Effects of early-life stress on probabilistic reversal learning and response perseverance in young adults

Abstract

Early life stress (ELS), including experiences with abuse and neglect, are related to several negative health outcomes in adulthood. One area that has received attention is the increased rate of substance abuse disorder in individuals who had experienced ELS. Given the critical role habitual behavior in the development of substance abuse, ELS may affect the trajectory of neural development such that habitual responding is more dominant than in individuals who did not experience ELS. Here, we examine learning of a probabilistic classification task (the Weather Prediction Task) in healthy young adults who reported significant ELS and those that did not. This task can be learned in a declarative, model-based manner, or in a more habitual, stimulus-response manner. Participants learned to choose the outcome (sun or rain) that was probabilistically associated with each cue combination through reinforcement on each trial. After 100 trials, the probabilities were reversed, and we conceptualized habitual behavior as perseverating responses based on the old probabilities. We also collected information about subjective socio-economic status (sSES), anxiety, depression, and substance use from participants. Using multiple regression, we found that our measure of habitual responding was correlated with reported alcohol use, suggesting that our measure of habit has validity for health behaviors. Furthermore, we found that some forms of early life stress led to greater response perseverance after contingencies were reversed. Overall, the results suggest that childhood adversity may contribute to the development of habit.

Keywords: Early-life stress; Habit; Human; Learning; Reversal learning.

Figure 1.. Task schematic.

Participants were tasked with viewing card combination stimuli and predicting the associated weather outcomes (rain, ‘q’; sun, ‘p’). Participants received either positive or negative feedback depending on response accuracy. Unbeknownst to participants, the card stimuli held probabilistic associations with the weather outcomes which were later switched during a probabilistic reversal phase (100 trials/probability phase, 200 trials total).

Figure 2.. Experimental group differences in self-reported stress.

Stress scores over the experimental time course separated by ELS and experimental group. Error bars denote standard error of the means.

Figure 3.. Differences in percent accuracy based on probability phase, ELS group, and experimental condition.

Significantly greater accuracy among the high compared to low ELS group during the initial probability phase and among the high compared to low ELS group in the stress condition. Error bars denote standard error of the means.

Figures 4A and 4B.. ELS and depression effects on reversal adjustment.

Partial regression plots showing that childhood sexual abuse exposure (A, top) and depressive symptomology (B, bottom) significantly predict phase accuracy subtraction scores, controlling for other variables at their means.

Figures 5A, 5B, 5C, and 5D.. ELS and subjective SES effects on response perseverance.

Partial regression plots showing that response perseverance is positively predicted by childhood emotional neglect (A), sexual abuse (B), and subjective SES (C) and negatively predicted by emotional abuse (D), controlling for all other variables at their means.

Figures 6A, 6B, and 6C.. Effects of childhood physical neglect exposure and response perseverance on frequency of risky alcohol use.

Partial regression plots showing significant positive effects of childhood physical neglect exposure on frequency of risky alcohol use in models accounting for reversal phase performance adjustment (A) or response perseverance (B), among other covariates. Response perseverance also significantly positively predicted alcohol use (C).