Local and network neural activations and their associations with sleep parameters during threat conditioning and extinction in persons with generalized anxiety disorder with and without insomnia disorder.
Seo J, Yuksel C, et al. • Psychiatry research. Neuroimaging • 2026
Better sleep quality promoted greater engagement of neural substrates of extinction learning, and the GAD+ID group failed to engage brain areas supporting extinction learning immediately following threat conditioning but did so when stimuli were again presented following a delay.
Key Findings
Results
Activations to the reinforced stimulus (CS+) that increased across threat conditioning were more extensive within the GAD+ID group compared to the GAD-ID group.
Participants had Generalized Anxiety Disorder (GAD) and were divided into those with moderate to severe Insomnia Disorder (GAD+ID) and those with absent or sub-threshold insomnia disorder (GAD-ID).
fMRI neural activations were measured during threat conditioning, extinction learning, and extinction recall.
The analysis used both a liberal cluster-determining threshold for network-level analysis and family-wise error correction for localized activations.
The reinforced stimulus (CS+) showed more widespread increasing activations across the conditioning phase in the GAD+ID group.
Results
Increased activations to the CS+ across extinction learning were greater within the GAD-ID group than the GAD+ID group, and were delayed by 24 hours in the GAD+ID group.
The GAD-ID group showed greater neural engagement during extinction learning immediately following threat conditioning compared to the GAD+ID group.
The GAD+ID group failed to engage brain areas supporting extinction learning immediately following threat conditioning.
The GAD+ID group did engage these brain areas when stimuli were again presented following a delay (extinction recall at 24 hours).
This suggests a temporal dissociation in extinction learning engagement between the two groups based on comorbid insomnia severity.
Results
Greater sleep efficiency was associated with decreased neural activations across threat conditioning.
Whole-sample correlations between sleep parameters and fMRI activations were examined.
Sleep efficiency showed a negative association with CS+ activations that increased across the threat conditioning phase.
This association was examined across the full combined sample of GAD participants with and without insomnia disorder.
Both local activations (family-wise error corrected) and network-level activations (liberal cluster-determining threshold) were examined.
Results
Greater sleep efficiency was associated with increased neural activations across extinction learning.
Sleep efficiency showed a positive association with CS+ activations that increased across the extinction learning phase.
This finding is in the opposite direction from the association between sleep efficiency and threat conditioning activations.
The authors interpreted this as indicating that better sleep quality promoted greater engagement of neural substrates of extinction learning.
Both sleep quality measures and sleep physiology measures were examined in relation to activations.
Background
Deficient extinction learning and memory are hypothesized mechanisms for pathological anxiety that are associated with sleep disturbance.
The study was designed to test whether comorbid insomnia disorder modulates fear extinction neural circuitry in GAD.
The study examined both local (region-specific) and network-level neural activations using large-scale brain network analysis.
Measures of both sleep quality (e.g., sleep efficiency) and sleep physiology were included as correlates of neural activation.
The paradigm included threat conditioning, extinction learning, and extinction recall (assessed after a 24-hour delay).
Methods
The study employed a dual analytical approach using both liberal cluster-determining thresholds for network-level analysis and family-wise error correction for localized activation identification.
Between-group comparisons (GAD+ID vs. GAD-ID) and whole-sample correlations with sleep parameters were examined in relation to large-scale brain networks using a liberal cluster-determining threshold.
Localized activations were subsequently identified using family-wise error correction to control for multiple comparisons.
fMRI was the primary neuroimaging modality.
The design allowed examination of both broad network engagement and precise anatomical localization of activations.
What This Means
This research suggests that people with Generalized Anxiety Disorder (GAD) who also have moderate to severe insomnia show different brain activation patterns during fear learning and extinction compared to people with GAD who sleep better. Specifically, those with comorbid insomnia showed more extensive brain activation while learning a fear association, but failed to engage the brain regions involved in learning to inhibit that fear (extinction learning) right after the fear was established. Interestingly, they did show this extinction-related brain engagement when tested again 24 hours later, suggesting the process was delayed rather than absent.
The study also found that across all participants with GAD, better sleep efficiency (spending more time asleep while in bed) was linked to less brain activation during fear learning, but more brain activation during extinction learning. This pattern suggests that good sleep may help the brain respond more efficiently during fear acquisition while simultaneously being better prepared to engage the neural machinery needed to suppress or extinguish fear responses.
These findings matter because extinction learning — the ability to learn that something previously threatening is no longer dangerous — is thought to be a key mechanism underlying anxiety disorders and their treatment, including exposure-based therapies. This research suggests that sleep problems in people with GAD may disrupt the timing and extent of extinction learning in the brain, which could help explain why insomnia and anxiety so often occur together and why treatment outcomes might differ between those with and without sleep problems.
Seo J, Yuksel C, Oliver K, Daffre C, Song H, Lasko N, et al.. (2026). Local and network neural activations and their associations with sleep parameters during threat conditioning and extinction in persons with generalized anxiety disorder with and without insomnia disorder.. Psychiatry research. Neuroimaging. https://doi.org/10.1016/j.pscychresns.2026.112152