1.1. A differentiated view of trust in science

Diverse social science disciplines have studied trust which is why many definitions and approaches co-exist [e.g., 32]. Commonly, research distinguishes between two different conceptualisations of trust [see 33]. One conceptualisation is behavioural trust as willingness to make oneself vulnerable [34], thus focusing on aspects of risk and uncertainty [33]. The second strand of research views trust as “confident positive expectations regarding another’s conduct” [35] by focusing on dimensions of trust [[15,36]; also referred to as trustworthiness perceptions, see [34,37]]. This study focuses on the second conceptualisation of trust and argues for the necessity to consider different dimensions to help developing specific communication strategies [15,34,36]. Nevertheless, rather than further fragmenting trust research, a meta-definition should combine both main ideas in the sense that trust can be understood as positive expectations towards an object of trust in the face of risks [33]. This means that trust is related to the willingness to be vulnerable to the risk that the object of trust does not meet addressed expectations [primarily based on 37–42]. In this study, we follow the understanding that trust in science is a key variable in the relationship between the public (subject of trust) and science (object of trust). We specifically consider epistemic trust in science [43], which refers to the dependence on scientific experts’ knowledge and the risk of being misinformed. Public trust in science is the public’s assumption that science offers reliable (and relevant) information [44]. To ensure this, scientific expertise, integrity and benevolence are required [45], referred to here as dimensions of trust in science. Based on the idea of public trust [37,46], communication research has pointed to the increased importance of media experiences with representatives of science, i.e., their transparency and openness to conversations. This is also reflected by calls of science communication research [15,36,47] and practice [48]. Especially against the backdrop of altered media environments, direct communication is increasingly enabled and insights into processes, contexts and backgrounds are provided that thereby become part of public expectations. Thus, the dimensions of transparency and dialogue orientation appear crucial for an epistemic trust in science and experts [21]. Based on sociological as well as communication research literature, trust in science can be placed on different levels including the macro-level (science as a functional system), the meso-level (scientific organisations) and the micro-level [scientists; see [38–40,42]]. This study focuses on the micro-level.

Rather than treating the public as a single entity [15,49], we highlight the importance to distinguish and examine different groups as suggested in current research on public perceptions of S&T that promotes segmentation approaches [for an overview, 50]. Based on either socio-demographic, psychographic or behavioural variables [51], previous studies have found four to six audiences of science communication in Europe or the USA [52–54]. Trust in science has been central for labelling or describing such population segments [54,55], despite neglecting it as a variable identifying the groups. However, considering group-specific effects regarding public trust in science as a key variable has the potential to inform tailored science communication strategies [16] and offers deeper empirical insights into the question of which theoretical types of trust in science are prevalent within the public – an issue extensively debated within science communication literature. For example, individuals’ trust in science may primarily be based on their perception of scientists’ expertise (competence trust [56,57]) or their benevolence (goodwill trust [56]or trust in motives [57]). Furthermore, from a different perspective, blind trust in science and scientific information [58], meaning complete and uncritical confidence, may not be ideal from a societal perspective. Informed or cognitive trust [44,59], as well as a general recognition of science and its function, accompanied by a fundamental trust in the performance of science, may be more desirable. In sum, a differentiated view on public trust in science is needed – which means considering various reasons, i.e., dimensions, and levels of trust by a group-specific approach.

In ZA, little is known regarding public trust in science; current research, however, provides relevant information on general perceptions of science. In contrast to Western countries such as the USA, where people are less likely to be concerned about the negative impacts of science the more optimistic they are [28,60], in ZA, beliefs about the promises of science are positively correlated with reservations about it [29,30]. Hence, South Africans with high levels of trust in science may believe the most in promises while also having strong reservations. Furthermore, religiosity plays a significant role in ZA [27], and many individuals are culturally distant from science [26]. These personal characteristics, which influence general attitudes towards science, may also be negatively associated to people’s trust in science. Culturally, however, ZA is characterised by a high score on Hofstede’s dimension of power distance, indicating that individuals with lower social authority tend to accept and anticipate an uneven distribution of power. Additionally, ZA is considered a collectivist society, emphasising collective needs over individualistic interests [61]. Based on these cultural dimensions, one might expect a high level of public trust in science within ZA. However, given the unique cultural context and the diverse potential assumptions regarding public trust, different sub-groups may exist that vary in their levels of trust in science, as well as in related factors such as attitudes toward science and the importance attributed to religiosity. Focusing on the South African public’s view on science during the COVID-19 pandemic, a time of high risk, we ask the following research question (RQ):

RQ1: Which groups of ZA online users can be distinguished regarding trust in science during the COVID-19 pandemic and how do they differ regarding socio-demographics and attitudes towards science?

1.2. Diverse types of contact with science

Access points form the relationship between science and the public [38]. Science communication and different ways of contact with science have, thus, been predicted to be related to trust in science, which is plausible because science communication not only provides information about science but also different trust cues [62]. At the same time, intermediaries of trust complicate the trust relationship between science and the public, as they themselves are objects of trust [37,62]. In public surveys, three main types of contact are commonly considered: direct or first-hand information, information obtained from intermediaries – namely, social agents such as friends and family – or media outlets, including journalistic sources [15,39]. Usually, the public most frequently encounters science-related information through media reporting or by talking to family and friends [63–65]. While direct contact (e.g., talking to scientists, visiting scientific events) is the least frequent, social media have become important sources [e.g., 66]. Nevertheless, people reported that they trusted direct information more than, for example, journalistic media [49,67,68]. For information about COVID-19 in the UK, health sector sources (e.g., scientists/researchers) were the most trusted. Family and friends were trusted more than journalists/the media, who were among the least trusted information sources [69]. Regarding social media news, scholars found either more sceptical attitudes towards it than other sources [70] or that (especially young) users valued social media as a trustworthy source of information [58].

Regarding the link between the frequency of science-related media use and trust in science, the state of research is inconsistent. While some scholars [71] did not find that the frequency of science-related media use predicted trust in science, others [72] showed that internet use was negatively linked to people’s perceptions of scientists and trust in the media. Further, more frequent use of social media has been linked to stronger beliefs in topic-related [73] and general conspiracy narratives [58], which may be connected to distrust in science. In contrast, others found a positive correlation between social media news use and trust in science in a study across 20 countries [22].

Previous segmentation studies considered the frequency of use of different types of science communication [74], often to describe the groups in post-hoc analyses [54]. Usually, the higher a group’s frequency of contact with science, the more positive their attitudes and trust in science [54,74]. For ZA, population segments with a higher frequency of contact tend to be more educated and have stronger beliefs about the benefits of science than groups with less intensive contact with science [75]. The use of social media as well as people’s trust in different sources of information, however, have not been considered to date. This is particularly interesting because – partly due to a young population – South Africans are characterised by active social media use [24]. And as a county that is described by cultural values of collectivism and power distance [61], social media use may be a particularly important predictor of public trust in science, as suggested by previous research [22]. Considering all access points and trust in them, we address:

RQ2: How do groups of trust differ regarding their contact with science and trust in these information sources?

1.3. Perceived changes in trust in science

As the first pandemic of its extent “in the era of wide-spread social media” [14] and as a unique time of globalised risks, COVID-19 may have had a negative impact on public trust in science. According to their secondary analysis of the global 2018 Wellcome Trust survey, Eichengreen et al. [14] found that only people who have encountered epidemics between the age of 18–25 trust scientists less (effects on the micro level). Besides age and the severity of the encountered epidemic, results showed that the loss of trust was greater in low-income countries with weaker medical care and connected with reduced exposure to science at school. In contrast to these results, surveys early in the pandemic showed Germans’ increased trust [17,76] and stable trust among the US public [77].

ZA has encountered several epidemics in the recent past. In the case of COVID-19, ZA was among the most affected African countries [78]. In the world-wide comparison, the number of infections was high, but mortality rather low [79]. The ZA government reacted rather quickly to the pandemic by a strict lockdown, which soon has been eased [12]. Considering the ‘infodemic’ and the assumption that media users may interpret scientific information differently depending on their level of trust in science [16], the COVID-19 pandemic may have led to divergent changes in public trust. Some segments of the population with initially high trust might have perceived an increase in their trust in science, while others with low trust may have experienced further decline or destabilisation. In both scenarios, factors such as the transparency of science communication – and the communication of scientific uncertainties – likely played a significant role.

So far, no longitudinal data for actual changes in trust in science for ZA existed at the time of this study’s data collection during the COVID-19 pandemic. However, even asking for self-perceived changes may give first insights. After all, the pandemic may have shaped respondents’ answers in this study. Thus, information on perceived changes may help to further characterise and interpret which types of trust in science are prevalent among the groups. Therefore, we want to investigate:

RQ3: How do groups of trust differ in the perceptions of and reasons for changes in trust in science due to the COVID-19 pandemic?

2.1. Data collection and sample

We conducted an online survey among South Africans over 18 years old from November 11 to December 7 of 2020 (the beginning of the second COVID-19 wave). We anticipated unequal group sizes with some groups comprising as little as 5–10% of the total sample. Based on these considerations, we set a target total sample size of n = 1500 to ensure sufficient statistical power for detecting small to medium-sized differences across up to seven groups. This sample size also allowed for adequate representation of smaller groups, with an expected minimum of 75–150 participants per group. It facilitates comprehensive descriptive and comparative analyses based on various characteristics.

Participants were recruited via an online access panel by Ask Afrika. Members of the panel were randomly invited to voluntarily participate, considering quotas for gender, age, province, population group and geographical setting (Table 1). Informed written consent was obtained by Ask Afrika. Participants had the autonomy to choose whether they wished to complete our survey and were free to end the survey at any time. The survey and our exploratory approach have not been preregistered but received ethical approval from the Research Ethics Committee Social, Behavioural and Education Research of Stellenbosch University (SU project number: 19084).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Socio-demographic information (frequencies, percentages) for total sample and quota plan.

https://doi.org/10.1371/journal.pone.0340881.t001

To achieve the desired sample size, the quotas for the characteristics of hard-to-reach individuals were slightly relaxed. The sample (n = 1,624; Table 1) is not representative for the general ZA population. Fewer respondents from the oldest age group were recruited compared to official ZA statistics [80]. Women and white individuals as well as some provinces are overrepresented. The sample comprises many educated people, which may have resulted from using an online survey [81].

2.2. Material and measures

The questionnaire was administered in English and employed five-point rating scales, including an additional option: ‘I don’t know/no response’. The development of the questionnaire involved an intensive, iterative process involving all authors, followed by testing, refinement, and finalisation through multiple feedback rounds with colleagues from South Africa. Validated scales were employed wherever possible. The primary focus was on assessing trust in science and the contact with science based on theoretical ideas [15] and aligned with a comparable project conducted in Germany [21,82].

2.2.1. Trust in science and other institutions.

Different theoretical understandings of trust exist which is acknowledged by this study. Based on a theoretical model [15], we distinguished between three levels of trust in science and five dimensions [see also 21]. First, trust in science on the macro-, meso- and micro-level was assessed with direct measures (see Table 2). Respondents were also asked to directly indicate their trust in other institutions and social elites (politics, the media, religion/the church, the military) and in other people in general to give further hints about the types of trust in science prevalent among the groups. We then captured the agreement to the different reasons for respondents’ trust in scientists, with two to three items per dimension (expertise, integrity, benevolence, transparency and dialogue orientation, Table 2) as our main focus of the survey and to be used to identify different groups of trust. Our scale is similar to a recently developed scale [21] that considers and combines items and ideas of established instruments (in particular: [36,45,66,85]). We conducted a confirmatory factor analysis (see Tables A and B in S1 Appendix) because the distinction of the five dimensions was based on theoretical ideas [15] and has been empirically tested, albeit only within a German sample [21]. We tested the five-factor solution against several alternative models with fewer factors. First, a single-factor model was evaluated. Subsequently, we examined a two-factor solution suggested by exploratory factor analysis (EFA) with varimax rotation (KMO = .94, χ^² = 9548.48, df = 66, p < .001, h^² = .40–76). In this two-factor solution, the first factor explained 55% of the variance and comprised items measuring benevolence, transparency and dialogue orientation. The second factor included all items related to expertise and integrity and explained approximately 8% of the variance. Despite two double-loadings, all expertise and integrity items were retained for comparison between the two-factor and five-factor solutions. Lastly, we tested a four-dimensional model used by previous research [36] against our five-factor model. The five-factor model demonstrated a significantly better fit than all alternative models, revealing acceptable fit indices, with the exception of the RMSEA, which was slightly above the conventional cut-off of.06 (RMSEA = .07[.06,.08]). We also developed two items to measure behavioural trust as willingness to be vulnerable [84] to engage with the current academic discourse [36]: ‘I would feel comfortable to rely on scientists’ efforts to find solutions to important problems, even if I could not monitor their actions’ and ‘I would be willing to let scientists have complete control over the future of our society’ (r = .54). Furthermore, direct measures are applied to different levels to allow a comparison of the micro-, meso- and macro-level. Lastly, in an open-ended question, respondents were asked to elaborate on the extent to which they had perceived a change in their trust in science due to the COVID-19 pandemic (RQ3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Measures of trust in science and descriptive statistics (mean, standard deviation).

https://doi.org/10.1371/journal.pone.0340881.t002

2.3. Procedure

After the first questions regarding socio-demographics (Table 1), the questionnaire captured interest, attitudes and knowledge, contact with and trust in science or other institutions. The survey ended with questions regarding religiosity, educational attainment, political orientation and monthly household income. Throughout the survey, we made comprehensive efforts to ensure high response quality. Particular emphasis was placed on clear instructions for questionnaire completion. Additionally, open-ended questions were incorporated regularly to enhance engagement and stimulate thoughtful responses. These strategies aimed to maintain participant interest and attentiveness, thereby promoting the overall quality of the data collected.

2.4. Analyses

Upon data cleansing, the dataset was screened for implausible response patterns prior to analysis. To answer RQ1, we conducted latent profile analysis (LPA) in RStudio using the package tidyLPA (model 1) with mean indices of expertise, integrity, benevolence, transparency and dialogue orientation (see Table C in S1 Appendix for correlation table). The aim was to examine which trust groups exist varying in their patterns of dimensions and, thus, reasons to trust in scientists. We chose LPA as an exploratory method of data analysis because we employed continuous data. In order to avoid imputation, 83 cases were excluded due to missing values resulting in a final sample of n = 1,541. We identified the appropriate number of groups with the help of fit indices and elbow criterion (see Table D and Figure A in S1 Appendix). We stepwise increased the number of groups to be calculated until the best solution was determined. An analytic hierarchy process, based on the fit indices Akaike information criteria (AIC), Approximate weight of evidence (AWE), Bayesian information criterion (BIC) and Kullback information criterion (KIC), suggested four groups as the best solution. Therefore, we chose this solution in which each subject was assigned to one of four profiles based on posterior group membership probabilities. An additional discriminant analysis in SPSS revealed 96% of correctly predicted cases for those four groups. No group was smaller than 5% of the total sample size (n = 116–532).

One-way analyses of variance (ANOVA) with post-hoc tests (Bonferroni) were calculated using SPSS to describe the groups regarding all measured variables and to examine statistically significant differences regarding the mean values across groups. First, we focused on group differences regarding trust in science (RQ1, between and within groups), before examining the links between group memberships and the contact types (RQ2). For RQ3, we coded the open-ended question on perceived changes of trust in science due to the COVID-19 pandemic according to the direction of change addressed (positive, negative, mixed/unspecific, no/not much change, no reference), to test for group-specific differences (chi-squared test). The intercoder reliability for two independent coders and n = 150 was good (κ = .74). In cases where the two coders’ coding did not initially overlap, the cases were discussed until consensus was reached. All remaining cases were coded by a single coder. To give a deeper insight into the material, we structured and systematically and thematically summarised the aspects regarding positive and negative changes, excluding statements without references to a change in trust, e.g., general praise or critique. We extracted prime examples to use as illustration in the results section.

3.1. The ZA groups of trust in science (RQ1)

The results of the latent profile analysis (Fig 1, Table 3) revealed that, according to the mean values, the groups mainly differ regarding the general levels of trust. Therefore, the group labels are primarily based on general trust levels ranging from highest to lowest trust. We found two groups with high levels of trust in science: (1) fully trusting and (2) highly trusting. Two groups, however, were very different in their levels of trust: (3) moderately trusting and (4) rather untrusting. Although the labels were chosen to indicate how much the groups trusted, the analyses also revealed differences within the groups regarding their patterns for the five trust dimensions measured. The effect sizes indicate that the biggest differences between the groups were regarding transparency (η² = .76), followed by benevolence (η² = .68), integrity (η² = .60), dialogue orientation (η² = .59) and expertise (η² = .54). Additionally, we found a weak positive correlation between respondents’ beliefs regarding the promises of science and their reservations against science for the total sample (r = .14; p < .001).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Description of four ZA groups of trust regarding their trust in science (M (SD)).

https://doi.org/10.1371/journal.pone.0340881.t003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Patterns for dimensions of trust in scientists of the four trust groups.

Note. n = 1,541. Figure depicts mean values and confidence intervals. See Table 3 or Table E in S1 Appendix for exact mean values and standard deviation.

https://doi.org/10.1371/journal.pone.0340881.g001

1. (1). The fully trusting (n = 465; 30%) were South Africans with (almost) complete trust in science across all levels and formed one of the largest groups. They strongly agreed with all dimensions as reasons to trust scientists – most strongly with benevolence and least strongly with expertise and dialogue orientation. Most differences within subjects exist between benevolence and all other dimensions except integrity. They also had the highest willingness to be vulnerable to science and their trust in other institutions was highest among all groups but lower than their trust in science (Table 4). The fully trusting had the highest interest in and knowledge about scientific topics and processes. They agreed particularly strongly with the promises of science but were also most critical of it. They were the only group for which a significant positive correlation between beliefs in promises and reservations about science was found (r = .18; p < .001). Compared to the other groups, the fully trusting were slightly older, particularly religious and conservative in their political views. In sum, the fully trusting can be described as individuals who have complete trust in scientists’ goodwill, but also their scientific performance and communicative abilities. They strongly engage with science both emotionally and intellectually.
2. (2). The highly trusting (n= 532; 35%) had the second-highest level of trust – particularly in academics at universities. They especially valued scientists for their benevolence, followed by their integrity. Within subjects, ratings of benevolence and dialogue orientation differed the most from other dimensions. Highly trusting individuals had the second highest willingness to grant science complete control over the societal future. Their trust in religion/the church was high, whereas they were undecided about their trust in the military and media and rather did not trust politics. They were highly interested and knowledgeable. Again, the largest proportion of this group had a university degree. They believed more in science’s promises than they had reservations about it; the two contrary beliefs were not significantly correlated (r = .08; p = .061). In brief, the highly trusting are individuals with strong, fundamental trust in scientists and who value scientists’ goodwill in particular believing that science brings a better future.
3. (3). The moderately trusting (n = 428; 28%) were the youngest group and had a moderate level of trust in science which was as high as their trust in religion/the church. They trusted scientists from private research institutions and scientists in general the least. Contrary to the first two groups, they tended to agree most with expertise as a reason to trust scientists. Within this group, the ratings of dimensions differed the most; expertise ratings in particular. Their willingness to be vulnerable by giving science complete control is rather low. They agreed slightly more to the promises of science than they had reservations about them; no significant correlation was found (r = .02; p = .695). Their level of interest in and knowledge about science was moderate. Many of them belonged to the centre of the political spectrum or are rather liberal. All in all, the moderately trusting trust in science but not at any costs. They value scientists’ expertise and balance science and religion.
4. (4). The rather untrusting (n= 116; 8% ) were the minority in the sample and the least trusting with the lowest willingness to give science complete control over the future of society. They were undecided about their trust in the functional system and scientists at universities but tended not to trust scientists from private institutions. They rather disagreed with the trust dimensions as reasons to trust scientists but agreed the most regarding scientists’ expertise (difference with all other dimensions). They had a higher trust in religion/the church than in science and (rather) did not trust politics, the media or the military. The rather untrusting were the least interested and knowledgeable. They were the only group that agreed more strongly with reservations about science than with its promises. A weak negative correlation was not significant (r = −.08; p = .393). They were particularly diverse in their political orientation. The rather untrusting, in sum, exhibit low overall trust in authorities. If at all, they believe in scientists’ expertise and prioritise religion over science.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Continued description of four ZA groups of trust regarding additional perceptions and socio-demographics (M (SD)).

https://doi.org/10.1371/journal.pone.0340881.t004

3.2. Types of contact with science and trust in information sources (RQ2)

The respondents used diverse sources for science-related information – most frequently online search engines, TV, video platforms, fictional content about science and homepages of scientific institutions. They reported speaking regularly with others; conversations with scientists were least common. The respondents stated to speak even less to scientists than they participate in the online discourse about science (Fig 2 and Table E in S1 Appendix).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Information use among the four trust groups.

Note. n = 1,541. Figure depicts mean values and confidence intervals. See Table E in S1 Appendix for exact mean values and standard deviation.

https://doi.org/10.1371/journal.pone.0340881.g002

For all variables, the groups exhibited the same order in their average frequencies as they previously did in their trust in science. This means, the fully trusting used all types of contact most frequently and many of them had even worked (or were working) in science and knew scientists personally. The highly trusting reported coming into contact with science the second-most frequently; again, many had met a scientist personally. The moderately trusting were less likely and the rather untrusting were the least likely to be in contact with science in their everyday lives. The vast majority of them had never worked in science. Thus, the general patterns of information use deviated only slightly per group. This was also reflected in the groups’ preferred sources of information. While the fully trusting and the highly trusting named a variety of preferred sources of information about science, including scientific journal articles and conversations with (acquainted) scientists, the rather untrusting mainly stated using ‘the internet’ and the moderately trusting also stated online search engines, websites, TV and books.

Similarly, the groups gradually deviated regarding trust in the different types of contact. Although direct experiences were generally the least frequent, direct information from scientists was trusted the most, followed by information from journalistic media or social contacts. The respondents reported trusting science-related information on social media the least (see Table E in S1 Appendix).

3.3. Perceived changes in trust in science (RQ3)

The answers to the mandatory, open-ended question about perceived changes in respondents’ trust in science showed that many used the option to share detailed information. For example, over 1,000 respondents provided answers ranging from ten words up to a maximum of 54 words, whereas only 190 participants responded with one to three words. However, less than half of the respondents directly referred to perceived changes in their trust in science in response to the COVID-19 pandemic. Those who did, mainly thought they had experienced a decrease in trust or no/not much change. Differences emerged across the four groups of trust (Fig 3). Most of the fully trusting indicated no (significant) change, and despite the second-biggest proportion of them perceiving a loss in trust, it was the lowest of all groups. The percentage of people indicating an increase was the highest. The distribution for the highly trusting was similar with only slightly more of them indicating a loss in trust. Compared to the highly trusting, a slightly higher proportion of the moderately trusting had perceived a negative shift and slightly fewer a positive shift in their trust in science. Proportionally, most of the rather untrusting indicated a loss in trust.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Changes in trust in science across the four groups (% per group).

Note. χ² (12, n = 1,541) = 57.95, p < .001. Classification was based on responses to the open-ended question “COVID-19 has had an impact on many people in most parts of the world. To what extent has the COVID-19 pandemic affected your personal trust in science? Please elaborate below”.

https://doi.org/10.1371/journal.pone.0340881.g003

Regarding positive changes in trust, the first three groups referred to scientists’ search for a cure and praised their commitment in the vaccine development.

My personal trust in scientist[s] has not decreased because of the COVID-19 pandemic; it had in fact heightened. This is because we saw scientist[s] working around the clock day-to-day to ensure that they find a vaccine again[st] the virus to prevent more people getting infected (highly trusting, 21, female).

Respondents also stated that science had generated and provided transparent, consistent and accurate information, predictions and recommendations to the public.

It has actually increased my trust in science. Scientists have been with us throughout this journey during COVID-19. […] They have helped educate our citizens through their findings. Everything that we have learnt about the virus can be credited to scientists (fully trusting, 28, female).

They also generally praised science and referred to the government as a negative contrast or some stated highly differentiated reasons for a perceived increase in trust and referred to dependence on science for finding solutions but the possibility that scientists could also contradict each other.

The reasons for perceived negative changes in trust were more detailed and diverse. Two reasons, however, were prevalent. The first was scientists’ perceived lack of competence in finding a cure. The process of vaccine development was criticised for taking too long which was interpreted as scientists’ failure.

The fact that […] so far, there is no confirmed vaccine, or scientific prevention of the pandemic […] has casted a lot of doubt and makes me feel that science is overrated (highly trusting, 42, male).

The second one was related to the cause of the virus, which puzzled people across groups but mainly the moderately trusting. They addressed rumours about and criticised the lack of clear information about the virus’ origins. Most groups claimed the lack of communication and changing, contradictory or false information as reasons for perceiving a loss of trust – first and foremost, the rather untrusting.

Much is not shared with the public, much is not properly explained, and this exacerbates my inherent mistrust of both political and scientific institutions (rather untrusting, 33, female).

Other reasons were the perceived lack of science’s independence from the government or industry, the effects of the pandemic, missed strategies to foresee and cope with the pandemic or perceived mistakes and lies told by scientists. People of the rather untrusting were most critical and even referred to a conspiracy narrative that scientists or other authorities had developed the virus.

My trust in science has been severely shaken due to the COVID-19 pandemic. I believe that this entire pandemic is a tool that is being used for the governments and other connected organizations around the world to control the population through fear and misinformation. I do not believe that COVID-19 is the deadly virus that we have all been told it is [but] that it is just a smoke-screen for a much larger, more sinister agenda. I believe that scientists created the virus and at the same time created the ‘vaccine’, and all the talk about the ‘vaccine’ being in development is a load of garbage (rather untrusting, 45, female).