Background
Lang’s tripartite model of anxiety suggests that it consists of three response domains: cognitive, behavioural, and physiological [
1], which together result in a state of apprehensive worry, hyperarousal to threat cues, avoidance behaviours and negatively-biased cognitions [
2]. Each of these domains is suggested to measure a separate element of response characteristics and potentially independent underlying mechanisms to the construct of anxiety [
3]. An influential model of anxiety sees it as reflecting the engagement of the Behavioural Inhibition System (BIS) of which the hippocampus is a key component [
4]. Briefly, in Gray’s original account the role of BIS is to govern avoidance behaviours in response to threat and punishment. Excessive activity in BIS when driven by enhanced reactivity to threat/punishment cues manifests as higher proneness to anxiety.
In support of this idea Gray reviewed the evidence in the animal literature that anxiolytic drugs impair hippocampal function, specifically septo-hippocampal theta, to suggest that the hippocampus was the key substrate of BIS [
4]. Subsequent revision of the theory has incorporated other regions, most notably the amygdala, as a part of the BIS network, with the amygdala and hippocampus mediating different aspects of anxiety [
5,
6], and with the BIS interpreted as a conflict mediator system biased toward fight/flight/freeze behaviours and using exploration to resolve conflict. Critically, ensuing empirical work has continued to implicate hippocampal theta in anxiety and anxiolytic drug effects (for examples, see [
7‐
14]). For example, Gray and McNaughton [
5] observe that anxiolytic drugs, despite their neurochemical dissimilarity, commonly reduce the frequency of reticular-elicited hippocampal theta in the anaesthetised animal. We recently showed that Gray and McNaughton’s central observation extends to the awake, freely moving rat, where anxiolytic drugs reduce the frequency of natural theta obtained during locomotion [
14].
Two commonly used and well-validated instruments designed to measure individual differences in Gray’s BIS are the BIS section of the BIS/Behavioural Activation System scales [
15] and the Sensitivity to Punishment (StP) subscale of the Sensitivity to Punishment and Sensitivity to Reward questionnaire [
16]. These instruments have been shown to predict clinical anxiety disorders (for examples, see [
17,
18]), and likely capture cognitive and affective, rather than somatic, aspects of anxiety [
16]. Using these instruments and other indicators of BIS activity, neuroimaging studies have begun to implicate the hippocampus and amygdala in behavioural inhibition. Hahn and colleagues [
19] found that StP scores predicted hippocampus-amygdala functional connectivity in a monetary loss anticipation task. Further, it is conceivable that hippocampal structure, as well as activity, may be partly heritable. This is supported by a study by Oler and colleagues [
20] who investigated ‘anxious temperament’ in monkeys using a three-part composite measure of anxiety consisting of two behavioural BIS measures and cortisol release. They found that anxiety was clearly heritable, and that both hippocampal and amygdalar activity predicted anxiety, but only the hippocampal anxiety-related activity was heritable.
Together these findings suggest that BIS-related anxiety may be associated with structural variations in the brain. To our knowledge, only three studies have specifically related brain volume measures to BIS self-report [
21‐
23]. Interestingly, two of these found that (para)hippocampal volume positively correlates with behavioural inhibition, one using voxel-based morphometry (VBM) and the StP questionnaire [
22], the other using volume measures based on manual tracings and the BIS scale [
21]. In the VBM study, the region correlating with StP scores was largely parahippocampal, but reportedly also included the right hippocampus proper [
22]. A similar but weaker correlation based on a largely middle-aged sample was found in the manual tracing study [
21].
A different approach to the BIS has been to look at neural asymmetry in human scalp electroencephalography (EEG), with right brain dominance, particularly prefrontal, associating with higher behavioural inhibition [
24,
25] and anxiety [
26‐
29]. Intriguingly, simply being left-handed, and thus more likely to be right-hemisphere dominant, predisposes to higher BIS activity and anxiety [
30]. Hippocampal activity cannot itself be detected by scalp EEG, but animal models suggest that hippocampal influences on prefrontal EEG are important in anxiety [
7].
In the present study we used an automated segmentation method to obtain gray matter volumes of both the hippocampus and amygdala in healthy adult students, with no current or past history of any mental health disorder. Restricting our sample to young, well-educated adults may be important in minimising confounding effects of depression, stress and education. Torrubia and colleagues [
16] suggest that StP implements Gray’s theoretical construct of anxiety more faithfully than Carver and White’s BIS scale. Notably, for instance, in Gray’s conceptual revision of Eysenck’s theory of personality, Gray theorised that anxious people would be both ‘introverted’ and ‘neurotic’. Consistent with this prediction, StP scores are positively correlated with Neuroticism and negatively correlated with Extraversion [
16], whereas scores on Carver and White’s BIS scale tend only to be positively correlated with Neuroticism [
15]. Torrubia and colleagues [
16] also suggest that their focus on the response to particular cues was more in line with Gray’s theory. Accordingly, in order to relate brain structure to Gray’s BIS, we asked participants to complete the StP subscale of the Sensitivity to Punishment and Sensitivity to Reward questionnaire [
16]. Participants also completed questionnaires assessing other measures potentially associated with hippocampal volume: depression with the Beck Depression Inventory (BDI)-II [
31], negative life events with the Life Experiences Survey (LES) [
32]; and two other measures of anxiety: Trait anxiety of the State and Trait Anxiety Inventory (STAI-T) [
33] and the Beck Anxiety Inventory (BAI) [
34], the latter thought to be particularly sensitive to panic symptomatology [
35]. These different measurement instruments approach anxiety differently, which is why we choose to use them in this study. For instance, StP likely captures cognitive and emotional, but not somatic, components of anxiety, while BAI certainly does tap the somatic component [
16,
35]; trait anxiety as measured by STAI-T is dissociable from anxiety as mediated by BIS [
21], and may predict depression and negative affect as much as, or even more than, anxiety
per se[
36,
37]. If StP were found to be significantly related to hippocampal volume, we aimed to be able to examine the potential selectivity of this relationship.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
LL, TH and CL designed the study and performed the statistical analysis, and were involved in the writing of the manuscript. CB made a significant contribution to the laterality analysis and first draft. CB, AH, ES and EP collected the data, contributed to the initial data analysis and to the first draft of the paper. LL wrote the first draft of the manuscript. All authors read and approved the final manuscript.