Authors: Corinna Y. Franco (Department of Psychology, University of California, Los Angeles, Los Angeles, California 90095, USA), Julieta Serobyan (Department of Psychology, University of California, Los Angeles, Los Angeles, California 90095, USA), Ovsanna Avetisyan (Department of Psychology, University of California, Los Angeles, Los Angeles, California 90095, USA), Barbara J. Knowlton (Department of Psychology, University of California, Los Angeles, Los Angeles, California 90095, USA)
Categories: Research Paper
Source: Learning & Memory
Early life racial/ethnic discrimination effects on behavioral control and health outcomes in young adults
Authors: Corinna Y. Franco, Julieta Serobyan, Ovsanna Avetisyan, Barbara J. Knowlton
Abstract
Early life trauma has been shown to facilitate habitual behavior, which may predispose individuals toward perpetuating maladaptive behaviors. However, previous investigations did not account for other traumatic childhood experiences like racial/ethnic discrimination exposure, nor have they examined the interaction of trauma and habits on real-world adverse outcomes. To examine these effects, we recruited 96 young adults (20.06 ± 1.89 years old) in a study probing early life racial/ethnic discrimination influences on habitual learning, and the conjunctive influences of early life discrimination and habit on disordered eating and substance use. To measure habit responses, participants completed a noise avoidance task during which they responded to abstract stimuli via associated keyboard presses to avoid an aversive screaming sound, after which they performed a devaluation test to measure avoidance habit responses. Participants then completed a series of questionnaires examining early life racial/ethnic discrimination exposure, disordered eating and substance use, and other psychological characteristics. Hierarchical regression results showed that certain early life discrimination subtypes, particularly threat/aggression experienced due to racial/ethnic background, significantly predicted habitual responding above and beyond the effects of psychological confounds. Additionally, overall early life discrimination exposure positively predicted binge eating, but no variables of interest predicted alcohol and drug use. These results expand on extant literature showing the negative impacts of childhood stressors on behavioral control and real-world outcomes.
Exposure to overt and subtle racial and ethnic discrimination, both in adulthood and early life, has been shown to increase risk of adverse health events, including substance misuse and disordered eating (Chae et al. 2002; Borrell et al. 2010; Yoo et al. 2010; Hunte and Barry 2012; Johnson et al. 2012; Hurd et al. 2014; Verissimo et al. 2014; Matsuzaka and Knapp 2019; Nelson et al. 2023). A potential reason this may occur is the role of racial/ethnic discrimination as a significant stressor among affected populations. Stress, whether acute or chronic, is shown to negatively impact decision-making processes, notably affecting behavioral control and promoting habitual responding (Schwabe and Wolf 2009; Soares et al. 2012; Patterson et al. 2013, 2019; Smeets et al. 2019; Gordon et al. 2020). This link between stress and altered behavioral control may be partly responsible in increasing susceptibility to intractable maladaptive behaviors like substance misuse and disordered eating among individuals experiencing adverse events, such as those exposed to discrimination.
Acute and chronic stress have wide-ranging impacts on neurocognitive functions. Studies have shown that stress exposure is associated with deficits in reward and feedback processing; instrumental and associative learning; and cognitive flexibility, among other alterations in decision-making processes (Bogdan and Pizzagalli 2006; Schwabe and Wolf 2012; Soares et al. 2012; Berghorst et al. 2013; Hanson et al. 2017; Raio et al. 2017; Harms et al. 2018; Wilkinson et al. 2021). Critically, researchers have also found that exposure to early life stress (childhood abuse and neglect) and other acute/chronic stressors leads to increased habitual responding, a type of behavioral control which relies on rigid stimulus-response (S-R) associations that are insensitive to changes in relational contingencies or outcomes (Dickinson 1985; Yin and Knowlton 2006; Schwabe and Wolf 2009, 2010; Soares et al. 2012; Patterson et al. 2013, 2019; Gordon et al. 2020; Zhou et al. 2020). This contrasts with goal-directed behavior, another type of behavioral control wherein an individual completes an (instrumental) action in the service of a particular goal (Balleine and Dickinson 1998; Yin and Knowlton 2006; Knowlton and Patterson 2018). Where habitual control is repetitive, automatic, and inflexible, goal-directed control is sensitive to contingency and outcome changes, adaptive, and reliant on complex environmental models informed by action-outcome (A-O) or response-outcome (R-O) associations (Balleine and Dickinson 1998). It has been posited that neurophysiological changes that occur when exposed to acute and chronic stressors promote a reliance on habit over goal-directed control. Such changes include dysregulation of the hypothalamic-pituitary-adrenal (HPA) axis, which regulates bodily homeostatic and stress responses, and the sympathetic-adrenal-medullary (SAM) axis, which modulates cardiovascular and immune system function (Stephens and Wand 2012; Berger and Sarnyai 2014). Moreover, acute and chronic stress exposure may be accompanied by frontal region glucocorticoid release, which alters higher-level cognitive processes (e.g., environmental modeling), and alterations in brain regions subserving goal-directed and habitual control, primarily deficits in caudate structure-function (goal-subserving region) and hypertrophy and increased activation in the putamen (habit-subserving region) (Arnsten 2009; Soares et al. 2012; Raio et al. 2017).
Numerous studies have supported the view of racial/ethnic discrimination as a significant psychosocial stressor. For instance, like early life stress and other types of chronic stress, racial/ethnic discrimination exposure has been shown to affect (typically elevate) cortisol levels (often time-dependent) (Zeiders et al. 2012; Korous et al. 2017; Peterson et al. 2020), flatten diurnal cortisol slopes (a measure indicating rate of cortisol decline throughout the day; Zeiders et al. 2014); and dysregulate HPA (and potentially SAM) axis function (Berger and Sarnyai 2014). Discrimination and chronic stress have been similarly linked to worse health outcomes (Thoits 2010; Williams and Mohammed 2013), including increased risk of cardiovascular disease (Kivimäki and Steptoe 2018; Panza et al. 2019), diabetes (Whitaker et al. 2017; Madhu et al. 2019; Gaston et al. 2021), adverse mental health effects (Brown et al. 2000; Anda et al. 2002; Schulz et al. 2006; Hwang and Goto 2008; Walker et al. 2017; Lee et al. 2020), disordered eating behaviors (Kelly et al. 2018; Beccia et al. 2020), and substance use (Borrell et al. 2007; Yoo et al. 2010; Gerrard et al. 2012; Verissimo et al. 2014). Researchers have also found alterations in cognitive function related to discrimination exposure. In particular, studies have shown deficits in executive function, including inhibiting, shifting, or updating mental representations, among those experiencing subtle racial/ethnic discrimination (Ozier et al. 2019); negative relationships between cognitive flexibility and working memory with recent racial discrimination exposure (Keating et al. 2022); and deficits in global cognitive function, episodic memory, and perceptual speed among older African American individuals reporting greater perceived racial discrimination (Barnes et al. 2012).
Given this evidence, it is likely that discrimination exposure experienced during key developmental periods (early life racial/ethnic discrimination) would affect other critical cognitive and decision-making behaviors into adulthood, as is the case for early life stress and other forms of chronic stress. Specifically, we may see a preponderance of habitual behavior among individuals reporting higher early life discrimination exposure. This increased habit responding, alongside other neurophysiological effects of early life discrimination exposure, may contribute to increased substance use and maladaptive health behavioral effects observed in discriminated populations. However, little work has been completed to investigate these complex discrimination-habit-health relationships.
Using an avoidance instrumental devaluation paradigm previously published by Patterson et al. (2019), we aimed to address this gap in the literature by (i) studying the effect of early life discrimination exposure on psychological measures and habitual responding, (ii) examining the conjunctive effects of habitual responding on early life discrimination exposure-health outcome (e.g., substance use and disordered eating behavior) associations, (iii) examining whether these effects are due to experiences of early life discrimination controlling for other variables, and (iv) whether the effects of early life discrimination specifically affect habitual responding and not other performance measures on habit responding, not learning or other responding behavior. Based on extant literature, we predicted that individuals reporting greater early life discrimination exposure would report higher depression, state/trait anxiety, and current stress scores compared to individuals with lower or no early life discrimination exposure. Higher early life discrimination exposed individuals would also exhibit greater habitual responding alongside no differences in instrumental learning or valued stimulus responding in an aversive instrumental devaluation paradigm, controlling for effects of depression, anxiety, and stress. Moreover, in an attempt to synergize stress-coping and habit models of addiction, we predicted that habitual responding would moderate the effect of early life discrimination exposure on substance use and disordered eating behaviors.
Results
We first characterized learning accuracy, habit responding, and valued stimulus responding to ascertain the validity of our avoidance instrumental devaluation task. Overall instrumental learning accuracy was high at 81.51% ± 12.71%, showing adequate learning of the stimulus-response relationships. On average, participants made 1.38 ± 2.41 habit responses during the test phase, ranging from 0 to 9, with 32.29% of participants making 1 or more habit responses on this task, illustrating a breadth of habitual control. Participants also exhibited an average of 8.05 ± 1.83 valued stimulus responses, indicating appropriate continued responding to the valued stimulus at test. These results are in line with previous work using the same paradigm (Patterson et al. 2019).
We also wanted to examine the effects of early life discrimination exposure on psychological variables, namely depression, state and trait anxiety, and current stress. As such, we conducted four multiple linear regression models using overall early life discrimination exposure as a predictor; overall early life stress exposure, age, and gender as covariates; and the aforementioned psychological variables as outcomes. The model predicting depression scores showed significant positive effects of overall early life discrimination exposure (b = 3.733, p = 0.047) and overall early life stress exposure (b = 0.221, p < 0.001), indicating that holding all other variables at their means, each unit increase in early life discrimination and early life stress exposure predicted 3.733 and 0.221 units greater depressive symptomology, respectively (see Table 1). No other terms were significant (ps > 0.070). The model predicting trait anxiety showed significant positive effects of overall early life stress exposure (b = 0.254, p < 0.001) and gender (b = 6.479, p = 0.022), indicating that holding all other variables at their means, individuals who reported greater early life stress exposure and who were women showed greater trait anxiety symptomology. Early life discrimination exposure and age were not significant predictors (ps > 0.345). Similarly, overall early life stress exposure (b = 0.095, p = 0.043) and gender (b = 4.998, p = 0.007) significantly positively predicted current stress, but early life discrimination exposure and age did not (ps > 0.194). No variables significantly predicted state anxiety (ps > 0.098). An additional Welch two-sample t-test was conducted to examine gender effects on early life discrimination, with results showing no significant differences in discrimination exposure between men (1.60 ± 0.85) and women [1.42 ± 0.43; t(21.65) = −0.915, p = 0.370]. Because we see evidence of early life discrimination exposure negatively impacting mental health outcomes, we used these psychological variables as covariates in the following analyses.
To determine if early life discrimination exposure significantly affected habitual responding, two hierarchical multiple regression analyses were conducted with two blocks of variables using habitual responding as the outcome. For both analyses, the first block included the following covariates overall early life stress exposure, depression, state and trait anxiety, perceived stress, age, and gender. For one analysis, overall early life discrimination exposure was included as a predictor in the second block, while for the other analysis early life discrimination subscales were included as predictors in the second block.
For the first block of the overall early life discrimination exposure analysis, only age had a significant positive effect on habitual responding (b = 0.278, p = 0.048). The addition of the overall early life discrimination variable in the second block, however, was not significant [ΔF (1, 87) = 0.242, p = 0.624, ΔR^2^ = 0.501], with the AIC selection method indicating that the simpler model excluding the early life discrimination variable best fitted the data (AIC block 452.572, AIC block 454.305).
For the early life discrimination subscale model, the same covariates as the overall model were included in the first block, while the second block included all five early life discrimination subscales. No terms were significant in the first block. The addition of the early life discrimination subscales, while not significant [ΔF (5, 83) = 2.313, p = 0.051, ΔR^2^ = 0.115], did result in a lower AIC (AIC block 452.572, AIC block 450.049). In the second block, threat/aggression due to race/ethnicity had a significant positive effect on habitual responding (b = 3.082, p = 0.010). Unfair police treatment had a significant negative effect (b = −1.319, p = 0.022), but due to low reporting of experiences in this variable, this result should be interpreted with caution. Age came out as a significant positive predictor (b = 0.343, p = 0.031; see Table 2 and Fig. 1A–C). These results indicate that, among those averaged on all other variables, exposure to threatening or aggressive behavior due to race/ethnicity in early life, alongside experiences with police maltreatment, altered habitual responding in young adults, predicting either 3.082 more or −1.319 fewer habitual responses for each unit increase in these measures, respectively. Moreover, despite the relatively small age range of the sample, increasing age appeared to enhance habitual responding.
To ensure effects of early life discrimination exposure on habit responding were due to influence on habit behavior rather than alterations in goal-directed responding or overall task performance, four additional multiple linear regression models were conducted to determine if overall or subscale early life discrimination exposure also affected responding to valued stimuli at test and overall instrumental learning accuracy. These models included the same covariates as the early life discrimination-habit models in block 1, and the addition of either overall or subscale early life discrimination exposure in block 2. Concerning valued stimulus responding, results showed no significant effects of the included covariates in block 1 or 2, with the addition of the early life discrimination variables indicating no significant change [overall ΔF (1, 87) = 0.601, p = 0.440, ΔR^2^ = 0.006; subscale ΔF (5, 83) = 1.91, p = 0.101, ΔR^2^ = 0.096]. AIC model selection favored the simpler model in the overall analysis (AIC overall block 398.835, AIC overall block 400.175) and showed no difference in the subscale analysis (AIC subscale block 398.835, AIC subscale block 398.38).
For instrumental learning task accuracy, the overall analysis revealed only significant effects of state anxiety (block b = −0.280, p = 0.028; block b = −0.280, p = 0.029) and gender (block b = 8.505, p = 0.008; block b = 8.363, p = 0.011), with no significant change after the addition of the overall early life discrimination variable [ΔF (1, 87) = 0.093, p = 0.762, ΔR^2^ = 0.001], and the simpler model having the best fit (AIC block 756.861, AIC block 758.759). The subscale analysis revealed a similar pattern of results, showing significant effects of state anxiety (block b = −0.280, p = 0.028; block b = −0.289, p = 0.030) and gender (block b = 8.505, p = 0.008; block b = 8.618, p = 0.011). The addition of the early life discrimination subscale variables indicated no significant change [ΔF (5, 83) = 0.421, p = 0.833, ΔR^2^ = 0.020], with the simpler model showing best fit (AIC block 756.861, AIC block 764.455). This indicates that holding all other variables at their means, women and individuals lower on state anxiety exhibited greater learning phase accuracy, with early life discrimination providing no significant contribution to the model.
To determine if early life discrimination exposure and habitual responding had conjunctive effects on substance use outcomes, two hierarchical multiple regression models were conducted with the first block composed of the aforementioned covariates; the second block including overall early life discrimination exposure, habitual responding, and their interaction as predictors; and either frequency of alcohol or drug use as the outcome. Results revealed only a significant effect of gender on alcohol use (block b = −2.788, p = 0.026; block b = −2.711, p = 0.036), indicating that men had higher rates of alcohol use than women. The addition of the early life discrimination, habit, and interaction terms showed no significant change [ΔF (3, 85) = 0.144, p = 0.934, ΔR^2^ = 0.005], with the simpler model showing better fit (AIC block 576.001, AIC block 581.516). Results additionally revealed a positive effect of early life stress exposure on drug use (block b = 0.049, p = 0.039; block b = 0.056, p = 0.024), indicating that holding all other variables at their means, participants reporting higher early life stress exposure exhibited greater drug use. There was no significant change when including the early life discrimination, habit, and interaction terms [ΔF (3, 85) = 0.519, p = 0.671, ΔR^2^ = 0.015], with the simpler model showing better fit (AIC block 509.616, AIC block 513.874).
Additionally, we wanted to examine whether early life discrimination exposure and habitual responding affected disordered eating patterns. As such, we conducted a hierarchical multiple regression analysis with the covariates added in block 1; overall early life discrimination exposure, habitual responding, and their interaction in block 2; and overall EDEQ scores as the outcome. Results revealed only significant positive effects of depression (block b = 0.047, p = 0.019; block b = 0.041, p = 0.043) and trait anxiety [block 2 (only): b = 0.040, p = 0.045] on overall EDEQ scores, indicating that holding all other variables at their means, individuals reporting greater depressive symptomology and trait anxiety exhibited greater disordered eating patterns. The inclusion of the early life discrimination, habit, and interaction terms showed no significant change [ΔF (3, 85) = 0.801, p = 0.497, ΔR^2^ = 0.020] and AIC favored the simpler model (AIC block 292.681, AIC block 296.004).
In an effort to determine early life discrimination or habit influences on precise aspects of disordered eating, additional hierarchical multiple regressions were conducted using the same predictors and covariates as the above model, but the EDEQ subscale scores (eating restraint, eating concern, weight concern, shape concern, and binge eating behavior) as the outcomes. Results revealed no significant changes in the eating restraint model (p = 0.740; AIC block 348.693, AIC block 353.284), the shape concern model (p = 0.496; AIC block 324.149, AIC block 327.471), or the weight concern model (p = 0.172; AIC block 323.814, AIC block 324.199), with the simpler models showing better fit. However, results revealed a significant negative effect of early life stress exposure (b = −0.013, p = 0.025) and significant positive effects of depression (b = 0.048, p = 0.002) and trait anxiety (b = 0.029, p = 0.049) on eating concern.
Concerning the binge eating model, we observed a significant change after including the early life discrimination, habit, and interaction terms [ΔF (3, 85) = 5.301, p = 0.002, ΔR^2^ = 0.114], with early life discrimination exposure showing a significant effect when included in the second block (b = 0.914, p < 0.001) and AIC selection showing the second model had better fit (AIC block 318.283, AIC block 307.812; see Fig. 2). However, neither habitual responding (p = 0.242) nor the interaction between early life discrimination exposure and habitual responding were significant (p = 0.073). This indicates that early life discrimination exposure, but not a propensity for responding habitually on this task, explains additional variance above and beyond the simple covariate mode l. Interestingly, results showed significant negative effects of early life stress exposure (block b = −0.019, p = 0.027; block b = −0.023, p = 0.006) and age [block 2 (only): b = −0.202, p = 0.003], a significant positive effect of depression [block 1 (only): b = 0.054, p = 0.018], and a significant effect of gender (block b = −0.812, p = 0.013; block b = −0.712, p = 0.022) on binge eating behavior (see Table 3). This indicates that individuals with lower early life stress exposure, higher depressive symptomology, younger individuals, and men also appear to report more binge eating when holding all other variables at their means.
While counter to hypothesized results, the negative relationships between early life stress and eating concern and binge eating may have resulted from the fact that we controlled for psychological factors such as anxiety and depression and used an overall measure of early life stress. Previous work has shown that the relationship between early life stress and disordered eating is dependent on specific subtypes of stress and is mediated by psychological variables like depression (Guillaume et al. 2016; Tabone et al. 2022). Our controlling for various types of psychological variables, including depression and anxiety, and inclusion of an overall early life stress measure in our analyses may have resulted in this negative association.
To determine if these patterns of results were primarily driven by underrepresented racial/ethnic groups reporting high discrimination scores, we examined descriptive data for the discriminatory subtypes across different racial/ethnic groups and conducted additional exploratory analyses restricted to Asian and White participants (see Supplemental Analyses and Tables S1, S2, S3, and S4). Descriptive analyses showed that most groups, including Asian participants, reported a wide range of different types of early life discriminatory experiences. Restricted exploratory analyses also revealed similar trends as in the larger sample, with threat/aggression exposure exhibiting a positive effect trajectory on habit responding (b = 2.023, p = 0.146) and early life discrimination exposure (b = 0.697, p = 0.039), early life stress exposure (b = −0.023, p = 0.013), and age (b = −0.236, p = 0.035) exhibiting significant effects on binge eating behavior. This suggests that while some groups report more discriminatory experiences and greatly contribute to discrimination-habit relationships, the effects of early life discrimination exposure can be generalized across different racial/ethnic groups.
Discussion
The present study aimed to examine the effects of early life racial/ethnic discrimination exposure on psychological outcomes and habitual responding, determine if such effects extended to instrumental learning or valued stimulus responding, and examine potential moderating effects of habitual responding on discrimination-health outcome relationships. We extended previous findings illustrating negative effects of perceived discrimination exposure on mental health outcomes, showing that retrospective reports of early life racial/ethnic discrimination related to depressive outcomes in young adulthood. In line with predictions, we found that early life discrimination exposure, specifically threat/aggression experienced due to race or ethnicity in early life, predicted habitual responding, controlling for related variables. This effect was limited to habitual responding, as early life discrimination exposure did not predict valued stimulus responding at test nor instrumental learning accuracy. Additionally, we found that while early life discrimination exposure and habitual responding did not predict substance use outcomes, early life discrimination exposure did positively predict binge eating behavior. Contrary to initial predictions, habitual responding did not significantly moderate early life discrimination-health outcome relationships.
The current findings partially coincide with a critical literature showing negative mental health effects related to racial/ethnic discrimination, in particular that experienced during childhood and adolescence. We supported previous findings showing greater risk of depression for individuals reporting higher racial/ethnic discrimination (Brown et al. 2000; Schulz et al. 2006; Hwang and Goto 2008), further showcasing that mental health effects of discrimination experienced in early life extend into adulthood. However, early life discrimination exposure did not predict current stress or state/trait anxiety symptomology, as has been reported in other studies (Hwang and Goto 2008; Hurd et al. 2014; Assari et al. 2017; Zong et al. 2022). This may be due to the nature of our early life discrimination measurement which, as a retrospective report, may not significantly influence present experiences of stress or anxiety in our sample. It may also be the case that effects of early life discrimination on anxiety or stress disorders may only extend to clinical samples, which were not measured here. Nevertheless, these findings illustrate the necessity of further probing consequential influences of racial/ethnic discrimination, including those experienced earlier in life, on psychological outcomes.
Critically, we observed significant positive effects of distinct dimensions of early life discrimination exposure on habitual responding in the current habit task. Specifically, threatening or aggressive experiences against one's person or property due to racial/ethnic discrimination during key developmental periods predicted more habit responding, controlling for relevant factors. These results coincide with and extend emerging literatures showing propensity toward avoidance and appetitive habit responding among individuals reporting higher early life stress exposure (Patterson et al. 2019; Gordon et al. 2020) and negative impacts on cognition among adults reporting perceived discrimination exposure (Barnes et al. 2012; Ozier et al. 2019; Keating et al. 2022). Early life discrimination exposure within this study has been conceptualized as a type of early life psychosocial stress. While early life stress effects on cognition and health outcomes are well-documented, the definition of an early life stressor within the literature typically includes early life experiences with domestic abuse and neglect only. Thus, we can extend findings concerning negative early life stress effects on behavioral control and health to other forms of early life stressors, including stressful experiences with racial/ethnic discrimination, which have thus far been neglected. What effects the current literature has found concerning racial/ethnic discrimination exposure on cognition and health outcomes have also focused on either adolescent or adult populations separately. Here we document evidence of lasting cognitive effects of early life racial/ethnic discrimination into adulthood, controlling for variables known to correlate with discrimination exposure and affect behavioral control. Importantly, we show that early life discrimination effects are limited to influences on habitual responding, given that early life discrimination exposure did not significantly predict responding to the still-valued stimulus at test or instrumental learning accuracy.
The results indicated that experiencing different subtypes of early life discrimination had different effects on habitual responding. The threat/aggression early life discrimination subtype was a positive predictor of habit responding, such that each unit increase in threat/aggression exposure predicted a substantial 3.082 more habit responses. It is possible that individuals commonly exposed to this early life discrimination subtype developed or enhanced a sensitivity to habitual behaviors expressing an avoidance phenotype, either due to the pronounced stress associated with this early life discrimination subtype or the benefit of employing habitual strategies to avoid profoundly negative (in this case, potentially dangerous) consequences. Future research should examine whether this and other early life discrimination subtypes also predict appetitive habitual responding (e.g., responding habitually to previously rewarding associations) and the extent to which threat/aggression exposure promotes avoidant habit responding.
Unfair police treatment exposure, however, was a negative predictor of avoidance habit responding, with each unit increase in unfair police treatment due to race/ethnicity in early life predicting 1.319 fewer habit responses on the current task. It remains unclear why this association was found, but we posit that the relatively low unfair police treatment exposure in this sample affected statistical analysis. Moreover, this discrimination subtype was composed of a singular question item that overlooks the complexity of police maltreatment in the real world. Considering that police maltreatment is a critical source of stress for vulnerable children and adults, researchers should interpret this result with caution and probe the cognitive and health consequences of such exposures more extensively.
Interestingly, only early life discrimination exposure positively predicted binge eating behavior in this sample. Some evidence has pointed to discrimination effects on disordered eating behaviors, including binge eating (Harrington et al. 2006; Assari 2018; Nelson et al. 2023). However, unlike our current findings, deficits in flexible response adaptation alongside greater self-reported and laboratory-derived habit-like behaviors have also been shown among individuals with disordered eating behaviors, coinciding with a habit theory of addiction posited for disorders marked by inflexible maladaptive patterns of behavior like disordered eating and substance misuse (Walsh 2013; Voon et al. 2014; Coniglio et al. 2017; Heriseanu et al. 2020). While these findings provide further support for the link between early life discrimination and disordered eating, we did not find a relationship between habitual responding and health behaviors that we have seen in our previous work (Franco and Knowlton 2023). This could be due to a variety of reasons, most notably the type of habit variable used here, which measured avoidance habits as opposed to appetitive habits which are more likely to predict behaviors associated with a positive outcome like binge eating patterns of behavior. Further investigation of this relationship is warranted.
Moreover, contrary to predictions, neither early life discrimination exposure nor habitual responding significantly predicted substance use and disordered eating behaviors other than binge eating, nor did habitual responding significantly moderate early life discrimination-health outcome relationships. Future work would benefit from greater sample size and diversity in early life discrimination exposure and health outcomes, the employment of alternate laboratory habit measures, and examination of other maladaptive health behaviors as we attempt to inform discrimination-habit-health models.
Notably, the racial/ethnic background characteristics of our participants were limited, consisting mostly of Asian and White participants. Women were also overrepresented. These factors may have impacted the variance of our measure of racial/ethnic discrimination experiences, given that individuals from backgrounds typically associated with higher discriminatory experiences were underrepresented. The relatively small number of individuals from underrepresented groups reporting more discriminatory experiences had a substantial influence on the effects we observed in this study. However, even within our sample, ranges of experiences with discrimination were reported within each racial/ethnic group. The categories used to measure race and ethnic identity in this study are also broad and included different subgroups that likely experienced differences in discrimination in their social and familial environments. Nevertheless, despite the modest levels of discrimination experienced by the sample at large, there were effects of discrimination experience on habitual behavior and binge eating that were apparent even after controlling for several covariates. Future work with more ethnically diverse samples is important, and we hypothesize that this would lead to stronger effects.
In conclusion, the objective of this study was to determine the effects of early life racial and/or ethnic discrimination on habitual responding as well as other psychological outcomes. Our initial prediction was that individuals who reported higher early life discrimination exposure would report higher levels of depression, anxiety, and current stress and exhibit higher levels of habitual responding. Our findings indicate that, while early life discrimination did not significantly predict psychological measures such as current perceptions of stress, state or trait anxiety, or most disordered eating behaviors, early life discrimination significantly predicted depression and habitual responding behavior. Of the criteria used to describe early life discrimination, threat/aggression due to race/ethnicity significantly predicted habitual responding. Our results also indicated that early life discrimination did not have a significant relationship with substance use but did predict binge eating behaviors.
The results of this study have several significant implications, such as the potential to increase our understanding of the relationships between discrimination, habitual behavioral control, and physical and mental well-being. Studying the potential neurobiological mechanisms that relate discrimination to habitual responding can help progress neurocognitive research and may be important for improving our knowledge of the effects of discrimination and stress on brain systems involved in higher decision-making. Such findings can also be useful in identifying novel interventions and optimizing policies that will help decrease negative health outcomes linked to discrimination, such as disordered eating or substance misuse.
Materials and Methods
Participants
In total, 105 participants (20.09 ± 1.86 years old, range = 18–33) were recruited from the University of California, Los Angeles (UCLA) undergraduate student community via flyers and online postings. Nine participants were excluded due to experimenter error (n = 3) and poor performance (responding to valued stimulus at test < 3; n = 6). The final sample consisted of 96 participants (20.06 ± 1.89 years old, range = 18–33). Notably, while efforts were made to advertise the study to cultural resource centers and areas associated with minoritized students, this sample was predominantly composed of Asian (n = 56) and White (n = 26) participants, somewhat consistent with UCLA population characteristics where Black and Indigenous students are severely underrepresented (see Table 4 for full demographic breakdowns). Participants provided written informed consent and were compensated $20 per hour of participation. The UCLA Institutional Review Board granted approval for all study procedures.
Design and procedure
Participants were taken into a private testing room and completed the behavioral task and subsequent questionnaires on an Apple Mac desktop computer. The behavioral task was created using PsychoPy while questionnaire responses were collected via Qualtrics. Testing sessions lasted approximately 1 h, after which participants received compensation for their participation.
To measure habitual responding, participants completed an aversive habit task previously conducted by Patterson et al. (2019), which is composed of a practice phase, an instrumental avoidance learning phase, and a test phase in extinction following instructed devaluation (see Fig. 3). During the learning phase, participants were given earphones and presented with 120 trials composed of three abstract stimuli (40 trials/stimulus), two of which were active while one functioned as a safe stimulus. The active stimuli were associated with either a left or right arrow key response, which was counterbalanced across all participants. While participants were not directly instructed to use certain hands or fingers to complete these keyboard responses, most participants used their right middle and index fingers when answering. They were instructed to make the correct key presses to the active stimuli; failure to do so resulted in the presentation of an aversive screaming sound (average of 75 dB) delivered to either the left or right earphone, in accordance with the direction of the correct answer key (e.g., if they failed to make a response to the right arrow key associated image, they would be presented with a screaming sound delivered to the right ear). Pressing the correct key while the stimulus was present on-screen prevented the presentation of the screaming stimulus. As the safe stimulus was not associated with any key press, no sound was delivered for any responses. Participants completed a practice phase composed of six trials (three trials/stimulus) using the same parameters to ensure their understanding of the task. Participants were allowed to redo the practice session, but few ultimately did so. Stimuli were presented for 500 msec, during which participants had to make their response, and followed by a 2000 msec ITI. The screaming stimulus was presented during the ITI for 1000 msec following response failure.
Of note, stimuli were presented rapidly due to extant literature showing that time pressures aid in the behavioral observation of habitual behavioral control. Various experimental studies have indicated that, when pressured to respond within a short time window, participants tend to perform more habitual responses, either due to an unmasking of a learned habit or speed/accuracy tradeoffs occurring between the “slow” but accurate goal-directed system and “fast” but potentially inaccurate habit system after initial learning (Betsch et al. 2004; Keramati et al. 2011; Luque et al. 2020).
After the learning phase, participants were told that one of the active stimuli was now devalued, meaning no screaming response would follow incorrect responding. The corresponding earbud was also removed to reinforce instructed devaluation. Devaluation for the left- and right-associated stimuli was counterbalanced across participants. Participants then completed the test phase, during which they were instructed to continue responding to the active, valued stimulus, but they no longer had to respond to the devalued stimulus, nor would any responses to this stimulus result in aversive consequences. Unbeknownst to participants, the test phase was performed in extinction, meaning that the screaming sound was no longer administered to any of the stimuli following incorrect responses to prevent extraneous learning effects. The test phase consisted of 30 trials (10 trials/stimulus) of the same abstract stimuli presented during the learning phase, using the same parameters. Participant responses to the devalued stimuli were taken to indicate habitual responding, recorded as the number of “correct” devalued stimulus responses out of 10. Because this paradigm promoted learned avoidance associated using an aversive S-R contingency, habitual responses were considered avoidant rather than appetitive (i.e., reliant positive outcome contingencies during initial S-R learning).
Following the behavioral task, participants also completed a series of questionnaires. To measure early life discrimination exposure, participants completed a modified version of the Brief Perceived Ethnic Discrimination Questionnaire Community Version (PEDQ–CV) (Brondolo et al. 2005), a 17-item questionnaire which assessed exposure to race and/or ethnicity-based discrimination within the first 16 years of life. Responses were recorded on a scale of 1–5, 1 indicating “Never happened,” and 5 indicating “Happened very often.” Early life discrimination exposure was assessed along five exclusion/rejection (degree of isolation or exclusion), stigmatization/devaluation (degree of demeaning or stigmatizing experiences), discrimination at work/school (degree of unfair treatment in school or work environments), threat/aggression (degree of threat and/or harm to their person or property), and unfair police treatment due to racial/ethnic background. Overall, early life discrimination exposure was calculated as the mean of all subscales. Participants were explicitly instructed to answer these questions in relation to their early life discriminatory experiences using similar language as other measures of early life adversity. Given that this sample was composed of young adults who very recently experienced these early life discriminatory exposures, we did not expect this change to affect psychometric properties of the questionnaire. Previous studies have supported the validity of this questionnaire in measuring racial/ethnic discrimination in Black, Latino/a, multiracial, and, importantly given our sample demographics, Asian populations (Brondolo et al. 2005; Kwok et al. 2011; Keum et al. 2018). To control for effects of early life stress (childhood trauma and abuse) exposure, depression, state and trait anxiety, and current stress, participants also completed a Childhood Trauma Questionnaire Short Form (CTQ-SF; Bernstein et al. 2003), the Beck Depression Inventory (BDI; Beck et al. 1996), the State and Trait Anxiety Forms Y1 and Y2 (STAI Forms Y1 and Y2) (Spielberger 1989), and the Perceived Stress Scale (PSS) (Cohen and Williamson 1988), respectively. These forms were used to calculate continuous measures of the aforementioned variables. Additionally, participants completed a health behaviors questionnaire to assess overall frequency of current and/or risky alcohol and drug use (e.g., how often do you have a drink containing alcohol?) similar to the approach used by Schilling et al. (2007). Finally, participants completed the Eating Disorder Examination Questionnaire (EDE–Q) (Fairburn and Beglin 1994, 2008), a 28-item questionnaire assessing the presence of multiple factors relating to disordered eating behaviors experienced within the past 28 days. From this questionnaire, overall EDEQ scores were calculated as the sum of reported disordered eating frequency, alongside five subscale eating restraint (frequency of food intake limitation), eating concern (frequency of preoccupation with food/eating and negative thoughts surrounding eating), weight concern (frequency of preoccupation and/or dissatisfaction with weight), shape concern (frequency of preoccupation and/or dissatisfaction with body shape), and binge eating behavior (confirmation and frequency of binge eating episodes).
Data analysis
Data were analyzed using R (R Core Team 2020). All continuous variables were mean centered, with gender dummy coded in all relevant analyses (0 = male, 1 = female). A significance level of 0.05 was utilized for all analyses. Many analyses used a hierarchical multiple regression approach where model fit was assessed through the Akaike information criterion (AIC), such that lower AIC values indicate greater model fit while taking into account parsimony.
Early life discrimination was measured using a modified questionnaire that calculated an overall discrimination exposure measure based on the mean of five exclusion/rejection (degree of isolation or exclusion), stigmatization/devaluation (degree of demeaning or stigmatizing experiences), discrimination at work/school (degree of unfair treatment in school or work environments), threat/aggression (degree of threat and/or harm to their person or property), and unfair police treatment due to racial/ethnic background experienced under 16 years of age. Other psychological variables, including depression, anxiety, current stress, and early life stress exposure, were also obtained (see Materials and Methods for more details).
Behavioral measures were acquired using an avoidance instrumental devaluation paradigm composed of an instrumental learning phase, where participants responded to two “active” cues to avoid an aversive scream outcome, accompanied by a test phase where participants had to continue responding to one “still valued” cue while the other was devalued via the removal of the scream outcome (see Fig. 3 for a diagram of the paradigm). Habitual responding was calculated as the number of responses made to devalued stimuli during the test phase, on a scale from 0 (no habitual responding) to 10 (out of 10 possible habitual responses). Responding to the still valued stimuli at test was also calculated, similarly along a scale of 0–10 possible valued stimuli responses. Additionally, the overall percent accuracy during the instrumental learning phase was calculated for the active stimuli only.
Results are visualized using partial regression plots, which illustrate the relationship between a predictor and outcome while controlling for the other predictors and covariates included in the regression analyses. Axes on these graphs will be labeled using a “Variable|Others” schema to indicate this.