Background and Theory

Denialist arguments and recent developments

Residential school denialism is a type of political misinformation that, as Justice and Carleton (Reference Justice and Carleton2021b) note, does not involve the outright denial of the residential school system’s existence, but “rather the rejection or misrepresentation of basic facts about residential schooling to undermine truth and reconciliation efforts.” Justice and Carleton (Reference Justice and Carleton2021b) highlight a number of rhetorical techniques used by residential school denialists to cast doubt on the harms of residential schools and the need for redress. One of the most common counterarguments is that the schools provided a rewarding education for Indigenous children. For instance, Frances Widdowson, a former political science professor at Mount Royal University, has repeatedly claimed that the residential schools had educational benefits (CBC News, 2022). This claim flies in the face of the schools’ explicit assimilationist goals as well as the extensive evidence that the schools offered little real academic or vocational training (Milloy, Reference Milloy1999; Truth and Reconciliation Commission of Canada, 2015).

Denialism also surfaces in arguments that church officials and staff at the schools had good intentions and that any harm was incidental. For example, just minutes after he was sworn in as Manitoba’s minister of Indigenous reconciliation in 2021, Progressive Conservative Member of the Legislative Assembly Alan Lagimodiere argued that “in retrospect, it’s easy to judge in the past, [b]ut at the time, they really thought that they were doing the right thing … the residential school system was designed to take Indigenous children and give them the skills and abilities they would need to fit into society as it moved forward” (quoted in Petz, Reference Petz2021). Comments to a university conservative club in 2020 by then federal Conservative leader Erin O’Toole made a similar argument (Zimonjic and Cullen, Reference Zimonjic and Cullen2020). In another high profile case, Senator Lynn Beyak was removed from the Conservative caucus in 2018 after giving a speech in the Senate defending the “abundance of good” that had come out of residential schools (quoted in Brake, Reference Brake2018). Her comments emboldened anti-Indigenous sentiment, as her office received numerous letters of support that were published to her website and posters appeared on university campuses calling on Canadians to “reject the anti-white narrative being pushed in media and academia” (Carleton, Reference Carleton2021).

Residential school denialist arguments also routinely seek to “balance” the few positive experiences reported by some survivors against the more extensive evidence of trauma suffered at the schools. For example, residential school denialists often point to how Cree writer Tomson Highway has credited his time at residential school for his career success, while ignoring the passages in his memoir describing the sexual abuse he and hundreds of other boys experienced while at Guy Hill Residential School (Cyca, Reference Cyca2023; Highway, Reference Highway2021). This selective choice of evidence ignores many other survivors’ first-hand accounts of negative experiences at the schools and the intergenerational trauma they endured as a result (for example, Fontaine, Reference Fontaine2010; Knockwood and Thomas, Reference Knockwood and Thomas1992; Sellars, Reference Sellars2013).

Since the announcements of suspected unmarked graves, denialist rhetoric has increasingly focused on questioning the credibility of the searches and the existence or extent of the unmarked graves. In 2022, on the 1-year anniversary of the first announcement of suspected unmarked graves near Kamloops, British Columbia, the New York Post ran a headline quoting political scientist Tom Flanagan, who called the unmarked graves “the biggest fake news story in Canadian history” (Kennedy, Reference Kennedy2022). In the same year, a group of academics and journalists created the Indian Residential Schools Research Group to ostensibly address “misconceptions” about residential schools; in reality, the group’s aims are to cast doubt on the residential schools’ harmful legacy. Maxime Bernier, the leader of the right-wing, populist People’s Party of Canada, tweeted his support for the endeavour, arguing it was “time to stop vilifying Canadian history and society” (Bernier, Reference Bernier2023). Soon after, Danielle Smith, Alberta Premier and leader of the United Conservative Party, decried the “fake news” of the unmarked graves on her social media (Carleton, Reference Carleton2023).

These arguments are based in part on the claim that radar technology is unreliable in identifying human burial sites, and that even if there are in fact remains at these sites, we cannot be sure they are Indigenous children, because the grave markers are missing. Statements from professional archaeology associations have pointed out the flawed logic in these claims (Canadian Archaeological Association, Society for American Archaeology, Canadian Association for Biological Anthropology and Canadian Permafrost Association, 2022). However, denialists continue to demand excavations of burial sites as proof the cemeteries at former school sites contain the bodies of Indigenous children, often against the wishes of Indigenous communities. At the former Kamloops Indian Residential School, denialists trespassed on the grounds of the school site at night and attempted to dig up suspected unmarked graves (Office of the Special Interlocutor, 2023, p. 98).

Denialism has existed alongside residential schools since their inception, but in recent years, as Canada began taking official steps to address the harmful legacy of residential schools, this rhetoric has come especially from right-wing academics, media and political actors. In 2013, a former speechwriter for Prime Minister Stephen Harper and Alberta Premier Jason Kenney wrote an article rejecting the “unchallenged narrative” and “bogus genocide story” of residential schools (von Scheel, Reference von Scheel2020). In 2018, the Frontier Centre for Public Policy, a right-wing think tank, aired a 2-minute advertisement on private radio stations in Saskatchewan claiming to debunk the “myths” of residential schools (Meloney, Reference Meloney2018). Of course, this rhetoric is not exclusive to conservative actors; in 2021, former Liberal Prime Minister Jean Chrétien denied being aware of abuse at residential schools during his time as Minister of Indian Affairs, despite archival evidence to the contrary (Barrera, Reference Barrera2021).

Members of Canada’s political and academic elite are not the only actors engaging in denialism. Debates over the existence and legacy of residential schools are, as with many other cases of misinformation, increasingly taking place among members of the public in online fora. To illustrate the rise in online denialism-related conversations, we collected nearly 40,000 comments posted to more than 600 residential-school-related threads on major Canadian politics Reddit communities between 2021 and 2024. We then identified comments that included words commonly appearing in denialist rhetoric, such as “cover up” and “hoax,” and calculated the proportion of all comments that included these words in each month (see the Supplementary Materials (SM) Section SM1 for details). As Figure 1 shows, the prevalence of denialism-related terms has increased more than fourfold during this time, from just over 3 per cent in the period immediately after suspected unmarked graves first became a national news story in May 2021 to more than 12 per cent at the end of 2023. While not all of these comments are endorsing denialism—in fact, many are accusing other users of denialism—these data demonstrate that online debates about the veracity of the unmarked graves and the history of residential schools have become increasingly common in the last 3 years.

Figure 1. Prevalence of Denialism-Related Terms in Reddit Comments, 2021–2024.

Note: This figure reports the percentage of comments including denialism-related words on Reddit posts related to residential schools that appeared on eight Canadian general and political subreddits for each month between 2021 and 2023 (see SM1 for details). Points are scaled by the total number of comments appearing each month. Note that the first suspected unmarked graves were announced in May 2021. The line of best fit is estimated from a generalized additive model weighted by the number of comments in each month.

The former chair of the TRC, Murray Sinclair, argues that misinformation represents one of the greatest obstacles to confronting the residential school history (Forester, Reference Forester2021). Accordingly, concerns about residential school denialism have triggered discussions about possible policy reforms. In response to the TRC’s Calls to Action related to updating educational curricula across Canada, most provinces have begun integrating the residential school history into elementary and high school education, which may serve as a bulwark against misinformation later in life. There have also been calls to criminalize residential school denialism as hate speech (Dufour, Reference Dufour2023; Stefanovich, Reference Stefanovich2023). What we aim to contribute to the debates over these proposals is empirical evidence regarding the nature and extent of residential school denialism among the Canadian public, as well as evidence assessing the effectiveness of interventions to counter denialist claims.

Theoretical Expectations

In this study, we conceptualize residential school denialism as a type of belief in political misinformation (Jerit and Zhao, Reference Jerit and Zhao2020).Footnote ³ The existing literature on this topic has generally not studied cases of misinformation linked to history, intergroup relations, and racism.Footnote ⁴ However, as discussed above, denialism shares commonalities with other forms of misinformation, namely in the language it uses to question evidence and its spread via online channels.

Kuklinski et al. (Reference Kuklinski, Quirk, Jerit, Schwieder and Rich2000) argue that believers of misinformation do not simply lack information, but rather “firmly hold the wrong information” (p. 792).Footnote ⁵ In our analysis, we consider not only whether people are misinformed, but also the very likely possibility that many Canadians lack information on the topic of residential schools. In Canada, public school curricula have historically either failed to teach students about residential schools or taught a sanitized account of this history (Bennett, Reference Bennett2021). While today 90 per cent of non-Indigenous Canadians say they have heard of residential schools and 63 per cent believe they are very or somewhat familiar with this history (Canadian Reconciliation Barometer, 2023; Environics, 2023), other research has shown that these self-reported indicators tend to overstate what respondents actually know (Boese, Neufeld and Starzyk, Reference Boese, Neufeld and Starzyk2017). We anticipate that many Canadians are uninformed about the issues underlying denialist claims and many others have been exposed to inaccurate information, which increases their propensity to endorse denialist misinformation. By extension, we expect that educating people about the true, factual history can reduce both ignorance and denialism. This expectation is in line with recent evidence showing that informational interventions can positively change attitudes toward Indigenous peoples in Canada (for example, Efimoff and Starzyk, Reference Efimoff and Starzyk2023; Neufeld et al., Reference Neufeld, Starzyk, Boese, Efimoff and Wright2022; Siemens and Neufeld, Reference Siemens and Neufeld2022).

Besides a lack of exposure to education about the relevant history, there are other reasons why individuals may adopt and retain denialist beliefs. A large literature in political science demonstrates that voters tend to adopt the policy positions espoused by politicians from their preferred party (for example, Broockman and Butler, Reference Broockman and Butler2017; Zaller, Reference Zaller1992), with some evidence that a similar dynamic may apply to belief in misinformation (Berinsky, Reference Berinsky2023; Van Duyn and Collier, Reference Van Duyn and Collier2019). Because denialist claims have been articulated more frequently by right-wing groups and opinion leaders in recent years in Canada, we expect that denialist beliefs will correlate with Conservative partisanship.

There are also reasons to suspect that denialism correlates with individuals’ predispositions toward beliefs in conspiracy theories. Much of the language appearing in denialist claims is suggestive of secretive and malevolent motives for presenting the public with allegedly fraudulent information about the searches for unmarked graves at former residential schools (for example, “hoax,” “scam”). Denialists also regularly accuse the mainstream media of intentionally misrepresenting the truth behind residential schools and the unmarked graves (Gerbrandt and Carleton, Reference Gerbrandt and Carleton2023). Social scientists have shown that some people are more prone to believe in conspiracy theories. “Conspiracy thinking” refers to the predisposition toward seeing events and circumstances in terms of conspiracies (Uscinski and Parent, Reference Uscinski and Parent2014; Uscinski et al., Reference Uscinski, Enders, Diekman, Funchion, Klofstad, Kuebler, Murthi, Premaratne, Seelig and Verdear2022). Those who have a tendency toward conspiracy thinking are more likely to endorse specific conspiracies, such as those related to 9/11, the moon landing, or U.S. President Barack Obama’s place of birth (Berinsky, Reference Berinsky2023; Brotherton, French and Pickering, Reference Brotherton, French and Pickering2013; Bruder et al., Reference Bruder, Haffke, Neave, Nouripanah and Imhoff2013; Uscinski and Parent, Reference Uscinski and Parent2014; Enders et al., Reference Enders, Uscinski, Seelig, Klofstad, Wuchty, Funchion, Murthi, Premaratne and Stoler2021). While not all residential school denialists are conspiracy theorists (and vice versa), we expect that a predisposition toward conspiracy thinking will correlate positively with beliefs in denialist misinformation.

Denialism may also be distinct from beliefs in other forms of misinformation because of its close association with individuals’ intergroup attitudes and identity attachments within the settler colonial system. In particular, one’s prior views toward Indigenous peoples may shape how non-Indigenous Canadians interpret the facts around Canada’s residential school history. Those who express greater Indigenous resentment may be more willing to endorse denialist claims, because this attitude is a strong predictor of opposition to policies that benefit Indigenous peoples (Beauvais, Reference Beauvais2022; Beauvais and Stolle, Reference Beauvais and Stolle2022). Residential school denialism, by minimizing the harms of residential schools, is tied to efforts to undermine reconciliation and weaken support for policies that would redress this history (Green, Reference Green, Starblanket and Long2025; Justice and Carleton, Reference Justice and Carleton2021a; Logan, Reference Logan, Hinton, Woolford and Benvenuto2014; Regan, Reference Regan2010).Footnote ⁶

Previous research has shown that one’s attachment to their racial/ethnic in-group is distinct from their attitudes toward a racial/ethnic out-group, with in-group identification exerting its own independent influence on political views (Beauvais and Stolle, Reference Beauvais and Stolle2022; Jardina, Reference Jardina2019). Among members of perpetrator or dominant groups, the invocation of historical injustices can create a significant threat to the esteem of one’s identity group (Branscombe et al., Reference Branscombe, Ellemers, Spears and Doosje1999; Doosje et al., Reference Doosje, Branscombe, Spears and Manstead1998; Tajfel and Turner, Reference Tajfel, Turner, Worchel and Austin1986). Due to their historical connection to Canadian identity and the settler colonial project, white Canadians offer a relevant subgroup that might experience a heightened sense of threat in response to evidence of past wrongdoing.Footnote ⁷ For this group, denying historical injustices can assuage uncomfortable emotions such as guilt or shame or fear that the settler status quo is insecure (Iyer, Leach and Pedersen, Reference Iyer, Leach and Pedersen2004; Knowles et al., Reference Knowles, Lowery, Chow and Unzueta2014; Maracle, Reference Maracle2017; Rotella and Richeson, Reference Rotella and Richeson2013; Tuck and Wayne Yang, Reference Tuck and Wayne Yang2012; Wohl, Branscombe and Klar, Reference Wohl, Branscombe and Klar2006). This psychological motivation to minimize harms against Indigenous peoples may help explain the persistence of “white ignorance” about past wrongdoing (Mills, Reference Mills, Sullivan and Tuana2007). To the extent that residential school denialism allows white Canadians to dismiss uncomfortable truths about their group’s historical treatment of Indigenous peoples, we should expect to see a greater endorsement of denialist claims among those that hold strong attachments to their white identity.

People of color (POC) also differ in the strength of their identity attachments. In our present work, we define POC as non-Indigenous members of non-European ancestry population groups (for a longer discussion of terminology, see SM2). Research shows that, among POC, attachment to a racial/ethnic identity (such as Asian or Black) is a strong predictor for whether an individual also identifies in solidarity with other non-white populations (Sanchez, Reference Sanchez2008; Pérez, Reference Pérez2021). A sense of commonality or “linked fate” among POC is in turn associated with political attitudes and behaviours in support of other marginalized groups (Chan and Jasso, Reference Chan and Jasso2023; Gershon et al., Reference Gershon, Montoya, Bejarano and Brown2019; Merseth, Reference Merseth, Aoki and Lien2020; Pérez, Reference Pérez2021). Many POC communities have themselves been displaced or otherwise negatively affected by European colonialism, which may inform opposition to colonial injustices in Canada (see Byrd, Reference Byrd2011; Lawrence and Dua, Reference Lawrence and Dua2005; Snelgrove, Dhamoon and Corntassel, Reference Snelgrove, Dhamoon and Corntassel2014). Recent evidence shows that, compared with white Canadians, POC perceive a higher degree of commonality with Indigenous people, which is in turn associated with their greater support for redressing historical injustices (Starzyk et al., Reference Starzyk, Neufeld, El-Gabalawy and Boese2019). We therefore expect that a strong attachment to one’s own racial/ethnic identity among POC can motivate solidarity with other marginalized communities, leading to lower endorsement of denialist claims.

Measuring Residential School Denialism

To develop our measures of residential school denialism, we drew on arguments appearing in online communities and media articles, and consulted secondary material describing common denialist claims (for example, Carleton, Reference Carleton2021). We also consulted with two representatives from the National Centre for Truth and Reconciliation (NCTR). On the basis of the NCTR staff’s recommendations, Dr. Beauvais met with an Indigenous Elder and survivor of the residential school system to listen to his experiences at residential school and with denialists (see SM3.6).

We sought to tap into two common types of denialism: (1) claims that the deaths or number of unmarked graves at former residential schools are false or exaggerated and (2) claims that the schools were well intentioned and had broadly positive impacts on Indigenous peoples’ lives. While claims in the former category are more amenable to empirical verification, the latter set of items related to the programme’s intentions and legacies are also based on extensive evidence. Experts have not only assembled documentation on the deaths and unmarked burials of children at the schools, but also on statements outlining the forced assimilation objectives of the program and on the government and churches’ efforts to conceal their failings (Carleton, Reference Carleton2021; McKenzie and Carleton, Reference McKenzie and Carleton2021; Milloy, Reference Milloy1999; Truth and Reconciliation Commission of Canada, 2015; Turnbull, Reference Turnbull2021).

We aimed to present a single denialist claim per item and to make the statements as simple as possible. Our nine items contain an average of 12.5 words per statement, and according to Flesch–Kincaid metrics, are readable at the 8th–10th grade level. All items are measured using a five-point Likert scale (strongly disagree, somewhat disagree, neither agree nor disagree, somewhat agree, strongly agree) and are (reverse) coded so that higher values indicate greater residential school denialism. We include positively and negatively worded items to guard against acquiescence bias (see SM5.1 for a discussion and tests for such bias).

For each item, respondents were able to respond with “don’t know” or “I haven’t thought too much about this.” Including this option increases the risk of satisficing, in which respondents offer a non-opinion to avoid expending the cognitive effort needed to form an opinion (Krosnick, Reference Krosnick1991). However, it is necessary in a survey on this topic because we ask about specific details related to the searches for unmarked graves and the residential school history, of which many non-Indigenous people have only a superficial knowledge. In this context, “don’t know” is an accurate reflection of many people’s position. In a pilot study, we did not include this option and found that it impacted the distribution of attitudes: for two out of the three items that appeared in both the pilot and full surveys, the proportion of respondents agreeing with denialist claims is notably higher when “don’t know” was not an option (see SM3.5). We recommend future users of this scale include a non-response option to ensure their estimates of denialism are not inflated.

The nine items in the residential school denialism scale are as follows:

1. 1. “The residential schools did more harm than good.” (reversed)

2. 2. “The people running Canada’s residential schools had good intentions.”

3. 3. “Indigenous children attending residential schools died at higher rates than other children because the conditions at residential schools were worse.” (reversed)

4. 4. “The purpose of residential schools was to help Indigenous people.”

5. 5. “Indigenous children died as a result of attending residential schools.” (reversed)

6. 6. “The suspected graves at former residential schools are probably tree roots or other debris, not graves.”

7. 7. “Radar technology can reliably locate Indigenous children’s graves at former residential schools.” (reversed)

8. 8. “The unmarked graves at former residential schools may not even contain Indigenous people.”

9. 9. “People saying that there are hundreds of unmarked graves at former residential schools are exaggerating.”

We construct a summated rating scale of residential school denialism by taking the average non-missing score for each respondent across our nine items tapping into this concept (Spector, Reference Spector1992). The items comprise a highly reliable scale (Cronbach’s α = 0.89) and a reliability analysis reveals that dropping any of the items would lower the overall reliability of the scale (see Table S5 for full results).

Plotting a scree plot using the reduced matrix eigenvalues (as recommended by Fabrigar and Wegener Reference Fabrigar and Wegener2012, p. 57) suggests that a single factor should be retained (Figure 2), and thus the scale items can be summarized using a single latent construct. For a longer discussion of dimensionality, see SM4.3.

Figure 2. Scree Plot of Residential School Denialism Items using Reduced Matrix Eigenvalues.

Note: A scree plot plotting the reduced matrix eigenvalues from largest to smallest to visually identify the number of eigenvalues to the left of the point where the eigenvalues level off (the “scree” of the graph). There is one point to the left of the scree for the observed data, suggesting a single factor best captures variation in the items.

Finally, we recognize that in larger surveys, future researchers may not always be able to include all nine items presented above. While a larger number of items generally helps to reduce measurement error, we recommend, in cases where the full scale is not feasible, using a four-item scale comprised of the following statements:

1. 1. “The residential schools did more harm than good.” (reversed)

2. 2. “The people running Canada’s residential schools had good intentions.”

3. 3. “Indigenous children attending residential schools died at higher rates than other children because the conditions at residential schools were worse.” (reversed)

4. 4. “The purpose of residential schools was to help Indigenous people.”

These items tap into the main aspects of denialism, but do not directly reference the unmarked graves. In SM4.5, we demonstrate that this shorter scale is also statistically reliable and shares many of the same measurement properties as the full scale. In the remainder of this study, we analyze the full nine-item scale.

Sample, Experimental Design, and Estimation

We contracted with Leger Opinion to recruit a sample of non-Indigenous adult Canadian citizens and permanent residents to complete an online survey in November 2023. The survey was available in both English and French. Leger uses quota-based sampling to generate samples that are reflective of the population in terms of gender, age, educational attainment, language and racial and ethnic population groups. Our final sample (n = 1,915) is broadly representative of the adult Canadian population, although our respondents skew slightly older on average (see SM3.1). We report results weighted by age, gender and region in SM3.2, finding that these estimates are nearly identical to those presented in the main text below. In addition to Leger’s internal efforts to ensure data quality, we excluded duplicate survey takers and likely bots. We also conducted data quality checks on less attentive respondents and find that our descriptive results are not sensitive to their inclusion in the sample and our experimental findings would in fact be stronger if we were to exclude these respondents (see SM3.3).

To assess the effect of an educational intervention on residential school denialism, we randomly assigned equal proportions of respondents to either receive factual information about the residential school history (the treatment group) or not (the control group), before subsequently asking about their attitudes toward denialist arguments. Randomization allows us to establish the causal effect of the educational treatment on attitudes by ensuring that information exposure is orthogonal to prior determinants of denialist beliefs.

In our study, respondents assigned to the treatment condition were tasked with reading a 250-word text describing the goals of the residential school system, conditions at the schools, facts about Indigenous children’s deaths, and details about the ongoing searches for unmarked graves (see SM6.2 for the full text). This information was accompanied by photos of a residential school and Indigenous children in a classroom, as well as a link to the Canadian Encyclopedia page on residential schools, from which much of the language for the intervention was borrowed.

The median respondent spent 42 seconds engaging with the content. We hypothesize that the educational treatment will reduce both “don’t know” responses and expressions of residential school denialism. Balance tests reveal that the treatment and control groups resemble one another on average across all pre-treatment covariates (see SM6.3).

The first portion of our analysis, which considers the prevalence and correlates of residential school denialism, is conducted strictly on the portion of the sample in the control group (n = 960). Responses in this condition capture the baseline attitudes of the Canadian public. At the end of the survey, we debriefed all respondents on the purpose of the study. To address ethical concerns, participants in the control group were given the same factual information as the treatment group as part of their debrief (that is, after they had responded to the denialism items).

The second portion of the analysis, which identifies the effect of education about residential schools on denialist attitudes, uses the full sample. We estimate average treatment effects (ATEs) using ordinary least squares (OLS) regression, which allows us to increase the precision of our estimates by controlling for covariates as well as to conduct subgroup analyses.

Our analyses rely on a number of pre-treatment sociodemographic and attitudinal variables, including age, gender, region, income, religion, partisanship, language, education, political knowledge, trust in media, racial/ethnic in-group identification (Jardina, Reference Jardina2019), Indigenous resentment (Beauvais, Reference Beauvais2021), conspiracy thinking (Uscinski and Parent, Reference Uscinski and Parent2014; Enders and Smallpage, Reference Enders and Smallpage2019), and factual knowledge about the residential school system (Boese, Neufeld and Starzyk, Reference Boese, Neufeld and Starzyk2017). Full coding rules for these variables can be found in SM3.4. To improve statistical efficiency in the estimation of treatment effects, we impute missing covariate values, first using the sample mode for categorical variables and then using multiple imputation by chained equations for continuous variables (King et al., Reference King, Honaker, Joseph and Scheve2001).Footnote ⁸ Approximately 31 per cent of our respondents had one covariate value imputed; covariate missingness is most significant for the religion variable, which appeared toward the end of the survey.Footnote ⁹ Imputed values are not used in the analysis of correlates.

The Correlates of Residential School Denialism

To investigate which Canadians are more likely to endorse denialism, the left panel of Figure 4 plots average scores on our denialism scale across a range of pre-treatment covariates. The right panel reports the percentage of “don’t know” responses to each of the denialism items for different covariate values.

Figure 4. Correlates of Residential School Denialism.

Note: This figure shows the distribution of denialism and “don’t know” responses among respondents in the control group (n = 960). In the left panel, the shading shows the distribution of denialism and the points indicate the mean denialism score (with 95% confidence intervals) among non-missing responses for each covariate. The right panel reports the percentage of “don’t know” responses. For continuous variables, respondents are grouped into “low,” “medium” and “high” categories on the basis of tercile. For the racial/ethnic identity variable, strong and weak identification with one’s racial/ethnic group is measured on the basis of whether a respondent’s racial/ethnic attachment score is above or below the median in their respective racial/ethnic category (that is, white or POC). See SM3.4 for additional details on covariate measurement.

The plots reveal that men are significantly more likely to agree with denialist claims, and less likely to indicate that they “don’t know.” This finding is consistent with research in other domains on a gender gap in both racial attitudes and propensity to select “don’t know” on political knowledge questions (e.g Miller, Reference Miller2019; Pratto, Stallworth and Sidanius, Reference Pratto, Stallworth and Sidanius1997).

Second, partisanship is an important correlate of denialist beliefs. Supporters of conservative parties are significantly more willing to endorse denialist claims compared with partisans of all other parties, especially those on the left. The average difference in scores on the denialism scale between Conservative Party and NDP supporters, for example, is around 0.9 standard deviations (or, equivalently, 0.8 points on the five-point scale in Figure 4). Those who do not consider themselves supporters of any of the major parties report denialist beliefs similar to Liberal partisans, but are significantly less likely to provide an opinion on these issues.

Denialism is linked to racial identity and prior attitudes toward Indigenous peoples. We find that those with the highest levels of Indigenous resentment express 1.3 standard deviations more denialism on average than those that score the lowest on this variable. Identification with one’s racial/ethnic in-group is also a significant predictor of denialism, but in a nuanced way. On average, white Canadians are no more receptive to denialist arguments than POC, but white Canadians who strongly identify with their racial in-group score much higher on the denialism scale (0.7 standard deviations) than those without a strong attachment to their white identity. For POC, the relationship is reversed: those that identify strongly with their racial/ethnic in-group are much less willing to endorse denialism.

We find that prior knowledge—measured using a series of factual questions—is also related to expressions of denialism. While general political knowledge does not covary with respondents’ opinions on the denialism items, scores on a three-item quiz about residential schools weakly predict denialist attitudes. Respondents who did not answer a single question correctly on this quiz (54% of the sample) reported roughly 0.3 standard deviations more denialism on average than those who scored better on the quiz.Footnote ¹¹ At the same time, both of these measures have a strong negative relationship with whether respondents answered with “don’t know”: non-opinions were about 20 percentage points more likely among those who failed to provide a single correct answer on the residential schools quiz.Footnote ¹²

Finally, denialist beliefs are weakly associated with conspiracy thinking. Those who express medium and high levels of agreement with statements in this scale, such as “much of our lives are being controlled by plots hatched in secret places,” tend to agree around 0.45 standard deviations more with the denialism items. Similarly, in supplementary analyses, we find that those who least trust the mainstream media and those who deny that the Holocaust occurred are both more likely to endorse denialism. In this sense, residential school denialism is similar to other types of misinformation in its positive relationship with individuals’ predisposition to believe in conspiracy theories, but these associations are smaller in magnitude than those related to partisanship or racial attitudes.

We also conduct a multivariate regression analysis of denialist beliefs (see SM5.2). Many of the bivariate correlations summarized in Figure 4 remain statistically significant after controlling for other demographic and attitudinal covariates, with the exception of the residential schools knowledge quiz scores and the conspiracy thinking scale. We also find that the ostensibly strong relationship between Indigenous resentment and non-response rates is attenuated after controlling for these other variables, suggesting that the expressive use of the “don’t know” option among those with greater anti-Indigenous attitudes is minimal. The “don’t know” response category appears to be primarily related to a genuine lack of exposure to the relevant information, and to a much lesser extent, negative views of Indigenous peoples. Finally, these supplementary models do not reveal any notable associations between denialism and age, region, or language, although religious adherents of all denominations do agree more with denialist claims than atheists and agnostics.

Educational Intervention

In Table 1, we present OLS estimates of the average treatment effect (ATE) of the educational intervention on residential school denialism and the proportion of responses indicating a non-opinion. For the expressed denialism outcome, we take the average of all responses to the denialism items (excluding “don’t know” responses), focusing only on respondents who offered an opinion on at least one of the items. We present estimates both with and without controlling for pre-treatment variables.

Table 1 Average Treatment Effects of Educational Intervention

Note: Table reports estimates from OLS models with HC2 standard errors. The outcome in the first two models is the average expressed denialism score (scaled by control group standard deviations) among respondents who expressed an opinion on at least one of the denialism items. The outcome for the third and fourth models is the proportion of a respondent’s denialism items that were responded to with a “don’t know” response. In models two and four, the following covariates are included in the model specification but not reported here: age, gender, party ID, region, language, visible minority status (and its interaction with racial/ethnic identity attachment), Bachelor’s degree, religion, household income, Indigenous resentment, conspiracy thinking, political knowledge, residential school factual knowledge, trust in media, and whether the respondent knows an Indigenous person). ^* p < .05

Our estimates suggest that the intervention reduces agreement with residential school denialism by just over 13 per cent of a standard deviation. The treatment also reduces the probability of non-opinions by nearly 7 percentage points, which is a large effect given the baseline rate of non-opinion is almost 20 per cent. These effects are both statistically significant at conventional confidence levels, regardless of whether we control for pre-treatment covariates. The estimates are also substantively meaningful: studies that focus on the persuasive effects of historical information on intergroup attitudes typically report effects of a similar magnitude (for example, Efimoff and Starzyk, Reference Efimoff and Starzyk2023; Fang and White, Reference Fang and White2022; Nyhan and Zeitzoff, Reference Nyhan and Zeitzoff2018b; Williamson, Reference Williamson2024).Footnote ^{13 ,} Footnote ¹⁴

The results in Table 1 are encouraging, but given the interdependence between non-opinions and endorsements of residential school denialism, it is also informative to examine these two outcomes simultaneously. In Figure 5, we plot the average prevalence of each response category across all nine denialism items separately for treated and control respondents (all items have been reversed as necessary so that greater agreement indicates greater endorsement of denialism). Figure 5 clarifies the relationship between the two sets of effects in Table 1: the intervention reduced the proportion of non-opinions and increased the number of respondents who strongly rejected residential school denialist statements. The increase in the percentage of respondents strongly disagreeing with residential school denialism (6–7 percentage points) was matched by an almost equal decline in non-opinions.Footnote ¹⁵ Of course, because respondents cannot be observed under both treatment and control conditions, we cannot say with certainty that the intervention caused those who otherwise would have said “don’t know” to shift directly to “strongly disagree.” The treatment may have, for example, also induced movement from those in the middle of the scale toward strong disagreement. Nonetheless, among those who are willing to offer an opinion on these issues, the proportion disagreeing with denialist claims is significantly greater in treatment condition.

Figure 5. Responses to Denialist Claims by Treatment Condition.

Note: Plot presents the average proportion of respondents, by treatment condition, providing each response type across all nine denialism items. (n = 1,915).

That being said, in the aggregate, roughly equal proportions of respondents in the treated and control conditions agreed with residential school denialist claims. Our experimental results therefore offer suggestive evidence that the beneficial effects of information largely function by reducing non-opinions, rather than persuading those who have already formed an opinion. The fact that the treatment does not reduce the overall percentage of respondents agreeing with denialism also suggests that our baseline estimates of the prevalence of denialism in the control group are fairly accurate. As discussed earlier, Kuklinski et al. (Reference Kuklinski, Quirk, Jerit, Schwieder and Rich2000) note that misinformation believers “firmly hold the wrong information.” In our study, 18 per cent of treated respondents read the factually correct information and still endorsed denialism, indicating that they are indeed willing to stand firm in their commitment to this misinformation.

To further investigate the patterns driving our main effects, we also consider conditional average treatment effects (CATEs). We focus on those who were more likely to endorse residential school denialism at baseline, including those who report greater Indigenous resentment and who identify as Conservatives and People’s Party (PPC) supporters. We re-estimate our models with an interaction between the treatment indicator and each of these moderators.Footnote ¹⁶ The CATEs are summarized in Figure 6 for both residential school denialism and the likelihood of offering a non-opinion.

Figure 6. Conditional Average Treatment Effects by Indigenous Resentment and Partisanship.

Note: These plots summarize four OLS models (two for each moderator) in which the treatment indicator is interacted with the Indigenous resentment indicator (binned by tercile) and the Party ID variable. Effect estimates for the denialism outcome (n = 1,822) are scaled in terms of control group standard deviations. The “don’t know” outcome (n = 1,915) is measured as the probability of giving that response type to a given denialist item. The models control for all covariates listed in the notes to Table 1.

We find that the CATEs on denialism scores are larger for those who expressed medium and high levels of Indigenous resentment (a difference of roughly 0.15 standard deviations versus those with low resentment). Effects are also substantively larger among Conservative and PPC identifiers than among non-partisans and supporters of more left-leaning parties. These plots demonstrate that the treatment does not induce backlash: even among those who might be most likely to redouble their endorsement of denialism after reading the treatment text, there is no evidence of effects in a counter-informational direction. However, the treatment had no conditional effect on non-opinions: information reduced the probability of offering a non-opinion by around 7 percentage points across all levels of Indigenous resentment and for all partisan types (for additional tests see SM6.7).