Anxious vision: Trait-like visual cortical hyperactivity in trait anxiety

Abstract

Sensory processing varies across individuals, often manifesting as distinct traits. Some traits, particularly sensory hypersensitivity, even to basic non-valenced stimuli, are closely linked to emotional traits and implicated in a diversity of psychiatric disorders. Extant theories attribute this sensory-emotion linkage to limbic and prefrontal modulation, but empirical support remains limited. Meanwhile, it is increasingly recognized that sensory cortex supports flexible, value-based processing beyond labeled-line sensory analysis. Assaying early visual cortical activity (using visual evoked potentials/VEPs) across four high-density electroencephalogram (hd-EEG) experiments, involving three independent samples (N = 150) and repeated assessments (up to three waves), we identify a reliable and generalizable pattern of visual cortical hyperactivity related to high trait anxiety. Notably, this effect was specific to the parvocellular (vs. magnocellular) pathway, emerging as early as 46 ms and localized to V1. Underscoring its trait-like stability, the effect was reproducible across different arousal states, neutral and negative stimuli, extended time intervals, and varied experimental paradigms. Moreover, its ecological validity was affirmed by its presence in response to both basic low-level stimuli (Gabor patches) and complex real-world images. Crucially, an excitation-inhibition (E/I) ratio (indexed by the resting-state EEG aperiodic exponent; 1/f slope) predicted parvocellular VEP magnitude in subjects with low but not high trait anxiety, suggesting disrupted E/I modulation in trait anxiety underpins this hyperactivity. We conclude that in trait anxiety, early visual processing deviates from strict “fidelity” to physical-world input or mere “subservience” to top-down modulation, instead adapting to an individual’s biological propensity at the earliest processing stage.

Sensory processing varies across individuals, often presenting as stable sensory traits that are also tied to emotional traits (1, 2). Among these, sensory hypersensitivity shows strong associations with anxiety-related traits, such as trait anxiety, behavioral inhibition, and neuroticism, as well as has been linked to clinical disorders (1–5). Sensory hypersensitivity is particularly well-documented in schizophrenia (6, 7) and autism (8). Recent research, including meta-analyses, indicated that sensory anomalies, especially sensory hypersensitivity, span a spectrum of psychiatric disorders, with anxiety disorders and posttraumatic stress disorder (PTSD) especially implicated (9–13).

Despite the above, we have limited insight into the precise nature of a link between sensory and emotional traits. Examination of its neural basis has focused predominantly on the limbic system, especially the amygdala and hippocampus, and, to a lesser extent, prefrontal regions (14–16). This aligns with prevailing clinical models, which attribute sensory anomalies in psychiatric disorders to dysfunctions in prefrontal and limbic regions, driven by their aberrant reentrant projections to sensory cortex (17–19). However, empirical support for a limbic-prefrontal origin of sensory traits is sparse and inconclusive (20). Paradoxically, despite its primary role in sensory processing, the sensory cortex has been largely overlooked in this research, rendering its role in this sensory-emotion linkage poorly understood.

A widespread assumption is that sensory cortex maintains high fidelity with respect to external stimuli. This implies its responses should be relatively consistent across individuals—unless modulated by higher-order (limbic and prefrontal) inputs. However, this conflicts with evolutionary evidence that sensory cortex emerged before limbic and prefrontal structures, and that ancient organisms (such as the ancestral amniote) depended on it to flexibly identify predators and prey (21). Recently, the idea that the sensory cortex encodes biological value, without higher-order neural input, and adjusts its responses accordingly, has gained traction (22, 23). Importantly, accruing evidence suggests that sensory cortical value encoding correlates with trait anxiety (24, 25) while sensory dysfunctions—such as sensory cortical hyperactivity and disinhibition—are implicated in disorders associated with high trait anxiety, including PTSD, schizophrenia, and autism (10, 22, 26–30).

We hypothesized that sensory hypersensitivity in trait anxiety involves a significant, active contribution from early sensory cortex. Mechanistically, sensory cortex operates through well-organized local circuits with rich feedforward and feedback excitation and inhibition, where the excitation-inhibition (E/I) ratio regulates sensory cortical activation levels (23, 31). Indeed, E/I imbalance has been associated with psychiatric disorders, especially those involving significant sensory anomalies, such as schizophrenia and autism (32). Therefore, on this basis, we also hypothesized that disrupted E/I modulation in trait anxiety contributes to sensory cortical hyperactivity.

To test these hypotheses, we characterized trait-like sensory cortical hyperactivity and examined its association with trait anxiety through four high-density Electroencephalogram (hdEEG) experiments. To ascertain reliability and generalizability of a sensory trait and its relation with trait anxiety, we included repeated assessments (up to three waves) and three independent samples. To address potential confounds, we implemented aversive conditioning between assessments (in Experiment 1) in order to evoke state anxiety and arousal and in so doing tease out potential effects of state (vs. trait) anxiety in heightening sensory cortical processing (33, 34). Early visual cortical activity was assayed using visual evoked potentials (VEPs), an electrophysiological feature emerging from the early (primary and secondary) visual cortex, at approximately 50–120 ms post stimulus onset (35, 36). In generating characteristic VEPs, visual processing in the primate brain is mediated by distinct pathways, especially the parvo- (P) and magno-cellular (M) visual pathways (35–37). These pathways serve specialized functions (37) and are differentially implicated in psychiatric disorders (6, 38, 39). To precisely delineate their contribution to sensory hyperactivity , we employed carefully calibrated M- and P-selective stimuli across all four experiments.

Results

Stable and trait-like visual cortical biases in trait anxiety

In Experiment 1, we applied two sets of stimuli known to selectively elicit M and P-pathway processing—M-pathway: achromatic, low-luminance, and low-contrast Gabor patches; P-pathway: chromatic (red/green) and isoluminant Gabor patches (37). Participants (N = 52; mean age, 19.5 years; 22 men) performed a simple orientation discrimination task during high-density (HD; 96-channel) EEG recordings (40) (Fig. 1). Sample size was determined based on a medium effect size typically observed for trait anxiety (i.e., r = .35, requiring a sample size of 46 to reach a power of .80).

Figure 1. Stable and trait-like visual cortical biases in trait anxiety (Experiment 1).

(A) Example stimuli and trial. The M-selective stimulus (Top) comprised a Gabor patch that contained a gray grating with a low spatial frequency of .68 cycles per degree (cpd) and low contrast of 6.9% (Michelson contrast). The P-selective stimulus (Bottom) comprised a Gabor patch that contained a green/red grating with high spatial frequency of 4 cpd. Notably, the green/red grating and the background were isoluminant (based on individual heterochromatic flicker photometry) such that magnocellular activation induced by the onset of these stimuli would be maximally inhibited. Participants performed an orientation discrimination task, which contained trials of two consecutively delivered Gabor patches of either the Same or Different orientation. M- and P-selective trials were randomly intermixed. (B-D) VEPs and their correlation with trait anxiety at Time 1 (B), Time 2 (C), and Time 3 (D). The orientation discrimination task was administered at three time points, including Time 1—the baseline, Time 2—immediately after aversive conditioning, and Time 3—15 days later. The M- and P-selective stimuli evoked characteristic P1 (black waveforms) and C1/N1 (red waveforms) potentials, respectively, both maximally distributed around Oz (indicated by the white dots in the topographic maps). In the scatter plots, the C1-N1 amplitude was inverted (as magnitude) to adjust for its negative polarity. * = p < .05. Black lines in the gray boxes at the bottom of the VEP waveforms reflect the windows of continuous significant correlation of the VEPs with trait anxiety.

To demonstrate stable and trait-like sensory cortical biases, we administered the task at three distinct time points—Time 1: before; Time 2: after a standard aversive conditioning manipulation; and Time 3: Day 16 (15 days after). Consistency in response between Time 1 and Time 2 would support reliability across different arousal and anxiety states, while consistency with Time 3 would attest to long-term reliability. At all three time points, trials for the conditioned and non-conditioned stimuli (CS and non-CS) were pooled. Additionally, we also analyzed CS and non-CS trials separately, which yielded virtually identical results (see details in SOM and Table S1). Trait anxiety was assessment with the Behavioral Inhibition Scale (BIS) (41), a neurobiologically motivated questionnaire with high reliability and high predictive validity for anxiety (42, 43). Importantly, the trait of behavioral inhibition is closely related to sensory sensitivity (44).

As established in the literature (35, 36), the M- and the P-selective Gabor patches evoked a positive-going (P1) component and a negative-going (C1/N1) complex, respectively (Fig. 1B–D). These VEPs were extracted from the Oz site (collapsed across 3 midline electrodes), where they were maximally distributed, and submitted to Pearson correlational analyses. We observed a positive correlation between C1/N1 magnitude (i.e., inverted amplitude to adjust for its negative polarity) and trait anxiety at all three time points (r’s > 0.37, p’s < 0.05), suggesting parvocellular visual cortical hyperactivity in trait anxiety. We also found a negative correlation between P1 amplitude and trait anxiety at all three time points (r’s < −0.33, p’s < 0.05), suggesting magnocellular visual cortical hypoactivity in trait anxiety.

In addition, at all three time points, point-by-point analysis of the waveforms throughout the early visual processing stage (i.e., 0–200 ms post-stimulus) confirmed continuous intervals encompassing the P- and M-evoked VEPs (grey boxes in Fig. 1), where the above correlation was reliably significant. Specifically, for the P-evoked waveforms, the continuous intervals were 86–175 ms; 86–171 ms; and 86–160 ms for Time 1, Time 2, and Time 3, respectively (see details in SOM). For the M-evoked waveforms, the continuous intervals were 72–200 ms, 42–200 ms, and 78–200 ms for Time 1, Time 2, and Time 3, respectively. Finally, using Hittner’s Monto Carlo evaluation method (45), we compared the correlation across the three time points and observed no effects of time (p’s > .44; see details in SOM), indicating reliability and robustness of this association across different arousal levels and over an extended time. Thus, these findings are consistent with stable, trait-like, visual cortical biases in trait anxiety.

An opposing pattern of M- and P- biases in trait anxiety was not predicted. Intriguingly, there were strong negative correlations between the magnitudes of VEPs for the two respective pathways, r’s < −0.62, p’s < .05 (see SOM Fig. S1). These negative correlations rule out alternative accounts of general differences in neural responsivity or sensory reactivity across individuals (as they would predict positive correlations). In addition, the finding might indicate a reciprocity between parvo- and magno-cellular functioning within an individual. However, given that M- and P-selective trials were intermixed and presented in rapid succession, it is also possible that a repetition-suppression type of effect between the two pathways might drive this reciprocity.

Reproducible and stable association of trait anxiety with parvocellular visual cortical hyperactivity

To adjudicate between the aforementioned explanations, we conducted Experiment 2 using an independent subject sample (N = 52; mean age, 20.0 years; 16 men). To minimize potential effects of repetition suppression between the two pathways, we presented the P- and M-selective Gabor patches in non-overlapping visual fields, thereby activating separate populations of retinal and visual cortical cells. Specifically, based on their relative sensitivity to foveal versus peripheral stimuli (37), we presented P- and M-selective Gabor patches at center and periphery (at 10° eccentricity), respectively (Fig. 2A). Similar methods (e.g., EEG recordings and preprocessing, trait anxiety measurement, statistical analysis) as in Experiment 1 were applied here and in the other experiments unless otherwise stated.

Figure 2. Reliable parvocellular visual cortical hyperactivity in trait anxiety (Experiments 2 and 3).

(A) Example stimuli and trials in Experiment 2. The M-selective stimulus (Top) comprised a Gabor patch that contained a gray grating with a low spatial frequency of .68 cpd and a low contrast of 6.9%. The P-selective stimulus (Bottom) comprised a Gabor patch that contained a green/red grating with high spatial frequency of 4 cpd. The green and red lines and the background were set isoluminant (15.4 cd/m²). Participants passively viewed the M- and P-selective stimuli presented in the periphery (10 degrees eccentricity) and at the fovea, respectively. (B) VEPs and their correlation with trait anxiety in Experiment 2. Peripheral M-selective stimuli evoked a P1 potential (black waveforms) maximally distributed at P7 and P8 (white dots in the topographic maps), contralateral to the stimuli presented in the right and left visual field (MR and ML). The central P-selective stimuli evoked a C1/N1 complex, as in Experiment 1. Significant association with trait anxiety was observed in the C1 potential primarily. (C) Example stimuli and trials in Experiment 3. The M-selective stimulus (Top) comprised a Gabor patch that contained a gray grating with a low spatial frequency of .68 cpd and a low contrast of 7%. The P-selective stimulus (Bottom) comprised a Gabor patch that contained a grey grating with high spatial frequency of 4 cpd and high contrast of 34%. Participants performed an orientation discrimination task with both M- and P-selective stimuli randomly intermixed and centrally presented (as in Experiment 1). (D) VEPs and their correlation with trait anxiety in Experiment 3. M-selective stimuli evoked a P1 potential (black waveforms) maximally distributed at P7 and P8 (white dots in the topographic maps). P-selective stimuli evoked a C1 potential maximally distributed at Oz (white dots in the topographic map), which correlated with trait anxiety. (E & F) Disrupted E-I modulation of parvocellular visual cortical processing in trait anxiety. The sample was divided by low (E) and high (F) trait anxiety groups based on median split. (E) Higher E-I ratio (i.e., bias towards excitation vs. inhibition; indexed by flatter/smaller aperiodic 1/f exponent of resting-state EEG power spectrum) correlated with larger C1 magnitude in the low trait anxiety group. Source-level analysis using eLORETA localized this association to the early visual cortex (V1/V2). Display threshold: p < .005 uncorrected. (F) No clear association between E-I ratio and C1 magnitude. In the scatter plots, the C1-N1 amplitude was inverted (as magnitude) to adjust for its negative polarity. * = p < .05; **= p < .01. Black lines in the gray boxes at the bottom of the VEP waveforms reflect the windows of continuous significant correlation of the VEPs with trait anxiety.

The P-selective stimuli evoked a C1/N1 complex at Oz (Fig. 2B bottom). The peripherally presented M-selective stimuli evoked a clear (albeit attenuated) P1 at the contralateral posterior sites (P7/P8) (Fig. 2B top). Importantly, we replicated a positive correlation between trait anxiety and C1 magnitude (r = 0.33, p < .05). A point-by-point analysis in the 200-ms post-stimulus window confirmed a continuous interval (58–105 ms; encompassing the C1 potential) where the correlation was reliably significant (see details in SOM). However, a negative correlation between trait anxiety and P1 observed in Experiment 1 was not replicated here (r = 0.03, p = .81), while point-by-point analysis in the 200-ms window failed to isolate any time points showing a correlation with trait anxiety (p > .10). These findings are again supportive of a selective association of trait anxiety with parvocellular visual cortical hyperactivity (but not with magnocellular activity).

However, an alternative explanation might be that hyperactivity reflects a restricted sensitivity for color processing, as opposed to a general parvocellular hyperactivity in trait anxiety. To exclude this explanation, and to ascertain retest reliability of this effect, we conducted Experiment 3 in the same sample (the sequence of Experiments 2 & 3 was counterbalanced across participants), where P-selective stimuli were achromatic but contained high spatial frequency and high contrast known to strongly activate the parvocellular pathway (Fig. 2C). These stimuli evoked a characteristic C1 potential (Fig. 2D), which again exhibited a positive correlation with trait anxiety (r = 0.38, p < .05). Similarly, a point-by-point analysis throughout the 200-ms post-stimulus window confirmed a continuous interval (46–94 ms; encompassing the C1 potential) where the correlation remained reliably significant (see details in SOM). The M-selective stimuli were centrally presented and evoked strong P1 potential at Oz. As in Experiment 2, we did not observe a significant correlation between trait anxiety and P1 (r = 0.04, p =.76) or any time point in the 200-ms window (p’s > .10). Finally, we found no difference in the correlation between the two experiments (p’s > .64), highlighting a constancy of this association despite changes in visual properties and tasks. Together, results in Experiments 2 & 3 confirm a reliable and replicable parvocellular visual cortical hyperactivity in trait anxiety, while negating the presence of a magnocellular hypoactivity in trait anxiety.

Disrupted E-I modulation of parvocellular visual cortical activity in trait anxiety

To gain mechanistic insight into visual cortical hyperactivity in trait anxiety, we examined resting-state EEG acquired from participants in Experiments 2 & 3. We extracted an index of excitation-inhibition (E/I) ratio to examine its modulatory effects on parvocellular visual cortical responses (i.e., C1 potentials). Specifically, we extracted the aperiodic exponent (1/f slope) across the EEG power spectrum (between 3–50 Hz) which has been shown to be tightly correlated with cortical E/I ratio, and where a less negative (numerically smaller) exponent is linked to a larger E/I ratio (i.e., biased towards excitation vs. inhibition) (46, 47). We then entered data from both Experiments 2 & 3 into a mixed-effects regression model with C1 magnitudes as the dependent variable and trait anxiety, experiment, aperiodic 1/f exponent, and their interaction as regressors.

We observed a significant interaction between the aperiodic 1/f exponent and trait anxiety in predicting C1 magnitude (coefficient = 0.46, SE = 0.20, p < 0.05), albeit the aperiodic exponent alone did not predict C1 magnitude (p’s > 0.09; see SOM for details). To unpack this interaction, we performed a median split based on trait anxiety level and applied a mixed-effects regression model on the aperiodic exponent and C1 magnitude in the high- and low-anxiety groups separately (Fig. 2E&F). In the low-anxiety group, we confirmed a negative association between C1 magnitude and aperiodic 1/f exponent (p = .04 one-tailed). This suggests that, in low anxiety, higher E-I ratio predicted stronger parvocellular visual cortical response (Fig. 2E). Leveraging the high-density EEG recordings, we also performed intracranial source analyses on the C1 component using exact low-resolution electromagnetic tomography (eLORETA) (48). In the low-anxiety group, an association between C1 magnitude and E/I ratio localized to early (primary and secondary) visual cortex (V1/V2; peak x = −10, y = −75, z = −10, r = −0.53, k = 100; FDR p < .05; Fig. 2E). By contrast, in the high-anxiety group, no significant association emerged from either the surface- (p = .12 one-tailed; and potentially trending in the opposite direction) or source-level analyses (Fig. 2F). Together, results here suggest that E/I modulation of the visual cortex was insufficient in high trait anxiety, potentially resulting in visual cortical hyperactivity among such individuals.

Ecological validity: Parvocellular visual cortical hyperreactivity to real-world images in trait anxiety

The above experiments used basic laboratory stimuli and, arguably, do not address ecological validity and real-world relevance of the observed association. A more general question is whether trait anxious individuals exhibit visual cortical hyperactivity navigating their natural environment. To address this, we examined visual cortical responses to natural scene images in association with trait anxiety. Specifically, we reanalyzed data from a prior study (N = 46; 19.3 +/−SD years; 22 men) using images conveying complex natural scenes (49). Critically, the images were filtered to contain high or low spatial frequency (HSF/LSF) that selectively activate the parvocellular or magnocellular pathway, respectively (Fig. 3A). Of note, these images were carefully calibrated to ensure that P- and M-selective images were comparable in luminance, contrast, wavelength energy, and visual complexity (edge density, entropy, compressed image size) (49). Finally, the images contained neutral, disgust, and fear emotion, allowing us to also examine the effect of emotion on the observed association.

Figure 3. Ecological validity: Parvocellular visual cortical hyperreactivity in trait anxiety applicable to natural scenes (Experiment 4).

(A) Example stimuli and trials in Experiment 2. The M- and P-selective stimuli were grey scale with LSF (< 3 cpd) and HSF (> 7 cpd), respectively. A visual search array containing a horizontal bar among vertical bars appeared after 150 ms, and participants were asked to indicate the quadrant containing the horizontal bar. (B) VEPs and their correlation with trait anxiety. Both stimuli evoked P1 potentials maximally distributed around Oz (white dots in the topographic maps). Significant association with trait anxiety was observed for the P1 evoked by P-selective stimuli. * = p < .05. Black lines in the gray boxes at the bottom of the VEP waveforms reflect the window of continuous significant correlation of the P1 with trait anxiety.

Both HSF and LSF stimuli evoked a P1 component. Amplitude of the P1 evoked by P-selective (HSF) images again correlated positively with trait anxiety (r = 0.38, p < .05), replicating the above findings with Gabor patches (Fig. 3B). Moreover, a point-by-point analysis of the 200-ms post-stimulus window again confirmed a continuous interval (101–140 ms; encompassing the P1 potential) where the correlation was reliably significant (see details in SOM). By contrast, amplitude of the P1 evoked by M-selective (LSF) images did not correlate with trait anxiety (r = 0.15, p = .35), while a point-by-point analysis also failed to identify any time points showing a significant correlation (p’s > .12). Finally, consistent with findings from Experiment 1, we observed no effect of emotion on the correlation for both P- and M-selective images (P-selective: p’s > .89; M-selective: p’s > .54), and the correlation coefficients were highly comparable across the three emotional categories (P-selective: r = 0.36–0.37; M-selective: r = 0.08–0.21). These results thus confirmed the association between trait anxiety and parvocellular visual cortical hyperactivity, validating its ecological validity.

Discussion

Across four experiments, and three independent subject samples, we show that early visual responses (VEPs) consistently manifest a reliable and generalizable pattern of parvocellular visual cortical hyperactivity linked to the presence of trait anxiety. The trait-like nature of this hyperactivity was reproducible across varying arousal states, neutral and negative stimuli, extended intervals, and diverse experimental paradigms. We also demonstrate ecological validity and real-world relevance to these findings using data involving processing of complex natural scenes, beyond simple laboratory stimuli. Mechanistically, an established metric of E/I ratio predicted parvocellular VEPs in low (but not high) trait anxiety, suggesting a linkage between visual cortex hyperactivity in trait anxiety and E/I modulation.

One implication of our findings is that in trait anxiety there is biased sensory encoding, of even basic, non-valenced, stimuli. Such biases could serve as a bottom-up mechanism, shaping downstream, higher-order processing and initiating a cascade of sensory, affective, and cognitive symptoms characteristic of anxiety and related disorders. While previous work has considered such sensory biases as state-dependent (e.g., linked to state anxiety) (33, 34), our work challenges this assumption. While it is known that arousal and state anxiety heighten sensory processing (33, 34), our experimental manipulation of arousal and anxiety states via aversive conditioning (Experiment 1) indicated an equivalent relationship between trait anxiety and VEPs before and after conditioning, effectively ruling out arousal or state anxiety as explanations for our findings. In addition, the observed association was selective, present for P (but not M) stimuli alone, a fact that excludes more general confounds such as attention or general neural reactivity. Finally, trait anxiety is known to upmodulate early sensory processing of threat cues (50–52). By including both neural and emotional stimuli (i.e., aversively conditioned Gabor patches in Experiment 1 or scenes depicting disgust or fear in Experiment 4), we show equivalent effects, indicating this sensory trait, and its association with trait anxiety, were independent of the emotional value of sensory stimuli.

While previous personality research has linked sensory traits to emotional traits (1–5), these studies focused predominantly on limbic and prefrontal regions as putative sites of origin, with inconclusive empirical support (20). This work has also emphasized a magnocellular primacy, wherein a fast magnocellular input to limbic and prefrontal areas is assumed to drive sensory cortical biases via reentrant feedback (17–19). This “magnocellular advantage” has been incorporated into theoretical conceptualization of diverse psychiatric disorders, wherein biased magnocellular processing contributes to a range of cognitive and affective symptoms (53).

Strikingly, our experiments consistently highlight parvocellular (but not magnocellular) hyperactivity in trait anxiety, contradicting a canonical magnocellular-dominant, top-down, account. Nonetheless, our findings align with evidence from autism research, which suggests an enhanced parvocellular (vs. magnocellular) hyperactivity in this disorder (39, 54). Notably, the parvocellular pathway operates at a slower speed compared to the magnocellular pathway, reaching the visual cortex (V1) at ~50 ms. Consistent with this, we detected parvocellular visual cortical hyperactivity in trait anxiety as early as 46 ms, highlighting its emergence in the course of an initial feedforward sweep (Figure 2D). In this regard, our study supports a “bottom-up” view wherein a sensory-emotion connection can originate in the early sensory cortex, biasing early sensory processing and influencing subsequent processes.

Our findings motivate a revised conceptualization of key features seen in disorders such as autism, anxiety, and PTSD. Anatomically, the parvocellular pathway is the dominant component of the visual system, constituting about 80% of the ganglion cell population and providing the primary input to the visual cortex (37, 53). As the gateway between brain and the external world, the sensory cortex plays a pivotal role in filtering sensory input (55). In individuals with high trait anxiety, sensory cortex hyperactivity (and disinhibition) would allow irrelevant or unwanted environmental input to evoke sensory cortical responses. We speculate that an ensuing excess of sensory output, or “neural noise,” could in turn overwhelm downstream, high-order brain areas, triggering a cascade of disruptions that manifest in the diversity of symptoms seen in PTSD, autism, and related disorders (29, 56, 57).

Sensory cortex is recognized as a critical storage site for threat memory (23, 58, 59), especially in individuals with high trait anxiety (24, 25). Consequently, hyperactivity in the visual cortex might facilitate over-activation of threat memory, intensifying threat detection and memory intrusions commonly seen in anxiety, PTSD, and autism (43, 44, 60). Furthermore, the parvocellular visual pathway plays a central role in detailed, fine visual analysis (37, 53), and hyperactivity might contribute to excessive fine-detail processing reported in anxiety and autism (61, 62). This bias could conceivably extend to heightened detection of minute and insignificant threat cues, a defining feature of anxiety (51, 63).

Sensory cortical activity is regulated at multiple levels. At a fine-grained microscopic level, it is influenced by excitatory and inhibitory inputs within local circuitry (23, 31). This E/I balance (or the lack thereof) is believed to play a fundamental role in psychiatric disorders (32). Specifically, sensory hypersensitivity (e.g., “superior low-level perceptual processing”), a hallmark symptom of autism (64), is linked to deficient E/I regulation of sensory cortex (56). Our findings related to E/I balance in visual cortex suggests a similar mechanism may be at play in trait anxiety. Thus, in low trait anxiety, an E/I ratio predicted the magnitude of parvocellular VEPs, demonstrating appropriate excitatory and inhibitory modulation of visual cortical activity. In contrast, this was not the case in high trait anxiety, suggesting a disruption of E/I modulation might drive visual cortical hyperactivity in these individuals.

GABAergic (inhibitory) and glutamatergic (excitatory) activity in local circuits are thought to directly maintain an E/I balance (31), though midbrain and hindbrain monoaminergic afferents are also considered to play a role. Although GABAergic and glutamatergic anomalies in trait anxiety are not yet well specified, monoaminergic dysfunctions are widely reported (56, 65). These monoaminergic imbalances, such as serotonergic deficiency, may contribute to the sensory-anxiety association by disrupting E/I regulation of the sensory cortex, thereby exaggerating sensory cortical activity (44). Conversely, anxiolytic drugs, including GABAergic treatments (such as diazepam and gabapentin) and monoaminergic treatments (such as serotonin selective receptor inhibitors), could conceivably alleviate anxiety by downregulating sensory cortical activity through enhancing cortical inhibition.

In summary, we provide novel empirical evidence for a direct connection between anxiety and sensory traits, arising at the earliest processing in primary and secondary sensory cortex. That the brain instantiates adaptable sensory regulation at the earliest staging post for external inputs has considerable relevance in relation to mental health. It opens the possibility that a fundamental neural mechanism underlying psychopathology involves early sensory pathophysiology (6, 10, 13, 29), promoting a trifecta conceptualization of psychiatric conditions that spans sensory, emotional, and cognitive domains. Further research is essential to elucidate genetic and environmental factors involved in the development of this intertwined sensory-emotion trait.

Methods

Experiment 1

Participants

As part of a larger study previouly reported in (40), we conducted an EEG experiment of visual discrimination and aversive conditioning consisting of three recording sessions separated by 15 (± 3.6) days. A total of fifty-two individuals (19.5 +/− 1.4 years; 22 men) participated in the first two sessions that took place on Day 1, and forty-two returned to participate in the third session on Day 16. Five and six participants were excluded from analysis for Day 1 and Day 16, respectively, due to excessive eye movements, severe EEG artifact and technical failures. All participants were right-handed with normal or corrected-to-normal vision and denied a history of severe head injury, psychological/neurological disorders or current use of psychotropic medication. All participants provided informed consent to participate in this study, which was approved by the University of Wisconsin Institutional Review Board. Sample size was determined based on a medium effect size typically observed for trait anxiety (i.e., r = .35, requiring a sample size of 46 to reach a power of .80).

Trait anxiety assessment

At the beginning of the first session, participants completed the Behavioral Inhibition Scale (BIS), a 7-item self-report measure (rated on a Likert scale of 1–4), assessing trait anxiety or the strength of the behavioral inhibition system (41, 50, 66, 67). This scale is neurobiologically motivated with high reliability and strong predictive validity of anxiety (42, 43) and recommended by the National Institute of Mental Health for assessing trait anxiety. Importantly, the trait of behavioral inhibition is found to be especially related to sensory sensitivity (44).

Stimuli

Two types of Gabor patches (sinusoidal gratings multiplied by a Gaussian envelope; 9°×9° in visual angles) were generated with specific visual properties that preferentially stimulate the magnocellular (M) and parvocellular (P) pathways, respectively. M-selective Gabor patches were of low spatial frequency (LSF; 0.67 cycles per degree/cpd), low luminance contrast (6.9% Michelson contrast), and achromatic. The lightest point of the Gabor patch was 23.14 cd/m² and the darkest point was 20.16 cd/m². P-selective Gabor patches were of high spatial frequency (HSF; 4 cpd) and chromatic (red-green isoluminant). By making the P-selective stimuli isoluminant, we would maximally inhibit the magnocellular activation evoked by the onset of these stimuli.

Individual red-green isoluminant point was experimentally determined by heterochromatic flicker photometry(68), with gray, red, or green squares alternatingly presented on a CRT monitor at a frequency of 30 Hz. With the luminance of the gray square fixed at 21.65 cd/m² (the same as the background luminance of the experimental tasks), participants first adjusted the intensity of the red gun, via button pressing, until minimal flicker was perceived between the alternating gray and the red squares. With the luminance of the red square set, participants then adjusted the intensity of the green gun until minimal flicker was observed between the alternating red and the green squares. After three repetitions, mean red and green values were computed for each participant to generate individualized red-green isoluminant chromatic Gabor patches.

Throughout the experiment, visual stimuli were presented on the same gray background on a CRT monitor, which was calibrated by first fitting a gamma function for each RGB channel based on sampled luminance values of each channel measured by a photometer, and then applying a reverse-gamma function on each RGB channel to achieve uniform steps of luminance increase using the Psychophysics Toolbox (69). Stimulus presentation was linked to the refresh rate (60 Hz) of the CRT monitor and delivered using Cogent2000 software (Wellcome Dept., London, UK) as implemented in Matlab (Mathworks, Natick, MA). Synchronization between stimulus display and data acquisition was verified using a photodiode placed at the center of the monitor screen.

Experiment procedure

The experiment consisted of four main phases: pre-conditioning (Time 1), conditioning, Day 1 post-conditioning (Time 2), and Day 16 post-conditioning phases (Time 3). Participants were seated ~60 cm from a CRT monitor in a dimly lit, electrically shielded room, and performed heterochromatic flicker photometry to determine their individual red-green isoluminant values, which were used to generate P-selective Gabor patches.

During conditioning, M- and P-selective CS were presented in two separate blocks, with block order counterbalanced. Within each conditioning block, 20 trials (10 for CS+, 10 for CS−) were presented. For 70% of the trials that were reinforced, a CS Gabor patch was centrally presented for 3000 ms, followed by simultaneous delivery of a UCS image for 2000 ms (fearful for CS+, neutral for CS−) and UCS sound for 1500 ms (fearful scream for CS+, pure tone for CS−). For the remaining 30% of trials, no UCS was presented following the CS. Before, immediately after, and 15 days after conditioning, participants performed an orientation discrimination task (see below).

For both M- and P-selective Gabor patches, two orientations were selected and differentially conditioned as CS+ and CS− via pairing with either aversive unconditioned stimuli (UCS) or neutral stimuli, respectively. To avoid orientation-specific effect, we employed two sets of Gabor patches (33° and 57° or 123° and 147° clockwise from the vertical meridian), which were counterbalanced across participants. The assignment of CS+ orientation within each set was further counterbalanced. Aversive and neutral UCS consisted of simultaneous presentation of images and sounds. Fearful and neutral images were selected from the International Affective Picture Set (IAPS; (70)) and internet sources, depicting threatening scenes (e.g. knife put to throat; gun pointed to head) and household artifacts (e.g. whistle, cabinet), respectively. Fearful sounds (i.e. screams) were obtained from the fear subset of human affective vocalizations (71). Neutral sounds consisted of pure tones of various frequencies.

Orientation discrimination task (ODT)

The orientation discrimination task (ODT) required participants to detect orientation differences between two serially presented Gabor patches. Each trial began with a centrally presented fixation cross with a jittered duration of 1600–1900 ms. The 1^st Gabor patch appeared for 400 ms, followed by a visual mask of 100 ms, before the 2^nd Gabor patch appeared for 400 ms, followed by another visual mask of 100 ms. The 1^st Gabor patch was always a CS (CS+ or CS−), while the 2^nd Gabor patch was either the same or different in orientation compared to the 1^st Gabor patch. ‘Same’ or ‘Different’ trials were presented with equal probability. On ‘Different’ trials, a Gabor patch 12° apart from both CS+ and CS− orientations (45° or 135°), was presented. Participants were required to make a speeded judgment to indicate whether the two serially presented Gabor patches were of same or different orientations. A total of 360 trials were randomly presented across three mini-blocks, with 90 trials for each experimental condition (i.e., M/P-selective CS+/CS−). In the current analysis, we pooled the CS+ and CS− trials, yielding 180 trials each for M- and P-selective Gabor patches.

EEG recording and analysis

EEG data were continuously recorded during all four phases of the experiment from a 96-channel (BioSemi ActiveTwo) system at a 1024 Hz sampling rate. Electrooculogram (EOG) was recorded at two eye electrodes at the outer canthi of each eye and one infraorbital to the left eye. EEG data were down-sampled to 256 Hz, digital bandpass filtered from 0.1 to 40 Hz, and then were referenced to the average of all 96 channels. EEG artifact detection and removal was achieved by the Fully Automated Statistical Thresholding for EEG artifact Rejection (FASTER) algorithm implemented in EEGLAB (72). FASTER first interpolates deviant channels from the continuous data using the EEGLAB spherical spline interpolation function. Data were then segmented into epochs 200 ms prior and 300 ms following the onset of the 1^st Gabor patch during orientation discrimination trials of each ODT. Epochs were rejected if their z-scores exceeded ±3 within parameters of amplitude range, variance, and deviation. Epoched data were then re-referenced to the average reference and submitted to independent component analysis (ICA) decomposition using the Infomax algorithm (73). Artefactual components (i.e., muscular artifacts, eye blinks and saccades, electrode “pop-offs”, etc.) were automatically detected and removed from the data. Lastly, deviant channels within individual cleaned epochs were interpolated again using the EEGLAB spherical spline interpolation function. Baseline correction was applied to the 200-ms pre-stimulus period.

We focused on visual event-related potential (VEP), the P1 and C1-N1 complex, which reflect early visual cortical processing and serve to dissociate magnocellular and parvocellular processing (35, 36). At site Oz, for all three time points, we extracted mean P1 amplitudes to M-selective and C1/N1 amplitudes to P-selective stimuli during 113–48 ms and 74–152 ms, respectively, centered on their peak latencies.

Experiments 2 & 3

Participants

Fifty-two right-handed individuals (20.0 +/− 3.18 years of age; 16 men) with normal or corrected-to-normal vision participated in this study, which consisted of a resting-state phase and Experiments 2 & 3. All participants denied a history of severe head injury, psychological/neurological disorders or current use of psychotropic medication. All participants provided informed consent to participate in this study, which was approved by the Florida State University Institutional Review Board. As in Experiment 1, based on the medium size effect of trait anxiety, we included 52 participants to ensure sufficient statistical power after potential exclusion due to artifacts and attrition.

Trait anxiety assessment

As in Experiment 1, we administered the BIS questionnaire to assess trait anxiety.

Stimuli

Two types of Gabor patches (sinusoidal gratings multiplied by a Gaussian envelope) were included, which contained visual properties that selectively stimulate the magnocellular (M) and parvocellular (P) pathways. In Experiment 2, M-selective Gabor patches were of LSF (0.67 cpd), low luminance contrast (6.9% Michelson contrast), and achromatic, which were presented 10° horizontally off-center (extending a 4°×4° visual angle). P-selective Gabor patches were of HSF (3 cpd), chromatic (red-green) and presented at the center (extending a 4°×4° visual angle). These patches were made isoluminant, by fixing the red and green luminance and the grey background luminance at 15.4 cd/m², such that magnocellular activation induced by the onset of these stimuli would be maximally inhibited.

In Experiment 3, stimuli consisted of elongated ovals (extending a visual angle of 4.8°× 9°) foveally presented in four different orientations (30°, 60°, 150°, 120° clockwise). Two types of achromatic filling gratings (tilting at 30°, 60°, 120°, 150° clockwise) were included: M-selective gratings were of LSF (.67 cpd) and low contrast (6.9% Michelson contrast) and P-selective gratings of HSF (3 cpd) and high contrast (34% Michelson contrast). As in Experiment 2, these physical properties were chosen to selectively activate the M- and P-selective pathways.

Stimulus presentation was linked to the refresh rate (60 Hz) of a CRT monitor and delivered using Cogent2000 software (Wellcome Dept., London, UK) as implemented in Matlab (Mathworks, Natick, MA). Synchronization between stimulus display and data acquisition was verified using a photodiode placed at the center of the monitor screen.

Experiment Procedure

Participants were seated ~60 cm from a CRT monitor in a dimly lit and electrically shielded room. Eyes-open resting-state EEG was first recorded when participants were asked to fixate on a crosshair on the screen. Participants then performed Experiments 2 & 3, the sequence of which was counter-balanced across participants.

In Experiment 2, a trial began with a blank screen for 100 ms, followed by a Gabor patch for 300 ms and then a gray fixation crosshair with a jittered duration of 700–1000 ms. No response was required from the participants, and twelve catch trials were included to ensure attention to the stimuli. The Gabor patches were either P-selective and presented centrally (100 trials) or M-selective and presented in the left/right visual periphery (10° off-center; 100 trials for each side). During a catch trial, a yellow fixation crosshair was presented for 1300 ms, to which participants were required to make a button press.

In Experiment 3, participants performed a similar orientation detection task as in Experiment 1. Each trial began with a centrally presented fixation crosshair with a jittered duration of 1600–1900 ms. The 1st oval appeared for 400 ms, followed by a visual mask of 100 ms, and a 2nd oval appeared for 400 ms, followed by another visual mask of 100 ms. The 2nd oval could be in the same orientation as the 1st oval or differ by 7° or 15°, and participants were required to make a speeded ‘Same’ or ‘Different’ response. A total of 560 trials, 280 trials each for M- or P-selective gratings, were randomly intermixed and presented across four blocks.

EEG recording and analysis

Recording procedures and analysis for Experiments 2 & 3 were largely the same as in Experiment 1. In Experiment 2, P1 potentials were evoked by M-selective Gabor patches presented in the right or left visual periphery, maximally distributed at the contralateral posterior temporal sites (P7 and P8). We extracted mean P1 amplitudes (collapsed across the two sites) during 125–160 ms at P7 and 121–156 ms at P8 (centered on the respective peak latencies). A C1-N1 complex was evoked by P-selective Gabor patches, maximal at Oz. We extracted mean C1 amplitudes during 62–86 ms interval at Oz (centered on the peak latency). In Experiment 3, P1 potentials were evoked by M-selective gratings and C1 potentials by P-selective gratings, both maximal at Oz. We extracted mean P1 and C1 amplitudes during 86–121 ms and 54–70 ms, respectively, centered on their peak latencies.

Resting-state EEG/EOG data were recorded for 2 minutes using the same recording protocol as described in Experiment 1. As in our previous studies (29, 74), resting-state EEG data were downsampled to 256 Hz, high-pass (1 Hz) and notch (60 Hz) filtered, and re-referenced to the average of all EEG channels before submission to the FASTER algorithm for further artifact correction. Next, resting-state data were segmented into 1-s epochs and converted into a power spectrum with Multitaper power spectral density estimate in signal processing toolbox in Matlab 2021b (The MathWorks, Inc, Natick, Massachusetts, United States). Zero padding was applied to provide a frequency resolution of 0.25 Hz in the 1-s epochs in time-domain EEG data. The aperiodic parameters (e.g., 1 /f spectral exponents) of the resting-state power spectrum (3–50 Hz) were calculated using the “Fooof” toolbox 1.0.0 (https://github.com/fooof-tools/fooof; (46)). The average R² of spectral fits was 0.97, with all individual R² exceeding 0.91, indicating strong overall and individual-level fit quality. According to (46), we extracted aperiodic 1/f spectral exponents from the Cz site (collapsed across the Cz and four surrounding electrodes).

Exact low-resolution electromagnetic tomography (eLORETA)

Based on VEP amplitudes from high-density EEG (hdEEG) in Experiment 2 & 3, we performed intracranial source analyses through exact low-resolution electromagnetic tomography (eLORETA; (75)). The eLORETA algorithm on hdEEG data has been increasingly used for intracranial source estimation in our lab and others (40, 66, 76–81), having been cross-validated in multiple studies combining EEG-based LORETA with fMRI (82–86), positron emission tomography (87, 88), and intracranial recordings (89).

Our solution space for intracranial source analyses consisted of 6239 cortical gray matter voxels with a spatial resolution of 5 × 5 × 5 mm in a realistic head model (90), which was registered to standardized space using a digitized MRI from the Montreal Neurologic Institute (MNI). We estimated voxelwise current density during the C1 and P1 windows, which was then submitted to voxelwise correlation with the aperiodic exponent (1/f slope). To minimize false-positive results in intracranial source localization, our laboratory has routinely applied two constraints in the analyses (40, 49, 66, 76, 80, 81, 91). First, eLORETA analysis were only applied to the time windows and tests that were significant in surface-level analysis (92). Accordingly, we only submitted the C1 sources to the analysis. Second, we applied Monte Carlo simulation based on the voxel spatial correlation inherent in the data to determine the statistical threshold of corrected p < 0.05 FDR. Specifically, using the Gaussian filter widths estimated from our data (FWHMx = 2.57mm, FWHMy = 2.56 mm, FWHMz = 2.38 mm), the voxel size (5 × 5 × 5 mm³), and a connection radius (5 mm), we derived a corrected threshold consisting of a voxel-level p < 0.005 over five contiguous voxels. All coordinates are reported in MNI space.

Experiment 4

Participants

This dataset was initially reported in (49). Forty-six right-handed individuals (19.3 +/− 2.3 years of age ; 22 men) with normal or corrected-to-normal vision participated in the study. Six participants were excluded from the analysis due to excessive eye movements and technical problems. All participants denied a history of severe head injury, psychological or neurological disorders, or current use of psychotropic medication. All participants provided informed consent to participate in this study, which was approved by the University of Wisconsin Institutional Review Board. Similar power analysis process, based on medium-size correlations of trait anxiety (r > .35), determined that a sample size of 46 would yield a statistical power of .80.

Trait anxiety assessment

As in Experiments 1–3, we administered the BIS questionnaire to assess trait anxiety.

Stimuli

Twenty-seven images were selected from the International Affective Picture Set (IAPS;(70)) and internet sources, depicting natural scenes/objects of fearful, disgusting, or neutral content (nine for each emotion category). These images were transformed to grayscale and equal size (256 × 256 pixels) and were low-pass filtered at 3 cycles/degree or high-pass filtered at 7 cycles/degree to generate low spatial frequency (LSF; M-selective) or high spatial frequency (HSF; P-selective) images, respectively. Spatial-frequency-filtered images were further normalized to equal luminance (17.11cd/m²) and luminance contrast by the SHINE Matlab toolbox (93). Next, wavelet analyses (94) were applied to filtered image sets to ensure equal wavelet energy for each emotion category in both H/LSF bands (49). Similarly, objective measures of visual complexity (edge density, entropy, compressed image size) were extracted (95) and submitted to ANOVAs (Emotion by SF), which revealed no effect of emotion or emotion-by-SF interaction (49).

Stimulus presentation was linked to the refresh rate (60 Hz) of a CRT monitor and delivered using Cogent2000 software (Wellcome Dept., London, UK) as implemented in Matlab (Mathworks, Natick, MA). Synchronization between stimulus display and data acquisition was verified using a photodiode placed at the center of the monitor screen.

Procedure

Participants were seated ~120 cm from a CRT monitor in an electrically shielded room and performed a visual search task adapted from a previous study (66). Each trial began with a centrally presented fixation crosshair with a jittered duration of 1150–1650 ms, followed by an image (7.2°× 7.2°) centrally displayed for 150 ms. Next, a search array in green was superimposed on the image for 500 ms, consisting of one horizontal bar (target) and seven vertical bars (distracters). Participants were required to make a button press to indicate the quadrant where the target was located while maintaining fixation. There was a total of 600 randomly intermixed trials presented in four experimental blocks, with 100 per each experimental condition (M/P-selective X Fear/Disgust/Neutral).

EEG recording and analysis

EEG were recorded from a 96-channel (BioSemi ActiveTwo) system at a 1024 Hz sampling rate, down-sampled to 256 Hz, and digital bandpass filtered from 0.1 to 40 Hz. Electrooculogram (EOG) was recorded at two eye electrodes at the outer canthi of each eye and one infraorbital to the left eye. The epoch was segmented from −200 to 300 ms post stimulus onset to focus on early visual processing. Trials with EEG/EOG voltages exceeding ±75 μV (relative to pre-stimulus baseline) were excluded from analysis. Inspection of grand average waveforms indicated clear P1s elicited by both M- and P-selective scenes, maximal at the central occipital site Oz, peaking at 119 ms and 115 ms post-stimulus, respectively. At site Oz, we extracted mean P1 amplitudes in the window of 100 – 135 ms, centered on the averaged peak latency of the two types of images.

Highlights:

• Hyperactivity in primary visual cortex is stable, trait-like

• It occurs as early as 46 ms post-stimulus, during initial feedforward sweep

• It covaries with trait anxiety, a key personality factor and risk predisposition for psychiatric disorders

• It is related to disrupted excitation-inhibition modulation of visual cortex in trait anxiety