Statistical Learning in a Stressful Environment: Autonomic Nervous System Reactivity Shapes Learning Probabilistic Patterns from Speech Streams

Abstract

Learning in adulthood is embedded in everyday social life, in which periods of psychosocial stress alternate with recovery. The autonomic nervous system regulates how the body responds to environmental demands, yet individuals differ markedly in this regulation. It remains unknown whether such individual differences in bodily regulation modulate the ability to learn probabilistic patterns from sensory input. Here, we investigated statistical learning of probabilistic patterns in speech streams (i.e., statistical language learning) in a six-hour experiment incorporating psychosocial stress and recovery to approximate everyday conditions. Sixty-five adults were exposed to novel speech streams in high- and low-stress contexts, with learning assessed immediately after exposure and following a rest period. Heart rate variability was recorded throughout the experiment to capture individual differences in autonomic reactivity to stress and recovery. From these measures, we constructed composite proxies of sympathetic (SNS) and parasympathetic (PNS) nervous system reactivity. Individuals with congruent SNS-PNS reactivity--either jointly high or jointly low--showed superior statistical learning outcomes across stress contexts. SNS reactivity preferentially supported encoding, whereas PNS reactivity supported consolidation. Moreover, the effect of SNS activation during speech exposure on statistical learning depended on individuals SNS reactivity profiles. These findings demonstrate that individual differences in bodily regulation are tightly linked to the ability to learn statistical dependencies in stressful environments. Overall, the findings highlight the essential role of brain-body-environment interactions in statistical learning.