About this paper symposium

---

Panel information
Panel 12. Methods, History, Theory

Paper #1
Adversity exposure and executive functioning: Decomposing the cognitive processes underlying performance differences
Author information	Role
Mr. Stefan Vermeent, Utrecht University, Netherlands	Presenting author
Anna-Lena Schubert, Johannes Gutenberg University, Mainz, Germany	Non-presenting author
Willem E. Frankenhuis, University of Amsterdam, Netherlands	Non-presenting author
Abstract
People who grew up in adverse conditions tend to show lower performance on executive functioning tasks. For instance, exposure to deprivation has been linked to slower response times on inhibition tasks. In contrast, recent studies suggest that people with more exposure to childhood adversity are faster at switching their attention between goals on attention shifting tasks. Both patterns of performance are typically interpreted as differences in EF ability. However, recent research from mathematical and cognitive psychology demonstrates that equating raw performance with EF ability is problematic: First, people may be slower because they are more cautious, not because their ability is lower. Second, raw performance across EF tasks is influenced by general processes, such as basic speed of processing. These psychometric issues pose an important challenge to adversity research: by only comparing people on their raw performance, it is unclear whether adversity affects specific EF abilities, general processing speed, or shapes the use of different strategies. In this preregistered study, we used cognitive modelling to move beyond raw performance to a better understanding of the underlying cognitive processes. Specifically, we investigated how exposure to threat and material deprivation in childhood were associated with EF performance in adulthood. We expected that people with more exposure to childhood adversity would (1) show lower general processing speed; (2) show lower inhibition ability but enhanced attention shifting ability, after accounting for general processing speed, and (3) show increased response caution. Our sample consisted of 1,055 Dutch adult participants drawn from the Dutch Longitudinal Internet Studies for the Social Sciences (LISS) panel. Participants first completed two inhibition tasks, three attention shifting tasks, and a basic processing speed task. Next, they reported on childhood and recent exposure to neighbourhood threat and material deprivation. We used Drift Diffusion Modeling to decompose raw performance into three cognitive processes: rate of evidence accumulation (information processing), response caution, and non-decision time (stimulus encoding and response execution). Next, we used structural equation modelling to estimate task-general and EF-specific latent factors for each cognitive process (Paper 1 - Figure 1). Task-general evidence accumulation was interpreted as general processing speed, and EF-specific evidence accumulation was interpreted as specific EF abilities. We then estimated associations with childhood adversity, controlling for biological sex. Our data only provided support for a task-general factor for each cognitive process. After accounting for task-general variance, no systematic variance related to specific EF abilities remained. People with more exposure to both childhood threat and deprivation showed lower general processing speed (β = -0.19, p < .001, and β = -0.12, p = .008, respectively, Paper 1 - Figure 2). In addition, exposure to childhood deprivation (but not threat) was associated with longer non-decision time (β = 0.13, p = .004, Paper 1 - Figure 2). Effects of childhood adversity were stronger and more consistent than effects of recent adversity, which is striking given that the average age of our sample was 39. These findings suggest that adversity researchers should be careful when interpreting raw performance differences on individual EF tasks.

Paper #2
Childhood Material Hardship and Attentional Problems in Adolescents: Application of Drift-Diffusion Modeling
Author information	Role
Yue Zhang, University of Michigan, United States	Presenting author
Alexander S. Weigard, University of Michigan, USA	Non-presenting author
Felicia A. Hardi, Yale University, USA	Non-presenting author
Sunghyun H. Hong, University of Michigan, USA	Non-presenting author
Edward Huntley, University of Michigan, USA	Non-presenting author
Colter Mitchell, University of Michigan, USA	Non-presenting author
Luke W. Hyde, University of Michigan, USA	Non-presenting author
Christopher S. Monk, University of Michigan, USA	Non-presenting author
Abstract
Introduction: Material hardship, which more commonly affects marginalized communities, is a risk factor for developing attentional problems in adolescence. However, the underlying cognitive mechanisms are unclear, partially due to poor reliability and lack of specificity of conventional task-based behavioral cognitive measures (e.g., reaction time). To address this limitation, computational models of cognition, such as drift-diffusion modeling, could be applied to better characterize latent processes underlying observed behavior. In particular, individuals’ efficiency of evidence accumulation (EEA), or information processing speed, for goal-relevant information during decision-making, has been suggested as a key mechanism associated with cognitive functioning and risk for psychopathology. The current project examines whether EEA 1) is impacted by early life exposure to material hardship; 2) reflects cognitive changes underlying attentional problems in a group of underrepresented adolescents; and 3) reflects associations above and beyond conventional performance measures. Methods: 187 adolescents (ages 15-17) from the Future of Families and Child Wellbeing Study (FFCWS), a population-based longitudinal cohort study, with substantial representation of marginalized youths (83% Black and Hispanic, 50% baseline household income < $15,000), were included in the analyses. Participants completed an MRI emotional-faces, gender-identification task. Reaction times (RT) and responses were recorded and fitted with a drift-diffusion model. Model parameters, including EEA, were estimated using the Dynamic Models of Choice package in R. Household material hardship (i.e., ability to pay for housing, utilities, food, and healthcare) was measured using a cumulative sum score of endorsed items from ages 1, 3, 5, and 9 from FFCWS. Adolescent attentional problems were measured using the Child Behavioral Checklist Attentional Problems Syndrome subscale at age 15. Multiple regressions were used to examine how material deprivation was associated with EEA, and whether EEA, in turn, was associated with attentional problems. Growth curve modeling was used to examine the influence of developmental trajectories of material hardship. Results: EEA was significantly associated with cumulative material hardship experience (r = -0.16, p =.027). A growth curve analysis indicated that the intercept, not slope, of material hardship was significantly associated with EEA (b = -0.17, p = 0.026, Paper 2 - Figure 1). EEA was also significantly associated with attentional problems in adolescents (r = -0.18, p = 0.014). Notably, these results were found with EEA but not conventional measures of cognitive performance using RT and accuracy (ps > .500). These results remained significantly after adjusting for demographic variables. Conclusion: We found that exposure to material hardship across childhood correlates with lower EEA, suggesting less effective information processing during adolescent decision-making. Growth curve modeling revealed that early exposure to material hardship, but not the trajectory of experience, has a lasting influence on cognitive processing in adolescents. Lower EEA, in turn, reflected attentional problems in adolescents. Our findings suggest that deprivation of essential living conditions in childhood may impact affective-cognitive processes underlying attentional problems in adolescents, which could inform future interventions addressing adversity in underrepresented populations. These findings also highlight the application of computational models of cognition in revealing specific cognitive mechanisms in adolescents that would otherwise be hidden under conventional behavioral measures.

Paper #3
Decoding Adolescent Risk Behaviors: Insights from Combining RL and EAMs
Author information	Role
Felix Pichardo, University of Minnesota, United States	Presenting author
Alexander S. Weigard, University of Michigan, USA	Non-presenting author
Edward Huntley, University of Michigan, USA	Non-presenting author
Daniel Keating, University of Michigan, USA	Non-presenting author
Timothy Pleskac, Indiana University, USA	Non-presenting author
Abstract
Background: Adolescence is marked by an increase in risky behaviors that is associated with poor health outcomes and mortality. Developmental changes in dopaminergic regions are theorized to overwhelm cognition during decision-making, driving risk-taking. Insights into this process could help mitigate these poor outcomes. This dopamine-driven reward-processing has been described by Reinforcement Learning (RL) models of decision-making, which provide insights into processes that have been linked with externalizing disorders and substance use – all highly associated with risky behavior in adolescence. Decision-making can also be modeled as a dynamic process of evidence accumulation in a noisy environment. These Evidence Accumulation Models (EAMs) describe choices and response times across a wide variety of decision tasks, and their parameters have also been linked to externalizing behaviors. Combining these modeling frameworks provides a novel and more comprehensive view of decision-making mechanisms driving adolescent risk behaviors. Methods: We will leverage data from the Adolescent Health Risk Behaviors study, a large, community-based study of high school students (N = 1,601; age M = 16.8; 55% female; 56% White, 21% Black, 23% other race/ethnicity). Adolescents completed assessments of their externalizing and risk behaviors and the Iowa Gambling Task (IGT), which is used to simulate risky decision-making. Performance on the IGT has been associated with both externalizing behaviors (e.g., substance use) and normative changes in adolescent decision-making. Computational models decompose raw performance into the underlying processes involved in decision-making, potentially improving sensitivity. Previous computational studies have relied on small, predominantly adult male samples and RL choice models alone. We will combine RL with Drift Diffusion Models (a type of EAM) using hierarchical Bayesian estimation to capitalize on both choice and response time data to more fully describe IGT performance and potentially improve the precision of RL parameter estimates. We plan to use these models to investigate associations between underlying decision-making processes with externalizing and real-world risk behaviors. Hypotheses/Preliminary Results: We hypothesize that the estimates of underlying decision-making processes will be more predictive of externalizing and real-world risk behavior than raw performance measures. Additionally, we hypothesize that motivational processes and evidence accumulation processes will be predictive of increased real-world risk behaviors. Our preliminary analyses (N = 200) suggest significant associations between reward and punishment sensitivity with self-reported disinhibition (Brief Sensation Seeking Scale; b = -2.30, p = .020, and b = -8.73, p = .011). These associations appear to differ between male and female participants (b = 4.86, p = .014, and b = 22.57, p = .001), potentially reflecting real-world gender differences in externalizing and risk-taking behaviors. Implications: Combining these modeling frameworks to assess adolescent risky decision-making provides a novel window into how adolescents learn from experience and accumulate evidence to come to decisions. This can elucidate the underlying processes in externalizing and risky behavior. Understanding how this peaks in adolescence and how it can persist into adulthood is essential in mitigating the detrimental outcomes associated with this by helping to guide targets for interventions.

⇦ Back to session