Anxiety Symptoms in Young Children Are Associated With a Maladaptive Neurobehavioral Profile of Error Responding

Abstract

Background: Childhood anxiety symptoms have been linked to alterations in cognitive control and error processing, but the diverse findings on neural markers of anxiety in young children, which vary by severity and developmental stage, suggest the need for a wider perspective. Integrating new neural markers with established ones, such as the error-related negativity, the error positivity, and frontal theta, could clarify this association. Error-related alpha suppression (ERAS) is a recently proposed index of post-error attentional engagement that has not yet been explored in children with anxiety.

Methods: To identify neurobehavioral profiles of anxiety in young children by integrating ERAS with the error-related negativity, error positivity, frontal theta, and post-error performance indicators, we employed K-means clustering as an unsupervised multimetric approach. For this, we first aimed to confirm the presence and scalp distribution of ERAS in young children. We performed event-related potentials and spectral analysis of electroencephalogram data collected during a Go/NoGo task (Zoo Task) completed by 181 children (ages 4-7 years; 103 female) who were sampled from across the clinical-to-nonclinical range of anxiety severity using the Child Behavior Checklist.

Results: Results confirmed ERAS, showing lower post-error alpha power, maximal suppression at occipital sites, and less ERAS in younger children. K-means clustering revealed that high anxiety and younger age were associated with reduction in ERAS and frontal theta, less negative error-related negativity, enlarged error positivity, more post-error slowing, and reduced post-error accuracy.

Conclusions: Our findings indicate a link between ERAS, maladaptive neural mechanisms of attention elicited by errors, and anxiety in young children, suggesting that anxiety may arise from or interfere with attention and error processing.

Keywords: Childhood anxiety; ERN; Error processing; Error-related alpha suppression; Pe.

Figure 1.

Left panel: Visual depiction of the Go/No-Go Zoo task. Children could respond during the stimulus presentation or during the 500 ms blank screen. Right panel: Representation of EEG electrodes placement.

Figure 2.. Response-locked wavelet frequency-decomposed spectra for all channels.

Each spectrogram represents a scalp electrode displayed in its corresponding position. Spectrograms show the difference between errors minus correct responses and colder colors represent a suppression in alpha power (black arrows). Electrode O2 is zoom-in to show that time zero represents response onset and that occipital alpha power for errors was maximally suppressed during the 200–500 ms time window (black square).

Figure 3.. Correlation between post-error alpha power and age.

The scatterplot illustrates the correlation between the average occipital alpha power following NoGo errors of channels O1, O2, Oz, and age in months (Spearman's rho = −0.346; p-value < .001).

Figure 4:. Distinct profiles in relation to ERAs, error monitoring and task performance.

The plot shows the standardized means across clusters obtained by K-means clustering. Compared to cluster 2, subjects from cluster 1 were characterized by less negative ERN, more positive Pe, higher post-error alpha power, lower post-error theta power, slower post-error reaction time and less post-error accuracy. Abbreviations: ERN: Error related negativity for electrode FCz. Pe: Error positivity for electrode Pz. Post-Error Alpha: Average occipital alpha power (8–14 Hz) following NoGo errors of channels O1, O2, Oz (200–500 ms). Post-Error Theta: Theta power (4–7 Hz) at channel FCz (100–300 ms post response). Post-Error RT: Reaction time for correct responses after errors of commission. Post-Error Acc: Post-Error Accuracy.

Figure 5.. Comparisons of anxiety levels between clusters.

A Mann-Whitney U test indicated that individuals from cluster 1 presented higher CBCL-Anxiety scores than their counterparts from cluster 2 (U = 3171, p = 0.007, rank biserial correlation effect size = 0.225).