Neuroticism is a personality trait that has been historically regarded as a stable trait (Cobb-Clark & Schurer, 2012) characterized by facets such as anxiety, hostility, impulsivity, and self-consciousness (Kim et al., 2017; Kurdek, 1997). Among personality traits, neuroticism is most predictive of negative mental health outcomes (Pereira-Morales et al., 2019), including depression (Jurczak et al., 2015; Nordfjærn et al., 2013) and anxiety (Üngür & Karagözoğlu, 2021), as well as physical health conditions such as cardiac disease (Suls & Bunde, 2005), asthma (Huovinen et al., 2001), and dementia (Wilson et al., 2011).

Individuals high in neuroticism are also high in susceptibility to stress (Francis, 1996) and stress reactivity (Costa & McCrae, 2008; Suls & Martin, 2005), which is often manifested by the presence of maladaptive coping strategies and intense negative affective responses when faced with daily stressors (Bolger & Zuckerman, 1995; Dunkley et al., 2014; Gunthert et al., 1999). Similarly, other studies have described a positive association between neuroticism and perceived stress (PS; Murberg & Bru, 2004), and research has found that PS may be an important link between neuroticism and other negative outcomes (Banjongrewadee et al., 2020; Smith & Gallo, 2001).

PS is defined as the degree to which an individual appraises life circumstances as stressful (Cohen et al., 1983), and presents negative thoughts or emotions (Banjongrewadee et al., 2020). PS is related to negative physical health outcomes (Hoferichter et al., 2014) including cardiovascular disease (O’Neal et al., 2015) and immunological issues (Assaf, 2013), and negative psychological outcomes including poor sleep (Charles et al., 2011), depression (Mohamadi Hasel et al., 2013; Zannas et al., 2012), substance use (Ng & Jeffery, 2003), and decreased quality of life (Assaf, 2013; Kondratowicz et al., 2021).

The relationship between neuroticism and PS is well documented, and early personality research hypothesized that individuals high in neuroticism are more likely to perceive life as stressful (McCrae, 1990). Subsequent work has supported this theory (Murberg & Bru, 2004), by demonstrating that neuroticism is a primary predictor of PS (Banjongrewadee et al., 2020; Conard & Matthews, 2008; Ebstrup et al., 2011; Mohamadi Hasel et al., 2013). Further research has suggested that increased neuroticism may result in higher PS and other negative outcomes by way of maladaptive stress reactivity (Banjongrewadee et al., 2020; Zuroff et al., 2004) as evidenced by dysregulated sympathetic nervous system (SNS) activity.

Facets of neuroticism such as aggression and hostility have been linked to nervous system functioning (Brummett et al., 2008), and neuroticism has been shown to positively relate to hyperarousal of the sympathetic nervous system (Eysenck, 1963; Eysenck & Eysenck, 1987) as indicated by increased skin conductance (Reynaud et al., 2012) and heightened activity and a lowered threshold in the limbic system (Ormel et al., 2013). A handful of studies have also utilized respiration rate (RR) as an indicator of physiological stress reactivity, and have found that individuals high in neuroticism have an increased RR during stressor tasks (Roslan et al., 2018; Wu et al., 2014).

Similar to neuroticism, PS is positively related to physiological stress reactivity (Javorka et al., 2018; Spodenkiewicz et al., 2018) and changes in RR in various settings (Grassmann et al., 2016; Naik et al., 2018; Tipton et al., 2017; Wheeler & Wilkin, 2008). Past studies of RR in response to stressful situations have concluded that RR is viable biomarker of individual perceptions of stress (Grossman, 1983). Further investigations of PS and respiration have evaluated interventions to lower PS via the manipulation of RR (Muthukrishnan et al., 2016; Naik et al., 2018). These studies found that such interventions reduce RR and PS concurrently (Huang et al., 2019; Muthukrishnan et al., 2016; Naik et al., 2018).

Despite the established relationship between neuroticism and PS, and the hypothesized role of stress reactivity as a link between neuroticism and PS, whether adaptive stress reactivity may buffer the relationship between neuroticism and PS has not been adequately explored. Though neuroticism and PS are ostensibly related, modifiers of this relationship require exploration. Particularly, it is unclear how RR during a stress task may impact this relationship. Such investigations are important in furthering our understanding of biomarkers of stress and personality, as well as illuminating potential buffers between neuroticism and PS which may serve as an important link to downstream outcomes. Further research in this area may also yield a clinical direction for individuals experiencing increased stress as the result of heightened neuroticism, and may particularly point to interventions that assist in regulating physiological stress reactivity and the SNS, such as meditation (Tang et al., 2009). The purpose of this study is two-fold: Primarily, we aim to investigate whether RR during a laboratory stress task moderates the relationship between neuroticism and PS. We hypothesize that participants with a low RR during the stress task will demonstrate a non-significant relationship between neuroticism and PS, whereas this relationship will be significant for participants who exhibit an elevated RR. Secondarily, we aim to elucidate the role of neuroticism in predicting RR during a stress task, and the role of RR during a stress task in predicting PS. We hypothesize that neuroticism will positively predict RR during a stress task, and that neuroticism and RR will each positively and significantly predict PS.


This investigation is a secondary analysis of data obtained from a parent study of an internet-based mindfulness course. Data were sampled from mildly stressed older adults (N = 64) at baseline in the parent study (for methodological details see Klee et al., 2020). The parent study was approved by the Oregon Health & Science University Institutional Review Board (IRB) and registered with (NCT02467660).


Participants were recruited from the Portland, Oregon area and completed various screens for the parent study (for a more detailed account of the recruitment and eligibility screening process, see Klee et al., 2020). Sixty-four generally healthy older adults were deemed eligible, completed baseline assessment, and are included in the present secondary analysis. Demographics may be found in Table 1.

Table 1


CharacteristicN = 64M ± SD (range)%
Age (years)59.97 ± 6.02
Education (years)16.27 ± 2.95
  African American23.12
  Native American0
Annual household income
  < $60,0003454.13
  > $60,0003046.87


Baseline RR was collected during an eyes-open auditory oddball paradigm. Participants listened to three types of pure tones presented at a randomized interval between 1500 and 2500 ms: low- (500 Hz), middle- (1000 Hz), and high-pitched (2000 Hz), comprising 10%, 80%, and 10% of the total 450 trials, respectively. Participants were instructed to ignore low and middle tones (non-targets), and to press a response button whenever they heard any high-pitched tones (target).

Participants then completed the Portland Arithmetic Stress Task (PAST; Atchley et al., 2017). The PAST was adapted from the Montreal Imaging Stress Task (Dedovic et al., 2005) to improve task parameterization, participant feedback, and timing accuracy. It presents mental arithmetic problems in tandem with distracting auditory feedback, and proceeds using an adaptive failure algorithm. Participants are asked to solve each problem mentally as quickly and accurately as possible, and then to use a stimulus-response box to select the correct answer. Performance is titrated to approximately 60% accuracy by the adaptive failure algorithm and participants are exposed to social pressure in the form of an on-screen marker for group average, which misleadingly indicates that average performance on the task is near ceiling. The PAST has been used previously to examine attentional and physiological responses to acute psychological stress (Atchley et al., 2017).

Utilization of the auditory oddball paradigm as a baseline for comparison when measuring RR during the PAST was constrained in part by the limited task set in the established parent protocol. The tones task was selected as a baseline for three primary reasons: 1) it is a relatively easy task (participants committed no errors, on average), 2) it is an approximate match in attentional engagement to the PAST (exogenous stimulus discrimination, decision making, and response action), and 3) Atchley et al. (2017) utilized a comparable paradigm for comparisons of physiological reactivity in a previous examination of the PAST. See our description of limitations in the Discussion section below for additional information.


Participants completed the following self-report measures at home. All study materials and data were exchanged between participants and study personnel securely using REDCap (Harris et al., 2009). REDCap utilization in this project was made possible by support from the National Institutes of Health (grant number: NIH UL1 TR002369).

NEO-FFI: Neuroticism. Neuroticism was measured using the NEO-Five Factor Inventory (NEO-FFI; Costa & McCrae, 2008). This scale contains 60 items divided across five facets: neuroticism, extraversion, openness to experiences, agreeableness, and conscientiousness. The neuroticism facet of the NEO-FFI demonstrated good internal consistency in the present sample (α = .86).

Perceived Stress Scale. Perceived stress (PS) was measured using the Perceived Stress Scale (PSS; Cohen et al., 1983). This scale contains 14 items that assess the degree to which one appraises life situations as stressful. The PSS-14 demonstrated good internal consistency in the present sample (α = .89).

Respiration rate. Respiration rate (breaths per minute) was measured during baseline and the PAST using an elastic piezoelectric belt (Ambu SleepMate) placed around the torso near the diaphragm. Data were recorded using BioSemi amplifiers and signal acquisition ActiView software and analysed offline in a BrainVision Analyzer (version, professional edition).


All statistical analyses were conducted using IBM SPSS Statistics version 24. No study variables demonstrated significant skewness or kurtosis, and no univariate or multivariate outliers were detected. Neuroticism and PS were assessed using the variance inflation factor (VIF) for problematic multicollinearity, and none was found. Means, SDs, and correlations for study variables may be found in Table 2. A post hoc power calculation was conducted to determine adequacy of sample size. Given a medium-to-large effect size (Cohen’s f 2), α = .05, and a sample size of N = 64, the present study achieved power of 0.84.

Table 2

Means, standard deviations, and correlations for study variables

VariableM (SD)12345
1. NEO-FFI-N22.81 (9.70)
2. PSS25.22 (8.31).72***
3. ST RR14.88 (3.62)–.18–.09
4. TT RR14.14 (2.80)–.05.07.53***
5. BC RR11.21 (3.89).***

[i] Note. NEO-FFI-N – NEO-Five Factor Inventory, neuroticism facet; PSS – Perceived Stress Scale; ST RR – stress task respiration rate; TT RR – tone task respiration rate; BC RR – tone task respiration rate; ***p ≤ .001.

To test our hypothesized moderation, we used the SPSS PROCESS macro (Hayes, 2018). We examined the interaction between the independent variable (neuroticism) and proposed moderator (RR during the PAST) in relation to the dependent variable (PS). A Johnson-Neyman output was generated, allowing us to investigate the significance of the relationship between neuroticism and PS at all levels of the moderator (RR during the PAST). A planned post hoc analysis using RR during a breath counting task was conducted. This sensitivity analysis allowed us to investigate the moderating role of stress task RR relative to two potential baselines and evaluate which may have been the better fit for the proposed model. Such an analysis was important in the present secondary analysis because neither the tone task nor breath counting was intended as a baseline in the parent study. A p value of .05 was used for tests of significance.


Correlations between study variables revealed significant correlations between PS and neuroticism (p < .001), and between tones task RR and stress task RR (p < .001). No study variables were significantly correlated with demographic variables. As shown in Table 3, the interaction between neuroticism and RR during the PAST was significant (p = .031), indicating a moderation effect. Contrary to our hypothesis, at pre-defined levels of the moderator (–1 SD, mean, +1 SD) the relationship between neuroticism and PS was significant. However, simple slope (see Figure 1) and Johnson-Neyman plots (see Figure 2) revealed that only participants with a very low (–1.78 SD; p = .072) RR exhibited a non-significant relationship between neuroticism and PS, whereas participants with a low (p < .001), medium (p < .001) or high level of stress task reactivity RR (p < .001) exhibited a statistically significant relationship between neuroticism and PS. As is visible on the Johnson-Neyman plot (see Figure 2), the confidence band surrounding the slope of neuroticism and perceived stress across values of ST RR includes zero at respiration rates of ≤ 9, indicating that participants who exhibited RR at or below this value also demonstrated a nonsignificant relationship between neuroticism and perceived stress. Unexpectedly, the main effect for stress task RR predicting PS was non-significant (p = .072). This finding is represented in Figure 1, which demonstrates the cross-over effect of the moderation. A post-hoc analysis examined a breath counting task RR as an alternative baseline to the tones task RR, and results did not differ (neuroticism × breath counting task RR interaction: p = .023).

Table 3

Results of a hierarchical regression model testing neuroticism, respiration rate, and their interaction in the prediction of perceived stress

ΔR2BSE Bβ95% CIp
Step 1.02
  Race Age (White)–.03
Step 2.00
  Race (White)–.02.39–.01–.79.75.954
Step 3.54
  Race (White)–.33.26–.11–.86.20.215
Neuroticism.< .001
Step 4.00
  Race (White) Age–.33
  TT RR.09.11.09–.12.31.391
  Neuroticism.< .001
  ST RR–.01.11–.01–.23.21.941
Step 5.04
  Race (White)–.41.26–.14–.93.12.123
  TT RR.11.10.11–.10.31.318
  Neuroticism.< .001
  ST RR–.04.11–.04–.26.18.712
  Neuroticism × ST RR.

[i] Note. TT RR – tone task respiration rate; ST RR – stress task respiration rate; LL – lower limit; UL – upper limit.

Figure 1

Simple slopes of neuroticism (NEO-FFI-N) in the prediction of perceived stress (PSS) at high (+1 SD), medium (mean), low (–1 SD), and very low (–1.78 SD) values of stress task respiration (ST RR)

Note. NEO-FFI-N – NEO-Five Factor Inventory, neuroticism facet.
Figure 2

Johnson-Neyman significance region

Note. ST RR – stress task respiration rate.


The primary goal of this study was to investigate whether physiological stress reactivity, as indicated by RR during a stress task, moderated the relationship between neuroticism and PS. We hypothesized that the moderation analysis (i.e., the interaction term neuroticism × stress task RR) would significantly predict perceived stress, and that participants with low RR during the laboratory stress task would demonstrate a non-significant relationship between neuroticism and PS.

Initial analyses were baseline correlations between study variables. Correlations revealed expected significant relationships between neuroticism and PS, and between stress task RR and tones task RR. However, stress task RR was not significantly correlated with either neuroticism or PS. The lack of correlation between neuroticism and stress task RR was surprising, given the established relationship between neuroticism and increased stress reactivity (Costa & McCrae, 2008; Suls & Bunde, 2005), and between neuroticism and SNS activity (Eysenck, 1963; Eysenck & Eysenck, 1987). Similarly, the lack of correlation between PS and stress task RR contradicts past research demonstrating a positive relationship between PS and physiological stress reactivity (Javorka et al., 2018; Spodenkiewicz et al., 2018), and between PS and SNS activity (Grassmann et al., 2016; Naik et al., 2018; Tipton et al., 2017). These null findings may be because respiration rate is a complex and multifaceted phenomenon (Grossman, 1983), and because neuroticism is composed of multiple components, such as aggression and hostility (Brummett et al., 2008). Due to the complex nature of these variables, it is possible that certain components of each are related, but not the singular constructs. It may also reflect the contradictory body of research regarding the directionality of the relationship between physiological stress reactivity and neuroticism (Evans et al., 2016; Ormel et al., 2013), and between physiological stress reactivity and PS (de Rooij et al., 2010; Dishman et al., 2000; Javorka et al., 2018). For example, although studies have demonstrated a positive relationship between neuroticism and stress reactivity (Evans et al., 2016), neuroticism has also been linked to chronic stress conditions such as burnout (Azeem, 2013; Goddard et al., 2004). Burnout is in turn related to blunted physiological stress reactivity (Penz et al., 2019; Wekenborg et al., 2019). Depending on the chronicity of their stress, some participants in the present study may have demonstrated a blunted physiological stress reactivity profile, while others may show an exaggerated profile, thereby accounting for the lack of correlation between neuroticism and stress task RR.

The present study found that physiological stress reactivity significantly moderated the relationship between neuroticism and PS. The relationship between neuroticism and PS was nonsignificant only for participants who demonstrated very low physiological stress reactivity. This finding aligns with previous research which has suggested that stress reactivity may be an important link between neuroticism and negative outcomes (Banjongrewadee et al., 2020; Smith & Gallo, 2001), and studies that have specifically related maladaptive SNS reactivity to both facets of neuroticism such as hostility and aggression (Brummett et al., 2008) and to PS (Grossman, 1983; Muthukrishnan et al., 2016; Naik et al., 2018). However, to our knowledge this is the first study to examine such a model using an index of physiological stress reactivity. This finding suggests that individuals with high neuroticism who are better able to regulate physiological responses to acute stressors may be less inclined to perceive their circumstances as stressful. Past research has found that dysregulated physiological stress reactivity is associated with PS (Ginty & Conklin, 2011) among samples vulnerable to stress (Obasi et al., 2017; Pierrehumbert et al., 2009; Seitz et al., 2019), and that interventions that target nervous system regulation such as mindfulness training improve physiological stress reactivity among healthy adults (Bullis et al., 2014), stressed adults (Lindsay et al., 2018), and patients with substance use disorder (Brewer et al., 2009). In light of the present study’s findings alongside this past research, future mechanistic research may investigate whether interventions such as mindfulness training may benefit stress-related outcomes such as PS via improved physiological stress reactivity among vulnerable populations. Our findings may also suggest that participants who decouple neuroticism and perceived stress are less physiologically reactive during acute stEarlier researchers alluded to the role of hyperventilation (e.g., elevated RR) in the development and maintenance of disorders related to neuroticism (Lum, 1975; Pfeffer, 1978). Past research found strengthened associations between neuroticism and apprehension, anxiousness, tension, and unhappiness following a period of purposeful hyperventilation (Clark & Hemsley, 1982). More recent research has found that individuals experiencing increased life stress and who have premorbid personality characteristics such as neuroticism are more susceptible to developing hyperventilation syndrome (Shu et al., 2007). Our finding that participants with higher-than-average RR demonstrated a significant relationship between neuroticism and PS aligns with this research. Our data suggest that participants with high levels of neuroticism were also likely to show elevated PS if their RR was increased during the stress task.

The secondary goal of this study was to elucidate the relationships between neuroticism, PS, and RR. We hypothesized that neuroticism and RR would independently significantly predict PS. Our hypothesis was partially supported: neuroticism significantly predicted perceived PS. This finding mirrors previous findings of the positive relationship between neuroticism and PS (Banjongrewadee et al., 2020; McCrae, 1990; Mohamadi Hasel et al., 2013). Contrary to our hypothesis, the relationship between physiological stress reactivity and PS was non-significant. As previously mentioned, research that has explored this relationship has been inconsistent. Some studies have described a significant relationship between respiration and PS (Dishman et al., 2000), while others have found that stress appraisals do not correspond to physiological measures (Schneider, 2004). This inconsistency may be the result of intra-individual variability in physiological stress reactivity. While some studies have found a relationship between exaggerated physiological stress reactivity and PS (Brodersen & Lorenz, 2020), others have found that PS is related to suppressed or blunted physiological stress reactivity (de Rooij et al., 2010). Further, both exaggerated and blunted physiological stress reactivity have been found to relate to negative outcomes (al’Absi et al., 2013; Brown et al., 2018; Carroll et al., 2017; Lovallo et al., 2016; Turner et al., 2020). This variability in dysregulated stress reactivity may be due to individual factors such as early childhood adversity (Repetti et al., 2002; Woody & Szechtman, 2011), social support (Closa León et al., 2007), and chronic stress (Kühnel et al., 2020) and related conditions such as burnout (Penz et al., 2019; Wekenborg et al., 2019).

There are several limitations of this study. The sample in the parent study was limited to older, mildly stressed, predominately white, highly educated female-identifying adults who were recruited from a small geographical area. These characteristics limit the generalizability of our findings. Further, use of the auditory oddball paradigm as a baseline for RR comparisons is imperfect. As mentioned previously, task selection was constrained by the parent study, which collected RR data during three phases: auditory oddball, breath counting, and the PAST. A true resting state is hard to define given that a simple rest state with no task is vulnerable to confounding related to simple arousal (Shields et al., 2016). Relatedly, the breath counting task was poorly matched to task in terms of attentional engagement, and it may have artificially lowered participants’ RR, as observed previously (Atchley et al., 2016). Though not perfectly matched to the PAST, the auditory oddball task more closely approximated cognitive engagement in terms of attention to exogenous stimuli, target discrimination, decision making, and stimulus-response actions. In addition to addressing these limitations, future studies may consider including ecological momentary assessment (EMA) measures of stress in order to elucidate the relationships between acute psychological stress and acute physiological stress reactivity. Prior studies have found higher levels of stress reported via EMA than collected during laboratory stressor paradigms in highly stressed populations such as dementia caregivers (Fonareva et al., 2012). Additionally, research has found that end-of-day ecological stress reporting methods such as journaling are highly internally reliable, and correspond uniquely with constructs such as general well-being and coping style (Ford et al., 2017, 2018). Similarly, physiological stress data collected via EMA methods (e.g., diurnal cortisol) have been associated with EMA PS (Lazarides et al., 2020). The unique benefits of ecological self-reported and physiological stress assessment, and the relationship between the two, warrant further investigation. This approach may yield novel insights into the relationship between “real world” stress and physiological stress reactivity, and personality traits.

In summary, using physiological monitoring of RR during different conditions, this study demonstrated that participants with very low physiological reactivity during a cognitive stress task experienced a buffered relationship between neuroticism and PS compared to those with higher levels of physiological reactivity.