Background
Exposure to discrimination produces severe consequences on health (for a recent review, see [
1]). It can increase risk of sleep disorders [
2‐
6], adverse physical and psychosocial conditions (e.g., cardiovascular disorders, psychological distress, and major depression [
7‐
11]), as well as harmful coping behaviors (e.g., substance use, such as tobacco and alcohol [
12‐
16]).
Given the relevant impact of discrimination on health, a crucial aspect for empirical research is to define best measures and methods to quantify exposure to discrimination [
17‐
19] in order to accurately estimate the population at risk and prevent potential negative health outcomes.
Over the last two decades, studies assessing discrimination primarily relied on explicit measures (i.e., self-reports) [
11,
17,
20,
21], which reflect conscious and controllable evaluations. Self-report data are thus potentially subject to intentional and social desirability processes [
22] that might prevent people from accurately reporting discrimination if they think this could be viewed negatively by others or unsafe for them. For example, individuals may be aware of having been a target of discrimination but they may be unwilling to disclose this information because they do not want to present themselves as weak and vulnerable or place themselves in potential danger.
To address the limitations of explicit measures, to our knowledge only four recent studies [
23‐
26] have assessed exposure to discrimination using the Implicit Association Test (IAT) [
27]. The IAT is a widely-used and validated implicit measure that infers automatic and spontaneous mental representations that exist in memory [
28]. Unlike explicit measures, the IAT assesses mental contents indirectly by measuring how quickly and accurately a person can categorize and associate stimuli related to two conceptual categories and two evaluative attributes. The underlying presumption is that categories and attributes that are strongly associated at a mental representation level show shorter latencies and fewer errors when classified together than when they are not [
27,
29]. The IAT thus is thought to be less influenced by conscious intentions and social desirability processes and to capture constructs that are outside of intentional and direct control [
28].
For example, in an IAT assessing exposure to racial/ethnicity discrimination, participants are asked to categorize stimuli representing the two conceptual categories – White people and Black people – and the two evaluative attributes – Target and Perpetrator – in two sorting conditions. In one condition, participants categorize stimuli representing the categories White people and Perpetrator with one response key, while categorizing stimuli representing Black people and Target by using another response key. In the other condition, participants categorize the same stimuli but with a different key configuration: this time stimuli of White people and Target are categorized with one key, whereas stimuli of Black people and Perpetrator are categorized with the other. Faster categorization in the first condition compared to the second condition indicates an implicit recognition of discrimination towards Black people than White people (implicit discrimination).
Studies using the IAT to measure exposure to discrimination have provided relevant insights into health research as well as in social and psychological sciences. For example, Carney et al. [
23] employed the IAT to investigate the well-known person-group discrimination discrepancy (PGDD [
30,
31];) typically observed on explicit measures. The PGDD refers to the tendency, showed by members of target groups, to report that other members of their social groups are discriminated against (group discrimination), but that they are not (person discrimination). Authors showed that although members of target groups do not explicitly report person discrimination, they do reveal person discrimination on the IAT. These results were observed in relation to sex and race/ethnicity discrimination [
23]. In addition, a series of three studies used the IAT to investigate the relation between exposure to discrimination and health outcomes. They showed that implicit measures of exposure to racial discrimination were associated with smoking [
25], elevated blood pressure, and risk of hypertension [
24,
26]. In other words, stronger implicit discrimination predicted an increased risk of physical conditions and behaviors harmful for health.
Taken together, these findings indicate that implicit and explicit measures of exposure to discrimination are not equivalent and research thus would benefit by using implicit measures to have an assessment of exposure to discrimination that is less influenced by intentional and conscious processes, such as social desirability. Indeed, research exploring the effects of social desirability on explicit and implicit measures of discrimination, showed that explicit measures (but not implicit measures) are associated with social desirability [
25]. Thus, investigating discrimination only by means of explicit measures might be inadequate because people may not accurately report exposure to discrimination at a conscious level.
The main goal of the present study is to advance measurement of exposure to discrimination by evaluating the application of implicit measures to different types of discrimination and by using a new brief validated version of these instruments to optimize the time required for their administration. Existing studies using the IAT for discrimination focused only on sex and race/ethnicity discrimination and employed exclusively the standard version of this instrument [
23‐
26], which requires around 13 min for administration. To our knowledge, no studies have used implicit measures of exposure to discrimination in relation to other social groups or used the new brief version of the IAT (Brief-Implicit Association Test, B-IAT, [
32,
33]) that only requires two minutes to be administered.
In this study, we performed such an investigation and conducted six experiments using the B-IAT to assess exposure to discrimination in relation to race/ethnicity, sex assigned at birth,
1 gender identity, age, sexual orientation, and weight. In addition, to have a more comprehensive view of this phenomenon, we (1) measured exposure to discrimination also by means of explicit measures (explicit self-reports of exposure to discrimination) and (2) assessed implicit and explicit attitudes towards different social groups of interest in each experiment in order to evaluate the existence of a possible association between discrimination and social attitudes.
Discussion
The present study is the first to employ the B-IAT to assess exposure to multiple types of discrimination. Specifically, we used the B-IAT to measure six types of discrimination based on race/ethnicity, sex, gender identity, sexual orientation, age, and weight.
Overall, we found implicit and explicit recognition of exposure to discrimination towards target groups. This result was observed both among individuals belonging to target groups and to dominant groups, but was stronger in the former. Implicit and explicit measures of discrimination tended to show no correlations. Similarly, no correlations were generally observed between implicit recognition of discrimination and implicit or explicit attitudes towards social groups of interest, indicating that implicit recognition of exposure to discrimination was not influenced by social preferences.
However, some exceptions emerged in experiments investigating discrimination in relation to race/ethnicity and weight. In the racism experiment, we found that White people, unlike people of Color, showed no implicit recognition of discrimination towards people of Color (i.e., Black, Asian, and Latinx). That is, no significant association of the categories
People of Color and
White People with the attributes
Target and
Perpetrator was observed. In contrast, a recognition of discrimination against people of Color emerged at the explicit level. In other words, results showed that although White people reported feeling that people of Color are discriminated against because of their race/ethnicity at a conscious level, no recognition of such discrimination was observed at an unconscious level. However, a low positive correlation was observed between implicit and explicit measures of discrimination, indicating that these measures assessed distinct but related constructs [
40]. In addition, we found that for White people, the implicit measure of discrimination was also negatively associated with implicit attitudes, i.e. stronger pro-White attitudes reduced the likelihood of implicit recognition of discrimination towards people of Color. These results expand existing research on implicit recognition of discrimination based on race/ethnicity. They suggest that explicit statements by White people about exposure to discrimination among people of Color may not be matched to their implicit recognition of such discrimination.
Of note, in contrast to our findings, the three prior studies using implicit measures to assess discrimination towards Black people, whose participants were recruited over a decade ago (i.e., between 2007 and 2010), showed that both Black and White individuals show implicit associations indicating a recognition of discrimination towards Black people [
23‐
26]. That is, discrimination towards Black people was recognized by both groups at an unconscious level. Here, we found instead that when implicit measures are used to assess the implicit recognition of discrimination towards people of Color (i.e., Black, Latinx, Asian), and not only towards Black people, White people show no implicit association. One possible reason for the difference in results observed in our study and the prior studies is that White people may show implicit recognition towards some specific race/ethnicity groups (e.g., Black people) but not towards other race/ethnicity groups that still belong to the category people of Color (e.g., Latinx and Asian). In the U.S., extensive research has documented both the specificities of anti-Black racism, as tied to enduring impacts of histories of enslavement, along with racism directed against other groups (e.g., American Indians/Native Americans in relation to histories of settler-colonialism, and racism against Latinx and Asian groups tied to histories of immigration) [
41‐
43]. Future studies, in which implicit recognition of exposure to discrimination towards each specific race/ethnic groups (i.e., Black, Latinx and Asian) is measured, would be thus useful to quantify and compare discrimination against different social groups categorized in the U.S. as people of Color. An alternative hypothesis is that explicit and implicit measures of discrimination and also the relationships between these measures may be influenced by societal context, which is tied to changes in race relations and race politics, and it has been shown to influence measures of racial attitudes [
44,
45]. Additionally, new research suggests that implicit biases may reflect structural inequalities, not just individual dispositions, and thus change as norms and practices of structural racism change [
46]. It is indeed relevant to note that the first three studies that used implicit measures to assess exposure to racial discrimination were conducted during the early years of the Obama Administration, which supported a public discourse regarding racial equality (even as policies did not always follow suit). In contrast, the current study was conducted during the Trump Administration, whose policies and pronouncements have been associated with an increase in racial intolerance and hate crimes [
47‐
51], along with partisan differences in expressing white grievance and racial resentment political views [
52]. The impact of changing political discourse and policies about race relations and racial justice on measures of both explicit and implicit discrimination warrant further study.
Similarly, in the fatphobia experiment, we found different results on implicit and explicit measures of discrimination and among participant groups. Specifically, on explicit measures, both heavy and not heavy groups reported feeling that fat people are discriminated against because of their weight, while results for implicit measures differed between participant groups. That is, heavy individuals showed no implicit association between the categories
Thin People and
Fat People and the attributes
Target and
Perpetrator, while not heavy individuals showed slight
Thin People + Target/Fat People + Perpetrator associations. In other words, although discrimination towards fat people was recognized at a conscious level, no recognition of this discrimination (heavy individuals) or opposite results (not heavy individuals) were observed at an unconscious level. These results are in line with previous studies using implicit and explicit measures to assess other psychological constructs based on weight, such as weight attitudes. These studies showed that implicit and explicit weight attitudes can display a large degree of dissociation or even opposite effects [
53‐
55]. For example, it has been shown that preferences between overweight/obese and underweight people differed when assessed by means of explicit or implicit measures. That is, pro-underweight preferences were observed at the explicit level, while pro-overweight/obese preferences were found at the implicit level [
54]. The lack of an implicit recognition of discrimination towards fat people found in the present study may reflect the assumption that overweight/obese people are responsible for their condition and that thus they cannot be considered victims of discrimination. Studies have shown indeed that obese/overweight people are perceived as lacking self-discipline and control and, for this reason, blamed for their weight [
56‐
58]. This belief may thus lead individuals to view fat people as not victims but even perpetrators of their condition and potential discrimination. This effect may emerge only on implicit measures as they are less influenced by intentional and social desirability processes. Indeed, although in Western societies, it is partially acceptable to denigrate obese and overweight people [
59], individuals may avoid reporting their evaluations at an explicit level because they may be viewed negatively from others.
Taken together, our findings have important applications for health research, in particular for those investigating the effects of discrimination on health. Our results show that for specific types of discrimination, implicit and explicit measures can display diverging results, indicating that these instruments can provide different information (but both valuable) about recognition of discrimination. Previous studies showed that implicit constructs (e.g., attitudes and stereotypes) are pervasive [
55,
60,
61] and predict variations in behavior across a variety of topics, in many cases above and beyond those assessed by means of explicit measures [
62‐
64] (for a meta-analytical comparison of the predictive power of implicit and explicit measures see Kurdi et al. [
65]). For example, it has been shown that implicit (but not explicit) race/ethnicity attitudes predicted physicians’ decisions to provide more thrombolysis recommendations for White than Black patients with acute coronary syndromes [
66]. Implicit measures of discrimination may thus predict health outcomes more accurately than explicit measures, which may underestimate the toll of discrimination on health. Further studies that include assessment of health outcomes (e.g., psychological distress, sleep disorders, and harmful coping behaviors) will be thus important to determine how the implicit recognition of discrimination found here for multiple types of social groups might contribute to health inequities.
In addition, our results have relevant implications for research in social cognition and for scientists interested in investigating discrimination as a psychological, cultural, and social phenomenon. For example, further research using the B-IAT may evaluate whether the implicit recognition of discrimination differs by region of the U.S. or world, and is influenced by brief interventions or specific contextual factors [
53,
67‐
71]. Recent research indeed showed that psychological constructs measured by means of implicit measures, such as social attitudes or stereotypes, can differ by geographic region and be associated with the societal context, including the complexities of political, cultural, and economic context [
53,
72]. Together, these societal features can powerfully shape the beliefs and behavior of its members [
73]. This novel study provides a brief and time-efficient tool to measure multiple types of implicit discrimination, including those that have not been conventionally measured (e.g., transphobia). The B-IAT improves measurement of discrimination at the implicit level and offers a sound methodology for subsequent studies. Future research would benefit from incorporating both implicit and explicit measures of discrimination.
Although the present study provides new relevant insights concerning multiple types of discrimination, some limitations should be noted. The samples used in our study may not represent a random selection of the general population. Selection biases might be present because the sample was comprised of people who learned about the Project Implicit website, volunteered to participate, had some interest in studies on implicit social cognition. Yet, we cannot identify any plausible reason why variation in selection biases across participants would explain the results found in our experiments. In addition, it is important to point out that by recruiting participants online through the Project Implicit website, samples are far more varied in ethnicity/race, age, class, and education level than most studies conducted in a specific location such as employees of a single hospital or students from a single university who also occupy a narrow age range. Nonetheless, replication with other sampling contexts will be useful to increase confidence in the observed results.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.