Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2025 Aug 23;16(1):7869.
doi: 10.1038/s41467-025-63078-x.

Neural correlates of human fear conditioning and sources of variability in 2199 individuals

Joaquim Radua #  1   2   3 Hannah S Savage #  4   5   6 Enric Vilajosana  1   2 Alec Jamieson  7 Birgit Abler  8 Fredrik Åhs  9 Tom Beckers  10   11 Narcís Cardoner  12   13   14 Josh M Cisler  15 Juliana B Diniz  16 Dominik R Bach  17   18   19 Sigrid Elsenbruch  20 Steven G Greening  21 Daphne J Holt  22   23   24 Antonia N Kaczkurkin  25 Andreas Keil  26 Merel Kindt  27 Kathrin Koch  28 Kevin S LaBar  29 Charlene L Lam  30   31 Christine L Larson  32 Tina B Lonsdorf  33   34 Christian J Merz  35 Katie A McLaughlin  36   37 Yuval Neria  38 Daniel S Pine  39 Carien M van Reekum  40 Alexander J Shackman  41   42 Carles Soriano-Mas  3   43   44 Victor I Spoormaker  45 Daniel M Stout  46   47   48 Benjamin Straube  49   50 Thomas Straube  51 Lauri Tuominen  52 Renée M Visser  27 Laura Ahumada  26 Volker Arolt  53   54 Marcelo C Batistuzzo  16   55 Paulo R Bazán  56 Emma E Biggs  10 Marta Cano  12   13 Pamela Chavarría-Elizondo  44   57 Samuel E Cooper  58 Udo Dannlowski  53 Víctor de la Peña-Arteaga  12   13   43 Stephanie N DeCross  36 Katharina Domschke  59   60 Mana R Ehlers  33   34 John L Graner  29 Alfons O Hamm  61 Martin J Herrmann  62 Ashley A Huggins  63 Adriane Icenhour  20 Asier Juaneda-Seguí  44 Markus Junghoefer  54   64 Tilo Kircher  49   50 Katja Koelkebeck  65 Manuel Kuhn  34   66   67 Franziska Labrenz  20 Shmuel M Lissek  68 Martin Lotze  69 Ulrike Lueken  60   70 Jürgen Margraf  71 Ignacio Martínez-Zalacaín  44   72 Robert Moeck  51 Jayne Morriss  73 María Ortuño  1 Andre Pittig  74 Daniel Porta-Casteras  12 Jan Richter  75 Isabelle C Ridderbusch  49   50 Winfried Rief  76 Kati Roesmann  64   77 Jörgen Rosén  78 Alena N Rußmann  34   79 Rachel Sjouwerman  34   80 Jennifer Spohrs  81   82 Andreas Ströhle  83 Benjamin Suarez-Jimenez  84   85 Martin Ulrich  8 Hans-Ulrich Wittchen  86 Xi Zhu  38   87 Lea Waller  83 Henrik Walter  60   83 Paul M Thompson  88 Janna Marie Bas-Hoogendam  89   90   91 Nynke A Groenewold  92 Dan J Stein  93 Nic J Van der Wee  90   91 Joseph E Dunsmoor  94 Andre F Marquand  4   5 Ben J Harrison  7 Miquel A Fullana  95   96
Affiliations

Neural correlates of human fear conditioning and sources of variability in 2199 individuals

Joaquim Radua et al. Nat Commun. .

Abstract

Pavlovian fear conditioning is a fundamental process in both health and disease. We investigate its neural correlates and sources of variability using harmonized functional magnetic resonance imaging data from 2199 individuals across nine countries, including 1888 healthy individuals and 311 with anxiety-related or depressive disorders. Using mega-analysis and normative modeling, we show that fear conditioning consistently engages brain regions within the "central autonomic-interoceptive" or "salience" network. Several task variables strongly modulate activity in these regions, contributing to variability in neural responses. Additionally, brain activation patterns differ between healthy individuals and those with anxiety-related or depressive disorders, with distinct profiles characterizing specific disorders such as post-traumatic stress disorder and obsessive-compulsive disorder. While the neural correlates of fear conditioning are highly generalizable at the population level, variability arises from differences in task design and clinical status, highlighting the importance of methodological diversity in capturing fear learning mechanisms.

PubMed Disclaimer

Conflict of interest statement

Competing interests: Dr Stein has received consultancy honoraria from Discovery Vitality, Johnson & Johnson,. Kanna, L’Oreal,Lundbeck, Orion, Sanofi, Servier, Takeda and Vistagen. The other authors declare no competing interests.

Figures

Fig. 1
Fig. 1. Neural correlates and individual-level heterogeneity in human fear conditioning.
Schematic indicating the levels of analysis (a). Significant brain functional activation (b) and deactivation (c) to the CS+ versus CS− determined by mega-analysis (n = 1888 healthy controls). Schematic of normative modelling framework (d). Normative probability maps illustrate the percentage of participants in the healthy control test sample who had positive (hot colours -right) or negative deviations (cool colours - left) >±2.6 within each voxel. Circle highlights frequent large deviations (both positive and negative) within the most ventral region of the vmPFC (e). AIC anterior insular cortex, AG angular gyrus, CN caudate nucleus, dACC dorsal anterior cingulate cortex, dlPFC dorsolateral prefrontal cortex, dPFC dorsal prefrontal cortex, dPons dorsal pons, dPrec dorsal precuneus, Hipp hippocampus, HYP hypothalamus, lOFC lateral orbitofrontal cortex, PCC posterior cingulate cortex, SI primary somatosensory cortex, SII secondary somatosensory cortex, SMA supplementary motor area, TG temporal gyrus, Thal thalamus, vmPFC ventromedial prefrontal cortex. Source data are provided as a Source Data file.
Fig. 2
Fig. 2. Robust influence of task variables on brain activation during fear conditioning.
Maps show the influence of pre-task instructions about CS-US contingency (a), type of US (b), number of CS used in paradigm (i.e. multiple CS+ or CS- or single CS+ or CS-) (c), pairing rate (d), and potential US confounding in CS + > CS- contrast (e) on mean activation (left; mega-analysis linear mixed-effects models) and relation to predicted activation (right; normative model structure coefficients). For the mega-analysis, warm colours indicate positive correlations (i.e., higher variable values associated with greater activation), while cool colours indicate negative correlations (i.e., higher variable values associated with reduced activation). For normative modeling, structure coefficient maps show the correlation coefficients (rho) thresholded by their respective coefficients of determination (rho2  >  0.3) of selected task variables. This can be interpreted as showing how predicted activation to the CS + > CS- contrast relates to the task variables included in the building of the normative models. Positive correlations (warm colours) indicate greater activation for higher values of the input variable and negative correlations (cool colours) greater activation for lower values of the input variable (note that some variables are dummy coded, e.g., pre-task instructions, type of US). CS Conditioned Stimulus; US Unconditioned Stimulus. For Pairing Rate (RR) in linear mixed-effects models, the figure shows significant results in the ANOVA comparing four categories (RR30, RR50, RR62, RR100). For the results of post-hoc tests, see Supplementary Figs. S6 and S7. Source data are provided as a Source Data file.
Fig. 3
Fig. 3. Differences between individuals with anxiety-related and depressive disorders and healthy controls in human fear conditioning.
Regions wherein individuals with anxiety-related and depressive disorders (n = 311) showed significantly increased functional activation to the CS+ versus CS − , as compared to healthy controls (a). Normative probability maps illustrate the percentage of participants in the sample (test controls – top; individuals with anxiety-related and depressive disorders – bottom) who had positive (hot colours – right) or negative deviations (cool colours – left) > ±2.6 within each voxel (b). Box plots show the distribution of the total number of large deviations (> ±2.6) per group. The centre line indicates the median; box bounds represent the 25 and 75th percentiles (interquartile range, IQR); whiskers extend to the smallest and largest values within 1.5 × IQR from the lower and upper quartiles. Sample sizes: control group n = 646; PTSD n = 55; OCD n = 68; GAD n = 48; SAD n = 31; total clinical group n = 202 (c). Normative probability maps illustrate the percentage of each clinical group who had positive (hot colours – right) or negative deviations (cool colours – left) > ±2.6 within each voxel (d). Confusion matrix for multi-class support vector machine differentiating patterns of deviations among clinical groups (e). ARDD anxiety-related and depressive disorders, GAD Generalised Anxiety Disorder, OCD Obsessive Compulsive Disorder, PTSD Post-traumatic Stress Disorder, SAD Social Anxiety Disorder. * = p < 0.05, ** = p < 0.01, *** = p < 0.0001. Kruskal–Wallis H-tests were used to test for main group effects (cases vs controls), with follow-up Mann–Whitney U tests false discovery rate (FDR) corrected at α = 0.05. Source data are provided as a Source Data file.

References

    1. Beckers, T. et al. Understanding clinical fear and anxiety through the lens of human fear conditioning. Nat. Rev. Psychol.2, 233–245 (2023). - PMC - PubMed
    1. Lonsdorf, T. B. et al. Don’t fear ‘fear conditioning’: methodological considerations for the design and analysis of studies on human fear acquisition, extinction, and return of fear. Neurosci. Biobehav. Rev.77, 247–285 (2017). - PubMed
    1. Fullana, M. A. et al. Neural signatures of human fear conditioning: an updated and extended meta-analysis of fMRI studies. Mol. Psychiatry21, 500–508 (2016). - PubMed
    1. Tovote, P., Fadok, J. P. & Lüthi, A. Neuronal circuits for fear and anxiety. Nat. Rev. Neurosci.16, 317–331 (2015). - PubMed
    1. LeDoux, J. The amygdala. Curr. Biol.17, R868–R874 (2007). - PubMed

MeSH terms

LinkOut - more resources