Throughout the spring and summer of 2020, there was extensive news coverage, social media debate, and public health messaging about mask wearing and COVID-19 (Godoy, 2020; Lopez, 2020). At the beginning of the pandemic in March and April, much of the debate and uncertainty was focused on whether face masks protect the wearer, or if a person who is positive for COVID-19 is less likely to infect others by wearing a mask (Esposito et al., 2020; Lopez, 2020). There was concern about global shortages of personal protective equipment for frontline healthcare workers (Esposito et al., 2020) amidst reports of members of the public stockpiling personal protective equipment, and turmoil in the United States around the production and distribution of masks (Nguyen, 2020). Some commentators believed that if the public were advised to wear face masks, people could accidentally infect themselves through improper donning and doffing of the mask, exacerbating rather than reducing overall spread of the virus (Howard, 2020; Matuschek et al., 2020).

By May 2020, the scientific consensus supported the idea that mask wearing can effectively limit the spread of COVID-19, largely by keeping infectious particles inside the mask and thus protecting others from becoming infected (Howard et al., 2020). Nonetheless, as health experts around the world began issuing formal recommendations for mask wearing, large segments of the public appeared to express confusion about this changing stance or “about-face” (Urback, 2020).

Even in May 2020, several months after the World Health Organization (WHO) had declared COVID-19 a Public Health Emergency of International Concern (WHO, 2020), there were still questions and gaps in the evidence base about whether cloth masks protect the people wearing them, and the degree of protection offered by different mask materials. There were no studies specifically examining the utility of cloth masks in the community (Clase et al., 2020). However, as Canada's chief public health officer Dr. Theresa Tam explained in April, although the science was not certain, encouraging members of the public to wear masks seemed to be “a sensible thing to do” based on emerging reports and studies (Tasker, 2020).

Indeed, COVID-19 has proven to be infectious before the onset of symptoms, and asymptomatic spread has been documented (Zhao et al., 2020). Masks are now widely viewed by the medical and scientific communities as valuable tools that will help reduce the spread of COVID-19, especially as communities around the world re-open and in contexts where physical distancing is difficult or not possible. Yet as many countries re-open and economies start back up, face masks have continued to be surrounded in controversy and public debate. In the U.S. and other parts of the world, mask wearing has become a political statement. Some of the so-called “anti-maskers” refute the growing evidence base that widespread adoption of mask-wearing in public can limit the spread of COVID-19. Others contend that mask mandates - laws or regulations that enforce mask-wearing in public spaces - are “authoritarian” and infringe on personal rights and freedoms (Stewart, 2020).

On September 12^th, 2020, several thousand people gathered in downtown Montreal to protest the provincial government's response to the pandemic. Protestors held signs that indicated support for conspiracy theories, such as a supposed link between COVID-19 and the 5G cellular network, or perceived corruption at the World Health Organization (Montpetit & MacFarlane, 2020). This protest fell on the same day that new police powers came into effect in the province, giving officers the authority to fine anyone not wearing a mask inside a public building. Other countries have seen similar protests motivated by anti-mask sentiment or opposition to other public health measures (Kim, 2020).

This opposition to masking is not uniform: public sentiment around masking appears to vary between different countries and may be connected to political affiliation (Alberga et al., 2020; Pew, 2020). In other words, public sentiment on mask wearing eludes easy characterizations. Much like the underlying evidence on the effectiveness of mask wearing as a public health intervention to contain COVID-19, public sentiment has seemingly evolved over time. Some segments of the public may be opposed to masking based on distrust of health authorities and other institutions, whereas others express doubt about the efficacy of facemasks, or appear to subscribe to conspiracy theories that state COVID-19 is a “hoax” or that the seriousness of the virus has been overblown (Stewart, 2020). Conversely, many Canadians have expressed disappointment in government leaders and health agencies for not recommending widespread masking earlier in the pandemic (Mahal, 2020).

In a short interval of time, competing narratives and explanations have emerged about the political motivations behind some “anti-masker” beliefs, and some research has explored the dynamics of misinformation spreading online (Kouzy et al., 2020) and the effectiveness of different public health messaging and strategies (Jordan et al., 2020).

Social media platforms such as Twitter have been viewed as sites that have the potential to shape public perceptions by exposing people to different messages and communities. In addition to possibly influencing beliefs, such platforms can also provide valuable insight into public sentiment and in turn, help develop more effective public health messaging and intervention strategies (Keim-Malpass et al., 2017; Scanfeld et al., 2010). This study attempts to better understand the public sentiment around mask-wearing and the dynamics of (mis)information spread on social media. We examined several prominent mask-related hashtags over the course of three months. Public perception is important because it reflects concerns, beliefs, and values, and identifies the information required to address gaps in understanding. Further, awareness of public opinion on health topics helps ensure that services and policies are aligned with the beliefs and priorities of the general public (Giles & Adams, 2015). This study documents public perceptions and beliefs surrounding mask-wearing during the COVID-19 pandemic and provides information about the nature and prevalence of different messaging strategies and the stated motivations for supporting or opposing mask-wearing.

Past research has extensively documented the spread of health-related misinformation, sensationalism, and public fear (Sell et al., 2020), noting the presence of underlying social anxieties concerning expanding transportation networks, mass immigration, the environment, and distrust of governmental authority (Tomes, 2000). Much work has described the sensationalist rhetoric and misinformation surrounding, for example, the 2002 Severe Acute Respiratory Syndrome (SARS) outbreak (Washer, 2004), Swine Flu epidemic (Klemm et al., 2016), and Ebola crisis (SteelFisher et al., 2015; Towers et al., 2015). In other words, media platforms are often viewed as sites that circulate and reinforce misinformation - whether because public perceptions of risk are susceptible to social and cultural factors that can lead to conflicts and confusion (Pidgeon et al., 2003; Slovic, 1987), or because social media conversations and news coverage is characterized by fear-mongering, sensationalism, and the de-contextualization of scientific information (SteelFisher et al., 2015; Towers et al., 2015; Washer, 2004), supposedly demonstrating the capacity for these media platforms to impact (or perhaps interfere with) public perceptions (Boyd et al., 2009).

This line of argument in some ways follows a simple transmission model of communication, with the assumption that media conversations and depictions of disease and healthcare directly influence public perceptions. Indeed, much of this research is overly preoccupied with “representations of disease’”issues, doing little to question the sites of mediation, forms of expertise and institutions, or problematizing the definitions of “disease” and “media.” Further, although there is increased attention towards institutions and sites of mediation beyond the news media, the literature on the mediation of diseases largely still subscribes to “media effects” assumptions. Various texts are read and inaccuracies, misinformation, narrative frames, and so forth are identified, typically gesturing towards an unreliable media (whether this is news, social, or entertainment) that obscures scientific fact, instilling and reproducing fear and ignorance in a susceptible, uncritical public. These kinds of assumptions, which focus on these media platforms as the sole or primary culprits, overlook other potential sources of misinformation. It positions the public as a passive audience that depends on receiving information from objective, infallible scientific and government authorities, and the only role of the media is to accurately translate or transmit this information. Further, the public communication of science is viewed as primarily occurring in the realm of the press, overlooking other sources of information, storytelling, and influence. Past research has made connections between media consumption and health-related beliefs (Allington et al., 2020; Yoo et al., 2018). Yet it is important to problematize notions of medical authority, mediation, and disease, and to avoid oversimplifying the potential role of media depictions of disease on shaping public perceptions.

Much of the literature on the mediation of disease is characterized by these limited or outdated frameworks, which could arguably be traced back to literature from the 1980s and 90s that analyzed media coverage and public understandings of HIV and AIDS. Broadly speaking, much of this research was focused on representations of people with HIV/AIDS in the news, connecting these depictions to larger societal responses surrounding the disease or “general notions of morality and .... medicine, health and illness” (Lupton, 1994b, p.37). Kreps et al. (1998) explored the development and history of the field of health communication, noting that many of these communication scholars worked on health promotion campaigns that sought to minimize public risks of contracting HIV/AIDS, as well as heart disease and cancer. This orientation towards health promotion was reflected in their scholarly work on HIV/AIDS, which typically considered the flow of health information through mediated channels and the depiction of diseases and health care in popular media.

More recent work has, notably, challenged some of these outdated, narrow conceptions of the public and sources of health-related information, or troubled distinctions between “risk” and “fear” (for examples, see: Abraham, 2011; Greenberg & Rainford, 2014; Levina, 2015; Muntean, 2009), or perhaps expanded the scope of analysis to other platforms and sources of disease information (for examples, see: Gerlach & Hamilton, 2014; Levina, 2015; Seltzer et al., 2015; Stellefson et al., 2014; Wonser & Boyns, 2016; Young et al., 2013). As supported by more recent science communication work on issues such as vaccine hesitancy, the public is not a passive audience with beliefs that are entirely determined by exposure to accurate (or inaccurate) information (Nyhan & Reifler, 2015).

More recent work problematizes this characterization of an ignorant or “panicky” public. Although Davis et al. (2014) note the prevalence of sensationalist depictions of contagious diseases in the media, they argue that the public’s engagement with and understanding of these images and representations is more complex than most critiques of the media would suggest, which typically contend that “the general public is a problem for efforts to manage global problems such as pandemic influenza” (513). Through focus groups and interviews with members of the public, Davis et al. (2014) found that most respondents were aware of media sensationalism and the amplification of risk.

This does not mean that mediatized discourses around health issues are irrelevant or unimportant, but rather, that public sentiment cannot be understood through models that focus exclusively on information transmission. Muntean (2009) posits that “social and political conceptual frameworks” work to discursively construct pandemic fear (199). Despite there being fundamental differences between, for example, viruses and bacteria, or between disease outbreaks and hazards such as radiation or terrorism, boundaries between these distinct issues are blurred through “epistemological influences” such as the 9/11 attack (199). In other words, disease threat may be socially constructed according to prominent events or “scripts” in the public consciousness: formative, “archetypal” events (such as the 9/11 attacks) provide frameworks that position government responses and public conceptions. Indeed, Ungar (1998) argued that diseases are constructed and understood through “interpretive packages” made up of “metaphors, exemplars, stories, visual images, moral appeals and other symbolic devices” (39).

Similarly, Serlin (2012) notes how public health is defined not only by imagery “drawn from biomedicine but also from the vast web of media forms with which public health intersects” (xxiii) and Wald’s (2010) conception of “outbreak narratives” demonstrates how the meaning of disease is articulated through shared understandings - tropes, storylines, and practices - that come about through past disease events and shape future health outcomes by stigmatizing certain groups, affecting economies, and influencing “how both scientists and the lay public understand the nature and consequences of infection” (3).

This study attempts to draw on this more nuanced understanding of public perceptions and media, in which the construction and articulation of health issues occur through a larger social imaginary that draws upon (and reproduces) certain narratives, ideologies, and practices. Disease is not just depicted or represented, something that is easily reduced to media imagery or frames, but rather is something constructed in the cultural imagination, informed by history, geography, and politics (Bashford & Hooker, 2001). Through this understanding of public perceptions and media, we can view social media platforms such as Twitter as sites where health and risk-related concepts are both represented and constructed, according to complex forms of patterning that emerge from a shared cultural repository of past events, media depictions, institutional actions, and so forth.

There is increasing attention on how social media can serve as a source of health-related (mis)information (e.g., Dyar et al., 2014; Fung et al., 2014; Mollema et al., 2015; Nagpal et al., 2015; Rodriguez-Morales et al., 2015; Towers et al., 2015). This research often argues that being exposed to other social media users’ expressions of fear can lead to misguided or exaggerated concerns (see Fung et al., 2014; Nagpal et al., 2015), or shows how prominent health crises or exposure to sensationalized news coverage may lead to bursts of activity online (for examples, see Dyar et al., 2014; Mollema et al., 2015; Rodriguez-Morales et al., 2015; Towers et al., 2015). Some of this research (for example, Dyar et al, 2014) presumes that a brief increase in online activity is a meaningful representation of societal engagement on a health issue, which may be overstating the results. Nonetheless, social media conversations may contribute to heightened public perceptions of risk, as platforms such as Twitter allow for the proliferation of (and exposure to) messages containing anxiety and fear, amplifying concerns and alarm (Fung et al., 2014). Other work has considered how organizations such as the CDC can use social media platforms to attempt to manage uncertainty during times of crisis. In an analysis of CDC tweets during the 2014 Ebola crisis, Dalrymple et al. (2016) found that there was an emphasis on providing background information and establishing credibility. Certain events, however, such as the transmission of Ebola from a patient to a nurse, can “serve as bifurcations or ‘flash points of change’ that require rapid redirection on social media” (457).

In response to concerns about misinformation propagating through social media, companies such as Facebook, Twitter, and Google have implemented changes to their policies, though information about the effectiveness of these efforts is limited (Anderson & Rainie, 2017; Levin, 2018, 2019). Past research on health-related issues and social media has called for the use of these platforms to help educate the public and challenge incorrect beliefs (Chou et al., 2009; Gesser-Edelsburg et al., 2018; Korda & Itani, 2013; Scanfeld et al., 2010).

Social media discourse may reveal public beliefs or misconceptions (Signorini et al., 2011), yet there are barriers to leveraging social media as a tool to educate the public (Bhattacharya et al., 2014; Heldman et al., 2013; Neiger et al., 2013). For example, so-called “filter bubbles” describe the isolation caused by algorithms and self-sorting into communities where users mostly encounter content that conforms to their pre-existing beliefs (Bozdag & van den Hoven, 2015; Pariser, 2011). However, some research challenges the existence of filter bubbles and their significance in society, rejecting the idea of filter bubbles from both a psychological and technical point of view (Dahlgren, 2021). Bruns contends that the focus on filter bubbles is merely an attempt to “absolve ourselves of the mess we are in by simply blaming technology” (2019, 7). Further, countering misinformation by attempting to circulate and expose people to accurate information harkens back to the “deficit approach” to science and health communication, which is now viewed as an ineffective and oversimplified model (Simis et al., 2016). Indeed, it has been found that mere exposure to corrective information does not necessarily persuade people to change their views on scientific and health issues (e.g. Corner et al., 2012; Nyhan & Reifler, 2015).

Past research has explored “hashtag activism” and the different actors and their narrative strategies for constructing, deconstructing, or circulating discourses (Wonneberger et al., 2020). A study on hashtags and online discourse around animal rights and seal hunts in Canada found that Twitter has failed “to generate a climate of genuine debate” as such platforms serve as echo chambers (Knezevic et al., 2018).

Despite the apparent ineffectiveness of the deficit model, public perceptions still matter: support for conspiracy theories and widespread misinformation can prolong disease outbreaks and hinder policy (Earnshaw et al., 2020). Understanding the themes and patterns of social media conversations can help reveal the interpretive systems that people are using (and circulating) as they engage with issues such as mask-wearing, and perhaps allow for more effective public health messaging that goes beyond information transmission and targets issues of trust, disenfranchisement, and political polarization.

Other work has examined public sentiment toward public health measures during COVID-19, using both surveys and Twitter data. Tsao et al (2022) collected English language Tweets from Ontario, Canada, and explored public sentiment about issues and measures including vaccination, lockdowns, and the impacts of COVID-19 on business; they found a slightly positive sentiment on average during the second wave of the COVID-19 pandemic in Ontario. A study on Twitter data and mask-wearing found that most of the mask Tweets were in favor or neutral towards mask-wearing, while a smaller percentage of Tweets were against mask-wearing (Cotfas et al., 2021). Other work used surveys to characterize attitudes toward mask-wearing and found that those who did not wear masks tended to have an aversion to being forced to wear masks and believe that masks are ineffective in preventing COVID-19 (Taylor & Asmundson, 2021). Using an online survey, Esmaeilzadeh (2022) identified and categorized the critical concerns of wearing masks, and found that discomfort barriers, external factors such as political beliefs, and usability issues are the primary concern categories. Another study that used Twitter data examined COVID-19 vaccine-related tweets to characterize the topics, themes, and sentiments of public Twitter users and found evidence of broader themes including vaccination policy, vaccine hesitancy, and postvaccination symptoms and effects. It was found that sentiment scores were negative for the topics “postvaccination symptoms and side effects,” but consistent positive sentiment scores were found for topics such as “vaccine efficacy,” “clinical trials and approvals,” and “gratitude toward health care workers” (Chandrasekaran et al., 2022). A similar study from fall 2021 that collected 300,000 geotagged Tweets about COVID-19 vaccines in the United States found an increasing trend in positive sentiment alongside a decrease in negative sentiment, reflecting rising confidence and anticipation towards vaccines. This suggested that the public trusts the vaccine and “critical social or international events or announcements by political leaders and authorities may have potential impacts on public opinion towards vaccines” (Hu et al., 2021).

As one of the most popular social media platforms in the world, Twitter provides rich data that can be used to understand public sentiment and discourses around an issue such as mask-wearing. Twitter has also increasingly been used to disseminate public health information, and to acquire real-time health data (Medford et al, 2020). Other studies have drawn on Twitter data to examine the spread of misinformation during the COVID crisis (Bolsover & Tizon, 2020; Kouzy et al., 2020), analyze discussions of COVID-19 conspiracy theories (Ahmed et al., 2020), and assess the credibility and prevalence of information relating to the pandemic (Yang et al., 2020).

Pei and Mehta (2020) performed a quantitative sentiment analysis on Tweets with racist hashtags about COVID-19, such as #Chinesevirus and #Chinavirus, across several stages of the pandemic. This kind of work provides valuable information about the kind of sentiment (positive or negative) associated with different hashtags, and how this sentiment might shift over time. Our study focuses on the issue of mask-wearing, and in contrast to studies that use automated sentiment analysis on larger datasets, our manual coding methodology (detailed below) goes beyond classifications of positive or negative sentiment and provides more detail about the kinds of framing and themes that are present, and thus a more granular understanding of how discourses around mask-wearing are being constructed and circulated.

This study draws on framing theory, which refers to how information is packaged and presented to the public; information is placed within certain contexts to encourage or discourage particular interpretations, exerting an influence over how people perceive reality (Tewksbury & Scheufele, 2009). In other words, “Framing refers to the process by which people develop a particular conceptualization of an issue or reorient their thinking about an issue” (Chong & Druckman, 2007, p. 104). According to Entman (1993), framing is an organizational and communication process in which some aspects of reality are given greater emphasis and messages can influence the perceptions of individuals. We applied framing theory to the collected Tweets in order to see how mask-wearing was being framed - for example, as a form of civic duty, to protect personal safety, or as an infringement on individual freedoms. This allowed us to understand the discourse around mask-wearing during specific time periods, and therefore how the wider public may have been perceiving the issue.

This study sought to understand the public sentiment around mask-wearing, the motivations behind beliefs and perceptions of this important and contentious public health intervention, and the rhetorical strategies and framing used by those expressing their stance on Twitter. By analyzing four prominent mask-related hashtags, this study found that anti- and pro-mask sentiments appear to be based on a range of motivations and expressions of these sentiments revealed a variety of themes, worldviews, and apparent rationales. This is consistent with the idea that changing health-related beliefs and behavior is difficult, and that interventions and public health messaging must account for the complexity of the social, political, and economic factors which influence belief and behavior. Our analysis attempts to not simply provide a snapshot of mask-related beliefs, but to unpack these complexities in the hopes of better understanding some of the dynamics that shape public sentiment about public health interventions, and the ways in which people communicate these sentiments.

The following sections propose some of the possible implications and opportunities for future research.

As the COVID-19 crisis and opposition to public health interventions continue to develop, mask-wearing will likely persist as an important and contentious issue throughout the pandemic. Non-pharmaceutical interventions provide simple and low-cost ways of reducing the transmission of communicable diseases, and thus understanding public sentiment could help the uptake of these kinds of interventions and inform the development of public health messaging (Teasdale et al., 2014). Social media platforms such as Twitter play a role in shaping public perceptions of health issues and the spread of misinformation and also provide insight into public beliefs.

A limitation of this study was the number of Tweets (600) analyzed. The method of manually reviewing Tweets ensured that the inaccuracies of automated sentiment analysis were avoided; labeling tweets with an emotion remains a subjective task due to factors such as sarcasm, slang, or ambiguous use of emojis (Chapman et al., 2018). However, this limited the volume of Tweets that could be examined. Future research could attempt to validate these findings using automated sentiment analysis tools.

Despite the promise of social media to provide spaces for education, community building, and protest, the framing of issues on sites like Twitter can be constrained by "institutional norms, local politics, and contextual realities” (Ofori-Parku & Moscato, 2018). Other researchers have used the term “hybrid media event” to describe the “complex intermedia dynamic” between social media, the news, and other actors and platforms (Sumiala et al., 2016; also see: Vaccari et al., 2015). Drawing on this, future research could explore public sentiment around mask-wearing as a hybrid media event: geotagged tweets could be collected and then analyzed alongside local news coverage, bylaws, and public health statements to understand the interplay between these different actors and channels.

Other areas of future research include exploring other time periods of the pandemic beyond the early months examined here. Other research could explore mask-wearing as an expression of identity, or as a rapidly developing social norm. Earlier in the summer, pro-mask Twitter users reportedly “hijacked” anti-mask hashtags such as #NoMaskSelfie (Petter, 2020), a practice known as counter-tweeting - using a hashtag in a way that is oppositional to its intended meaning. However, there has been debate around whether this practice resulted in greater outreach for pro-mask messages, or if this actually boosted the anti-mask movement by amplifying their message and raising the profile of the hashtag, and potentially introducing more people to conspiracy theories (Dotto & Morrish, 2020).

This study has unpacked some of the apparent motivations, themes, and appeals used by people tweeting pro- and anti-mask messages, providing a starting point for exploring this phenomenon in future research.