and rather positive attitudes (p < .001) toward the NIPM. Consistent with our hypothesis vaccinated participants showed positive attitudes regarding the IPM and negative attitudes regarding the NIPM (p < .001; see Figure 2B). In sum, these findings were consistent with the idea of a polarized society in which unvaccinated and vaccinated people express opposite attitudes regarding the IPM and NIPM. However, results from the OA measure only indicated clearly univalent attitudes for unvaccinated people toward the IPM.
Based on these findings, we were interested in the results of the SA measurement. Literature (e.g., van Harreveld et al., 2015) and the MAID model (van Harreveld et al., 2009) proposed that OA is a prerequisite for SA. Therefore, we would not expect SA in the groups either for the IPM or the NIPM.
Subjective Ambivalence
Inconsistently, we found significant differences in SA. Results indicated that unvaccinated participants experienced more ambivalence than vaccinated participants, F(1, 304) = 9.05, p = .003, 
= .018. There was no significant difference between the IPM and the NIPM (p = .409) but the interaction of vaccination status and preventive measures was significant, F(1, 304) = 142.45, p < .001, 
= .155. Inconsistent with the hypothesis that OA is a prerequisite for SA, unvaccinated people showed more SA toward the IPM than the NIPM (p < .001). In contrast, vaccinated people experienced more SA toward the NIPM than the IPM (p < .001). Thus, even though unvaccinated people exhibited univalent negative attitudes (i.e., negative OA and negative difference score) toward the IPM, they experienced high SA toward the IPM. Their SA toward the IPM was significantly higher than the SA of vaccinated participants toward the IPM (p < .001) and the SA of vaccinated participants toward the NIPM (p = .03). The SA for vaccinated IPM and unvaccinated NIPM did not differ (p = .42; all other ps < .05; see Figure 2C).
Multilevel Analysis
The results of the OA and SA measures seem inconsistent with the idea that OA is a prerequisite for SA (e.g., van Harreveld et al., 2009, 2015). To further investigate the relationship between OA and SA, vaccinated and unvaccinated people were examined separately in two maximum likelihood multilevel model analyses. Participants were included as random intercept factors. Building the model stepwise, first OA was added as a predictor for SA, followed by preventive measures and the interaction of OA and preventive measures. Interestingly, only the addition of OA as a predictor for SA did not improve the model compared to the model with the random intercepts (p = .24) for unvaccinated participants whereas it significantly improved the model for vaccinated participants (p < .001). The model with OA, preventive measures, and the interaction of OA and preventive measures predicting SA was significantly better compared to the other models for both, vaccinated and unvaccinated people (see Table S2).
In the final models, the main effects for preventive measure (IPM as reference category) replicate the results of the ANOVA with SA as a dependent variable. That is, for vaccinated people IPM elicited lower SA (b = −22.43, t[1,226.29] = −15.2, p < .001), and for unvaccinated people the IPM elicited higher SA (b = 21.18, t[1,558.94] = 11.65, p < .001) compared to the NIPM. OA did not significantly predict SA (p = .104) for vaccinated participants, however, for unvaccinated participants, OA predicted SA (b = 0.19, t[1,626.17] = 7.36, p < .001). The interaction of OA and preventive measures significantly predicted SA for vaccinated (b = 0.14, t[1,273.82] = 3.31, p < .001) and unvaccinated participants (b = −0.3, t[1,580.8] = −6.76, p < .001).
To investigate the interactions, simple slopes of low (−1 SD), average, and high (+1 SD) OA and for IPM and NIPM were explored (see Table S3 for full simple slopes results). For vaccinated participants, simple slopes for the IPM (b = 0.19, t[1,358.7] = 6.01, p < .001) indicated that SA increased with OA. However, the effect for NIPM was not significant (p = .104; see Figure 3A). For unvaccinated participants, all simple slopes were significant. As expected, SA increased with OA for the NIPM (b = 0.2, t[1,626.17] = 7.36, p < .001). Inconsistent with the assumption that OA is a prerequisite for SA, SA decreased when OA increased for the IPM (b = −0.11, t[1,612.22] = −2.77, p = .006; see Figure 3B).
In sum, even though adding OA as a predictor for SA was significantly better than the random intercepts-only model, in the final model OA did not predict SA for vaccinated participants. Simple slopes only revealed the simultaneous increase of OA and SA for the IPM, but not for the NIPM. For unvaccinated participants, the results were reversed. Adding OA to the model did not improve the prediction. However, in the final model, OA predicted SA. Simple slopes of the interaction revealed that OA and SA increased simultaneously for the NIPM, however, this effect was reversed for the IPM – when OA increased SA decreased.
Discussion
In the present study, we examined the attitudinal ambivalence toward the (non-)implementation of COVID-19 preventive measures in vaccinated and unvaccinated German citizens. Preventive measures like lockdowns came with many constraints like the freedom of movement, however, they simultaneously protected health – especially for unvaccinated people. Therefore, we expected unvaccinated people to have ambivalent attitudes toward the IPM as well as the NIPM and vaccinated people to have univalent positive attitudes toward the IPM and univalent negative attitudes toward the NIPM. Specifically, we investigated whether people simultaneously hold positive and negative associations (i.e., OA) and if they experienced conflict (i.e., SA) toward the IPM and NIPM.
In contrast to the hypothesis (H1) that unvaccinated people would express higher OA toward the IPM and NIPM, we found that unvaccinated people expressed rather neutral attitudes toward the NIPM. Even though we expected positive aspects (e.g., protection of their health) as well as negative aspects (e.g., reduced freedom of movement) to influence the evaluation of the IPM, unvaccinated people only showed strong negative attitudes toward the IPM. This is consistent with the public view perceiving unvaccinated people as strongly opposing preventive measures. Consistent with our second hypothesis, vaccinated people showed rather positive evaluations toward the IPM and rather negative evaluations toward the NIPM in the difference score, however, the OA measure indicated rather neutral attitudes. This neutrality can be explained by the limited effects the (non-)implementation of the restrictions has for them – almost no restriction of freedom of movement independent of the IPM or NIPM. Inconsistent with our third hypothesis, we did not find that unvaccinated people show higher SA toward the IPM and NIPM compared to vaccinated people. We only found that unvaccinated people showed higher SA toward the IPM compared to vaccinated people. Regarding the NIPM unvaccinated people showed lower SA compared to vaccinated people.
Our findings are not only inconsistent with our hypothesis but also with the assumption that OA is a prerequisite for SA (van Harreveld et al., 2009, 2015). Even though the OA measure indicated univalent negative attitudes for unvaccinated people toward the IPM, the SA measure indicated high ambivalence. Furthermore, OA results indicated quite neutral attitudes of vaccinated people toward the IPM and the NIPM, however, vaccinated people showed higher SA for the NIPM. Thus, we find a dissociation between OA and SA for vaccinated and unvaccinated participants. Simple slopes analysis to explore the significant interaction in the multilevel models replicated this dissociation of OA and SA. For vaccinated people, they did not indicate a significant relation between OA and SA for the NIPM. For unvaccinated participants when OA increased SA decreased for the IPM.
These dissociations between OA and SA for the IPM in unvaccinated people and for the NIPM in vaccinated people can be explained by personal relevance (e.g., Thompson & Zanna, 1995). The personal relevance of the preventive measures is particularly evident in the decline in well-being at the beginning of the pandemic (Lukács, 2021; Metin et al., 2021) when social isolation was one of the only preventive measures. However, at the time of our data collection most preventive measures were associated with vaccination, unvaccinated people were more strongly affected when the preventive measures were implemented (i.e., restriction of freedom of movement) than when the preventive measures were not implemented. Inversely, vaccinated people were more strongly affected when the preventive measures were not implemented (i.e., increased risk of infection) than when they were implemented. Thus, it can be assumed that SA rather than OA increases with the personal costs the IPM or NIPM poses.
Limitations
Even though measuring positivity and negativity on separate unipolar scales and combining them into an index has clear advantages it also has limitations. For example, there are multiple formulae to combine positivity and negativity into an ambivalence index (Thompson et al., 1995). The SIM-Index is the most used (e.g., Buttlar et al., 2023; Schneider et al., 2021; Schneider & Mattes, 2021) and recommended index (Thompson et al., 1995), however, it has also been associated with producing moderator effects that are statistical artifacts (Ullrich et al., 2008). Additionally, people often think in bipolar terms (Thompson et al., 1995) – they view positivity and negativity as complementary to each other. Because people strive for consistency (Festinger, 1957), they might rate an object low on positivity, if they rate it high on negativity. This was also reflected in some comments, for example, “the positivity and negativity question in the beginning was redundant”. Because we excluded participants indicating they did not understand the measure, we believe this has limited influence on the results. If all people viewed positivity and negativity as complementary, we would expect univalence in all conditions, however, data only indicate univalence in unvaccinated people in the IPM. Therefore, the benefits of measuring positivity and negativity on separate unipolar scales, that is, a differentiation between univalence, neutrality, and ambivalence, outweigh the limitations.
Closely related to the measurement is the proposed causality of the effects. Even though theory (van Harreveld et al., 2009, 2015) proposes a causal relationship between OA and SA, the proposed causal relationship cannot be clearly determined in this study. An experimental study would be necessary to determine a causal effect.
Additionally, during the time of data collection, nationwide preventive measures were adapted three times (Die Bundesregierung, 2022). This might have led to different evaluations of the IPM and NIPM at different time points. However, the first relaxation of the measures were implemented in March 2022. Because we terminated data collection in February 2022, we believe that this has only limited effects on our data.
The sample of the current study was German and cultural differences might limit the generalization of the results. For example, Kriesi and Oana (2023) found that even though the preventive measures implemented by the government of France and Germany were similar, in Germany were more protests compared to France. Another factor shown to influence compliance with preventive measures is collectivism versus individualism. For example, Feng and colleagues (2023) compared 79 countries and regions and found that people in individualistic countries showed higher mobility (e.g., visiting parks) despite social distancing rules. The higher compliance with preventive measures in collectivistic countries might be due to the higher normative influence in these cultures (e.g., Savani et al., 2015). According to the Theory of Planned Behavior, norms influence intentions which in turn influence behavior (Ajzen, 1991). The influence of norms should be amplified if people experience ambivalence because they use norms to reduce ambivalence and show behavioral intentions consistent with the norm (Hohman et al., 2016). Because Germany is a rather individualistic culture (e.g., Darwish & Huber, 2003), this effect might be limited. Thus, participants from collectivistic cultures might experience less SA because they use norms to reduce it. In sum, cultural (e.g., collectivistic vs. individualistic orientation) as well as context factors (e.g., mortality rate) can differ between countries and influence acceptance of and compliance with preventive measures. Even though cultural and context factors might have influenced our results, ambivalence has been shown for multiple health-related behaviors in different countries (e.g., Armitage & Conner, 2000; Hohman et al., 2014; Lipkus & Noonan, 2017; Schneider et al., 2015, 2021).
Implications for Behavior Interventions and Health Psychology
As outlined before, investigating ambivalence in the public health context is particularly relevant because of its behavioral implications (Cooke & Sheeran, 2004). Especially in the health context, people often are aware of the negative consequences of a behavior, but perform the behavior anyway (e.g., Trofor et al., 2019). Ambivalence might help explain this inconsistency between attitude and behavior and might be a good starting point for behavior interventions. For example, in motivational interviewing ambivalence indicates potential for change, and resolving this ambivalence should lead to behavior change (Miller & Rollnick, 2002). Additionally, ambivalence theory and research indicate that higher SA might indicate the potential for attitude (Petty et al., 2006) and behavior change (Armitage & Conner, 2000). Thus, the high SA toward the IPM in unvaccinated people might indicate a potential to increase compliance in the future. Even though the OA indicated negative attitudes, high SA is associated with people preparing behavior (Armitage & Arden, 2007). Thus, highlighting the positive consequences of adhering to preventive measures might help unvaccinated people reduce their SA and thus might lead to behavior change. In line with the idea of motivational interviewing that clients must resolve their ambivalence autonomously, research indicated that voluntarily adhering to preventive behavior (e.g., wearing a mask voluntarily) is favorable to implementing mandatories (e.g., mask mandatory). For instance, an imagined COVID-19 vaccination mandatory was associated with reactance which in turn was associated with reduced uptake of other voluntary vaccinations (Sprengholz et al., 2021, 2022).
The measures to prevent the spread of COVID-19 are one example of health behavior in which decisions have high personal costs. This is symptomatic of many decisions in the health context that elicit involvement but also possibly detrimental consequences and therefore ambivalence. For instance, undertaking an operation might lead to a better quality of life but simultaneously risk complications. Thus, the decision of whether to undertake an operation or not might lead to ambivalence (Kuhn et al., 1989). From ambivalence toward condom use (Kane, 1990) and cancer (Abhyankar et al., 2011) to ambivalence toward a global pandemic (Schneider et al., 2021), ambivalence seems to be omnipresent in the health-related context.
Because of the omnipresence of ambivalence in the health-related context and based on our findings, three clear-cut implications for health psychology regarding ambivalence can be derived. Firstly, the results show that attitudes toward the IPM and the NIPM are not (always) symmetric – unvaccinated people are univalent negative toward the IPM but rather neutral toward the NIPM. Therefore, not only the attitude toward an intervention or behavior should be measured but also the attitude toward the non-implementation of an intervention or behavior should be measured, especially if the non-implementation has severe health consequences, like the non-implementation of COVID-19 preventive measures. Similar effects of the non-implementation are ubiquitous in other health areas. For example, a smoker might evaluate smoking positively, however, this does not automatically mean that s/he evaluates not smoking as negative.
Secondly, the present results highlight the benefits of measuring positivity and negativity on separate unipolar scales, thus allowing differentiation of univalence, neutrality, and ambivalence. This is important because univalence, neutrality, and ambivalence influence behavior differently. Furthermore, health behavior has high personal involvement, and no health decision is without possible side effects. This means that the existence of ambivalent attitudes is very likely.
Finally, the results show that OA and SA are not always linked, therefore OA as well as SA should be measured. High personal involvement in health decisions is inevitably linked with potentially high personal costs. These high personal costs might lead to a dissociation between OA and SA. For example, for unvaccinated people when OA increased SA decreased for the IPM. When only OA or only SA is measured such a dissociation cannot be found. Therefore, not only the simultaneous positive and negative associations (i.e., OA) should be measured, but also the experienced conflict (i.e., SA).
The authors wish to thank all participants for their valuable contributions to this study as well as Charlotte Gudat and Hanna Knisch for their help in conducting the study. We are also thankful for the helpful feedback on previous drafts of the manuscript by Benjamin Buttlar and Tarini Singh.
References
(2011). Why values elicitation techniques enable people to make informed decisions about cancer trial participation. Health Expectations, 14(s1), 20–32. https://doi.org/10.1111/j.1369-7625.2010.00615.x
(1991). The theory of planned behavior. Organizational Behavior and Human Decision Processes, 50(2), 179–211. https://doi.org/10.1016/0749-5978(91)90020-T
(2007). Felt and potential ambivalence across the stages of change. Journal of Health Psychology, 12(1), Article
1 . https://doi.org/10.1177/1359105307071749(2000). Attitudinal ambivalence: A test of three key hypotheses. Personality and Social Psychology Bulletin, 26(11), Article
11 . https://doi.org/10.1177/0146167200263009(2023). The meat ambivalence questionnaire: Assessing domain-specific meat-related conflict in omnivores and veg*ans. Collabra: Psychology, 9(1), Article
73236 . https://doi.org/10.1525/collabra.73236(2021). Food for thought: Investigating communication strategies to counteract moral disengagement regarding meat consumption. Environmental Communication, 15(1), Article
1 . https://doi.org/10.1080/17524032.2020.1791207(2019). Understanding ambivalence toward organ donation and transplantation: An exploratory study of nursing students. Nurse Education Today, 76, 191–195. https://doi.org/10.1016/j.nedt.2019.02.008
(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), Article
2 . https://doi.org/10.1348/0144666041501688(2003). Individualism vs. collectivism in different cultures: A cross-cultural study. Intercultural Education, 14(1), 47–56. https://doi.org/10.1080/1467598032000044647
. (2022). Corona: Basis-Schutz und Hotspot-Maßnahmen | Bundesregierung
[Corona: Basic protection and hotspot – Measures | Federal Government] . Die Bundesregierung informiert | Startseite. https://www.bundesregierung.de/breg-de/themen/coronavirus/corona-regeln-und-einschrankungen-1734724(2023). Cultural antecedents of virus transmission: Individualism is associated with lower compliance with social distancing rules during the COVID-19 pandemic. Journal of Personality and Social Psychology, 124(3), 461–482. https://doi.org/10.1037/pspa0000322
(1957). A theory of cognitive dissonance. Stanford University Press.
(2021). Alles Covidioten? Politische Potenziale des Corona-Protests in Deutschland
[All covidiots? Political potential of the Corona-protests in Germany] (WZB Discussion Paper No. ZZ 2021-601) . Wissenschaftszentrum Berlin für Sozialforschung (WZB). https://www.econstor.eu/handle/10419/234470(2023). Ambivalence towards the implementation of preventive measures in (un-)vaccinated German citizens [Data, Materials, Supplements]. https://osf.io/7emh8
(2022). Polarisierung auf der Straße – Die politischen Folgen der Dynamiken von Protest und Gegenprotest
[Polarization on the street – The political consequences of the dynamics of protest and counter-protest] . Forschungsjournal Soziale Bewegungen, 35(2), 336–350. https://doi.org/10.1515/fjsb-2022-0023(2016). Attitude ambivalence, social norms, and behavioral intentions: Developing effective antitobacco persuasive communications. Psychology of Addictive Behaviors, 30, 209–219. https://doi.org/10.1037/adb0000126
(2014). Attitude ambivalence, friend norms, and adolescent drug use. Prevention Science, 15(1), Article
1 . https://doi.org/10.1007/s11121-013-0368-8(1990). AIDS, addiction and condom use: Sources of sexual risk for heterosexual women. The Journal of Sex Research, 27(3), 427–444. https://doi.org/10.1080/00224499009551570
(2023). Protest in unlikely times: Dynamics of collective mobilization in Europe during the COVID-19 crisis. Journal of European Public Policy, 30(4), 740–765. https://doi.org/10.1080/13501763.2022.2140819
(1989). Ambivalence in cardiac transplantation candidates. The International Journal of Psychiatry in Medicine, 18(4), 305–314. https://doi.org/10.2190/ARHL-YGD4-T9F5-5UYV
(2017). Association between felt ambivalence and the desire to quit waterpipe use among college students. Journal of Health Psychology, 22(14), 1780–1788. https://doi.org/10.1177/1359105316636948
(2021). Mental well-being of university students in social isolation. European Journal of Health Psychology, 28(1), 22–29. https://doi.org/10.1027/2512-8442/a000065
(2009). Review of motivational Interviewing in promoting health behaviors. Clinical Psychology Review, 29(4), 283–293. https://doi.org/10.1016/j.cpr.2009.02.001
(2020). Effect of risk messages on risk appraisals, attitudes, ambivalence, and willingness to smoke hookah in young adults. Health Psychology and Behavioral Medicine, 8(1), 96–109. https://doi.org/10.1080/21642850.2020.1730844
(2021). Subjective well-being and resilience during COVID-19 pandemic in Turkey. European Journal of Health Psychology, 28(4), 152–160. https://doi.org/10.1027/2512-8442/a000081
(2002). Motivational interviewing: Preparing people for change (2nd ed.). Guilford Press.
(2002). A motivational interviewing perspective on resistance in psychotherapy. Journal of Clinical Psychology, 58(2), 185–193. https://doi.org/10.1002/jclp.1142
(2019). Ambivalence toward healthy and unhealthy food and moderation by individual differences in restrained eating. Appetite, 140, 309–317. https://doi.org/10.1016/j.appet.2019.05.033
(2006). Implicit ambivalence from attitude change: An exploration of the PAST model. Journal of Personality and Social Psychology, 90(1), Article
1 . https://doi.org/10.1037/0022-3514.90.1.21(1994). Stages of change and decisional balance for 12 problem behaviors. Health Psychology, 13(1), 39–46. https://doi.org/10.1037/0278-6133.13.1.39
. (2017). R: A language and environment for statistical computing. R Foundation for Statistical Computing. https://www.r-project.org/
(2015). When norms loom larger than the self: Susceptibility of preference–choice consistency to normative influence across cultures. Organizational Behavior and Human Decision Processes, 129, 70–79. https://doi.org/10.1016/j.obhdp.2014.09.001
(2021). Ambivalence and self-reported adherence to recommendations to reduce the spread of COVID-19. Social Psychology, 52(6), 362–374. https://doi.org/10.1027/1864-9335/a000465
(2021). Mix is different from nix: Mouse tracking differentiates ambivalence from neutrality. Journal of Experimental Social Psychology, 95, Article
104106 . https://doi.org/10.1016/j.jesp.2021.104106(2015). The path of ambivalence: Tracing the pull of opposing evaluations using mouse trajectories. Frontiers in Psychology, 6, Article
996 . https://doi.org/10.3389/fpsyg.2015.00996(2004). Predictors and predictive effects of ambivalence. British Journal of Social Psychology, 43(3), 371–383. https://doi.org/10.1348/0144666042037980
(2021). Reactance revisited: Consequences of mandatory and scarce vaccination in the case of COVID-19. Applied Psychology: Health and Well-Being, 13(4), 986–995. https://doi.org/10.1111/aphw.12285
(2022). Vaccination policy reactance: Predictors, consequences, and countermeasures. Journal of Health Psychology, 27(6), 1394–1407. https://doi.org/10.1177/13591053211044535
(1995). Let’s not be indifferent about (attitudinal) ambivalence. Attitude Strength: Antecedents and Consequences, 4, 361–386.
(1995). The conflicted individual: personality-based and domain specific antecedents of ambivalent social attitudes. Journal of Personality, 63(2), Article
2 . https://doi.org/10.1111/j.1467-6494.1995.tb00810.x(2019). Knowledge of the health risks of smoking and impact of cigarette warning labels among tobacco users in six European countries: Findings from the EUREST-PLUS ITC Europe Surveys. Tobacco Induced Diseases, 16, Article
10 . https://doi.org/10.18332/tid/99542(2008). The moderator effect that wasn’t there: Statistical problems in ambivalence research. Journal of Personality and Social Psychology, 95(4), Article
4 . https://doi.org/10.1037/a0012709(2015).
Chapter Five - The ABC of Ambivalence: Affective, Behavioral, and Cognitive Consequences of Attitudinal Conflict . In J. M. OlsonM. P. ZannaEds., Advances in experimental social psychology (Vol. 52, pp. 285–324). Academic Press. https://doi.org/10.1016/bs.aesp.2015.01.002(2009). The agony of ambivalence and ways to resolve it: Introducing the MAID model. Personality and Social Psychology Review, 13(1), Article
1 . https://doi.org/10.1177/1088868308324518(2018). Using motivational interviewing to manage process markers of ambivalence and resistance in cognitive behavioral therapy. Cognitive Therapy and Research, 42(2), 193–203. https://doi.org/10.1007/s10608-017-9857-6
