MULTIMODAL SPATIOTEMPORAL DYNAMICS OF FRONTOMEDIAL THETA AND BOLD SIGNAL REVEAL FUNCTIONAL ROLES IN UPDATING AND SUPPRESSING AVERSIVE MEMORY DURING FEAR EXTINCTION

MULTIMODAL SPATIOTEMPORAL DYNAMICS OF FRONTOMEDIAL THETA AND BOLD SIGNAL REVEAL FUNCTIONAL ROLES IN UPDATING AND SUPPRESSING AVERSIVE MEMORY DURING FEAR EXTINCTION | Authorea try { document.documentElement.classList.add('js'); } catch (e) { } var _gaq = _gaq || []; _gaq.push(['_setAccount', 'G-8VDV14Y67G']); _gaq.push(['_trackPageview']); (function() { var ga = document.createElement('script'); ga.type = 'text/javascript'; ga.async = true; ga.src = ('https:' == document.location.protocol ? 'https://ssl' : 'http://www') + '.google-analytics.com/ga.js'; var s = document.getElementsByTagName('script')[0]; s.parentNode.insertBefore(ga, s); })(); Skip to main content Preprints Collections Wiley Open Research IET Open Research Ecological Society of Japan All Collections About About Authorea FAQs Contact Us Quick Search anywhere Search for preprint articles, keywords, etc. Search Search ADVANCED SEARCH SCROLL Psychophysiology This is a preprint and has not been peer reviewed. Data may be preliminary. 17 July 2025 V1 Latest version Share on MULTIMODAL SPATIOTEMPORAL DYNAMICS OF FRONTOMEDIAL THETA AND BOLD SIGNAL REVEAL FUNCTIONAL ROLES IN UPDATING AND SUPPRESSING AVERSIVE MEMORY DURING FEAR EXTINCTION Authors : Arslan Gabdulkhakov 0000-0003-4857-3601 , Matthias F. J. Sperl 0000-0002-5011-0780 , Christian J. Merz 0000-0001-5679-6595 , Laura-Isabelle Klatt 0000-0002-5682-5824 , Christoph Fraenz , and Erhan Genç [email protected] Authors Info & Affiliations https://doi.org/10.22541/au.175274364.47542664/v1 Published Psychophysiology Version of record Peer review timeline 1420 views 164 downloads Contents Abstract Information & Authors Metrics & Citations View Options References Figures Tables Media Share Abstract Fear and extinction learning are fundamental processes shaping the regulation and expression of fear in both healthy individuals and patients with anxiety-related disorders. Pavlovian fear conditioning serves as a powerful model for these mechanisms; however, the precise spatiotemporal neural dynamics underlying fear and extinction learning in humans still remain unclear. Theta oscillations have been implicated in these learning processes, yet their precise relationship with blood-oxygen-level-dependent (BOLD) signals of corresponding brain networks remains poorly understood. This study employed simultaneous electroencephalography and functional magnetic resonance imaging (EEG-fMRI) in fifty healthy humans to investigate the role of frontomedial theta oscillations (4-8 Hz) in fear learning. Participants underwent a one-day differential fear conditioning paradigm with concurrent EEG and fMRI recordings. We assumed that theta power variations would correspond to distinct activation patterns in the fear and safety networks during fear acquisition and extinction training. To test this hypothesis, we extracted frontal-midline theta power across three trial segments (0–2 s, 2–4 s, 4–5.5 s after the onset of conditioned stimuli, CS) and integrated these measures into whole-brain fMRI analyses. Results revealed a significant increase in differential (CS+ vs. CS−) theta power towards the end of the trial during fear acquisition training, aligning with prior findings of theta ramping-up before unconditioned stimulus onset. EEG-driven fMRI analyses during fear acquisition showed distinct theta-BOLD co-activations in cuneal and precuneal cortices and visual areas at 2–4 s and 4–5.5 s trial segments. Notably, during extinction training, the theta activity of the mid-trial segment (2–4 s post-stimulus) was co-activated with the BOLD signal in vmPFC, suggesting a role of theta during extinction learning in suppressing aversive memory representations. Our findings support the hypothesis that theta oscillations may contribute to the temporal encoding of threat expectation during fear learning, but also to memory updating through fear response suppression during extinction learning. Interestingly, theta modulation was linked to distinct brain regions in different learning phases. Critically, our results integrate previous findings from different neuroimaging modalities and extend our understanding of the spatiotemporal neural dynamics underlying fear and extinction learning. 1. INTRODUCTION The prevalence of anxiety-, trauma-, and stressor-related disorders in modern society underscores the importance of understanding the neural mechanisms underlying fear learning and its regulation (Somers et al., 2006; VanElzakker et al., 2014). Fear conditioning serves as a well-established model for studying these processes, providing key insights into how fear responses are acquired and extinguished (Lissek et al., 2005; Milad & Quirk, 2012; Sehlmeyer et al., 2009; Wake et al., 2024). In this paradigm, a previously neutral stimulus is paired with an aversive unconditioned stimulus (US) and becomes a conditioned stimulus (CS+), which then leads to a conditioned response (CR), while another neutral stimulus (CS−) acts as an unpaired control stimulus (Lonsdorf et al., 2017). To induce extinction learning, the same CS+ is repeatedly presented without the US. Over time, this diminishes the CR to the CS+ by creating a new inhibitory memory trace, competing with the initially acquired CS+/US association (Bouton, 2002). Contemporary neuroimaging research has identified key brain circuits involved in fear acquisition and extinction and their functional role in these processes (Knight et al., 2004; LaBar et al., 1998 ; Battaglia et al., 2020). These structures include but are not limited to the amygdala, hippocampus, insular cortex, dorsal anterior cingulate cortex (dACC), and ventromedial prefrontal cortex (vmPFC), constituting the fear and extinction network (Fullana et al., 2016; Sehlmeyer et al., 2009). Yet critical questions remain regarding their precise temporal and spatial dynamics, as well as the relationship between findings from different imaging modalities. For decades, behavioural and electrophysiological studies in animal models dominated the field of fear conditioning, disentangling the functional role of specific anatomical regions related to the conditioning and extinction of fear. Rodent prelimbic prefrontal cortex (PL; homologous to the human ACC) and infralimbic prefrontal cortex (IL; homologous to the human mPFC) were shown to be involved in fear expression and suppression, relaying sensory information to the amygdala through so-called “cortical” and “subcortical” pathways (VanElzakker et al., 2014; Blair et al., 2001; Pitkanen 2000; Milad and Quirk, 2002; Burgos-Robles and Vidal-Gonzalez, 2009). Particularly, theta frequency oscillations were reported to support the communication between the basolateral amygdala and mPFC (Karalis et al., 2016; Likhtik et al., 2014). One of the earliest electrophysiological studies on fear conditioning reported that the hippocampal CA1 region and the lateral amygdala display synchronized theta oscillations during conditioned fear, suggesting that coherent theta activity in amygdalo-hippocampal circuits may promote the synaptic plasticity mechanisms - such as LTP and LTD - underlying the consolidation and retention of fear memory (Seidenbecher et al., 2003). More recent work has identified the nucleus reuniens (RE) of the thalamus as a critical hub for coordinating mPFC-hippocampus theta synchrony during extinction retrieval: inactivation of RE disrupts both mPFC-hippocampus coherence and extinction memory, whereas theta-paced stimulation of RE restores context-appropriate fear suppression (Totty et al., 2023). More recently, studies utilising functional magnetic resonance imaging (fMRI) in awake rodents and avian models have expanded our understanding of large-scale fear networks during learning. Brydges et al. (2013) demonstrated that CS elicit activation in regions such as the lateral amygdala, insular cortex, and hypothalamus in awake rats. Behroozi et al. (2020) used event-related 7 Tesla fMRI in awake, behaving pigeons to map not only hippocampal-NCL (avian PFC homolog) interactions, but also the concurrent activation of primary olfactory, tactile, auditory, and visual sensory areas, along with motor regions, the amygdala, and the dopaminergic ventral tegmental area during multimodal sensory operant conditioning tasks. By capturing whole-brain BOLD responses within a single experiment, this ultra-high-resolution approach transcends the spatially restricted scope of traditional electrophysiological studies, which can typically sample only one or two discrete sites, and reveals a distributed network of sensory, motor, and executive regions engaged during multimodal operant conditioning. However, the integration of electrophysiological and hemodynamic data in animal models remains limited due to technical challenges, leaving open important translational questions about how oscillatory activity relates to blood-oxygen-level-dependent (BOLD) signals across species and modalities. Functional neuroimaging in humans has implicated the amygdala, hippocampus, insula, and prefrontal regions (including the dACC and vmPFC) in the acquisition and extinction of conditioned fear (Büchel & Dolan, 2000; Fullana et al., 2016; Kim & Jung, 2006; Sehlmeyer et al., 2009; LaBar et al., 1998; Phelps et al., 2004; Milad & Quirk, 2012). During fear acquisition, increased BOLD activity is typically observed in the amygdala and dACC, reflecting the encoding and expression of threat-related information. In contrast, extinction learning and recall are associated with enhanced activation of the vmPFC and hippocampus, which are thought to support context-dependent safety learning and inhibition of the previously learned conditioned response (Milad et al., 2007; Kalisch et al., 2006; Fullana et al., 2016, 2018). However, despite numerous individual findings from animal electrophysiology and human neuroimaging studies that helped identify regions constituting the fear and safety networks, taken together, they indicate the difficulty of integrating all the known results into a single cohesive interpretation of the involvement of these structures due to the high heterogeneity in the findings from study to study. Andres et al. (2024) discuss this issue, particularly in relation to the meta-analysis by Fullana et al. (2018), which did not find consistent vmPFC activation during fear extinction across 31 studies involving 1,074 participants. Similarly, meta-analyses have struggled to consistently confirm robust amygdala activation during fear acquisition and extinction (Fullana et al., 2016, 2018). Among other experiment-specific factors, Andres et al. (2024) attribute it to the transient nature of extinction learning activity that might happen only during the first few trials. Thus, although fMRI offers excellent spatial resolution for identifying these distributed circuits, its limited temporal resolution constrains the investigation of rapid neural dynamics involved in fear updating and regulation. Additionally, not much is known about the temporal specificity of respective brain structures during learning. It remains unclear if CS presentations evoke a synchronised BOLD response across all structures or if the functional activation of individual structures change dynamically throughout each trial. Complementary to the spatial insights offered by fMRI, electroencephalography (EEG) studies in humans have provided valuable information about the fast temporal dynamics of fear processing. Event-related potentials (ERPs) such as the late positive potential (LPP) and contingent negative variation (CNV) have been linked to conditioned fear responses and their modulation during extinction (Flor et al., 1996; Miskovic & Keil, 2012; Sperl et al., 2021). More recently, oscillatory analyses have revealed the involvement of frontomedial theta activity in the flexible updating of threat associations, with theta power increases often observed during extinction learning and recall (Bierwirth et al., 2021, 2023; Chien et al., 2017; Mueller et al., 2014). An intracranial EEG study in humans reported that theta oscillations in the prefrontal cortex covaried with the progression of learning, thus reflecting the learned association strength (Chen et al., 2021). Additionally, theta frequency oscillations were in synchrony between the amygdala and mPFC, with an increased theta power after successful learning and the latency of these theta oscillations shifted to earlier time points as learning progressed. However, Clarke et al. (2018) reported opposing results, where the frontal-midline theta power decreased as the association strength increased in an associative learning task. While the works above focused on dynamics and progression of learning across trials, a scalp EEG study by Starita et al. (2023) analysed temporal dynamics of theta power within trials. Specifically, they offered a novel approach in discretising the theta power into three 2-second-long trial segments: 0–2 s, 2–4 s, and 4-6 s relative to CS onset within the trial. Their findings, source-localised to midcingulate cortex and vmPFC, showed distinct dynamics of the midcingulate theta power across these trial segments in acquisition and reversal phases, with a prominent increase of the differential CS+/CS− effect towards the end of the trial, which could be attributed to threat expectation. An earlier study by DeLaRosa et al. (2014) also reported dynamics of the distinct topographical distribution of theta and beta power at different timepoints throughout the trial. Taken together, these findings implicitly hint at the dynamic interplay of various cortical and subcortical structures throughout the learning phase. While EEG source reconstruction provides a practical balance between temporal and spatial resolution, the spatial resolution of EEG source reconstruction is still inferior to that of fMRI (for a broader review of capabilities of EEG source imaging, see Michel et al., 2004). Hence, a direct link between the transient and temporally specific dynamics of EEG theta power and BOLD signal remains elusive. To date, only a few studies have employed simultaneous EEG-fMRI as part of the fear conditioning paradigm (Sperl et al., 2019; Yin et al., 2020). For example, Sperl et al. (2019) linked oscillatory EEG activity to BOLD responses during extinction recall and reported that frontal-midline theta power was significantly reduced for extinguished compared to non-extinguished stimuli, and that this reduction was positively correlated with decreased amygdala BOLD activation. However, the theta-BOLD relationship during fear acquisition and extinction training remains unknown. Although EEG and fMRI studies have independently advanced our understanding of the temporal and spatial aspects of fear learning, integrated multimodal findings remain scarce, limiting a unified interpretation of underlying neural mechanisms. Simultaneous EEG-fMRI enables a direct validation of modality-specific effects and offers a framework to link temporally defined oscillatory events to spatially distributed brain networks. This integration is especially crucial in learning paradigms, where rapid neural dynamics interact with large-scale circuit activity to shape behaviour. The understanding of precise temporal and spatial dynamics of oscillatory activity and brain regions could help establish more robust treatment protocols and inform future brain stimulation studies. To address this gap, we conducted a simultaneous EEG-fMRI study including fear acquisition and extinction training. By leveraging the complementary strengths of both imaging modalities, we aimed to confirm and extend prior findings by Sperl et al. (2019) and Starita et al. (2023), with a dual focus: (1) Tracking the temporal evolution of theta power across distinct segments within a trial during fear and extinction training, and (2) linking these dynamics to BOLD activation patterns across fear acquisition and extinction training. To enhance the generalisability of our findings across studies with varying CS presentation times, we adopted the approach of Starita et al. (2023) and analysed frontomedial theta across three trial segments: 0–2 s, 2–4 s, and 4–5.5 s post-CS onset. We assumed that the theta power in the three trial segments would involve distinct brain regions in the fMRI BOLD signal, thereby showing precise temporal and spatial dynamics of fear and extinction learning. 2. MATERIALS AND METHODS 2.1 Participants A sample of 50 (24 women, 26 men) right-handed participants aged between 18 and 26 (M = 22.38 years, SD = 2.35 years) years was recruited. The exclusion criteria at the recruitment stage were a history of mental health conditions, substance or psychoactive medication use, and standard exclusion criteria for fMRI examinations at a 3 Tesla scanner. All participants had normal or corrected-to-normal vision and were able to understand the provided written and oral instructions. All participants were naive to the purpose of the study and had no prior experience with the fear conditioning paradigm used for the experiment. The study was approved by the local ethics committee of the Faculty of Psychology at Ruhr University Bochum (application number 327). All participants provided written informed consent before participation and were treated in accordance with the Declaration of Helsinki. 2.2 Experiment The experiment was conducted at a 3-Tesla Philips Achieva scanner (Philips Healthcare, Best, The Netherlands) equipped with a 32-channel head coil at the Bergmannsheil Hospital in Bochum, Germany. The participants underwent fear acquisition and extinction training on the same day, separated by an 8-minute rest period inside the scanner. The stimuli and procedures for these paradigms were adapted from Milad, Wright, et al. (2007). The stimuli were presented using an MR-compatible display positioned at the back of the MRI bore (BOLDscreen 24 LCD; Cambridge Research Systems, Cambridge, UK). Participants were instructed to pay close attention to the images presented during fear acquisition and extinction training. They were also told that electrical stimulation may or may not be present during the experiment. Fear acquisition training consisted of 32 trials (16 CS+ and 16 CS−), with the CS+ paired with the aversive US in 10 of its 16 trials (62.5 % reinforcement rate), followed by extinction training with 8 CS+ (without the US) and 8 CS− trials. The number of acquisition and extinction trials was selected to be consistent with common experimental designs in human fear conditioning, as documented across numerous studies reviewed by Fullana et al. (2018), ensuring the comparability and applicability of our findings. A single trial commenced with a context presentation for 1 s (AB design, lamp on an office desk during fear acquisition training, lamp on a bookshelf during extinction training). This was followed by a 6 s CS presentation, namely the lamp lighting up in one of two colours: blue or yellow. The US was delivered via a constant voltage stimulator (STM2000 BIOPAC systems, CA, USA) and two electrodes attached to the fingertips of the index and middle fingers of the right hand. The stimulation started at 5.9 s after CS onset and co-terminated with the CS+ presentation. The stimulation lasted for 100 ms, consisting of 1 ms pulses at 50 Hz. The US intensity was calibrated individually to be rated by participants as “very unpleasant but not painful”, starting from 30 V and increasing in steps of 5 V. The trials were presented in a pseudo-randomised order, without repeating the same type of trial more than twice in a row. After fear acquisition and extinction training, the participants were asked to rate the contingency of the CS+, CS−, and US. This report included the ratings of: (1) the total number of electrical stimulations during the experiment; (2) the rate in percent of electrical stimulation following the blue and yellow lamp. Figure 1. Experimental outline of fear acquisition (left panel) and extinction (right panel) training. US = unconditioned stimulus. Both fear acquisition and extinction training start with an intertrial interval (ITI) represented as a fixation cross, which was jittered from 6.8-9.5s between the trials. Following the ITI, a context - office table with a lamp in acquisition, and a bookshelf with a lamp in extinction - was presented for 1 second. Then, the lamp colour, which lit up with either blue or yellow colour for a duration of 6 s, signalled either CS+ or CS− trial. The US, represented by an electrical stimulation to the right hand, was delivered for a duration of 100 ms at 5.9s, and terminated with the offset of CS+. We used a pseudorandomized partial reinforcement rate of 62.5%; thus, 10 out of 16 CS+ trials in the acquisition training were paired with the US. 2.3 Skin conductance response recording and analysis Subjects’ skin conductance responses (SCRs) were measured using an additional channel (GSR-MR sensor; Brain Products, Munich, Germany) of the EEG amplifier (see below). Two Ag/AgCl electrodes, each containing an isotonic 0.05 NaCl electrolyte solution, were placed on the hypothenar area located below the little finger of the left hand. Data were recorded using the Brain Vision Recorder software (Brain Products, Munich, Germany) with a sampling rate of 5,000 Hz. The SCR data were preprocessed to remove scanner artefacts (the specific steps described in the next section) and exported into MATLAB (version R2022b; MathWorks, Natick, MA, USA) for further preprocessing in EDA-App (Otto, et al., 2023). The SCRs were defined as the maximum amplitude values starting within a 1 to 6.5 s window relative to CS onset. The conditioned response (CR) was defined as the difference between the across-trial average SCR values for CS+ and CS− for each participant. After preprocessing, the SCR data were exported for subsequent statistical analyses. 2.4 EEG recording and analysis The electroencephalography data (EEG) were recorded using an MR-compatible 64-channel BrainCapMR system (BrainProducts, Munich, Germany) with an additional electrocardiogram (ECG) channel. The electrode array consisted of the following channels, which were positioned according to the 10-20 system: AF3, AF4, AF7, AF8, C1, C2, C3, C4, C5, C6, CP1, CP2, CP3, CP4, CP5, CP6, CPz, Cz, F1, F2, F3, F4, F5, F6, F7, F8, FC1, FC2, FC3, FC4, FC5, FC6, Fp1, Fp2, Fpz, FT10, FT7, FT8, FT9, Fz, O1, O2, Oz, P1, P2, P3, P4, P5, P6, P7, P8, PO3, PO4, PO7, PO8, POz, Pz, T7, T8, TP10, TP7, TP8, and TP9. The AFz and FCz electrodes were used as ground and reference electrodes, respectively. Data were recorded at a 5,000 Hz sampling frequency. A SyncBox (Brain Products, Munich, Germany) was used to synchronise the scanner clock with the EEG system, to ensure a precise gradient marker estimation for correct gradient artefact removal. The recorded EEG data were cleaned from gradient (GA) and the cardioballistic artefacts (BCG) using BrainVisionAnalyzer (Brain Products, Munich, Germany), following procedures described by Allen et al. (1998, 2000). The GA was removed using the average artefact subtraction method (AAS). We visually compared the two approaches using the recorded scanner volume markers (TR = 2500 ms) and the estimated volume markers with a constant interval of 125 ms. The latter resulted in better GA removal and thus better signal-to-noise ratio, which we then adopted. The template artefact for GA correction was constructed using the default 21 intervals for sliding average calculation. Following the GA removal, the ECG channel was used to find the R-peaks semiautomatically with BrainVisionAnalyzer to correct for BCG artefacts. We used a range of 40-90 pulses per minute for the detection, and adjusted this range for higher values for certain subjects with a higher heart rate. All the marked R-peaks were visually inspected for all subjects and manually adjusted if necessary. Following the BCG correction, the EEG channels were high-pass filtered at 0.5 Hz with a 50 Hz notch filter to remove line noise. The Infomax Independent Component Analysis (ICA) was fitted to detect and remove eye-related artefacts (blinks and saccades) manually. Bad channels were identified manually, rejected and interpolated. All channels were then re-referenced to the average reference. Trials were visually inspected for remaining artefacts and rejected if necessary (M Acquisition = 1.82 trials, SD Acquisition = 2.46; M Extinction = 0.88 trials, SD Acquisition = 1.02). One participant in the acquisition and two participants in the extinction were rejected at this stage due to more than half of either CS+ or CS− trials being noisy. Using the wavelet method (5-cycle width) as implemented in the FieldTrip package (version 20220104; Oostenveld et al., 2011), participants’ time-frequency representations (TFR) were computed from -4 to 8 s (in steps of 0.05 s) and from 2 to 30 Hz (in steps of 0.5 Hz). All the trials were then dB-normalised with a baseline period of -2.3 to -1.3 s relative to CS onset. The per-subject frontal-midline theta values were defined as the average of F1, Fz, and F2 electrodes in the range of 4 to 8 Hz (Mueller et al., 2014; Sperl et al., 2019). They were computed for 3 distinct trial time segments relative to CS onset: 0 to 2 s, 2 to 4 s, and 4 to 5.5 s. The last segment was chosen to be 0.5 s shorter to avoid potentially confounding effects of the US, which was delivered at 5.9 s post-CS onset. The per-subject frontal-midline theta averages for 16 CS+ and 16 CS− trials, respectively, were exported for further analysis and modelling with the fMRI General Linear Model (GLM). These nearly 2-second-long windows of the theta power also allowed us to link the EEG theta signal to the fMRI BOLD signal. To this end, we incorporated respective theta averages as a parametric modulation regressor in whole-brain fMRI general GLM analyses, allowing us to link transient electrophysiological activity with slower-evolving BOLD responses.To ensure high data quality in all modalities, we visually inspected whether participants’ data exhibited strong movement artefacts in EEG and/or fMRI, or poor ECG quality. As a result, a total of 15 participants’ fear acquisition data and 11 participants’ fear extinction data had to be excluded from further analyses. In addition to the main analysis described above, we performed an exploratory analysis, in which we split fear acquisition and extinction training into early (first half) and late (second half) sub-phases and tested for differential theta for CS+ vs. CS− separately for early and late acquisition and early and late extinction. These results are reported in detail in the supplementary materials. 2.5 (f)MRI recording, analysis and EEG-fMRI integration For coregistration, T1-weighted high-resolution MP-RAGE anatomical images were acquired with the following parameters: TR = 8.2 ms, TE = 3.7 ms, flip angle = 8°, 220 slices, matrix size = 240 mm x 240 mm, resolution = 1 mm x 1 mm x 1 mm. Scanning time was around 6 minutes. fMRI BOLD responses during fear acquisition and extinction training were obtained using echo planar imaging with the following parameters: TR = 2500 ms, TE = 35 ms, flip angle = 90°, 40 slices, matrix size = 112 mm x 112 mm, resolution = 2 mm x 2 mm x 3 mm. The scanning time of fear acquisition training was around 8 minutes and 4 minutes for extinction training. BOLD data were preprocessed using FEAT from the FSL toolbox (http://www.fmrib.ox.ac.uk/fsl, version 6.0.1). Preprocessing steps included correcting for head motion (MCFLIRT) and slice timing, with a 6 mm FWHM Gaussian kernel for spatial smoothing and a high-pass filter with a cutoff of 50 s. Each BOLD image was linearly registered to the participant’s high-resolution T1-weighted anatomical scan, followed by linear registration to the Montreal Neurological Institute (MNI) standard template with 12 degrees of freedom. A total of four participants with strong movement or suboptimal coverage were removed from consecutive analyses. For the whole-brain EEG-informed fMRI modelling, we implemented two distinct approaches: (1) we used per-subject theta averages (collapsed across trials within condition and time segment) as parametric modulation (PM) regressors at the first level; and (2) we used the same per-subject average theta values as covariates at the second level, following the method outlined by Sperl et al. (2019). As a reference, we also conducted a conventional fMRI analysis without incorporating EEG data. FEAT preprocessing parameters were identical across all approaches. In each model, we included the following task-related regressors in the GLM: context; fixation cross following a reinforced CS+ trial (acquisition only); fixation cross following an unreinforced CS+ trial; fixation cross following a CS− trial; and onsets of CS+ and CS− trials. The last two regressors were contrasted at the first-level GLM to examine condition-related BOLD differences in both EEG-fMRI approaches and the fMRI-only approach. The specifics and the key differences of the two EEG-driven-fMRI analyses are described below. In the first approach, we averaged the frontal-midline theta values within CS+ and CS− trials for each subject and used the average CS+ and average CS− theta values (i.e., one value per condition per subject per trial segment) as PM regressors on the first-level GLM (see Figure 2). Specifically, we ran three distinct analyses for fear acquisition and extinction training for the following three post-CS time windows: 0–2 s, 2–4 s, and 4–5.5 s. The event onsets in FEAT were shifted to match the trial timing from which the EEG data were taken. The event duration was set to 0.1 s, consistent with the event-related design. The regressors for CS+ and for CS− were convolved with the canonical hemodynamic response function (HRF) and modulated by each participant’s theta value. All second-level analyses were performed using FLAME (FMRIB’s Local Analysis of Mixed Effects) to estimate random effects. The second approach followed the EEG-fMRI integration methodology outlined by Sperl et al. (2019), using subject-level, trial-segment-averaged theta value differences as a second-level covariate for the CS+ > CS− and CS− > CS+ first-level contrasts. As in the first approach, we ran three first-level models for fear acquisition and extinction training, corresponding to the 0–2 s, 2–4 s, and 4–5.5 s trial segments, with the difference that we did not use the EEG theta values as PM regressors. For the second-level group analysis, we entered the mean normalised CS+/CS− difference theta values per subject as a covariate for the corresponding BOLD contrast. This approach did not yield any significant results for any of the contrasts and thus is not reported in the results section below. After first-level analysis the second-level models (group level) were analysed with FLAME 1, applying cluster thresholding with a Z-statistic threshold of 3.1 (Eklund et al., 2016) and p < .05. The significant clusters were mapped to brain areas using the Jülich Histological Atlas and the Harvard-Oxford Cortical and Subcortical Atlases, as implemented in FSL. The significant statistical maps are visualised using the MRIcroGL tool (version 2022-07-20; Rorden & Brett, 2001). jabbrv-ltwa-all.ldf jabbrv-ltwa-en.ldf Figure 2 . Overview of the EEG-fMRI analysis and integration pipeline as a first-level parametric modulation regressor in fMRI GLM, referred to as a first approach. The figure outlines analysis steps for a single participant, which are repeated for all subjects before the group analysis. Please note that for visualisation purposes, the topographic plot, the differential time-frequency plot, and the average of trial-segments plot show the grand-average of all subjects during extinction training; however, they were modelled with fMRI GLM on the first level in the actual analysis. (A) marks the start of EEG-analysis part; (B) gradient-, cardioballistic-, and ocular artefacts correction and filtering; (C) time-frequency decomposition was performed on data averaged across frontal-midline electrodes (F1, Fz, F2); (D) theta power (4–8 Hz) was averaged across three trial segments post-CS onset (0–2 s, 2–4 s, 4–5.5 s) for CS+ and CS− conditions separately (E). The individual participants’ average theta power values were convolved with the hemodynamic response function (HRF), separately for CS+ and CS− trials, to generate parametric regressors; (F) simultaneously acquired fMRI data were preprocessed; (G) EEG-theta derived regressors were included in general linear models (GLMs) as predictors; (H) resulting first-level statistical maps per participant are then analysed on the group level separately for each trial-segment. 2.7 Statistical analyses All statistical analyses were conducted using the SciPy (Virtanen et al., 2020), Statsmodels (Seabold and Perktold, 2010) and Pingouin (Vallat, 2018) Python packages. The normality of the mean differential frontal-midline EEG theta power within each trial segment and the mean differential SCR data per trial was tested with Shapiro-Wilk tests and additionally inspected using Q-Q plots. Variables and time points that violated the assumption of normality were modelled with one-sided Wilcoxon signed-rank tests, for CS+ > CS−. The data points that did not violate the assumption of normality were modelled with one-sided t -tests, also for CS+ > CS−. All p -values from parametric and non-parametric post-hoc tests were corrected with Benjamini-Hochberg false discovery rate (FDR) correction (Benjamini and Hochberg, 1995). The ANOVA results for points that violated the assumption of sphericity were corrected with the Greenhouse-Geisser correction. 3. RESULTS 3.1 Contingency self-report scores The analysis of self-report scores obtained after fear acquisition training revealed successful fear learning. Participants reported receiving an average of 10.12 ( SD = 3.41) electrical stimulations, aligning with the 10 stimulations over 16 CS+ trials. Likewise, the average reported percentage of electrical stimulation (US) received after the CS+ was 54.5 % ( SD = 18.19) and 7.98 % ( SD = 15.53) after CS−. This difference was statistically significant (W = 27.5, p < .001) and the mean reported values closely approximated the actual reinforcement rates of 62.5% for CS+ and 0% for CS−. For extinction training, all participants correctly reported that they received 0 electrical stimulations. jabbrv-ltwa-all.ldf jabbrv-ltwa-en.ldf 3.2 Skin conductance responses (SCRs) A repeated-measures ANOVA with Greenhouse-Geisser correction was conducted to assess the effects of CS type (CS+ vs. CS−), time (Early vs. Late), and their interaction on SCRs during fear acquisition training. The analysis revealed a significant main effect of CS type, F (1, 35) = 15.93, p < .001, partial η² = .057, and a significant main effect of time, F (1, 35) = 13.79, p < .001, partial η² = .030. However, the interaction between CS type and time was not significant, F (1, 35) = 0.64, p = .427, partial η² < .001 (see Figure 3, left). Follow-up Wilcoxon signed-rank tests showed that SCRs were significantly greater for CS+ compared to CS− during both the early (W = 71.0, p corr < .001) and the late ( W = 106.0, p corr = .003) fear acquisition training. For extinction training, analogous repeated-measures ANOVA showed no significant main effect of CS type, F (1, 42) = 1.01, p = .320, partial η² = .002, which is consistent with successful fear extinction learning. A significant main effect of time was observed, F (1, 42) = 9.11, p = .004, partial η² = .051, indicating that SCRs decreased across extinction training. The CS type × time interaction was not significant, F (1, 42) = 0.02, p = .890, partial η² < .001. Wilcoxon tests during extinction showed no significant difference between CS+ and CS− responses in either the early ( W = 411, p corr = .801) or late ( W = 250, p corr = .256; see Figure 3, right) extinction training. Figure 3. Skin conductance responses (SCRs) for CS+ and CS− during early and late fear acquisition and extinction training. The error bars indicate 95% confidence intervals. Asterisks indicate significant differences (*** p < .001, ** p < .01, n.s. not significant). 3.3 EEG: frontal-midline theta power During fear acquisition training, a 2 (CS type: CS+, CS−) × 3 (trial segment: 0–2 s, 2–4 s, 4–5.5 s) repeated-measures ANOVA on frontal-midline theta power revealed a significant main effect of CS type, F (1, 34) = 7.95, p = .008, partial η² = .070, and a significant interaction between CS type and trial segment, F (2, 68) = 4.57, p = .013, partial η² = .014 (see Figure 5, left). The main effect of trial segment was not significant, F (2, 68) = 0.35, p = .708, partial η² = .001. Follow-up paired-sample t -tests comparing CS+ and CS− responses across successive trial segments indicated a progressively increasing difference from early to late trial segments: 0–2 s : t (34) = 1.44, p corr = .158; 2–4 s : t (34) = 2.67, p corr = .012; 4–5.5 s : W = 150.0, p corr = .017. This suggests that differentiation between CS+ and CS− was not yet significant at the early trial segment (0–2 s) but became statistically significant and even stronger at later trial segments. To further examine whether the CS+ vs. CS− difference in theta power increased over time (“ramping-up effect”), we computed per-subject difference scores (CS+ minus CS−) separately for each time window (0–2 s, 2–4 s, and 4–5.5 s). We then performed paired one-tailed t-tests between consecutive time windows to test for an increase in effect size over time. During the acquisition phase, the CS+ vs. CS− difference increased from 0–2 s to 2–4 s ( t (34) = 1.48, p corr = .074), from 2–4 s to 4–5.5 s (W = 413.0, p corr = .074), and significantly from 0–2 s to 4–5.5 s ( t (34) = 2.67, p corr = .017). These results are in accordance with the theta ramping-up effect described by Starita et al. (2023) during fear acquisition. Figure 4. Topographic distribution of EEG theta power (4–8 Hz) during fear acquisition training and extinction learning. Scalp maps are shown separately for CS+, CS−, and their difference (CS+ − CS−) across three trial time windows (0–2 s, 2–4 s, and 4–5.5 s post-CS onset). The three black dots indicate the F1, Fz, and F2 electrodes, which define the frontal-midline region used for averaging and subsequent analyses. Colour intensity represents theta power in decibels (dB), with consistent colour scales across time windows within each condition to allow comparability (ranges from -0.5 to 1 for the differential (CS+) - (CS−) topographies, and from -3.5 to -1.75 for CS+ and CS− in both acquisition and extinction). Note that although frontal and posterior midline activity is clearly visible, fixed colourbar limits (to ensure comparability) may cause some of these effects to appear visually attenuated. For extinction training (Figure 5, right) an analogous 2 × 3 repeated-measures ANOVA revealed no significant main effect of CS type, F (1, 38) = 2.21, p = .145, partial η² = .018, no significant main effect of trial segment, F (2, 76) = 1.04, p = .359, partial η² = .002, and no CS type × trial segment interaction, F (2, 76) = 0.58, p = .565, partial η² = .001. Follow-up paired-sample t -tests comparing CS+ and CS− responses at each trial segment revealed no statistically significant differences, although effect sizes increased across time: 0–2 s : t (38) = 1.06, p corr = .294; 2–4 s : t (38) = 1.32, p corr = .293; 4–5.5 s : t (38) = 1.63, p corr = .293. Complementary ramp-up analyses revealed no significant increase in the CS+ vs. CS− difference across time windows during extinction ( 0–2s to 2–4 s : t (38) = 0.35, p corr = .366; 2–4 s to 4–5.5 s : t (38) = 0.71, p corr = .362; 0–2 s to 4–5.5 s : t (38) = 0.984, p corr = .362). Figure 5. dB-normalised EEG theta dynamics from frontal-midline electrodes (F1, Fz, and F2) during fear acquisition and extinction training across three trial segments (0 to 2, 2 to 4, and 4 to 5.5 s post-CS). The error bars indicate 95% confidence intervals. The significant CS+ > CS− differences in specific trial segments, and the significant increase in differential theta effect between trial segments are marked with asterisks (* p < .05). Prior studies suggest that participants’ behaviour as well as neurocognitive correlates change from early to late fear acquisition and extinction training (for example, see Åhs et al., 2015; Graner et al., 2020). To further investigate whether differential CS+/CS− theta power is more prominent during early or late fear acquisition and extinction training, respectively, we split the trials into early and late (see Supplementary Figure S1). We found the difference between CS+ and CS− to be significant for the early but not the late fear acquisition training. In the case of extinction training, we observed the opposite pattern: the difference between the CS+ and CS− was significant in the late, but not in the early extinction training. 3.4 Conventional fMRI analysis Conventional fMRI analysis revealed a significant activation in the dACC for the CS+ > CS− contrast in the acquisition training, with the peak activation voxel at x = 2; y = 22; z = 32, and a z-statistic value at peak voxel 4.21 (see Supplementary Figure S2). No other contrasts were significant as a result of this analysis, i.e., no CS− > CS+ contrasts, neither in acquisition nor in extinction and also not CS+ > CS- contrast in extinction. 3.5 Simultaneous EEG-fMRI The EEG-driven fMRI analysis revealed significant theta-BOLD co-activation spanning the known fear and safety networks during fear acquisition (see Figure 6, and Table 1) and extinction training (see Figure 6, and Table 1). During fear acquisition training, for the trial segment at 0–2 s post CS onset, we did not find any significant theta-BOLD co-activation. For the trial segment at 2–4 s, the bilateral cuneal and precuneal cortices, as well as regions from the left visual cortex (V2-V4), were significantly co-active for the CS+ > CS− contrast (see Figure 6, and Table 1). For the trial segment at 4–5.5 s, a significant cluster in the right primary somatosensory cortex was co-active with the EEG theta for the CS+ > CS− contrast (see Figure 6 and Table 1). For the CS− > CS+ contrasts we did not find significant co-activation in all trial segments. Figure 6. EEG-theta-BOLD co-activation for CS+ > CS– during fear acquisition training, across the 2 to 4 s and 4 to 5.5 s trial segments, plotted on the transparent MNI152 template. Red-to-yellow (left colourbar) indicates a significant cluster during the 2–4 s trial segment for the CS+ > CS− contrast. Blue-to-green (right colourbar) indicates a significant cluster during the 4–5.5 s trial segment for the CS+ > CS− contrast. For the 0 to 2 s trial segment, no significant theta-BOLD co-activation occurred. The colour bars code the thresholded z-statistic value from 3.1 to 4.1. Table 1. Overview of the MNI coordinates for the peak cluster activation points related to theta-BOLD co-activation . The areas highlighted in bold are the brain regions according to anatomical atlases (Jülich Histological Atlas and the Harvard-Oxford Cortical and Subcortical Atlases). vmPFC = ventromedial prefrontal cortex. MNI Coordinates Phase Time Contrast Area x y z z-stat Acquisition 0 to 2 s CS+ > CS– NOT SIGNIFICANT - - - - Acquisition 0 to 2 s CS– > CS+ NOT SIGNIFICANT - - - - Acquisition 2 to 4 s CS+ > CS– Left and Right Primary Motor Cortex (alternatively Left and Right Medial portions of Pre- Postcentral Gyrus; Extending to Right Posterior Cingulate) 1.9 -35.6 57.9 4.4 Left V2-V4 -25.5 -54.0 -5.8 4.3 Right Inferior parietal lobule PGp 46.3 -72.2 34.6 4.3 Left Inferior parietal lobule PGp / 7A -22.4 -76.2 34.3 4.2 Left Cuneal Cortex / Left Precuneus Cortex -12.3 -72.0 30.0 4.0 Acquisition 2 to 4 s CS– > CS+ NOT SIGNIFICANT - - - - Acquisition 4 to 5.5 s CS+ > CS– Right Primary Motor Cortex (alternatively Left and Right Medial portions of Postcentral Gyrus) 4.2 -19.7 53.9 4.2 Left Inferior parietal lobule PGp / 7A -21.9 -74.0 32.1 4.7 Acquisition 4 to 5.5 s CS– > CS+ NOT SIGNIFICANT - - - - Extinction 0 to 2 s CS+ > CS– NOT SIGNIFICANT - - - - Extinction 0 to 2 s CS– > CS+ Left Precentral Gyrus -24.0 -10.3 49.9 3.9 Extinction 2 to 4 s CS+ > CS– Left vmPFC -11.5 40.3 -4.2 4.1 Extinction 2 to 4 s CS– > CS+ NOT SIGNIFICANT - - - - Extinction 4 to 5.5 s CS+ > CS– NOT SIGNIFICANT - - - - Extinction 4 to 5.5 s CS– > CS+ NOT SIGNIFICANT - - - - jabbrv-ltwa-all.ldf jabbrv-ltwa-en.ldf During extinction training, we found for the trial segment 2–4 s that the vmPFC exhibited a theta-BOLD co-activation for the contrast CS+ > CS− (see Figure 7, and Table 1). In the case of the vmPFC, this might point towards a role of frontomedial theta in the extinction of aversive memories. We did not observe significant theta-BOLD co-activations for the 0–2 s nor for the 4–5.5 s for CS+ > CS− contrast. Figure 7. EEG-theta-BOLD co-activation for CS+ > CS– contrast during extinction training in the 2 to 4 s trial segment, plotted on the transparent MNI152 template. Red-to-yellow indicates a significant cluster during the 2–4 s trial segment for CS+ > CS−. This cluster spans across the left vmPFC. No significant effects occurred in the other trial segments. The colour bar codes the thresholded z-statistic value from 3.1 to 4.1. We also found a significant theta-BOLD co-activation for the CS− > CS+ contrast in the extinction training at 0−2 s trial segment in the left pre-central gyrus (see Figure 8, and Table 1). The CS− > CS+ contrasts in other trial segments were not significant. Figure 8. EEG-theta-BOLD co-activation for CS– > CS+ contrast during extinction training in the 0 to 2s trial segment, plotted on the transparent MNI152 template. Red-to-yellow indicates a significant cluster during the 0–2 s trial segment for CS− > CS+. This cluster is centered at the left precentral gyrus. No significant effects occurred in the other trial segments. The colour bar codes the thresholded z-statistic value from 3.1 to 4.1. 4. DISCUSSION In this study, we investigated the temporal dynamics of frontal-midline theta during fear acquisition and extinction learning using simultaneous EEG-fMRI for the first time. By discretising each trial into three segments, we first confirmed the effects reported by Starita et al. (2023) of theta power differences between CS+ and CS− ramping-up during fear acquisition trials, and we extended these findings by showing theta dynamics across trial segments during extinction learning. By using individual subject theta values as parametric modulation regressors, we found that during fear acquisition training, theta-BOLD co-activations shifted from parietal and occipital areas (2–4 s post-CS) to motor regions (4–5.5 s), mirroring threat stimulus encoding. During extinction training, theta-driven BOLD coupling was confined to the vmPFC in the mid-segment (2–4 s), pointing to a temporally specific role for frontal-midline theta in updating aversive memory representations. Together, these results confirm our assumptions about the temporal specificity of functional activation of different brain structures during fear conditioning. Self-report ratings and SCRs both confirmed that participants successfully acquired and extinguished conditioned fear. As expected, SCR results showed a robust CS+ > CS− differentiation during early and late acquisition and no effect in both halves of extinction. While we did not find a significant CS+ > CS− effect in the EEG theta averaged across all trials during extinction training, an exploratory approach, analysing the early and late EEG trial segments separately, showed a similar theta ramping-up effect during late but not early extinction training (see Supplementary Materials Figure S1), with the 4–5.5 s trial segment being significant for CS+ > CS−.Our interpretation of this finding is that participants have to experience the absence of a US in the new context for several trials before recognizing the shift; by late extinction trials (trials 4–8), their fear response is suppressed and/or a new safety memory is being formed, as indicated by a return of the theta ramping-up effect. Hence, we propose that this theta effect facilitates CS-US association during early acquisition training and supports the formation of a new safety memory trace during late extinction training. Since this effect emerged only after separating the early and late phases, we speculate that this reflects dynamic nature of extinction learning, where initial associations must be inhibited while new ones are learned (Bouton, 2002). This is likely because the increase in theta power (CS+ > CS−) was predominantly present only during late extinction training. The idea of the transient nature of fear and extinction learning, taking place in a matter of just a few trials, has also been discussed in a recent review by Anders et al. (2024). Similarly, after splitting fear acquisition training into early and late halves in EEG analysis, it became apparent that the effect of differential theta ramping-up across the trial segments appears rather in the early half, and is not significant in the late half. This finding is in line with the results of Clarke et al. (2018), who report that frontal-midline theta power drops as association strength increases. As suggested by Starita et al. (2023), the theta ramping-up effect during fear learning might suggest a temporal coding of the threat stimuli (i.e., threat expectation). However, based on earlier electrophysiological work in animal models (Karalis et al., 2016; Likhtik et al., 2014; Seidenbecher et al., 2003) and our own (exploratory) finding that frontal-midline theta is significantly different in the early but disappears in the late fear acquisition training, we hypothesise that theta functions as a wiring mechanism – facilitating the formation of the CS+/US association, despite their onsets being temporally disconnected. On the other hand, we acknowledge that the absence of a differential CS+/CS− effect for EEG theta values during late fear acquisition training might also be attributed to a mere habituation of the participants to the US, but the significant differential CS+/CS− effect of SCR data during early and late fear acquisition training counters this alternative interpretation. Albeit the limited trial numbers per early and late experimental halves warrant caution, this differential CS+/CS− effect during late fear extinction training might indicate a potential role of frontal-midline theta in fear regulation and suppression that goes beyond threat expectation. The findings of the conventional fMRI-only analysis showed a rather classical map, consistent with earlier studies (see Fullana et al., 2016 or 2018, for example) for the dACC being active for the CS+ > CS− in acquisition. With EEG-driven fMRI analysis, we found distinct activation patterns across the brain at different trial segments during fear acquisition and extinction training for CS+ > CS− contrast. Specifically, activation was observed in the left cuneal and precuneal regions, the left and right segments of the inferior parietal lobule, as well as the right visual and bilateral motor cortices at 2 to 4 seconds during fear acquisition training. The theta-modulated BOLD responses at the 4–5.5 second trial segment were exhibited in the right primary motor cortex and the left inferior parietal lobule. This partially overlapped with the activation observed at the 2–4 second segment; however, other structures active at 2–4 seconds were no longer engaged. The involvement of cuneal and precuneal cortices during the 2–4 s trial segment, together with visual cortex structures, might indicate multimodal sensory integration and again confirm the interpretation of Starita et al. (2023) on the role of theta in threat expectation during fear learning. However, in contrast to our findings, this area has primarily been reported in previous studies during extinction training (Phelps et al., 2004; Fullana et al., 2016; Wen et al., 2021). For instance, Wen et al. (2021) interpreted precuneus activation - part of the default mode network (DMN) - as reflecting the integration of otherwise segregated sensory and cognitive systems to support the complex demands of safety learning. However, as suggested by Anders et al. (2024), the functional role of brain structures previously believed to be specific to threat or safety processing may need to be re-evaluated. For example, Battaglia et al. (2020) reported that participants with vmPFC lesions failed to produce conditioned physiological responses as a result of acquisition training - suggesting a re-evaluation of the role of vmPFC beyond extinction training. Though our findings do not overlap with the source-localised regions reported by Starita et al (2023), this discrepancy is not surprising since we explored the co-activation of theta values and the BOLD signal across all voxels in the whole brain. Likewise, by using EEG theta as a parametric modulation regressor in the fMRI BOLD analysis, we aimed to directly link the two imaging modalities, whereas source localisation techniques (as applied by Starita et al., 2023) serve a different purpose - mapping scalp-level EEG activity to the anatomical locations of its underlying generators. Importantly, the shift from parietal and occipital brain structures at 2–4 s to motor regions at 4–5.5 s might reflect the initiation of motor responses for fight-or-flight behaviour as an expectation of threat (see Mobbs et al., 2007; 2009). During extinction training, we found a significant theta-modulated BOLD response in the vmPFC solely during the 2–4 s trial segment, which highlights a potential temporal specificity of vmPFC activation relative to CS onset. Taken together with our finding that the differential theta power effects emerged during the latter half of extinction training, these results may align with the 6–9 Hz effects reported by Totty et al. (2023), which were shown to mediate mPFC-hippocampal connectivity in mice. Additionally, the inconsistency of vmPFC activation during extinction learning (Anders et al., 2024; Fullana et al., 2018) may be attributable to the temporal specificity of vmPFC engagement. Thus, our data suggest that studies failing to detect vmPFC involvement during extinction likely used analysis windows that did not encompass the critical interval at 2–4 s post-CS onset in which theta-modulated vmPFC activation emerges. A recent review of non-invasive brain stimulation studies in fear conditioning by Zhang et al. (2025) emphasises the role of the vmPFC in facilitating extinction learning, while the dlPFC, which is interconnected with the vmPFC, also plays a key role in reducing fear responses. Interestingly, the theta-BOLD co-activation in the precentral gyrus at the 0–2 s trial segment of extinction training was the only significant effect among all trial segments across both experimental phases for the CS− > CS+ contrast. Given the heightened SCR to CS− trials in early extinction, this effect may reflect an expectancy of CS-US contingency reversal in a novel context. Taken together with the robust theta-BOLD co-activation for CS+ > CS− and the timing of these effects, our findings suggest that frontal midline theta is more closely tied to threat-related processing than to safety signalling. Although this study aimed to combine the temporal precision of EEG with the spatial resolution of fMRI to leverage the strengths of both modalities, each method individually presents challenges for directly linking EEG signals to the BOLD response (for a review please see Abreu et al., 2018). Of particular importance is the low signal-to-noise ratio of the EEG and the necessity of having a large number of repetitions, which are then averaged to enhance the signal quality (Luck, 2014). While this method works well in sensory-motor experiments, learning processes, which are characterized by a dynamic change between trials during the course of each learning phase, might be too transient and be averaged out over the course of many repetitions (Sperl et al., 2021), and participants may even habituate to the aversiveness of the stimuli (Sperl et al., 2016). Earlier work in the associative learning experiment by Hanslmayr et al. (2011), who attempted to relate trial-by-trial theta activation to the BOLD signal with parametric modulation, did not yield any significant results. Therefore, to establish a direct relationship between the BOLD signal and EEG spectral components, we averaged across all trials for each participant, trading off across-trial temporal resolution in favour of capturing within-trial dynamics in the EEG-fMRI integration. Also, our additional approach to adapt the methodology of Sperl et al. (2019) - linking EEG theta activity to the BOLD signal at the second-level group analysis - did not yield any significant results. However, in light of our combined EEG-fMRI results and the previous work using EEG (for example, Starita et al., 2023), we see the necessity of studying the EEG-fMRI relationship on a longer time scale, beyond the short periods post CS-onset, as the activation pattern changes dynamically throughout the trial. By adopting the approach from Starita et al. (2023), we artificially split data into three 2-second segments, but the true theta-BOLD relationship might be on even narrower time segments. This warrants future research to explore indirect fusion methods for the two modalities, applying techniques such as representational similarity analysis (Kriegeskorte et al., 2008). Specifically, this method could enable the linking of similarity measures from EEG data on time scales both between and within trials - shorter than the 2 s trial segments - while also allowing for a division of trials into early and late phases. On the other hand, an extension of our approach of EEG-fMRI integration as a parametric modulation, but with EEG-theta averaged from temporally shorter segments, could possibly reveal even finer, temporally specific activation in the BOLD signal. Likewise, future studies combining EEG and fMRI simultaneously during fear acquisition and extinction training at higher field strengths, which offer higher spatial resolution, may help clarify the directionality of the effects we reported. For example, whether vmPFC activation - similar to results from animal models - enables the extinction of fear memories by suppressing amygdala activation, remains elusive, and region of interest (ROI) analysis, as opposed to whole brain analyses in our study, could help in answering this question. Lastly, future fMRI-only studies in a bigger sample might find distinct activation patterns across the brain at different trial segments relative to the CS-onset. This could potentially untangle the brain structures comprising the fear- and safety-networks and link activation of individual brain structures to specific time-points throughout the experiment. In conclusion, this study demonstrates that frontal-midline theta power during fear and extinction learning reflects temporally dynamic neural processes that are linked to distinct brain regions. Ramping of theta power (for CS+ relative to CS− stimulation) corresponds to somatosensory engagement during threat anticipation in acquisition training, while theta-vmPFC co-activation during extinction learning suggests involvement in updating aversive memory representations. Our findings highlight the value of combining EEG and fMRI to uncover the spatiotemporal mechanisms underlying both fear and extinction learning. Future work may build on this foundation by investigating neuromodulatory or clinical interventions targeting these processes. Author contribution Arslan Gabdulkhakov: methodology; software; data processing, analysis and visualization; writing - original draft, review and editing. Matthias F. J. Sperl: methodology; investigation; writing - review and editing. Christian J. Merz: methodology; experimental design; SCR analysis; writing - review and editing. Laura-Isabelle Klatt: EEG data curation; writing - review and editing Christoph Fraenz: data collection and curation; writing - review and editing. Erhan Genç: project administration; supervision; methodology; conceptualization; funding acquisition; data curation; writing - review and editing Acknowledgements We would like to thank our student assistants, Patrick Friedrich and Helene Selpien, for supporting data recordings. We also would like to thank Tobias Otto and Dorothea Metzen for assistance with the analysis of SCR data and Marie-Christin Fellner for the establishment of simultaneous EEG/fMRI recordings. Funding information The project was funded by the Deutsche Forschungsgemeinschaft (DFG) grant (project number 31680338), as part of the A03 project in the SFB 1280 Extinction Learning Collaborative Research Centre. Conflict of Interest The authors declare no conflict of interest. Data Availability Statement The preprocessed EEG, EDA, cotingency ratings and the fMRI statistical maps are available at https://osf.io/w4vsh. Raw data and other materials are available upon request. Use of AI Generated Content (AIGC) and tools We used ChatGPT (OpenAI) to improve grammar, enhance clarity, and check for cohesion at the sentence level in the manuscript. ORCID Arslan Gabdulkhakov https://orcid.org/0000-0003-4857-3601 Matthias F. J. Sperl https://orcid.org/0000-0001-9525-0368 Christian J. Merz https://orcid.org/0000-0001-5679-6595 Laura-Isabelle Klatt https://orcid.org/0000-0002-5682-5824 Christoph Fraenz https://orcid.org/0000-0003-0653-2385 Erhan Genç https://orcid.org/0000-0001-6514-5479 REFERENCES Abreu, R., Leal, A., & Figueiredo, P. (2018). EEG-Informed fMRI: A review of data analysis methods. Frontiers in Human Neuroscience , 12 , 29. https://doi.org/10.3389/fnhum.2018.00029 Åhs, F., Kragel, P. A., Zielinski, D. J., Brady, R., & LaBar, K. S. (2015). Medial prefrontal pathways for the contextual regulation of extinguished fear in humans. NeuroImage , 122 , 262–271. https://doi.org/10.1016/j.neuroimage.2015.07.051 Allen, P. J., Polizzi, G., Krakow, K., Fish, D. R., & Lemieux, L. (1998). Identification of EEG events in the MR scanner: The problem of pulse artifact and a method for its subtraction. NeuroImage , 8 (3), 229–239. https://doi.org/10.1006/nimg.1998.0361 Allen, P. J., Josephs, O., & Turner, R. (2000). A method for removing imaging artifact from continuous EEG recorded during functional MRI. NeuroImage, 12(2), 230–239. https://doi.org/10.1006/nimg.2000.0599 Andres, E., Meyer, B., Yuen, K. S. L., & Kalisch, R. (2024). Current state of the neuroscience of fear extinction and its relevance to anxiety disorders. In Current Topics in Behavioral Neurosciences . Springer. https://doi.org/10.1007/7854_2024_555 Battaglia, S., Garofalo, S., Pellegrino, G. di, & Starita, F. (2020). Revaluing the role of vmPFC in the acquisition of Pavlovian threat conditioning in humans. Journal of Neuroscience , 40 (44), 8491–8500. https://doi.org/10.1523/JNEUROSCI.0304-20.2020 Behroozi, M., Helluy, X., Ströckens, F., Gao, M., Pusch, R., Tabrik, S., Tegenthoff, M., Otto, T., Axmacher, N., Kumsta, R., Moser, D., Genc, E., & Güntürkün, O. (2020). Event-related functional MRI of awake behaving pigeons at 7T. Nature Communications , 11 (1), 4715. https://doi.org/10.1038/s41467-020-18437-1 Benjamini, Y., & Hochberg, Y. (1995). Controlling the False Discovery Rate: A practical and powerful approach to multiple testing. Journal of the Royal Statistical Society: Series B (Methodological) , 57 (1), 289–300. https://doi.org/10.1111/j.2517-6161.1995.tb02031.x Bierwirth, P., Antov, M. I., & Stockhorst, U. (2023). Oscillatory and non‐oscillatory brain activity reflects fear expression in an immediate and delayed fear extinction task. Psychophysiology , 60 (8), e14283. https://doi.org/10.1111/psyp.14283 Bierwirth, P., Sperl, M. F. J., Antov, M. I., & Stockhorst, U. (2021). Prefrontal theta oscillations are modulated by estradiol status during fear recall and extinction recall. Biological Psychiatry: Cognitive Neuroscience and Neuroimaging , 6 (11), 1071–1080. https://doi.org/10.1016/j.bpsc.2021.02.011 Blair, H. T., Schafe, G. E., Bauer, E. P., Rodrigues, S. M., & LeDoux, J. E. (2001). Synaptic plasticity in the lateral amygdala: A cellular hypothesis of fear conditioning. Learning & Memory , 8 (5), 229–242. https://doi.org/10.1101/lm.30901 Bouton, M. E. (2002). Context, ambiguity, and unlearning: Sources of relapse after behavioral extinction. Biological Psychiatry , 52 (10), 976–986. https://doi.org/10.1016/S0006-3223(02)01546-9 Brydges, N. M., Whalley, H. C., Jansen, M. A., Merrifield, G. D., Wood, E. R., Lawrie, S. M., Wynne, S.-M., Day, M., Fleetwood-Walker, S., Steele, D., Marshall, I., Hall, J., & Holmes, M. C. (2013). Imaging conditioned fear circuitry using awake rodent fMRI. PLOS ONE , 8 (1), e54197. https://doi.org/10.1371/journal.pone.0054197 Büchel, C., & Dolan, R. J. (2000). Classical fear conditioning in functional neuroimaging. Current Opinion in Neurobiology , 10 (2), 219–223. https://doi.org/10.1016/S0959-4388(00)00078-7 Burgos-Robles, A., Vidal-Gonzalez, I., & Quirk, G. J. (2009). Sustained conditioned responses in prelimbic prefrontal neurons are correlated with fear expression and extinction failure. Journal of Neuroscience , 29 (26), 8474–8482. https://doi.org/10.1523/JNEUROSCI.0378-09.2009 Chen, S., Tan, Z., Xia, W., Gomes, C. A., Zhang, X., Zhou, W., Liang, S., Axmacher, N., & Wang, L. (2021). Theta oscillations synchronize human medial prefrontal cortex and amygdala during fear learning. Science Advances , 7 (34), eabf4198. https://doi.org/10.1126/sciadv.abf4198 Chien, J. H., Colloca, L., Korzeniewska, A., Cheng, J. J., Campbell, C. M., Hillis, A. E., & Lenz, F. A. (2017). Oscillatory EEG activity induced by conditioning stimuli during fear conditioning reflects salience and valence of these stimuli more than expectancy. Neuroscience , 346 , 81–93. https://doi.org/10.1016/j.neuroscience.2016.12.047 Clarke, A., Roberts, B. M., & Ranganath, C. (2018). Neural oscillations during conditional associative learning. NeuroImage , 174 , 485–493. https://doi.org/10.1016/j.neuroimage.2018.03.053 DeLaRosa, B. L., Spence, J. S., Shakal, S. K. M., Motes, M. A., Calley, C. S., Calley, V. I., Hart, J., & Kraut, M. A. (2014). Electrophysiological spatiotemporal dynamics during implicit visual threat processing. Brain and Cognition , 91 , 54–61. https://doi.org/10.1016/j.bandc.2014.08.003 Eklund, A., Nichols, T. E., & Knutsson, H. (2016). Cluster failure: Why fMRI inferences for spatial extent have inflated false-positive rates. Proceedings of the National Academy of Sciences , 113 (28), 7900–7905. https://doi.org/10.1073/pnas.1602413113 Fullana, M. A., Albajes-Eizagirre, A., Soriano-Mas, C., Vervliet, B., Cardoner, N., Benet, O., Radua, J., & Harrison, B. J. (2018). Fear extinction in the human brain: A meta-analysis of fMRI studies in healthy participants. Neuroscience & Biobehavioral Reviews , 88 , 16–25. https://doi.org/10.1016/j.neubiorev.2018.03.002 Fullana, M. A., Harrison, B. J., Soriano-Mas, C., Vervliet, B., Cardoner, N., Àvila-Parcet, A., & Radua, J. (2016). Neural signatures of human fear conditioning: An updated and extended meta-analysis of fMRI studies. Molecular Psychiatry , 21 (4), 500–508. https://doi.org/10.1038/mp.2015.88 Kalisch, R., Korenfeld, E., Stephan, K. E., Weiskopf, N., Seymour, B., & Dolan, R. J. (2006). Context-dependent human extinction memory is mediated by a ventromedial prefrontal and hippocampal network. The Journal of neuroscience : the official journal of the Society for Neuroscience, 26(37), 9503–9511. https://doi.org/10.1523/JNEUROSCI.2021-06.2006 Karalis, N., Dejean, C., Chaudun, F., Khoder, S., Rozeske, R. R., Wurtz, H., Bagur, S., Benchenane, K., Sirota, A., Courtin, J., & Herry, C. (2016). 4-Hz oscillations synchronize prefrontal-amygdala circuits during fear behavior. Nature Neuroscience , 19 (4), 605–612. https://doi.org/10.1038/nn.4251 Kim, J. J., & Jung, M. W. (2006). Neural circuits and mechanisms involved in Pavlovian fear conditioning: A critical review. Neuroscience & Biobehavioral Reviews , 30 (2), 188–202. https://doi.org/10.1016/j.neubiorev.2005.06.005 Knight, D. C., Smith, C. N., Cheng, D. T., Stein, E. A., & Helmstetter, F. J. (2004). Amygdala and hippocampal activity during acquisition and extinction of human fear conditioning. Cognitive, Affective, & Behavioral Neuroscience , 4 (3), 317–325. https://doi.org/10.3758/CABN.4.3.317 Kriegeskorte, N., Mur, M., & Bandettini, P. A. (2008). Representational similarity analysis—Connecting the branches of systems neuroscience. Frontiers in Systems Neuroscience , 2 . https://doi.org/10.3389/neuro.06.004.2008 LaBar, K. S., Gatenby, J. C., Gore, J. C., LeDoux, J. E., & Phelps, E. A. (1998). Human amygdala activation during conditioned fear acquisition and extinction: A mixed-trial fMRI study. Neuron , 20 (5), 937–945. https://doi.org/10.1016/S0896-6273(00)80475-4 Likhtik, E., Stujenske, J. M., Topiwala, M. A., Harris, A. Z., & Gordon, J. A. (2014). Prefrontal entrainment of amygdala activity signals safety in learned fear and innate anxiety. Nature Neuroscience , 17 (1), 106–113. https://doi.org/10.1038/nn.3582 Lissek, S., Powers, A. S., McClure, E. B., Phelps, E. A., Woldehawariat, G., Grillon, C., & Pine, D. S. (2005). Classical fear conditioning in the anxiety disorders: A meta-analysis. Behaviour Research and Therapy , 43 (11), 1391–1424. https://doi.org/10.1016/j.brat.2004.10.007 Lonsdorf, T. B., Menz, M. M., Andreatta, M., Fullana, M. A., Golkar, A., Haaker, J., Heitland, I., Hermann, A., Kuhn, M., Kruse, O., Meir Drexler, S., Meulders, A., Nees, F., Pittig, A., Richter, J., Römer, S., Shiban, Y., Schmitz, A., Straube, B., … Merz, C. J. (2017). Don’t fear ‘fear conditioning’: Methodological considerations for the design and analysis of studies on human fear acquisition, extinction, and return of fear. Neuroscience & Biobehavioral Reviews , 77 , 247–285. https://doi.org/10.1016/j.neubiorev.2017.02.026 Luck, S. J. (2014). An introduction to the event-related potential technique (2nd ed.). MIT Press. Michel, C. M., Murray, M. M., Lantz, G., Gonzalez, S., Spinelli, L., & Grave de Peralta, R. (2004). EEG source imaging. Clinical Neurophysiology, 115(10), 2195–2222. https://doi.org/10.1016/j.clinph.2004.06.001 Milad, M. R., & Quirk, G. J. (2002). Neurons in medial prefrontal cortex signal memory for fear extinction. Nature , 420 (6911), 70–74. https://doi.org/10.1038/nature01138 Milad, M. R., & Quirk, G. J. (2012). Fear extinction as a model for translational neuroscience: Ten years of progress. Annual Review of Psychology , 63 (1), 129–151. https://doi.org/10.1146/annurev.psych.121208.131631 Milad, M. R., Wright, C. I., Orr, S. P., Pitman, R. K., Quirk, G. J., & Rauch, S. L. (2007). Recall of fear extinction in humans activates the ventromedial prefrontal cortex and hippocampus in concert. Biological Psychiatry , 62 (5), 446–454. https://doi.org/10.1016/j.biopsych.2006.10.011 Miskovic, V., & Keil, A. (2012). Acquired fears reflected in cortical sensory processing: A review of electrophysiological studies of human classical conditioning. Psychophysiology , 49 (9), 1230–1241. https://doi.org/10.1111/j.1469-8986.2012.01398.x Mobbs, D., Petrovic, P., Marchant, J. L., Hassabis, D., Weiskopf, N., Seymour, B., … & Frith, C. D. (2007). When fear is near: threat imminence elicits prefrontal-periaqueductal gray shifts in humans. Science, 317(5841), 1079-1083. https://doi.org/10.1126/science.1144298 Mobbs, D., Marchant, J. L., Hassabis, D., Seymour, B., Tan, G., Gray, M., … & Frith, C. D. (2009). From threat to fear: the neural organization of defensive fear systems in humans. Journal of Neuroscience, 29(39), 12236-12243. https://doi.org/10.1523/JNEUROSCI.2378-09.2009 Mueller, E. M., Panitz, C., Hermann, C., & Pizzagalli, D. A. (2014). Prefrontal oscillations during recall of conditioned and extinguished fear in humans. Journal of Neuroscience , 34 (21), 7059–7066. https://doi.org/10.1523/JNEUROSCI.3427-13.2014 Oostenveld, R., Fries, P., Maris, E., & Schoffelen, J.-M. (2011). FieldTrip: Open source software for advanced analysis of MEG, EEG, and invasive electrophysiological data. Computational Intelligence and Neuroscience , 2011 , 156869. https://doi.org/10.1155/2011/156869 Otto, T., Wolf, O.T. & Merz, C.J. (2023). EDA-Analysis App (5.11). Zenodo. https://doi.org/10.5281/zenodo.7965376 Phelps, E. A., Delgado, M. R., Nearing, K. I., & LeDoux, J. E. (2004). Extinction learning in humans. Neuron , 43 (6), 897–905. https://doi.org/10.1016/j.neuron.2004.08.042 Pitkanen, A. (2000). Connectivity of the rat amygdaloid complex. In J. P. Aggleton (Ed.), The Amygdala: A functional analysis (p. 0). Oxford University Press. https://doi.org/10.1093/oso/9780198505013.003.0002 Schiller, D., Levy, I., Niv, Y., LeDoux, J. E., & Phelps, E. A. (2008). From fear to safety and back: Reversal of fear in the human brain. Journal of Neuroscience , 28 (45), 11517–11525. https://doi.org/10.1523/JNEUROSCI.2265-08.2008 Seabold, S., & Perktold, J. (2010). Statsmodels: Econometric and statistical modeling with Python. In Proceedings of the 9th Python in Science Conference (pp. 92–96). https://doi.org/10.25080/Majora-92bf1922-011 Sehlmeyer, C., Schöning, S., Zwitserlood, P., Pfleiderer, B., Kircher, T., Arolt, V., & Konrad, C. (2009). Human fear conditioning and extinction in neuroimaging: A systematic review. PLoS ONE , 4 (6), e5865. https://doi.org/10.1371/journal.pone.0005865 Seidenbecher, T., Laxmi, T. R., Stork, O., & Pape, H.-C. (2003). Amygdalar and hippocampal theta rhythm synchronization during fear memory retrieval. Science , 301 (5634), 846–850. https://doi.org/10.1126/science.1085818 Smith, S. M., Jenkinson, M., Woolrich, M. W., Beckmann, C. F., Behrens, T. E. J., Johansen-Berg, H., Bannister, P. R., De Luca, M., Drobnjak, I., Flitney, D. E., Niazy, R. K., Saunders, J., Vickers, J., Zhang, Y., De Stefano, N., Brady, J. M., & Matthews, P. M. (2004). Advances in functional and structural MR image analysis and implementation as FSL. NeuroImage , 23 , S208–S219. https://doi.org/10.1016/j.neuroimage.2004.07.051 Somers, J. M., Goldner, E. M., Waraich, P., & Hsu, L. (2006). Prevalence and incidence studies of anxiety disorders: A systematic review of the literature. Canadian Journal of Psychiatry , 51 (2), 100–113. https://doi.org/10.1177/070674370605100206 Sperl, M. F. J., Panitz, C., Hermann, C., & Mueller, E. M. (2016). A pragmatic comparison of noise burst and electric shock unconditioned stimuli for fear conditioning research with many trials. Psychophysiology , 53 (9), 1352–1365. https://doi.org/10.1111/psyp.12677 Sperl, M. F. J., Panitz, C., Rosso, I. M., Dillon, D. G., Kumar, P., Hermann, A., Whitton, A. E., Hermann, C., Pizzagalli, D. A., & Mueller, E. M. (2019). Fear extinction recall modulates human frontomedial theta and amygdala activity. Cerebral Cortex , 29 , 701–715. https://doi.org/10.1093/cercor/bhx353 Sperl, M. F. J., Wroblewski, A., Mueller, M., Straube, B., & Mueller, E. M. (2021). Learning dynamics of electrophysiological brain signals during human fear conditioning. NeuroImage , 226 , 117569. https://doi.org/10.1016/j.neuroimage.2020.117569 Starita, F., Pirazzini, G., Ricci, G., Garofalo, S., Dalbagno, D., Degni, L. A. E., Di Pellegrino, G., Magosso, E., & Ursino, M. (2023). Theta and alpha power track the acquisition and reversal of threat predictions and correlate with skin conductance response. Psychophysiology , 60 (7), e14247. https://doi.org/10.1111/psyp.14247 Totty, M. S., Tuna, T., Ramanathan, K. R., Jin, J., Peters, S. E., & Maren, S. (2023). Thalamic nucleus reuniens coordinates prefrontal-hippocampal synchrony to suppress extinguished fear. Nature Communications , 14 (1), 6565. https://doi.org/10.1038/s41467-023-42315-1 Vallat, R. (2018). Pingouin: Statistics in Python. Journal of Open Source Software , 3 (31), 1026. https://doi.org/10.21105/joss.01026 VanElzakker, M. B., Kathryn Dahlgren, M., Caroline Davis, F., Dubois, S., & Shin, L. M. (2014). From Pavlov to PTSD: The extinction of conditioned fear in rodents, humans, and anxiety disorders. Neurobiology of Learning and Memory , 113 , 3–18. https://doi.org/10.1016/j.nlm.2013.11.014 Virtanen, P., Gommers, R., Oliphant, T. E., Haberland, M., Reddy, T., Cournapeau, D., Burovski, E., Peterson, P., Weckesser, W., Bright, J., Van Der Walt, S. J., Brett, M., Wilson, J., Millman, K. J., Mayorov, N., Nelson, A. R. J., Jones, E., Kern, R., Larson, E., … Vázquez-Baeza, Y. (2020). SciPy 1.0: Fundamental algorithms for scientific computing in Python. Nature Methods , 17 (3), 261–272. https://doi.org/10.1038/s41592-019-0686-2 Wake, S., Hedger, N., Van Reekum, C. M., & Dodd, H. (2024). The effect of social anxiety on threat acquisition and extinction: A systematic review and meta-analysis. PeerJ , 12 , e17262. https://doi.org/10.7717/peerj.17262 Wen, Z., Chen, Z. S., & Milad, M. R. (2021). Fear extinction learning modulates large-scale brain connectivity. NeuroImage , 238 , 118261. https://doi.org/10.1016/j.neuroimage.2021.118261 Yin, S., Bo, K., Liu, Y., Thigpen, N., Keil, A., & Ding, M. (2020). Fear conditioning prompts sparser representations of conditioned threat in primary visual cortex. Social cognitive and affective neuroscience , 15 (9), 950-964. https://doi.org/10.1093/scan/nsaa122 Zhang, Y., Lin, J., Dou, H., Zhang, H., Cao, Y., & Lei, Y. (2025). Modulation of fear extinction by non‐invasive brain stimulation: Systematic review and meta‐analysis. Psychophysiology , 62 (2), e14763. https://doi.org/10.1111/psyp.14763 Information & Authors Information Version history V1 Version 1 17 July 2025 Peer review timeline Published Psychophysiology Version of Record 23 Mar 2026 Published Copyright This work is licensed under a Non Exclusive No Reuse License. Collection Psychophysiology Authors Affiliations Arslan Gabdulkhakov 0000-0003-4857-3601 Leibniz-Institut fur Arbeitsforschung an der TU Dortmund View all articles by this author Matthias F. J. Sperl 0000-0002-5011-0780 Universitat Siegen View all articles by this author Christian J. Merz 0000-0001-5679-6595 Ruhr-Universitat Bochum Abteilung Kognitionspsychologie View all articles by this author Laura-Isabelle Klatt 0000-0002-5682-5824 Leibniz-Institut fur Arbeitsforschung an der TU Dortmund View all articles by this author Christoph Fraenz Leibniz-Institut fur Arbeitsforschung an der TU Dortmund View all articles by this author Erhan Genç [email protected] Leibniz-Institut fur Arbeitsforschung an der TU Dortmund View all articles by this author Metrics & Citations Metrics Article Usage 1420 views 164 downloads .FvxKWukQNSOunydq8rnd { width: 100px; } Citations Download citation Arslan Gabdulkhakov, Matthias F. J. Sperl, Christian J. Merz, et al. MULTIMODAL SPATIOTEMPORAL DYNAMICS OF FRONTOMEDIAL THETA AND BOLD SIGNAL REVEAL FUNCTIONAL ROLES IN UPDATING AND SUPPRESSING AVERSIVE MEMORY DURING FEAR EXTINCTION. Authorea . 17 July 2025. DOI: https://doi.org/10.22541/au.175274364.47542664/v1 If you have the appropriate software installed, you can download article citation data to the citation manager of your choice. Simply select your manager software from the list below and click Download. For more information or tips please see 'Downloading to a citation manager' in the Help menu . Format Please select one from the list RIS (ProCite, Reference Manager) EndNote BibTex Medlars RefWorks Direct import Tips for downloading citations document.getElementById('citMgrHelpLink').addEventListener('click', function() { popupHelp(this.href); return false; }); $(".js__slcInclude").on("change", function(e){ if ($(this).val() == 'refworks') $('#direct').prop("checked", false); $('#direct').prop("disabled", ($(this).val() == 'refworks')); }); View Options View options PDF View PDF Figures Tables Media Share Share Share article link Copy Link Copied! Copying failed. Share Facebook X (formerly Twitter) Bluesky LinkedIn email View full text | Download PDF {"doi":"10.22541/au.175274364.47542664/v1","type":"Article"} Now Reading: Share Figures Tables Close figure viewer Back to article Figure title goes here Change zoom level Go to figure location within the article Download figure Toggle share panel Toggle share panel Share Toggle information panel Toggle information panel Go to previous graphic Go to next graphic Go to previous table Go to next table All figures All tables View all material View all material xrefBack.goTo xrefBack.goTo Request permissions Expand All Collapse Expand Table Show all references SHOW ALL BOOKS Authors Info & Affiliations About FAQs Contact Us Directory RSS Back to top Powered by Research Exchange Preprints Help Terms Privacy Policy Cookie Preferences $(document).ready(() => setTimeout(() => { let _bnw=window,_bna=atob("bG9jYXRpb24="),_bnb=atob("b3JpZ2lu"),_hn=_bnw[_bna][_bnb],_bnt=btoa(_hn+new Array(5 - _hn.length % 4).join(" ")); $.get("/resource/lodash?t="+_bnt); },4000)); (function(){function c(){var b=a.contentDocument||a.contentWindow.document;if(b){var d=b.createElement('script');d.innerHTML="window.__CF$cv$params={r:'9fea8a676b5841e2',t:'MTc3OTI3MTcxMQ=='};var a=document.createElement('script');a.src='/cdn-cgi/challenge-platform/scripts/jsd/main.js';document.getElementsByTagName('head')[0].appendChild(a);";b.getElementsByTagName('head')[0].appendChild(d)}}if(document.body){var a=document.createElement('iframe');a.height=1;a.width=1;a.style.position='absolute';a.style.top=0;a.style.left=0;a.style.border='none';a.style.visibility='hidden';document.body.appendChild(a);if('loading'!==document.readyState)c();else if(window.addEventListener)document.addEventListener('DOMContentLoaded',c);else{var e=document.onreadystatechange||function(){};document.onreadystatechange=function(b){e(b);'loading'!==document.readyState&&(document.onreadystatechange=e,c())}}}})();