Skip to content

Mechanism and models

TL;DR — There is no mechanistic account of GAD that is both specific and validated. The psychological literature has converged on one idea from five directions — worry is avoidance — but the models disagree about what is avoided: somatic/imaginal arousal (avoidance model), uncertainty (intolerance-of-uncertainty model), worry itself (metacognitive model), emotional intensity (emotion dysregulation), or a negative emotional shift (contrast avoidance) (Behar 2009, PMID 19700258; Newman 2013, PMID 23537486). The best-supported construct, intolerance of uncertainty, is also demonstrably transdiagnostic: across 181 studies and 52,402 participants its association with symptoms is r=0.51, and its correlation with GAD (0.57) is barely above its correlation with major depression (0.53) or OCD (0.50) (McEvoy 2019, PMID 31678816; Gentes 2011, PMID 21664339). Biologically, heritability is ~30% (Kendler 1992, PMID 1558460; Hettema 2001, PMID 11578982), and the largest anxiety GWAS to date — 122,341 cases — found 58 loci, brain-wide enrichment and a GABAergic signal, but for anxiety disorders pooled, not GAD (Strom 2026, PMID 41634414). The one neuroimaging finding that distinguishes GAD from fear-based disorders is structural rather than functional: reduced left insula and lateral/medial prefrontal volume, increased right putamen volume, in GAD specifically (Liu 2022, PMID 36151073). The gap between the psychological models — which have generated effective treatments — and the biology — which has generated none — is the defining feature of this page.

Five psychological models of worry

Model Core claim What worry does Key evidence Weakness
Avoidance model of worry (Borkovec) Worry is verbal-linguistic thought that suppresses aversive imagery and somatic arousal Blunts autonomic response to feared images, negatively reinforcing itself One of five models reviewed and critiqued in Behar 2009 (PMID 19700258) Explains maintenance, not onset
Intolerance of uncertainty model (Dugas/Ladouceur) Uncertainty is experienced as intolerable; worry is an attempt to eliminate it Drives approach to hypothetical futures In the original test, IU was pivotal in discriminating 24 GAD patients from 20 controls; the four-component discriminant function classified 82% correctly (Dugas 1998, PMID 9613027) IU is transdiagnostic: r=0.57 with GAD, 0.53 with MDD, 0.50 with OCD (Gentes 2011, PMID 21664339)
Metacognitive model (Wells) Type-1 worry is followed by meta-worry — negative beliefs about worrying — which maintains the disorder Worry about worry converts a normal process into a disorder Structural test in 312 participants: metacognitions → worry → meta-worry in the hypothesised order gave good model fit, with meta-worry the most influential path (Nordahl 2023, PMID 36321682) Analogue samples; cross-sectional
Emotion dysregulation model (Mennin) GAD reflects heightened emotional intensity plus poor understanding/management of emotion Worry is a maladaptive regulation strategy Reviewed and contrasted with the metacognitive model (Behar 2009, PMID 19700258); an empirical head-to-head found the emotion-dysregulation model adds beyond the metacognitive model (Deleurme 2022, PMID 36185761) Overlaps with depression constructs
Contrast avoidance model (Newman/Llera) People with GAD are averse not to negative emotion but to sharp negative emotional shifts; worry keeps negative affect elevated so no contrast can occur Worry sustains, rather than reduces, negative affect — by design Reviewed and synthesised in Newman 2011 (PMID 21334285) and Newman 2013 (PMID 23537486); prospective test in Crouch 2017 (PMID 28577589); incremental validity over IU and negative problem orientation (Llera 2023, PMID 36958137) Also observed in MDD and social anxiety in a transdiagnostic test (Newman 2023, PMID 36565682)

The models converge on a shared prediction — worry is functional for the worrier — and Behar's critique groups them into cognitive (IU, metacognitive), emotional/experiential (emotion dysregulation, acceptance-based) and integrated (avoidance) clusters (Behar 2009, PMID 19700258). The contrast avoidance model is the one that reverses the usual assumption: worry does not reduce negative affect, it maintains it, so that a sudden shift never occurs (Newman 2011, PMID 21334285).

The specificity problem. None of these constructs has established GAD specificity. IU shows a broad transdiagnostic association across 335 effect sizes (McEvoy 2019, PMID 31678816), and contrast avoidance appears in MDD and social anxiety as well as GAD (Newman 2023, PMID 36565682). These models may therefore describe processes shared across the internalizing spectrum rather than disorder-exclusive mechanisms (the diagnostic boundary).

Where the models earn their keep is treatment. Across 26 studies and 1,199 GAD patients, psychological treatment produced large within-group reductions in intolerance of uncertainty (g=0.88 post, g=1.05 follow-up) and worry (g=1.32, g=1.45), with a large between-group effect on IU (g=1.35); crucially, CBT that explicitly targeted IU outperformed general CBT on both IU and worry at post-treatment (p<0.01 each), with meta-regression showing effect size increasing with time spent on IU — although the advantage was not maintained at follow-up (Wilson 2023, PMID 37271039). This is the closest thing in the GAD literature to a mechanism→treatment validation loop (cognitive behavioural therapy, other psychological therapies).

Learning and threat processing

  • Overgeneralization of conditioned fear. In a fear-generalization paradigm with rings of graded size, 22 GAD patients showed abnormally shallow generalization gradients on fear-potentiated startle and self-reported risk relative to 26 controls — i.e., the conditioned fear response degraded less as stimuli became less similar to the danger cue (Lissek 2014, PMID 24001473). This is the mechanistic candidate that best explains why anxiety in GAD is "generalized": the environment fills with cues.
  • Induced versus pathological anxiety. A meta-analysis of 181 fMRI studies (693 induced-anxiety participants; 2,554 patients and 2,348 controls) found shared activation in bilateral insula and cingulate/medial prefrontal cortex — but when split by disorder, specific phobia resembled induced anxiety most and GAD least (Chavanne 2021, PMID 33054384). Experimentally induced anxiety is a poor model for GAD specifically.
  • Parent-to-child transmission is at least partly environmental. Children may acquire generalized threat-responding through verbal information and modelling, with parents conveying that the world is unsafe, uncertainty is intolerable and worry helps — a route distinct from genetic transmission (Aktar 2017, PMID 28867938).
  • Childhood adversity is a large population-attributable component. In 60,719 university students across 18 countries, 64.9% reported ≥1 childhood adversity, and population attributable risk proportions for 12-month anxiety and mood disorders were 40.7–61.0%; the association was with lifetime risk rather than persistence (Husky 2025, PMID 40541041). Pooled anxiety/mood, not GAD-specific.

Neurobiology

Structure and connectivity. The most complete synthesis screened 4,645 records and included 85 (35 structural, 33 functional-connectivity, 42 task-based): reduced hippocampal, ACC and amygdala volume; reduced dlPFC–amygdala and ACC–amygdala functional connectivity; greater pre- and postcentral volume with reduced supplementary motor area volume; altered cerebellar connectivity (Kolesar 2019, PMID 31835287). An earlier systematic review of DSM-IV-defined GAD reached a compatible summary — abnormal amygdala and prefrontal activation with decreased functional and structural connectivity between them — while noting neuroendocrine findings were less consistent, with increased noradrenergic reactivity and perturbed cortisol secretion reported but not settled (Hilbert 2014, PMID 24655775).

A GAD-specific comparative structural estimate. A seed-based d-mapping meta-analysis contrasted GAD with fear-related anxiety disorders and MDD: GAD showed decreased left insula and lateral/medial prefrontal volume and increased right putamen volume relative to fear disorders; GAD and MDD both showed decreased prefrontal volumes; no shared abnormalities across all three groups were found (Liu 2022, PMID 36151073). These group-level meta-analytic differences support—but do not prove—a neurobiological distinction between GAD and fear disorders and some overlap with depression.

Autonomic. Resting-state parasympathetic HRV is reduced across anxiety disorders (99 studies, 4,897 patients vs 5,559 controls; Hedges' g=−0.39), with significant reductions specifically in GAD, panic, social anxiety and PTSD; HRV reactivity showed no group differences (Cheng 2022, PMID 35340102). A network meta-analysis of 42 studies (n=4,008) went further and found GAD patients had significantly lower SDNN than panic disorder patients (SMD −0.60, 95% CI −1.09 to −0.11) — the only quantitative physiological discriminator between two anxiety disorders located in this build (Wang 2023, PMID 36914118).

Sleep as a bidirectional mechanism. Insomnia is among the most pervasive GAD features; the proposed shared substrate is HPA hyperactivity, amygdala hyperreactivity, reduced prefrontal control and GABAergic deficit, and CBT for insomnia reduces anxiety severity without directly targeting anxiety (Xue 2025, PMID 40318600). Review-level rather than trial-level for the causal claim.

Cognitive and electrophysiological findings

Domain Finding Source
Executive function Across 32 studies and 13,084 participants, GAD is associated with poorer cognitive flexibility and working memory but not inhibitory control; effects are similar in youths and young/middle-aged adults, absent in older adults, stronger in clinical than non-clinical samples and stronger for self-report than performance indices. GAD showed both slower reaction times and poorer accuracy — which contradicts attentional control theory's prediction that anxiety degrades efficiency, not accuracy Nguyen 2025, PMID 40513710
Longitudinal cognition No consistent evidence of sustained cognitive decline across 13 longitudinal studies and RCTs Li 2026, PMID 42191132
Event-related potentials 37 studies, 1,086 GAD patients vs 1,315 controls, 10 ERP components; GAD showed decreased P3 amplitude (g=−0.54, 95% CI −0.70 to −0.38, I²=20%) and increased ERN amplitude (g=−0.42, −0.72 to −0.12), plus delayed P3 (g=0.43), N2 (g=0.36) and MMN (g=0.63, 0.52–0.75, I²=0%) latencies; 25 studies low risk of bias, 12 moderate Xie 2025, PMID 39675130
Metacognition Across 47 studies (3,772 patients, 3,376 controls) using the MCQ, negative beliefs about the uncontrollability and danger of thoughts were most prevalent in GAD specifically; heightened cognitive self-consciousness was more characteristic of OCD; GAD, OCD and eating disorders had the most similar metacognitive profiles Sun 2017, PMID 28763680
Social cognition Emotion recognition and theory of mind were impaired in social anxiety disorder (SMD −0.32 and −0.44) but results for GAD were inconclusive with only two studies per domain Baez 2023, PMID 37275989
Inflammation 14 studies, 1,188 GAD patients vs 10,623 controls, 16 cytokines evaluated: CRP, interferon-γ and TNF-α raised in two or more studies; only CRP had enough data to pool — Cohen's d=0.38 (0.06–0.69), comparable in size to schizophrenia, with high heterogeneity (I²=75%). Five of 14 studies found no difference in at least one cytokine Costello 2019, PMID 31326932

The metacognition result is the closest thing in this literature to a GAD-specific psychological finding: of the MCQ dimensions, the one about uncontrollability and danger of thoughts discriminates GAD from other disorders (Sun 2017, PMID 28763680). It converges with the criterion-level evidence that uncontrollability — not excessiveness — carries incremental validity (Hallion 2013, PMID 23713499) and with metacognitive therapy's head-to-head performance (other psychological and non-drug therapies).

The inflammation finding should be read cautiously: a pooled CRP effect of d=0.38 with I²=75% from studies differing in medication use, comorbid depression and sampling methodology is a signal that GAD is under-studied in neuroimmunology, not a demonstration that inflammation causes GAD — which is what the authors say (Costello 2019, PMID 31326932).

Infrastructure note. The ENIGMA-Anxiety/GAD working group has assembled a multi-site structural-imaging mega-analysis and published its methodological approach, including the problems of cross-country subject-level data transfer and quality control at that scale (Zugman 2022, PMID 32596977). That consortium is the most likely source of a replicated GAD imaging finding.

Genetics

Level Finding Source
Twin heritability, GAD ~30%; familial aggregation is genetic rather than shared-environmental; not explained by GAD occurring only during depression or panic episodes; similar with 1- vs 6-month duration Kendler 1992, PMID 1558460
Meta-analytic heritability 0.32 for GAD (vs 0.43 panic); family-study OR 4–6 for first-degree relatives across anxiety disorders; possible role for shared familial environment in women only Hettema 2001, PMID 11578982
GAD–MDD genetic correlation +1.00 (women), +0.74 (men) in 37,296 Swedish twins Kendler 2007, PMID 17121688
Multivariate structure GAD loads with major depression on the second genetic factor, separate from phobia/panic/bulimia Kendler 1995, PMID 7726718
Early GWAS 9 samples, >18,000 individuals: rs1709393 (3q12.3) for case-control; rs1067327 in CAMKMT (2p21) for factor scores; both require replication Otowa 2016, PMID 26754954
Current GWAS 122,341 anxiety cases / 729,881 controls: 58 loci, 66 genes, 51/58 replicated in 1.18M self-report cases; enrichment across all major brain regions; GABAergic signalling highlighted Strom 2026, PMID 41634414
Fear vs GAD 4 genome-wide loci for GAD (≤60,879 cases) vs 1 for fear disorders; fear–GAD rg differed from unity only under phenotype-specific control screening (rg=0.87, p=9.32×10⁻³); education, cognitive ability and CAD correlated more with fear, bipolar I, anorexia and neuroticism more with GAD Ter Kuile 2026, PMID 42374129

The genetics tells a consistent story that is awkward for the diagnosis: a broad internalizing liability with modest disorder-specific components, in which GAD's nearest genetic neighbour is depression rather than the other anxiety disorders (Kendler 1995, PMID 7726718; Ter Kuile 2026, PMID 42374129). Redefining phenotypes hierarchically has been proposed explicitly as the fix for underpowered diagnosis-level genetics (Waszczuk 2020, PMID 31804095).

What is missing

  1. No mechanistic biomarker. Nothing in the imaging, autonomic or endocrine literature has been tested prospectively as a diagnostic or predictive marker in GAD. The HRV SDNN difference between GAD and panic (Wang 2023, PMID 36914118) is the only candidate discriminator with an effect size attached, and it is a group difference, not a classifier.
  2. No mechanism→drug loop. The GABAergic GWAS signal (Strom 2026, PMID 41634414) points at a system already targeted by benzodiazepines and pregabalin (pregabalin, benzodiazepines and others), which were developed without it.
  3. Model-to-model comparison is rare. Head-to-head tests exist (Deleurme 2022, PMID 36185761; Llera 2023, PMID 36958137) but are small, cross-sectional and analogue-sample.

Open questions

  • Is any psychological mechanism GAD-specific, or is the model literature largely describing internalizing psychopathology? IU is broadly transdiagnostic (McEvoy 2019, PMID 31678816), and contrast avoidance appears across three disorders (Newman 2023, PMID 36565682).
  • Does the GAD-specific insula/prefrontal/putamen signature (Liu 2022, PMID 36151073) replicate prospectively, and does it move with treatment?
  • Why does GAD resemble experimentally induced anxiety least of the anxiety disorders (Chavanne 2021, PMID 33054384)? If induced anxiety is the translational model used in drug development, GAD may be the disorder that model fits worst.
  • Is conditioned-fear generalization a reliable GAD marker? Lissek 2014 reported overgeneralization in 22 patients (PMID 24001473), but a later study of 28 GAD patients and 30 controls found no group difference on most indices (Tinoco-González 2015, PMID 26459843). The small, conflicting experimental literature has not established a clinically useful trait or state marker.
  • Does targeting IU directly produce durable rather than transient advantage? The advantage disappears by follow-up (Wilson 2023, PMID 37271039).

References

  1. Behar E, et al. Current theoretical models of generalized anxiety disorder (GAD): conceptual review and treatment implications. J Anxiety Disord. 2009;23:1011-23. PMID 19700258
  2. Newman MG, et al. Worry and generalized anxiety disorder: a review and theoretical synthesis of evidence on nature, etiology, mechanisms, and treatment. Annu Rev Clin Psychol. 2013;9:275-97. PMID 23537486
  3. Newman MG, Llera SJ. A novel theory of experiential avoidance in generalized anxiety disorder: a review and synthesis of research supporting a contrast avoidance model of worry. Clin Psychol Rev. 2011;31:371-82. PMID 21334285
  4. Dugas MJ, et al. Generalized anxiety disorder: a preliminary test of a conceptual model. Behav Res Ther. 1998;36:215-26. PMID 9613027
  5. Gentes EL, Ruscio AM. A meta-analysis of the relation of intolerance of uncertainty to symptoms of generalized anxiety disorder, major depressive disorder, and obsessive-compulsive disorder. Clin Psychol Rev. 2011;31:923-33. PMID 21664339
  6. McEvoy PM, et al. The impact of methodological and measurement factors on transdiagnostic associations with intolerance of uncertainty: A meta-analysis. Clin Psychol Rev. 2019;73:101778. PMID 31678816
  7. Nordahl H, et al. An empirical test of the metacognitive model of generalized anxiety disorder. Scand J Psychol. 2023;64:263-267. PMID 36321682
  8. Deleurme KA, et al. Generalized Anxiety Disorder: Does the Emotion Dysregulation Model Predict Symptoms Beyond the Metacognitive Model? J Ration Emot Cogn Behav Ther. 2022. PMID 36185761
  9. Crouch TA, et al. Prospective Investigation of the Contrast Avoidance Model of Generalized Anxiety and Worry. Behav Ther. 2017. PMID 28577589
  10. Llera SJ, Newman MG. Incremental validity of the contrast avoidance model: A comparison with intolerance of uncertainty and negative problem orientation. J Anxiety Disord. 2023. PMID 36958137
  11. Newman MG, et al. A transdiagnostic evaluation of contrast avoidance across generalized anxiety disorder, major depressive disorder, and social anxiety disorder. J Anxiety Disord. 2023. PMID 36565682
  12. Wilson EJ, et al. The impact of psychological treatment on intolerance of uncertainty in generalized anxiety disorder: A systematic review and meta-analysis. J Anxiety Disord. 2023;97:102729. PMID 37271039
  13. Lissek S, et al. Generalized anxiety disorder is associated with overgeneralization of classically conditioned fear. Biol Psychiatry. 2014;75:909-15. PMID 24001473
  14. Chavanne AV, Robinson OJ. The Overlapping Neurobiology of Induced and Pathological Anxiety: A Meta-Analysis of Functional Neural Activation. Am J Psychiatry. 2021;178:156-164. PMID 33054384
  15. Aktar E, et al. Environmental transmission of generalized anxiety disorder from parents to children: worries, experiential avoidance, and intolerance of uncertainty. Dialogues Clin Neurosci. 2017;19:137-147. PMID 28867938
  16. Husky MM, et al. Childhood adversities and their associations with mental disorders in the World Mental Health International College Student surveys initiative. Psychiatry Res. 2025;351:116585. PMID 40541041
  17. Kolesar TA, et al. Systematic review and meta-analyses of neural structural and functional differences in generalized anxiety disorder and healthy controls using magnetic resonance imaging. Neuroimage Clin. 2019;24:102016. PMID 31835287
  18. Hilbert K, et al. Neural structures, functioning and connectivity in Generalized Anxiety Disorder and interaction with neuroendocrine systems: a systematic review. J Affect Disord. 2014;158:114-26. PMID 24655775
  19. Liu X, et al. Pathological fear, anxiety and negative affect exhibit distinct neurostructural signatures: evidence from psychiatric neuroimaging meta-analysis. Transl Psychiatry. 2022;12:405. PMID 36151073
  20. Cheng YC, et al. Heart rate variability in patients with anxiety disorders: A systematic review and meta-analysis. Psychiatry Clin Neurosci. 2022;76:292-302. PMID 35340102
  21. Wang Z, et al. Heart rate variability in generalized anxiety disorder, major depressive disorder and panic disorder: A network meta-analysis and systematic review. J Affect Disord. 2023;330:259-266. PMID 36914118
  22. Xue Y, et al. Sleep disturbances in generalized anxiety Disorder: The central role of insomnia. Sleep Med. 2025;132:106545. PMID 40318600
  23. Kendler KS, et al. Generalized anxiety disorder in women. A population-based twin study. Arch Gen Psychiatry. 1992;49:267-72. PMID 1558460
  24. Hettema JM, et al. A review and meta-analysis of the genetic epidemiology of anxiety disorders. Am J Psychiatry. 2001;158:1568-78. PMID 11578982
  25. Kendler KS, et al. The sources of co-morbidity between major depression and generalized anxiety disorder in a Swedish national twin sample. Psychol Med. 2007;37:453-62. PMID 17121688
  26. Kendler KS, et al. The structure of the genetic and environmental risk factors for six major psychiatric disorders in women. Arch Gen Psychiatry. 1995;52:374-83. PMID 7726718
  27. Otowa T, et al. Meta-analysis of genome-wide association studies of anxiety disorders. Mol Psychiatry. 2016;21:1391-9. PMID 26754954
  28. Strom NI, et al. Genome-wide association study of major anxiety disorders in 122,341 European-ancestry cases identifies 58 loci and highlights GABAergic signaling. Nat Genet. 2026;58:275-288. PMID 41634414
  29. Ter Kuile AR, et al. Genome-wide genetic overlap between fear-based disorders and generalised anxiety disorder. Mol Psychiatry. 2026. PMID 42374129
  30. Waszczuk MA, et al. Redefining phenotypes to advance psychiatric genetics: Implications from hierarchical taxonomy of psychopathology. J Abnorm Psychol. 2020;129:143-161. PMID 31804095
  31. Nguyen L, et al. Executive functioning in individuals with generalized anxiety disorder: A systematic review and meta-analysis. J Affect Disord. 2025;389:119683. PMID 40513710
  32. Li J, et al. Longitudinal Association Between Generalized Anxiety Disorder and Cognitive Change: A Meta-Analysis. Psychiatry Investig. 2026;23:588-596. PMID 42191132
  33. Xie C, et al. The characteristics of event-related potentials in generalized anxiety disorder: A systematic review and meta-analysis. J Psychiatr Res. 2025;181:470-483. PMID 39675130
  34. Sun X, et al. Dysfunctional metacognition across psychopathologies: A meta-analytic review. Eur Psychiatry. 2017;45:139-153. PMID 28763680
  35. Baez S, et al. Performance in emotion recognition and theory of mind tasks in social anxiety and generalized anxiety disorders: a systematic review and meta-analysis. Front Psychiatry. 2023;14:1192683. PMID 37275989
  36. Costello H, et al. Systematic review and meta-analysis of the association between peripheral inflammatory cytokines and generalised anxiety disorder. BMJ Open. 2019;9:e027925. PMID 31326932
  37. Zugman A, et al. Mega-analysis methods in ENIGMA: The experience of the generalized anxiety disorder working group. Hum Brain Mapp. 2022;43:255-277. PMID 32596977
  38. Hallion LS, Ruscio AM. Should uncontrollable worry be removed from the definition of GAD? A test of incremental validity. J Abnorm Psychol. 2013;122:369-75. PMID 23713499
  39. Tinoco-González D, et al. Conditioned Fear Acquisition and Generalization in Generalized Anxiety Disorder. Behav Ther. 2015;46:627-639. PMID 26459843