Rumination mediates associations between microaggressions and sleep quality in Black Americans: The toll of racial microstressors

Abstract

Disparities in health outcomes between Black and White Americans are well-documented, including sleep quality, and disparities in sleep may lead to disparities in health over the life course. A meta-model indicates that cognitive processes may underly the connection between race and poor sleep quality, and ultimately, health disparities. That is, there are race-specific stressors that disproportionately affect Black Americans, which are associated with poor health through biological, cognitive, and behavioral mechanisms (e.g., sleep). Among these race-specific stressors is discrimination, which has been linked to poor sleep quality, and there is a body of literature connecting perseverative cognition (e.g., rumination and worry or vigilance) to poor sleep. Microaggressions, a more subtle but pervasive form of discrimination, are another race-specific stressor. Although less research has considered the connection of microaggressions to perseverative cognition, there are some studies linking microaggressions to health outcomes and sleep. Therefore, using a cross-sectional survey, we tested the following hypotheses: racism-related vigilance and rumination would mediate the relationship between discrimination and poor sleep as well as between microaggressions and poor sleep among Black Americans (N = 223; mean age = 35.77 years, 53.8% men, 86% employed, 66.8% with college degree or higher education). Results of seven parallel mediation models showed that neither rumination nor racism-related vigilance mediated a relationship between discrimination and poor sleep quality. However, rumination partially mediated relationships between the six microaggression sub-scales and poor sleep quality: there were significant indirect effects for Foreigner/Not Belonging (β = .13, SE = 0.03, 95% CI: 0.08, 0.20), Criminality (β = .11, SE = 0.03, 95% CI: 0.05, 0.17), Sexualization (β = .10, SE = 0.03, 95% CI: 0.05, 0.17), Low-Achieving/Undesirable (β = .10, SE = 0.03, 95% CI: 0.05, 0.15), Invisibility (β = .15, SE = 0.04, 95% CI: 0.08, 0.23), and Environmental Invalidations (β = .15, SE = 0.04, 95% CI: 0.08, 0.23). Overall, these findings indicate support for the meta-model, demonstrating a specific pathway from racial microstressors to poor sleep quality. Furthermore, these results suggest the importance of developing clinical and community approaches to address the impact of microaggressions on Black Americans’ sleep quality.

Introduction

Researchers have long documented health disparities between Black and White Americans, ranging from differential mortality to higher rates of chronic disease and greater illness severity for Black Americans compared to White Americans (Centers for Disease Control and Prevention, 2013; Hopko et al., 2003; Williams, 2012). Notably, short or long sleep duration and poor sleep quality are risk factors for many of the noted disparities in health conditions (Chen et al., 2014; Gangwisch et al., 2007; Gangwisch et al., 2006; Patyar & Patyar, 2015; Shankar et al., 2010), and Black Americans largely report poorer sleep quality compared to White Americans (Knutson et al., 2010; Lauderdale et al., 2006; Mezick et al., 2008; Ruiter et al., 2011; Thomas et al., 2006). Therefore, Black Americans may face worse health outcomes in part due to having lower quality sleep than White Americans.

Myers (2009) proposed a meta-model, which accounts for observed health disparities as the result of complex relationships between race/ethnicity and socioeconomic status (SES), that are mediated through psychological, behavioral, and biological pathways. In the United States, centuries of oppression and racist policies have limited social, economic, and educational opportunities and privileges for Black Americans; this historical background has resulted in documented disparities in SES (Williams, 2012). Myers’ (2009) meta-model posits that observed health disparities in the Black population are the ultimate result of increased exposure to psychosocial adversities; these adverse experiences are related to higher proportions of the population being of low SES and to independent associations with stressors directly related to race—namely, discrimination. These stressors, then, are associated with cognitive processing and emotion regulation (e.g., rumination, worry, anxiety, depression; Myers, 2009). Cognitive and emotional processes are theorized to relate to health behaviors (e.g., sleep), which then contribute to cumulative biopsychosocial vulnerabilities and resistances—ultimately leading to disparities in health status and outcomes (e.g., morbidity and mortality; Myers, 2009).

Disparities in SES between Black and White Americans do not explain disparities in sleep (Knutson et al., 2010; Ruiter et al., 2011); indeed, Black Americans of higher SES are more likely to face sleep disparities (Jackson et al., 2013). Rather, there is another pathway in Myers’ (2009) meta-model to explain differences in sleep: race-specific stressors, including racial discrimination. Discrimination disproportionately affects racial/ethnic minority groups and is particularly burdensome to Black Americans (Lee et al., 2019). Furthermore, Black Americans of higher SES may be more likely to face certain kinds of discrimination than those of lower SES, perhaps because they are likely to live and work in more integrated settings, increasing their chances of experiencing some types of discrimination (Lewis & Van Dyke, 2018).Therefore, race-specific stressors, such as discrimination, may contribute to observed patterns of sleep disparities. Researchers have documented associations between discrimination and poor health in various domains; these include associations between discrimination and poor mental health, chronic conditions, and poor self-reported health (Pascoe & Smart Richman, 2009; Williams & Mohammed, 2009). Furthermore, there are consistent reports of associations between measures of discrimination and self-reported sleep quality and duration (Fuller-Rowell et al., 2017; Gaston et al., 2020; Johnson et al., 2021; Slopen et al., 2016). Therefore, discrimination may impact health outcomes through its relationship with poor sleep. For example, Ong and Williams reported that greater lifetime discrimination is associated with higher inflammation burden (Ong & Williams, 2019), a potential precursor to health conditions such as diabetes and cardiovascular disease, and that sleep quality mediates this association. Furthermore, others have reported links from discrimination to type II diabetes (Gaston et al., 2021; Whitaker et al., 2017).

Some research has begun to address how discrimination affects sleep (Beatty et al., 2011; Hicken et al., 2013; Hoggard & Hill, 2018). Myers’ meta-model (2009) predicts that cognitive processing (e.g., perseverative cognition) may link discrimination to sleep. Empirically, two forms of perseverative cognition, rumination and evening worry, have been linked to poor sleep (Guastella & Moulds, 2007; McGowan et al., 2016; Thomsen et al., 2003; Zoccola et al., 2009), and studies suggest that perseverative cognition mediates the link between discrimination and sleep. For example, Beatty and colleagues (2011) reported that nightly worry mediated the association between unfair treatment and a range of sleep quality measures in a diverse sample of adults, and Hoggard and Hill (2018) reported that rumination, but not worry, mediated the association of discrimination to poor sleep quality in African American college students. Furthermore, Hicken et al. (2013) reported that race-related vigilance mediated the association of race with sleep problems; that is, Black participants reported higher levels of vigilance, and vigilance in turn was positively associated with sleep problems.

Discrimination is one among a variety of race-specific stressors to consider in Myers’ (2009) model. Harrell (2000) offered a multidimensional conceptualization of race-specific stressors which includes episodic, discrete experiences which are generally captured by measures of discrimination (e.g., being harassed by law enforcement) and day-to-day, chronic racial microstressors. While much research has considered the role of discrimination in health disparities, microaggressions (Torres-Harding et al., 2012) are likely to occur with more frequency than discrete episodes of racial discrimination, and may be a more chronic form of racial stress (Harrell, 2000). Racial and ethnic minority individuals also experience both overt forms of race-specific stress as well as subtle forms of verbal or nonverbal invalidation that are more ambiguous in nature (Meyers et al., 2020). Microaggressions function as a form of race-specific stress of this kind (Torres-Harding et al., 2012), and racial/ethnic microaggressions have been organized into several commonly occurring themes (Sue et al., 2008; Sue et al., 2007; Torres-Harding et al., 2012). These include being made to feel as a foreigner or not belonging, being ascribed intelligence based on race, color blindness and denial of individual racism (e.g., statements that indicate a desire not to acknowledge race or equating racial oppression with other forms of oppression), and the assumption of criminal status based on race (Sue et al., 2007; Torres-Harding et al., 2012). Further categories are invalidation of interethnic differences (e.g., lumping all members of a racial or ethnic group into one stereotype) as noted by (Sue et al., 2008), as well as being exoticized or overly sexualized based on race (Torres-Harding et al., 2012) and the myth of meritocracy (e.g., denying structural barriers to achievement or ascribing achievement to unfair benefits to minorities) reported by Sue et al. (2007). Other themes include pathologizing cultural values and communication styles, second-class status or implications that racial minorities occupy low social status, environmental invalidations, such as lack of racial representation in higher education or media, and invisibility, or being ignored or devalued because of race (Sue et al., 2007; Torres-Harding et al., 2012).

As a race-specific stressor faced by Black people, microaggressions are hypothesized to be associated with cognitive processes and health behaviors (i.e., perseverative cognition and sleep; Myers, 2009), as are episodic, discrete episodes of race-specific stress (i.e., discrimination). There is an emerging body of research on microaggressions and health outcomes and behaviors (in particular, sleep) in samples of racial and ethnic minority individuals. Some researchers have found associations between microaggressions, health behaviors, and mental health outcome in ethnic and racial minority groups (Lilly et al., 2018; Ong et al., 2017; Sittner et al., 2018), and others have found associations between microaggressions and poor sleep outcome in Latinx and African American college students and African American women (Davenport et al., 2021; Erving et al., 2023). These initial findings indicate a need for studies to replicate the connection between microaggressions and sleep quality in Black Americans, as well as the potential mediators of this association.

In summary, sleep may be one intermediate link between race and poor health outcomes, as sleep is a risk factor for a range of poor health outcomes (Gangwisch et al., 2007; Gangwisch et al., 2006; Shankar et al., 2010), and there is documented disparity in sleep quality between Black and White Americans (Knutson et al., 2010; Lauderdale et al., 2006; Mezick et al., 2008; Ruiter et al., 2011; Thomas et al., 2006). Race-specific stressors like discrimination and microaggressions (Erving et al., 2023; Johnson et al., 2021; Ong et al., 2017) are connected to poor sleep (Fuller-Rowell et al., 2017; Gaston et al., 2020; Slopen et al., 2016); perseverative cognition—including rumination (Hoggard & Hill, 2018), vigilance (Hicken et al., 2013), and nightly worry (Beatty et al., 2011)—has been linked to poor sleep and is a likely mediator of relationships between discrimination, microaggressions, and poor sleep quality in Black Americans. While general measures of worry report null associations with sleep quality (Hoggard & Hill, 2018), racism-specific vigilance has been associated with poor sleep (Hicken et al., 2013). Rumination, however, has support as a mediator of discrimination and poor sleep quality among African American college students (Hoggard & Hill, 2018). Furthermore, while many studies examine discrimination (Beatty et al., 2011; Hicken et al., 2013; Hoggard & Hill, 2018), few examine the relationship of more chronic racial stressors such as microaggressions to sleep among Black Americans (Davenport et al., 2021; Erving et al., 2023). Finally, since perseverative cognition may mediate a negative relationship between discrimination and sleep quality, it may also mediate a negative relationship between microaggressions and sleep quality. Therefore, this project tested the hypotheses that racism-related vigilance and rumination would mediate the relationship between discrimination and poor sleep (Hypotheses #1 and 2), as well as between microaggressions and poor sleep (Hypotheses #3 and 4), among Black Americans.

Methods

Participants

This project collected data from participants recruited via the Amazon Mechanical Turk (MTurk) program, an online platform that allows researchers to post “human intelligence tasks” (HITs) that workers (i.e., individuals with accounts on the platform) can sign up for and complete for compensation provided by the researcher (Amazon Mechanical Turk, n.d.). That is, MTurk workers are not employees of Amazon, but rather individuals who use the service to access and complete HITs. Recent research identified that samples of workers on MTurk are generally similar to the U.S. population in terms of income and race, though women are over-represented compared to men, and nearly one-half have a college education (Moss et al., 2023); since workers typically are active on the platform from 12–28 months (Difallah et al., 2018), samples vary in their exact makeup but have remained generally stable in demographics (Moss et al., 2020). All data were collected between February and July 2021. A screener survey with questions regarding inclusion criteria was advertised to all workers on MTurk; those who took this survey and met inclusion criteria were then provided access to the full study survey. Inclusion criteria included: 1) current U.S. residency, 2) age 18 years to 65 years (given that experiences of discrimination or microaggressions may be more overt in older populations, we limited the age of participants to 65 years old) 3) Black racial/ethnic identity (including participants who identified as biracial or multiracial), 4) having been born in the United States, and 5) having at least one parent born in the United States, as research suggests that experiences of discrimination may differ across immigration status (Krieger et al., 2005).

Procedures

Screening.

Participants completed a brief screening survey with questions regarding inclusion criteria, as recommended by McDuffie (2019); a total of 4,832 participants were screened, with 311 (6.93%) meeting inclusion criteria.

Survey completion.

Those participants who met inclusion criteria were assigned credentials in MTurk that allowed them to access the full survey. Of the 311 participants who qualified for the study based on their responses to the screener, 264 (84.9%) completed the full survey. The final analytic sample consisted of the total who completed the survey (264) minus 22 (8.3%) who were eliminated for nonsensical answers, such as answers to free-text items that were non-responsive to the question or which included times of day that are not possible (e.g., 39), seven (2.6%) duplicate entries, and 12 (4.5%) who indicated in the main survey that they did not meet criteria (i.e., these participants indicated meeting all inclusion criteria in the screener, but in the main survey indicated that they did not identify with Black racial or ethnic identity). Therefore, the total N was 223. Participants completed online questionnaires measuring study variables, described below, as well as a demographic questionnaire.

Compensation and ethics approval.

Participants who completed study materials were paid $3.02 for their participation, a rate equivalent to federal minimum wage, which was $7.25 per hour at the time of data collection (U.S. Department of Labor, n.d.), based on estimated time of task completion of six questions per minute (total questions: 150; total estimated time: 25 minutes; Cloud Research, 2020). Study procedures were approved by the Institutional Review Board of the University of Missouri-Kansas City. Prior to survey questions, participants were provided a description of study procedures, voluntary nature of procedures, potential risk and benefits of participation, and contact information for investigators.

Participant characteristics

Table 2 presents descriptive information on the study sample (N = 223); participants ranged in age from 20 to 65 years, with a mean age of 35.7 (SD = 9.75). In the sample, 53.4% indicated they were assigned male sex at birth; 53.8% of the sample identified as men, while 44.8% identified as women and 1.3% identified as non-binary. Three participants indicated that they identified as a gender discordant with the sex they were assigned at birth (e.g., assigned female at birth and identify as male or non-binary). Most participants identified as Black/African American (78.2%), and most were partnered (61.4%). The largest percentage of participants had completed a four-year college degree (37.2%), and most were employed (86.0%); annual income in the sample ranged widely, but the highest proportion of participants reported annual incomes between $30,001 and $60,000 (36.3%). In terms of region of residence, the largest percentage of participants indicated residence in the southern region of the United States (48.2%).

Table 2.

Descriptive statistics and correlations for all study variables

	1	2	3	4	5	6	7	8	9	10	11	12
1. Poor Sleep Quality	--
2. Foreigner/Not Belonging	.37**	--
3. Criminality	.36**	.47**	--
4. Sexualization	.30**	.49**	.54**	--
5. Low-Achieving/Undesirable	.29**	.39**	.64**	.55**	--
6. Invisibility	.38**	.59**	.66**	.57**	.66**	--
7. Environmental Invalidation	.18**	.36**	.40**	.43**	.50**	.54**	--
8. Discrimination	.29**	.39**	.57**	.45**	.50**	.58**	.41**	--
9. Rumination	.45**	.44**	.34**	.31**	.30**	.47**	.24**	.23**	--
10. Vigilance	.14*	.15*	.28**	.16*	.34**	.33**	.08	.28**	.20**	--
11. Pandemic Stress	.28**	.32**	.29**	.26**	.20**	.31**	.17*	.25**	.25**	.25**	--
12. Income	−.15*	.06	.13	−.05	.08	−.03	.06	.01	.02	.00	−.02	--
M	7.53	3.03	5.04	3.01	12.94	8.55	7.35	10.66	50.54	23.16	19.70	5.28
SD	3.83	2.71	3.58	2.83	6.84	6.36	4.06	7.86	14.45	5.97	16.85	2.75
Range	0 – 21	0 – 9	0 – 12	0 – 9	0 – 27	0 – 24	0 – 15	0 – 36	22 – 88	8 – 36	0–95	1–12
Skew	0.50	0.57	0.12	0.54	−0.08	0.32	−0.02	0.05	0.15	−0.40	0.44	2.05
Kurtosis	0.18	−0.79	−1.11	−0.91	−0.79	−0.98	−0.87	0.59	−0.53	−0.44	−0.16	1.34
Cronbach’s alpha	.87	.81	.86	.85	.87	.90	.78	.78	.93	.82	.88	-

Note

^**p < 0.01

^*p < .05

the mean income represented by 5.28 would fall between $40,000 to $60,000 per year; the range includes values from $0 to over $100,000 per year.

Measures.

This study employed the measures described below, as well as a general demographic information form. All measures are available to the public free of charge online or through contacting the respective authors. Observed internal consistency reliabilities of all scales and subscales are reported in Table 1. A measure of pandemic stress was included as a potential covariate, given the context of data collection during the COVID-19 pandemic.

Table 1.

Descriptive statistics of participant characteristics (N=223).

Participant characteristics	M (SD) or %
Age (years) a	35.77 (9.75)
Sex Assigned at Birth
Male	53.4%
Female	46.6%
Gender Identity
Identify as man	53.8%
Identify as woman	44.8%
Identify as non-binary	1.3%
Racial/Ethnic Identity
Black/African American	78.2%
Multiracial b	10.6%
Did not provide race/ethnicity c	11.2%
Relationship Status
Partnered	61.4%
Not Partnered	38.4%
Education
High school diploma/GED	8.5%
Technical school or some college	24.6%
4-year college degree	37.2%
Graduate degree	29.6%
Employment
Employed	86.00%
Not employed	6.70%
Other (student, self-employed and working at home)	7.10%
Annual Income
$0 – $30,000	30.00%
$30,001 – $60,000	36.30%
$60,001 – $90,000	21.00%
$91,001 +	5.80%
Region d
South	48.2%
Northeast	17.8%
Midwest	16.9%
West	16.4%

Note:

Percentages in some categories do not sum to 100 as not all participants provided data for those categories.

^aAge calculated from 2021 - year born.

^bParticipants categorized as multi-racial self-identified as Black/African American and one or more other racial/ethnicities including Native American/Alaska Native (0.4%), Asian (1.3%), Latino/Hispanic (2.2.%), and White (6.3%).

^cAlthough study participants could decline to self-identify race or ethnicity, the screener survey only provided full study survey access to potential participants who identified as Black/African American, therefore ensuring the study sample was comprised of Black/African American participants.

^dStates within regions include: Midwest = Illinois, Ohio, Wisconsin, Minnesota, Missouri, Michigan, Indiana, South Dakota, North Dakota, Kansas; Northeast = New York, New Jersey, Massachusetts, Pennsylvania, Connecticut, Rhode Island. South = Florida, Texas, Georgia, Alabama, Louisiana, Virginia, North Carolina, Maryland, South Carolina, Tennessee, Mississippi, Oklahoma, Kentucky, Washington, D.C., Arkansas; West = California, New Mexico, Arizona, Washington, Wyoming, Oregon, Montana, Hawaii. Only states in which participants reported living are listed here. States are listed in descending order of frequency within regions.

Racial/Ethnic Discrimination.

The Experiences of Discrimination Scale (EOD; Krieger et al., 2005) was used to assess experiences of discrimination across nine settings (school, work, housing, medical care, stores and restaurants, financial services, public spaces, and police or courts). Questions in this nine-item measure ask participants whether they have “ever experienced discrimination, been prevented from doing something, or been hassled or made to feel inferior in any of the following situations because of your race, ethnicity, or color?” in settings such as school, work, or housing (Krieger et al., p. 1590). Participants can respond “yes” or “no” to each situation. If participants respond affirmatively, they are asked to rate how many times this occurred, from 1 = once to 3 = four or more times. Standard scoring was applied; scores can range from 0 to 45, with higher scores indicating more lifetime discrimination.

Microaggressions.

The Racial Microaggressions Scale (RMAS; Torres-Harding et al., 2012) was used to measure experiences of microaggressions. This 32-item tool was developed in a sample of ethnic/racial minorities (N = 377; 39.5% Black), and it comprises six sub-scales, measuring distinct themes of microaggressions frequently reported in the literature and reported by participants: Foreigner/Not Belonging (e.g., being treated as if one does not belong); Criminality (e.g., treated as if one is dangerous, violent, or a criminal); Sexualization (e.g., being sexualized or exoticized due to one’s race/ethnicity; Low-Achieving/Undesirable (e.g., treatment indicating members of one’s race/ethnicity are “interchangeable, uniformly incompetent, incapable, low achieving, and dysfunctional, and as if successes are due to unfair entitlements and special treatment” (Torres-Harding et al., 2012, p. 157)]); Invisibility (e.g., being treated as if one is not a “real” person or of lower status); and Environmental Invalidations (e.g., lack of members of one’s racial/ethnic group being represented in work, school, community settings or in positions of power). As the authors (Torres-Harding et al., 2012) reported that variance was accounted for by high internal consistency within sub-scales, it is best practice to consider them individually. Participants indicate how frequently they experience examples of microaggressions on a Likert type scale (0 = never, 1 = a little/rarely, 2 = sometimes/a moderate amount, 3 = often/frequently). Global scores range from 0–96, and scores range on the subscales as follows: Foreigner/Not Belonging, 0 to 9; Criminality, 0 to12; Sexualization, 0 to 9; Low-Achieving/Undesirable, 0 to 27; Invisibility, 0 to 24; and Environmental Invalidations, 0 to 15.

Rumination.

The Ruminative Responses Scale (RRS; Nolen-Hoeksema & Morrow, 1991) was used to assess participants’ tendency to ruminate. The RRS is a 22-item scale which asks participants how frequently they engage in ruminative cognitions (1 = never to 4 = almost always). Items are summed; scores range from 22 to 88, and a higher score indicates a higher tendency to ruminate.

Vigilance.

The Racism Related Vigilance Scale (RRV; Clark et al., 2006) was used to assess participants’ tendency to anticipate and prepare for instances of racism. This scale reflects themes which Black participants expressed when asked about how they prepare for day-to-day experiences of racism. This measure contains six items (1 = Almost Every day to 6 = Never) assessing the frequency of participants’ anticipatory or preparatory cognitions related to experiences of discrimination. Item responses are reverse scored so that a higher score indicates more racism-related vigilance, and scores range from 6 to 36.

Sleep quality.

Sleep quality over the past month was assessed with the 19-item Pittsburgh Sleep Quality Index (PSQI; Buysse et al., 1989). The PSQI comprises seven components of sleep quality—duration, sleep efficiency, sleep latency, disturbance, restfulness, daytime dysfunction, and use of sleeping medication. Component scores are summed to create a global sleep quality score, with possible scores ranging from 0 to 21. Higher scores indicate more poor global sleep quality; scores greater than 5 are considered indicative of clinically significant poor sleep.

Pandemic stress.

The Pandemic Stress Questionnaire (PSQ; Kujawa et al., 2020) was used to assess the impact of the COVID-19 pandemic on participant wellbeing. The PSQ is a 25-item measure that assesses exposure to stressful events related to the pandemic in six domains: general life disruption, interpersonal, financial, education/professional goals, health-self, and health-others. Items are rated as “yes” or “no”; those items that are endorsed are then rated in terms of how severely (i.e., negative impact, frequency, and duration) they have impacted the respondent (1 = not at all bad to 5 = extremely bad). Scores range from 0 to 25, and higher scores indicate greater exposure to pandemic-related stressors.

Data analysis

Descriptive statistics were run using SPSS 26.0 (IBM Corp, 2019). Data were screened for completeness and assumptions prior to hypothesis testing. We conducted correlation analyses and independent samples t-tests to identify covariates, screening multiple demographic variables: age, education, income, pandemic stress, sex assigned at birth, employment, and relationship status. Variables that were significantly associated with the sleep quality were included as covariates.

Descriptive statistics and assumptions.

A total of 223 participants met inclusion criteria, had complete data, and were thus included in the analytic sample. Inspection of histograms and skew and kurtosis values revealed that study variables were roughly normally distributed with no univariate outliers. Mahalanobis distance values and plots of residuals revealed no multivariate outliers. Descriptive statistics for study variables and scale correlation coefficients are summarized in Table 2.

Analyses.

Hypotheses were tested using the PROCESS macro (Hayes, 2018). Specifically, one model used parallel mediation to test Hypotheses #1 and #2. That is, rumination was expected to mediate an association between discrimination and sleep quality, while racism-related vigilance was also expected to mediate that association (Figure 1). A second series of six parallel mediation models were used to test Hypotheses #3 and #4: in each, rumination was expected to mediate associations between each microaggression subscale and sleep quality, and racism-related vigilance was also expected to mediate those associations. The PROCESS macro provides point estimates and confidence interval estimates of direct and indirect effects; evidence of mediation would be found if the 95% CI’s of the indirect paths estimated in the above models do not contain zero. Hochberg’s False Discovery Rate (FDR) correction was applied to reduce the risk of Type I error, following procedures described by Cribbie (2000) to control for FDR.

Figure 1.

Diagram for the parallel mediator models including the total effect and direct effects of discrimination and microaggressions on poor sleep quality; beta-coefficients for the effects of each path are included (ß = standardized coefficient). Income and pandemic stress were included in the model as covariates but not depicted in the diagram. * = significant with the Hochberg’s False Discovery Rate correction applied. Results reflect outcomes after controlling for pandemic stress and annual income.

Power analysis.

Initial power analyses indicated that the sample sizes necessary to detect small and medium effects in R² deviation from 0 and R² change with .80 and .70 power at four different alpha levels ranged from 44 to 1037. Using G*Power (Faul, 2008), we calculated observed post-hoc power for overall models, direct effects, and indirect effects; we used alpha of .05, a sample of 223, and observed effect size ranges for each parameter. Power for each of the overall models had observed power > .99. Power for individual direct effects ranged from .05 for effects sizes of 0 with three predictors to .99 for effect sizes of .19 with five predictors. Power for indirect effects ranged from .05 for effects sizes of 0 with five predictors to .36 for effect sizes of .023 with five predictors. In sum, this project achieved power that was sufficient to detect large effects in the overall models (i.e., >.99), but was insufficient to detect small direct or indirect effects (.05 to .36).

Results

Covariates

We examined potential covariates coded as continuous variables—age, education, income, pandemic stress—with correlation analyses (Table 1). Results showed that pandemic stress and income were significantly related to poor sleep quality. Therefore, these two variables were entered as covariates in the models estimated. We also conducted t-tests to examine potential covariates with dichotomous variables: sex assigned at birth, gender¹, employment, and relationship status. However, none of these significantly differed by sleep quality. Therefore, none were used as a covariate in the models estimated.

Hypothesized models

Hypotheses #1 and #2.

One parallel mediation model tested whether Hypotheses #1 (rumination) and #2 (racism-related vigilance) would mediate the association between discrimination and poor sleep quality. That is, we predicted that discrimination would be positively associated with rumination and racism-related vigilance, both of which in turn would be positively associated with poor sleep quality. Parameters for pathways through both mediators were estimated in Model 1 (Table 3, Figure 1). Significant parameters after the FDR correction are bolded.

Table 3.

Models 1–7: Direct Effects Coefficients of Discrimination and Microaggressions on Poor Sleep Quality.

Antecedent → Consequent	Path	b	SE	95% CI	t (df = 219)	p
Model 1: Discrimination (DIS)
X (DIS) →M1 (RUM) F(3, 216) = 6.26, p = .0004, R2 = .08	a 1	0.29	0.14	0.02, 0.55	2.12	.04
X (DIS) →M2 (VIG) F(3, 216) = 8.42 p < .0001, R2 = .10	a2	0.19	0.05	0.08, 0.30	3.44	.0007
X (DIS) → Y (PSQ)	c′	0.08	0.03	0.01, 0.14	2.32	.02
M1 (RUM) → Y (PSQ)	b1	0.10	0.03	0.07, 0.14	6.51	<.0001
M2 (VIG) → Y (PSQ) F(5, 214) = 16.80 p < .0001, R2 = .28	b2	0.00	0.04	−0.08, 0.08	0.03	.98
Model 2: Foreigner/Not Belonging Microaggressions (FOR)
X (FOR) →M1 (RUM) F(3, 216) = 18.76, p < .0001, R2 = .21	a 1	2.16	0.34	1.49,2.84	6.30	<.0001
X (FOR) →M2 (VIG) F(3, 216) = 4.73, p = .003, R2 = .06	a2	0.18	0.15	−0.12, 0.48	1.17	.24
X (FOR) → Y (PSQ)	c′	0.30	0.09	0.12, 0.48	3.22	.002
M1 (RUM) → Y (PSQ)	b1	0.09	0.02	0.05, 0.12	5.10	<.0001
M2 (VIG) → Y (PSQ) F(5, 214) = 18.16, p < .0001, R2 = .30	b2	0.02	0.01	−0.06, 0.09	0.47	.65
Model 3: Criminality Microaggressions (CRIM)
X (CRIM) →M1 (RUM) F(3, 216) = 11.60, p < .0001, R2 = .14	a 1	1.18	0.27	0.65, 1.70	4.42	<.0001
X (CRIM) →M2 (VIG) F(3, 216) = 8.48, p < .0001, R2 = .10	a2	0.39	0.11	0.17, 0.61	3.46	.0006
X (CRIM) → Y (PSQ)	c′	0.22	0.07	0.08, 0.35	3.17	.0018
M1 (RUM) → Y (PSQ)	b1	0.09	0.02	0.06, 0.13	5.80	<.0001
M2 (VIG) → Y (PSQ) F(5, 214) = 18.07, p < .0001, R2 = .30	b2	−0.00	0.04	−0.08, 0.07	−0.10	.92
Model 4: Sexualization Microaggressions (SEX)
X (SEX) →M1 (RUM) F(3, 216) = 10.70 p < .0001, R2 = .13	a 1	1.38	0.33	0.72, 2.04	4.12	.0001
X (SEX) →M2 (VIG) F(3, 216) = 5.34, p = .0014, R2 = .07	a2	0.25	0.14	−0.03,0.53	1.76	.08
X (SEX) → Y (PSQ)	c′	0.21	0.08	0.04, 0.37	2.48	.014
M1 (RUM) → Y (PSQ)	b1	0.10	0.02	0.07, 0.13	5.96	<.0001
M2 (VIG) → Y (PSQ) F(5, 214) = 16.92, p < .0001, R2 = .28	b2	0.01	0.04	−0.06, 0.09	0.32	.74
Model 5: Low-Achieving/Undesirable Microaggressions (LOW)
X (LOW) →M1 (RUM) F(3, 216) = 10.58 p < .0001, R2 = .13	a 1	0.56	0.14	0.29, 0.83	4.08	.0001
X (LOW) →M2 (VIG) F(3, 216) = 12.58, p < .0001, R2 = .15	a2	0.27	0.06	0.16, 0.38	4.85	<.0001
X (LOW) → Y (PSQ)	c′	0.08	0.04	0.01, 0.15	2.34	.02
M1 (RUM) → Y (PSQ)	b1	0.10	0.02	0.07, 0.13	6.07	<.0001
M2 (VIG) → Y (PSQ) F(5, 214) = 16.92, p < .0001, R2 = .28	b2	−0.01	0.04	−0.09, 0.07	−0.16	.86
Model 6: Invisibility Microaggressions (INV)
X (INV) →M1 (RUM) F(3, 216) = 21.38 p < .0001, R2 = .23	a 1	0.99	.14	0.70, 1.27	6.86	<.0001
X (INV) →M2 (VIG) F F(3, 216) = 11.25, p < .0001, R2 = .14	a2	0.28	0.06	0.16, 0.40	4.46	<.0001
X (INV) → Y (PSQ)	c′	0.11	0.04	0.03, 0.19	2.59	.01
M1 (RUM) → Y (PSQ)	b1	0.09	0.02	0.06, 0.12	5.22	<.0001
M2 (VIG) → Y (PSQ) F(5, 214) = 17.16, p < .0001, R2 = .29	b2	−0.01	0.04	−0.08, 0.07	−0.14	.88
Model 7: Environmental Invalidations Microaggressions (ENV)
X (ENV) →M1 (RUM) F(3, 216) = 8.14, p < .0001, R2 = .10	a 1	0.73	.23	0.27, 1.19	3.13	.0001
X (ENV) →M2 (VIG) F(3, 216) = 4.48, p = .005, R2 = .06	a2	0.08	0.10	−0.11, 0.04	1.10	0.27
X (ENV) → Y (PSQ)	c′	0.05	0.06	−0.06, 0.16	0.90	.37
M1 (RUM) → Y (PSQ)	b1	0.11	.02	0.07, 0.14	6.42	<.0001
M2 (VIG) → Y (PSQ) F(5, 214) = 15.56, p < .0001, R2 = .27	b2	0.02	0.04	−0.06, 0.10	0.50	.62

Note. Parameters in bold font are significant with the Hochberg’s False Discovery Rate correction applied. Results reflect outcomes after controlling for pandemic stress and annual income. RUM = rumination; VIG = racism-related vigilance; PSQ = poor sleep quality.

The overall model was significant. Path a₁, estimating the relationship between discrimination and rumination was significant, as well as Path a₂, estimating the relationship between discrimination and racism-related vigilance. In addition, Path b₁, estimating the path from rumination to poor sleep quality was significant. Path b₂, however, was not significant. Furthermore, path c’, the relationship of discrimination to poor sleep quality after controlling for paths through rumination and racism-related vigilance, remained significant. There were significant total effects of the model, but no indirect effects were significant (Table 4). Therefore, neither Hypotheses #1 and #2 were supported.

Table 4.

Models 1–7: Total and Indirect Effects of Discrimination and Microaggressions on Poor Sleep Quality

Model 1: Discrimination	Unstandardized Coeff	SE	95% CI	Standardized total effect	t (df = 219)	p
Total effect	0.11	0.03	0.04, 0.17	.22	3.07	.002
	Unstandardized Coeff	SE	95% CI	Standardized β	SE	95% CI
Total indirect effect	0.03	0.02	0.00, 0.07	.06	0.03	0.00, 0.13
Through RUM	0.03	0.02	0.00, 0.07	.06	0.03	0.00, 0.13
Through VIG	0.00	0.00	−0.02, 0.02	.00	0.02	−0.03, 0.04
Model 2: Foreigner/Not Belonging	Unstandardized Coeff	SE	95% CI	Standardized total effect	t (df = 219)	p
Total effect	0.49	0.09	0.31, 0.76	.35	5.44	<.0001
	Unstandardized Coeff	SE	95% CI	Standardized β	SE	95% CI
Total indirect effect	0.19	0.05	0.12, 0.29	.14	0.03	0.08, 0.20
Through RUM	0.19	0.05	0.11, 0.29	.13	0.03	0.08, 0.20
Through VIG	0.00	0.01	−0.01, 0.03	.01	0.01	−0.01, 0.02
Model 3: Criminality	Unstandardized Coeff	SE	95% CI	Standardized total effect	t (df = 219)	p
Total effect	0.33	0.07	0.19, 0.46	.31	4.76	<.0001
	Unstandardized Coeff	SE	95% CI	Standardized β	SE	95% CI
Total indirect effect	0.11	0.03	0.05, 0.18	.10	0.03	0.05, 0.17
Through RUM	0.11	0.03	0.05, 0.19	.11	0.03	0.05, 0.17
Through VIG	−0.00	0.02	−0.03, 0.03	−.00	0.02	−0.03 0.03
Model 4: Sexualization	Unstandardized Coeff	SE	95% CI	Standardized total effect	t (df = 219)	p
Total effect	0.35	0.09	0.18, 0.52	.25	3.87	.0001
	Unstandardized Coeff	SE	95% CI	Standardized β	SE	95% CI
Total indirect effect	0.14	0.05	0.07, 0.24	.10	0.03	0.05, 0.17
Through RUM	0.14	0.05	0.06, 0.24	.10	0.03	0.05, 0.17
Through VIG	0.00	0.01	−0.02, 0.04	.00	0.01	−0.01, 0.03
Model 5: Low-Achieving/Undesirable	Unstandardized Coeff	SE	95% CI	Standardized total effect	t (df = 219)	p
Total effect	0.14	0.04	0.07, 0.21	.25	3.87	.0001
	Unstandardized Coeff	SE	95% CI	Standardized β	SE	95% CI
Total indirect effect	0.05	0.02	0.02, 0.09	.10	0.03	0.03, 0.17
Through RUM	0.06	0.02	0.02, 0.09	.10	0.03	0.05, 0.16
Through VIG	−0.00	0.01	−0.02, 0.03	−.00	0.02	−0.04, 0.05
Model 6: Invisibility	Unstandardized Coeff	SE	95% CI	Standardized total effect	t (df = 219)	p
Total effect	0.20	.04	0.12, 0.27	.33	5.06	<.0001
	Unstandardized Coeff	SE	95% CI	Standardized β	SE	95% CI
Total indirect effect	0.09	.02	0.05, 0.14	.15	0.04	0.08, 0.22
Through RUM	0.09	0.02	0.05, 0.14	.15	0.04	0.08, 0.23
Through VIG	−0.00	0.01	−0.02, 0.03	−.00	0.02	−0.04, −.04
Model 7: Environmental Invalidations	Unstandardized Coeff	SE	95% CI	Standardized total effect	t (df = 219)	p
Total effect	0.13	0.06	0.01, 0.25	.14	2.14	.03
	Unstandardized Coeff	SE	95% CI	Standardized β	SE	95% CI
Total indirect effect	0.08	0.03	0.03, 0.14	0.08	0.03	0.03, 0.15
Through RUM	0.08	0.03	0.03, 0.14	0.08	0.03	0.03, 0.15
Through VIG	0.00	0.01	−0.01, 0.02	0.00	0.01	−0.01 0.02

Note. Indirect effects are calculated based on bootstrapped samples of 5000. Parameters in bold font are significant with the Hochberg’s False Discovery Rate correction applied. RUM = rumination; VIG = racism-related vigilance.

Hypotheses #3 and #4.

Hypotheses #3 and #4 predicted that both rumination and racism-related vigilance would mediate the association between microaggressions and poor sleep quality. That is, we predicted that microaggressions would be positively associated with rumination and racism-related vigilance, both of which in turn would be positively associated with poor sleep quality. As microaggressions are not a unitary construct,(Torres-Harding et al., 2012) and to protect against multicollinearity, six separate parallel mediation models were run to estimate relationships between six subscales of the RMAS (Torres-Harding et al., 2012) and poor sleep quality. As above, significant parameters reflect the application of the FDR correction.

Models 2–7 tested parallel mediation of each microaggression subscale through rumination and racism-related vigilance to poor sleep quality (Table 3, Figure 1). There was a consistent pattern of direct effects from microaggressions to rumination, path a₁, and from rumination to poor sleep quality, path b₁. However, direct effects from microaggressions to racism-related vigilance were less consistent. For models testing associations between microaggression subscales of Criminality, Low Achieving/Undesirable, and Invisibility, there was a significant association between the microaggression subscale and racism-related vigilance, path a₂. In contrast, for those models testing associations between microaggression subscales of Foreigner/Not Belonging, Sexualization, and Environmental Invalidations, there were no significant direct paths from the microaggression subscale to racism-related vigilance. Finally, there were no significant direct effects from racism-related vigilance to poor sleep quality, path b₂.

There were significant total effects in Models 2–6 (Table 4), and significant total indirect effects in Models 2–7. That is, there were significant indirect effects from in each model from each microaggressions subscale to poor sleep quality. Specifically, there were significant indirect effects (95% CI’s) from each microaggressions subscale to poor sleep quality through rumination. However, there were no significant indirect effects from microaggressions to poor sleep quality through racism-related vigilance. In sum, Hypothesis #3 was supported, while Hypothesis #4 was not.

Discussion

Using Myers’ (2009) meta model as a theoretical framework, this project investigated relationships between experiences of race-specific stress (e.g., discrimination and microaggressions), perseverative cognition, and poor sleep quality. We posited that rumination (hypothesis 1) and racism-related vigilance (hypothesis 2) would mediate the link between discrimination and poor sleep quality in Black Americans. Additionally, we hypothesized that rumination (hypothesis 3) and racism-related vigilance (hypothesis 4) would mediate a relationship between microaggressions, another form race-specific stress, and poor sleep quality. Seven parallel mediation models tested whether rumination and racism-related vigilance mediate the relationship between experiences of discrimination and poor sleep quality, as well as models to test whether rumination and racism-related vigilance mediate relationships between six different subscales of microaggressions (Foreigner/Not Belonging, Criminality, Sexualization, Low-Achieving/Undesirable, Invisibility, and Environmental Invalidations) and poor sleep quality. Results were mixed. Contrary to predictions, neither rumination nor racism-related vigilance mediated relationships between experiences of discrimination and poor sleep quality. However, rumination partially mediated a significant indirect relationship between all six microaggression subscales and poor sleep quality. Finally, racism-related vigilance did not mediate any relationships between microaggressions and poor sleep quality. Despite these mixed findings, each of the overall models resulted in large effect sizes, indicating that discrimination, microaggressions, and perseverative cognition, in addition to pandemic stress and income, accounted for a large proportion of the variance in poor sleep quality in this sample.

Comparison to other samples

Relative to results in other studies, the current sample demonstrated similar sleep quality, but higher frequencies of microaggressions and discrimination. (Beatty et al., 2011; Hoggard & Hill, 2018; Krieger et al., 2005; O’Keefe et al., 2015; Torres-Harding & Turner, 2015; Torres-Harding et al., 2012) The sample reported higher levels of rumination compared to similar samples (Guastella & Moulds, 2007); direct comparisons with vigilance were not possible due to differing sample characteristics.

Discrimination

Consistent with the extant literature (e.g., Hoggard & Hill, 2018), we found direct effects from discrimination to rumination and from rumination to poor sleep quality; likewise, results indicate that discrimination is positively related to racism-related vigilance, which has been observed in other samples (Keum & Li, 2022; Pichardo et al., 2021). Unlike previous studies, we did not find indirect effects from discrimination to sleep (Hoggard & Hill, 2018) or an association between vigilance and sleep (Hicken et al., 2013). However, we employed a more focused measure of discrimination that assessed potentially infrequent events (e.g., getting bank loans; Krieger et al., p. 1590) compared to previous research (Hoggard & Hill, 2018), who assessed a wider range of discriminatory experiences (e.g., being ignored or excluded. Therefore, null indirect effects from discrimination to poor sleep quality through rumination may indicate that the more limited range of experiences we assessed may not impact sleep through perseverative cognition (e.g., rumination or vigilance). Furthermore, vigilance may be differentially related to distinct sleep domains. Nevertheless, the positive relationship of discrimination with vigilance and rumination highlights a race-related burden, as theorized by Myers (2009), wherein Black Americans’ cognitive resources are depleted by discrimination. That is, the associations between discrimination and rumination and vigilance may imply that these cognitive processes related to experiences of discrimination divert Black Americans’ cognitive resources to interpreting past discrimination and preparing for future instances; this process may reduce their ability to engage in meaningful activities and lead to poor mental health outcomes. Furthermore, we found an independent association between discrimination and sleep quality, consistent with a large body of literature (Gaston et al., 2020; Slopen et al., 2016); as sleep is a risk factor for chronic disease, our findings amplify existing literature suggesting the importance of assessing the burden of discrimination on the health of Black Americans.

Microaggressions

We also observed an overall pattern of relationships between six subscales of microaggressions—Foreigner/Not Belonging, Criminality, Sexualization, Low-Achieving/Undesirable, Invisibility, and Environmental Invalidations—and sleep quality: direct effects from microaggressions and indirect effects through rumination. These findings align with previous research on discrimination and sleep (Hoggard & Hill, 2018) as well as on microaggressions and sleep among Black women (Erving et al., 2023) and college students (Davenport et al., 2021), providing evidence of the importance of racial microstressors, as theorized by Harrell (2000). Importantly, these findings suggest that chronic, low intensity race-related stress may have implications for the development of chronic disease, inasmuch as these stressors impact health maintaining processes such as sleep (Shankar et al., 2010). Furthermore, this finding points to a potential mechanism—rumination—that links these stressors to sleep. Rumination may be considered an emotion-focused coping approach, in which a person focuses their attention on negative moods or events (Nolen-Hoeksema, 1991). While this coping approach may be employed to process emotion and make sense of experiences, it can interfere with sleep, particularly for those prone to rumination. That is, for Black Americans who cope through rumination, they may be likely to experience poor sleep the more they experience microaggressions. Therefore, these findings provide partial support for Myers’ (2009) meta-model linking race/ethnicity to health outcomes via psychosocial adversities (e.g., microaggressions), cognitive processing (rumination) and emotional regulation, and health behaviors (sleep).

We found that a sub-set of microaggressions—Criminality, Low-Achieving/Undesirable, and Invisibility—are positively related to racism-related vigilance. As with experiences of discrimination, this finding indicates that experiences of certain microaggressions tax the cognitive resources of Black Americans, who may engage in racism-related vigilance to avoid future microaggressions. Again, it appears that the overall severity of the experience (i.e., discrimination or microaggressions) is not important in terms of its cognitive toll. That is, both major experiences of discrimination and microaggressions of these categories are sufficiently stressful for the target person to engage in cognitive efforts (i.e., vigilance) to avoid future instances. Therefore, while microaggressions are less overt, and may result in less severe immediate harm to the target than overt experiences of discrimination (e.g., subtle implications of criminality may not have the same material impact as unfair treatment by the police), their relatively high frequency has an outsized impact on the cognitive resources of Black Americans.

As noted by the scale developers (Torres-Harding et al., 2012), each sub-scale on the microaggressions measure assesses distinct experiences, which aligns with our finding that certain experiences (e.g., being treated as though one is dangerous, unimportant, or of second-class status) are related to increased vigilance, while others are not as important. In our sample, all participants were born in the U.S., which may have affected the null association between Foreigner/Not Belonging microaggressions; it is possible that samples of Black Americans from immigrant backgrounds would reveal an association not seen in our study. In terms of Sexualization microaggressions, there may be an intersectional effect; future research should investigate whether for Black women or gender minorities, these microaggressions are associated with race-related vigilance. Regarding Environmental Invalidations, this subscale primarily assesses how much participants experience a lack of representation of others of their race in workplaces, educations, and positions of power. While these microaggressions may lead to rumination, as found by this project, it may be that vigilance is a less frequent response to these experiences. That is, these microaggressions are more related to what an individual sees in their environment rather than how they are perceived or treated by others, which then may not lead to increased self-monitoring to avoid unwanted treatment.

Implications

Clinically, our findings suggest several important implications. First, perseverative cognition could be a useful target for clinical intervention in Black Americans who have poor sleep quality—rumination focused on experiences of microaggressions interferes with sleep quality. Therefore, while clinicians should validate an individual’s experiences and reactions, therapeutic interventions may be tailored to address the toll these experiences take on a patient’s sleep quality. For instance, it may be helpful to equip Black Americans who are targets of microaggressions with responses that can buffer their impact in the moment; this may include formulating effective verbal or behavioral responses which enhance their sense of control and validate their perspective (Sue et al., 2019). It may be valuable to employ therapeutic approaches that focus on reducing the impact of rumination, which can include social withdrawal and depressed mood (Roberts et al., 1998), rather than seeking to alter participants’ perceptions of microaggressions. In addition, CBT for Insomnia (CBT-I) is well supported as a treatment for insomnia (Wang et al., 2005). Future research should investigate the potential benefit tailoring the intervention to address the connection between discriminatory experiences and perseverative cognition with poor sleep. Research should be conducted with Black Americans to identify which aspects of CBT-I are most beneficial and what changes would be most appropriate. Additionally, affirming a positive racial identity and encouraging patients to pursue accountability for those involved in racist incidents (Forsyth & Carter, 2012), as well as other individual and collective coping approaches may be beneficial for Black Americans to cope with race-related stress more generally (Jones et al., 2020).

Beyond individual-level interventions to mitigate the effects of race-related stressors, researchers should consider systems-level interventions to address their effects as well as their root cause. Some have suggested that mentoring and support programs for Black Americans in professional disciplines, such as medicine and psychology, may help mitigate the effects of discrimination and microaggressions, perhaps through enhancing a sense of belonging (American Psychological Association, 2023; Velazquez et al., 2022); however, it should be noted that there may be limited availability of mentors in such fields, and future research should evaluate their utility. Furthermore, initiatives across societal sectors should be developed that directly address and eliminate discrimination and microaggressions (Campbell & Tan, 2023; Fattoracci & King, 2023; Sandoval et al., 2020)

Limitations

It is important to acknowledge some limitations of this project. Since data were cross-sectional, we cannot make causal or temporal inferences about the relationships observed. Mediation analyses with cross-sectional data leave open a range of interpretations; that is, variables in the model that are presumed to be predictors (e.g., microaggressions) may or may not temporally precede dependent variables in the model (e.g., rumination or poor sleep quality). For example, it is possible that rumination and vigilance may contribute to perceptions of discrimination or microaggressions. Conceptually, it is plausible that poor sleep quality could influence cognitions (e.g., rumination) which may amplify perceptions of exclusion or discrimination. Therefore, future research should verify the project’s findings with longitudinal designs. Furthermore, the use of bootstrapping to create confidence intervals of indirect effects limits inferences of these parameters beyond the current sample. In addition, this project had low power to detect indirect effects, and future research should seek to collect data from a larger number of participants. Methodologically, we relied exclusively on self-report measures, and it is important to validate our findings with objective measures of sleep quality. Likewise, as the study did not recruit participants older than 65 years, the results may not generalize to older Black Americans or those recruited in other settings. In addition, given the relatively high levels of income and education in this sample, these results may not generalize to Black Americans with lower SES; as noted by Lewis and Van Dyke (2018), it is likely that the impact of discrimination and microaggressions on Black Americans’ sleep quality is greater for those of higher SES. Furthermore, since this sample was collected from online workers, the results may not apply to Black Americans who do not have reliable access to the internet. Finally, there may be residual confounding not accounted for by controlling for the identified covariates.

Conclusions

With predictions based on Myers’(2009) meta-model, this project found significant associations between two forms of racial stressors (i.e., microaggressions and experiences of discrimination) and rumination, and between rumination and poor sleep quality among Black Americans. Furthermore, results demonstrated that rumination mediated relationships between microaggressions and poor sleep. Additionally, three kinds of microaggressions were found to be significantly related to racism-related vigilance (Criminality, Low-Achieving/Undesirable, and Invisibility). These results amplify literature documenting the toll of racial stressors on Black Americans and suggest possible mechanisms that may lead to health damaging behaviors and poor health outcomes. This project indicates the need for interventions to prevent and ameliorate the biopsychosocial impact of racial stressors through individual, clinical, and community level approaches.