Psychological and Physiological Changes during Basic, Underwater, Demolition/SEALTraining

Abstract

This longitudinal study examines the growth of psychological characteristics and adaptation of physiological stress markers during a six-month assessment and selection course for U.S. Navy SEALs. Resilience, hardiness, and grit instruments were used to evaluate the psychological characteristics. Blood samples were taken to determine physiological markers related to stress adaptation, specifically evaluating DHEA, DHEA-to-cortisol ratio, BDNF, NPY, and cortisol. Data was collected at four time points throughout the assessment and selection course from 353 students over three classes. Results indicated that resilience and hardiness grow after an initial decline, DHEA and DHEA-to-cortisol increased, suggesting physiological adaptation. However, psychological and physiological markers do not exhibit the same growth patterns for participants in the course. This study enhances the understanding of psychological growth and physiological adaptation in a high-stress environment over an extended duration.

1. Introduction

Special Operations Forces (SOF) are elite military personnel with specialized skill sets that enable them to conduct high-risk missions under arduous conditions (Shamir & Ben-Ari, 2018). As such, SOF personnel undergo a grueling selection process and extensive training to develop attributes that enable mission success. Despite making up 3% of the joint forces (Vergun, 2021), the U.S. military relies heavily on SOF to conduct complex worldwide missions (Benedict et al., 2022). Moreover, it takes years and millions of dollars to fully train each special operator. Given the time and resources involved in recruiting, training, and retaining SOF operators, advances in our understanding of the development of characteristics that make candidates successful throughout their careers are needed. The knowledge of these characteristics and how they develop can assist in cultivating an environment that supports greater growth, which may benefit the special operations community and military, as well as other high-performing organizations.

Our research examines the psychological and physiological growth of Navy, Sea, Air, and Land Teams (SEALs) candidates in Basic Underwater Demolition SEAL (BUD/S) training, commonly acknowledged as one of the most demanding SOF schools in the U.S. military (Benedict et al., 2022; Ledford et al., 2020). SEALs face extreme and high-pressured environments both in their training and in combat (Smith et al., 2020). BUD/S is designed to assess and select service members equipped to deal with the physical and psychological challenges that SEALs will face in their careers (Ledford et al., 2020). In this study, we examine growth—change in a beneficial direction—in psychological characteristics and physiological markers of stress commonly associated with some psychological characteristics. For psychological characteristics (resilience, hardiness, and grit), growth refers to an increase in the respective characteristic. For physiological markers, growth can refer to a change in biomarker concentration that leads to a favorable adaptation to stress by either increasing or decreasing. For example, a decrease in cortisol with repeated exposure to the same stressor may indicate stress habituation (Gifford et al., 2019) and considered growth in the context of the present study.

1.1. Psychological Indicators of Performance

For decades, researchers have examined how psychological factors operate and enhance one’s ability to perform in various settings, including, but not limited to, academics (e.g., Ayala et al., 2018; Duckworth et al., 2011; Kotzé et al., 2013), management and business (e.g., Jordan et al., 2019; McManus et al., 2007, Mooradian et al., 2016), athletics (Sarkar & Fletcher, 2014, Madrigal et al., 2016), and the military (e.g., Bartone, 1999; Zakin et al., 2003). Research indicates three human characteristics that contribute to performance in these settings: resilience (e.g., Ledford et al., 2020), hardiness (e.g., Beasley et al., 2003; Westman, 1990), and grit (e.g., Caza & Posner, 2019; Maddi et al., 2017). These psychological characteristics are particularly important for those operating in high-stress environments, especially those in the military (Bartone, 2006; Eid & Morgan, 2006; Kelly et al., 2014; Matthews, 2008).

Resilience is the ability to adapt and recover from adversity (Southwick et al., 2014). Psychological resilience has proven to be important in military contexts because of its value for preparing for combat stress (Bonanno, 2005). Hardiness is consistently recognized as a personality trait composed of a sense of control, commitment, and an openness to challenge. Hardiness is also important in high-stress military situations, such as combat, because of its relationship with post-traumatic stress (Escolas et al., 2013), depression (Dolan & Adler, 2006), and success in training (Bartone et al., 2008). Grit is understood as working passionately toward a long-term goal despite setbacks and obstacles (Duckworth et al., 2007). Studies of grit in military settings consistently indicate that higher levels of grit lead to higher levels of performance (e.g., Maddi et al., 2017) and retention (e.g., Eskreis-Winkler et al., 2014; Kelly et al., 2014). Although resilience, hardiness, and grit are positively correlated with one another (e.g., Georgelous-Sherry & Kelly, 2019, Ledford et al., 2021; Martin et al., 2015; Matthews et al., 2019), Georgelous-Sherry and Kelly (2019) and Ledford et al. (2021) concluded they are distinct characteristics. Thus, each characteristic may play a unique role and operate in a specific way as one persists through difficult training environments, such as SOF training.

Much of the literature on resilience, hardiness, and grit focuses on the relationship between these factors and performance outcomes. However, there are indications that these psychological characteristics grow or can be developed. Specifically, resilience scholars highlight four responses to adversity (1) self-destructive response - health and wellbeing decline in combination with self-destructive behaviors, (2) increased fragility/less resilience - health and wellbeing decline post-disruption, (3) return to normal - return to the same state of health and wellbeing prior to the event, and (4) growth / more resilience - grow in the face of adversity; the event provides an opportunity to gain strength (Bonanno, 2004, 2005; Connor & Davidson, 2003; Harms et al., 2018; Masten et al., 1990).

One’s ability to either recover (return to normal) or even grow as a result of stressors may be fostered intentionally; Seligman (2007, 2011) highlights how resilience and grit can be developed over time. The U.S. military studied and focused on creating programs to help service members develop resilience to deal with the challenges of deployments (Meredith et al., 2011). The U.S. Army uses a Master Resilience Training program, developed from tenets of positive psychology, to train soldiers to be more resilient (Reivich et al., 2011). Bartone (2006) argued that in military settings, leaders can boost an individual’s hardiness, which in turn helps increase the resilience of a group. In a longitudinal study examining grit and growth mindset in adolescents, growth mindset was found to predict increases in grit over a two-year period (Park et al., 2020), thus, supporting the idea that grit can develop or grow over time. While these studies suggest that these characteristics can be developed intentionally (e.g., Bartone, 2006; Park et al., 2020; Seligman, 2007, 2011), there is limited empirical evidence to support or deny the possibility that resilience, hardiness, and grit can grow without specific intervention.

1.2. Physiological Indicators of Stress

While there are psychological indicators of performance that may be developed and help one progress through challenging situations, it is also important to consider physiological indicators of stress which may hinder or enhance one’s persistence through difficult situations. In the presence of a perceived threat, the brain initiates a response that engages the immune and cardiovascular systems via neuroendocrine mechanisms, including the secretion of cortisol and dehydroepiandrosterone (DHEA) (Osorio et al., 2017). Cortisol is essential to stress adaptation by mobilizing and replenishing energy stores, suppressing nonessential anabolic activity, and increasing arousal and cardiovascular tone (Charney, 2004; Morgan et al., 2000a; Osorio et al., 2017; Russo et al., 2012), though prolonged cortisol exposure can have negative effects on the body (Kamin & Kertes, 2017). DHEA modulates the effects of cortisol by providing neuroprotective and anti-inflammatory effects, including stimulation of neural stem cells and enhancement of immune cell production (Charney, 2004; Kamin & Kertes, 2017; Russo et al., 2012). As a result, the ratio of DHEA-to-cortisol is often used to represent the balance between anabolic and catabolic processes (Lennartsson et al., 2012).

Other neuroendocrine markers that have a protective role in the stress response include neuropeptide-Y (NPY) and brain-derived neurotrophic factor (BDNF). NPY is a pancreatic polypeptide that inhibits excessive activation of the stress response and possibly acts as an endogenous anxiety-reducing agent to buffer the effects of stress in the brain (Hirsch & Zukowska, 2012). BDNF is also released both centrally and peripherally to protect neurons and mobilize energy resources during stress (de Assis & Gasanov, 2019). Given their role in the stress responses, circulating biomarkers such as those discussed here are commonly used to monitor the physiological status and measure the impact of operational stress during military training (Henning et al., 2014; Morgan et al., 2000a; Nindl et al., 2012; Nindl et al., 2003; Nindl et al., 2007; Szivak et al., 2018). Following eight weeks of U.S. Army Ranger training, cortisol concentration increased by 21%–50% from baseline levels (Friedl, 2000; Henning, 2014; Nindl, 2007), with a concomitant 33% decrease in BDNF concentrations (Henning, 2014). During Survival, Evasion, Resistance, and Escape (SERE) school, a three-week training course designed to help individuals who may be captured, cortisol and NPY increased to 253% and 229%, respectively, from baseline (Morgan et al., 2002).

While intense military training can have a negative impact on circulating biomarkers, several studies have demonstrated favorable neuroendocrine adaptations that may suggest physiological growth (i.e., improvement) can occur during these scenarios. Notably, higher DHEA concentrations were associated with fewer stress-induced symptoms of dissociation (detachment from surroundings) among soldiers following a highly stressful underwater navigation task (Morgan et al., 2009). U.S. Army SOF soldiers had significantly higher concentrations of NPY following interrogation stress than non-SOF soldiers, suggesting the soldiers who had undergone SOF training had a better anxiolytic response to the stressful interrogation (Morgan et al., 2000b). Evidence for a growth process was also evident in female U.S. Marines completing 13 weeks of recruit training (Lieberman et al., 2009). Compared to pre-training, there was a 21% decline in cortisol accompanied by favorable changes in metabolic status, mood, and body composition (Lieberman et al., 2009). Taken together, the assessment of circulating biomarkers may be important indicators of growth during high-stress military training (Jensen et al., 2019).

1.3. Psychological and Physiological Indicators in Combination

While many studies have examined relationships between psychological or behavioral metrics with stress biomarkers during an acute period (Farina et al., 2019; Ledford et al., 2020; Morgan et al., 2000a; Morgan et al., 2009), few studies have investigated the growth process of psychological and physiological markers across military training, or the extent to which these processes co-occur. Jensen et al. (2020) observed that Marines receiving mental skills or mindfulness training exhibited significantly lower cortisol and epinephrine concentrations in response to ambush training than those receiving standard military training. Gifford et al. (2019) examined the adaptation of women during infantry-based training in the U.K. They found that psychological resilience decreased over a 44-week training program, while cortisol initially increased in training and then decreased. This result indicates habituation to the training relative to the physiological markers but a decrease in psychological capabilities. However, few other studies have examined the adaptation or growth of psychological performance indicators and physiological markers to understand how they develop or diminish in strenuous environments.

Psychological performance indicators, such as resilience, hardiness, grit, and physiological factors, are important for elite warfighters to endure the stressors of BUD/S (Ledford et al., 2020; 2021). The course takes place over a six-month period which involves several phases designed to task the SEAL candidate’s mentally and physically to assess their stamina of mind and body. Thus, BUD/S provides an ideal environment to examine how one psychologically and physiologically develops within a strenuous environment. The purpose of this study was to identify growth patterns in self-reported psychological measures and physiological indicators (i.e., circulating blood biomarkers) during BUD/S selection training. We hypothesized that participants at BUD/S training would experience growth in resilience, hardiness, and grit throughout the program. We also hypothesized that participants would experience growth adaptations in neuroendocrine responses, specifically increases in DHEA, DHEA-to-cortisol ratio, BDNF, and NPY, accompanied by a decrease in cortisol. Given the resources required to train and operate in high-stress environments (Taylor et al., 2006), results from the study can elucidate the importance of monitoring individuals’ psychological and physiological growth trajectories in arduous training programs.

2. Methods

2.1. Study Overview

The accession and selection program for the U.S. SEAL Teams resides at the Naval Special Warfare Center in Coronado, California, and lasts approximately 28 weeks (Ledford et al., 2020). BUD/S consists of one pre-phase, (0) Basic Orientation (four weeks) focused on familiarizing students with standard operating procedures and with the training evolutions, and three phases: (1) Phase 1 (eight weeks) dedicated to physical conditioning; (2) Phase 2 (eight weeks) dedicated to combat diving; and (3) Phase 3 (eight weeks) dedicated to weapons and demolition. Each phase has a grueling week designed to maximize stress and friction between participants. There are usually six BUD/s classes per year that begin BO with approximately 200 to 300 students. The attrition rate for first-time participants is approximately 70% to 85% (Ledford et al., 2022). This accession and selection process is designed to ensure that students meet the minimum mental and physical toughness threshold. This study conducted psychological surveys and blood collection at the start of training (at the end of BO) as well as the end of each of the two-month phases of BUD/S. The corresponding author’s institutional IRB approved the study.

2.2. Participants

The present study followed participants from three BUD/S classes. All participants were active duty men in the U.S. Navy, specifically screened for their athleticism, mental stability, and intelligence. All participants were briefed in person about the study. Volunteers were provided written informed consent prior to participation. Participants could choose not to participate or withdraw from the study at any time.

In general, candidates can progress through the training in several ways: (1) candidates can go straight through the six-month course; (2) candidates can quit on their own accord (Drop on Request - DOR); (3) candidates can be “dropped” by the BUD/S cadre for poor performance; or (4) candidates can be “rolled” meaning they have a setback (often medical) that requires them to heal/recover and then continue with a later class. Candidates can be rolled into a different class multiple times. In addition, candidates that rolled may reappear in the next class or a subsequent class months later.

2.3. Psychological Measures

Resilience

This study used 16 items from the Connor-Davidson Resilience Scale (CD-RISC; Connor & Davidson, 2003). The CD-RISC is a self-assessment instrument that uses five-point Likert scaled items (zero-not true at all- through four-true nearly all the time) to measure resilience based on previously identified characteristics shared among resilient individuals (e.g., adaptability, self-efficacy, tolerance of negative affect). The scores could range from zero to 64 with higher scores indicating higher resilience. The CD-RISC has been tested in the general population as well as clinical samples with good internal consistency and test-retest reliability and has been shown to be valid compared to other measures (Connor & Davidson, 2003; Davidson, 2019). The 16 items used in this study were validated in previous research on this unique population of SEAL candidates (Ledford et al., 2021).

Hardiness

Hardiness was measured with 14 items from Bartone et al.’s (1989) 15-item Dispositional Resilience Scale (DRS), which is comprised of three factors: commitment, control, and challenge. DRS is a self-assessment instrument that uses four-point Likert scaled items (zero-not at all true- through three-completely true). The scores could range from zero to 42, where higher hardiness scores indicate greater hardiness. Six of 14 items are negatively worded (i.e., agreement with the item indicates lower levels of hardiness); these items were reverse coded for the analyses.

Grit

Grit was measured using the ten items from Duckworth et al.’s (2007) Grit Scale that was validated in a sample of SEAL candidates at BUD/S (Ledford et al., 2021). Items are measured on with five-point Likert scaled items (one-not like me at all- through five-very much like me). The scores could range from 10 to 50 with higher grit scores indicating greater grit. Four of the ten items were negatively worded and reverse coded for the analyses.

2.4. Biological Collection Procedures

A 21-gauge needle (BD Vacutainer ^® Eclipse and Vacutainer ^® one-use holder, Becton, Dickinson and Company, Franklin Lakes, NJ) was used to collect 12 mL of blood at each blood collection. Due to the nature of training and logistics, sample collection time points varied between classes but were kept consistent within each class. Specifically, samples from two of the three classes were collected in the morning across all time points, whereas samples from the third class were collected in the afternoon across all time points. Serum was obtained from SST tubes (Becton, Dickinson and Company) by allowing the blood to clot for 30 minutes at room temperature, then centrifuged at 1500 × g for 15 minutes. Plasma was obtained from blood drawn into EDTA tubes (Becton, Dickinson and Company) and subsequently centrifuged immediately after collection at 1500 × g for 15 minutes at room temperature. The supernatant of plasma and serum samples were aliquoted into 500 mL cryotubes, packed on dry ice, and shipped overnight to the University of Pittsburgh. Samples were stored at −80°C until ELISA assays were conducted for DHEA (Eagle Biosciences, Amherst, NH), cortisol (ALPCO, Salem, NH), BDNF (EMD Millipore, Burlington, MA), and NPY (R&D systems, Minneapolis, MN). All assays were measured in duplicate, and coefficients of variation were below manufacturers’ reported variance and above the assay level of sensitivity, which was 0.15 ng/mL for DHEA, 0.4 μg/dl for cortisol, and 0.23 pg/mL for BDNF. Sensitivity data was unavailable for NPY.

2.5. Statistical Analysis.

This study used growth modeling methods for assessing patterns of change in repeated measures in longitudinal studies. Growth models are hierarchical models in which repeated measures are viewed as arising from intercepts and growth factors (i.e., slopes) that are unique to each individual (Bollen & Curran, 2006; Lynch & Taylor, 2016; Raudenbush & Bryk, 2002). A basic growth model can be specified with “level-1” and “level-2” equations. In a simple growth model, the level 1 equation is:

$y_{it}=b_{\mathit{0}i}+b_{\mathit{1}i}{t}_{it}+b_{\mathit{2}}{x}_{it}+e_{it},$

where $y_{it}$ is the outcome, $y$, at time $t$ for individual $i,b_{\mathit{0}i}$ is an individual-specific intercept, $b_{\mathit{1}i}$ is an individual-specific (here, linear) growth factor reflecting change across time, $t_{it}$ is the measure of time for individual $i$ at time $t,b_{\mathit{2}}$ is the effect of a time-varying covariate, $x_{it}$ and $e_{it}$ are individual- and time-specific error terms.

Level 2 equations decompose the individual-specific intercepts and growth factors as functions of time-invariant covariates and random effects:

$b_{\mathit{0}i}=g_{\mathit{0}}+g_{\mathit{1}}{z}_i+u_i$

$b_{\mathit{1}i}=d_{\mathit{0}}+d_{\mathit{1}}{z}_i+v_i,$

where $g_{\mathit{0}}$ is an adjusted mean for the intercepts, $g_{\mathit{1}}$ is the effect of a time-invariant covariate $z_i$ on the intercept, $u_i$ is the unexplained component of $b_{\mathit{0}i}$, and $d_{\mathit{0}},d_{\mathit{1}}$, and $v_i$ are the counterparts of $g_{\mathit{0}}$, $g_{\mathit{1}}$, and $u_i$ for the growth factors. $u$ and $v$ are usually assumed to follow a bivariate normal distribution with variances ${\tau }_{\mu }$ and ${\tau }_{\mu }$ and covariance ${\tau }_{\mu \nu }$. Both the individual-specific intercepts and growth factors and the variances, $\mathrm{\tau }$, are called the “random effects” depending on the discipline, statistical approach, and software used (Allison, 2009). The Level 2 equations can be substituted into the Level 1 equation to produce a reduced form equation with an error term that consists of a clustered component, a heteroscedastic component, and an independent and homoscedastic Level 1 residual.

Growth models can be estimated using specialized software for hierarchical models or procedures built into most general statistical packages or using structural equation modeling (SEM) software. Here, we use Stata’s mixed procedure 16.1 (StataCorp, LLC, College Station, TX). The data consisted of $n=692$ records, such that each row in the data set represented a single measurement occasion for a BUD/S trainee. Given that we observed each candidate after each phase of training and prior to the start of the first phase, each candidate could contribute at most four records to the data.

We estimated a series of growth curve models for each psychological measure and biomarker $\left(y_{it}\right)$. We report results from our final models here (other results are available upon request). Our final models included an indicator for the beginning of training because a preliminary investigation of individual patterns in the various outcomes showed that psychological and physiological responses prior to training were substantially different than (inconsistent with) the linear pattern observed after each phase. Thus, our reduced-form model is:

$y_{it}=b_{\mathit{0}}+b_{\mathit{1}}I\left(p_{it}=\mathit{0}\right)+b_{\mathit{2}}{p}_{it}+b_{\mathit{3}}ag{e}_i+w_{it}$

where $p_{it}$ represents phase of training (i.e., at the $\mathrm{s}\mathrm{t}\mathrm{a}\mathrm{r}\mathrm{t}\mathrm{o}\mathrm{f}\mathrm{t}\mathrm{r}\mathrm{a}\mathrm{i}\mathrm{n}\mathrm{i}\mathrm{n}\mathrm{g}\left.=0,1,2,3\right)$, $I\left(p_{it}=0\right.$) is an indicator for whether the person-phase record represents measurement at the beginning of training, , $ag{e}_j$ is the age of the trainee at entry into the program, $w_{it}$ contains the (Level 2) random effects $u$ and $v$, and the Level 1 residual, $e$.

There were at least two reasonable approaches for measuring time in this study: the phase of BUD/S and the total number of phases a candidate had attended prior to the current one, including repeats. Most candidates who successfully complete BUD/S end up repeating at least one training phase. On the one hand, one could argue that those who repeat multiple phases have greater room for growth so the total number of phases under their belts should be the appropriate clock for our growth models. On the other hand, those who do not complete a phase do not spend the full time in the phase. Ultimately, we estimated models using both potential clocks and obtained nearly identical results, most likely because most candidates who did not complete BUD/S on the first try only repeated one phase, so that phase and the total number of phases are extremely highly correlated (r=.94).

Although our data contained several potentially important covariates, including BUD/S class, we included only age at the start of training in our model because it is potentially relevant both to psychological maturity and physiological ability. Additional models, including various combinations of these additional covariates, did not affect our results. Further, we excluded measures of BUD/S class, race, region of residence, and family income because there is relatively little variation in these items, and none is consistently related to any psychological measure, biomarker, or completion of the program.

3. Results

In all, we surveyed candidates from three classes across eight months and obtained a total of 692 individual responses from 380 individuals. Data were collected by class, with classes A, B, and C contributing 123, 119, and 138 initial usable responses, respectively. Because candidates progressed through the program in many ways, it was impossible to exactly map each candidate’s progress; rather, we describe the broad approach we took to collect data. Class C could not be surveyed a fourth time (as we were able to do with classes A and B) due to COVID-19 restrictions. Of the 380 candidates in our study, 151 ultimately completed BUD/S, while 158 DOR’d, 39 were dropped for performance, and 32 were medically dropped or rolled. Among the 151 successful candidates in our study, 54 of them (36%) completed BUD/S in the class they started, 68 (45%) completed BUD/S over two classes, 22 (15%) over three classes, six (4%) over four classes, and one (<1%) over five classes. Descriptive statistics (Table 1) and model results report varying sample sizes due to incomplete information for some variables.

Table 1.

Descriptive statistics for BUD/S candidates

Factor		Mean ± sd or Frequency (%)
Total Sample (N)		380
Age		23.32 ± 3.04
Years of College		2.51 ± 1.84
White vs Non-White		323 (87.29%)
Officer vs Enlisted		43 (11.59%)
Region of US: South vs Non-South		134 (36.21%)
Partnered vs Unpartnered		43 (11.59%)
Independent Variable
Completed BUD/S		151 (39.73%)
DOR		158 (41.57%)
Performance/medical Drop		71 (18.68%)
Outcome Variable
Completed BUD/S		151 (39.73%)
Usable Responses by Class		Usable responses by Phase
BUD/S Class 1	123	Phase 1	371
BUD/S Class 2	119	Phase 2	145
BUD/S Class 3	138	Phase 3	110

Abbreviations: BUD/S = Basic, Underwater, Demolition/SEAL. DOR = Dropped on Request.

Table 2 presents the results of the growth models for all three psychological outcomes and five biomarkers. Additional models, including various combinations of two additional covariates—officer vs. enlisted status and educational attainment—did not differ from those reported; therefore, these two covariates were not included in the final model. The upper half of Table 2 shows results for all sample members, including those who dropped out of the program at some point. The bottom half of the table shows results for only those who successfully completed BUD/S. The results of the two sets of models are substantively identical; thus, we only discuss results for the full sample. Each column of the table represents a distinct outcome.

Table 2.

Results of growth models for psychological and biomarker outcomes.

Full Sample
Parameter	Resilience	Grit	Hardiness	Cortisol	DHEA	Ratio	BDNF	NPY
Fixed Effects
Intercept	53.7(2.3)***	39.2(1.7)***	34.5(1.5)***	14.1(2.6)***	45.1(5.3)***	3.4(.52)***	12.2(12.1)	32.6(6.7)***
Slope	.59(.24)*	.12(.16)	.44(.15)**	−.57(.39)	2.32(.91)*	.40(.12)**	3.80(2.8)	.44(.45)
Phase=BO	1.78(.44)***	.80(.34)*	1.38(.31)***	.34(.72)	−5.88(1.66)***	−.47(.21)*	13.96(4.9)**	.55(.91)
Age	.08(.10)	.11(.07)	−.07(.06)	.02(.10)	−.16(.21)	−.01(.02)	.32(.46)	−.38(.28)
Random Effects
SD (intercept)	5.10(.19)	3.73(.19)	3.38(.16)	6.06(.34)	10.42(.86)	.41(.24)	5.62(.)	15.04(.64)
SD (slope)	1.09(.24)	.30(.30)	.42(.19)	2.43(.27)	4.89(1.03)	.56(.24)	12.5(.)	1.39(.43)
Corr (I,S)	.15(.18)	.05(.44)	.25(.35)	−.8(.05)	−.62(.11)	−.33(.37)	1(.)	.23(.26)
SD (residual)	2.64(.13)	2.12(.10)	1.88(.09)	3.96(.21)	9.24(.54)	1.23(.07)	28.3(.)	4.73(.25)
Model Wald (df)	17.27(3)***	10.58(3)*	22.25(3)***	8.54(3)*	83.65(3)***	72.39(3)***	9.65(3)*	2.87(3)
Records (n)	687(368)	687(368)	687(368)	621(347)	603(334)	594(332)	603(343)	583(334)
Successful Candidates Only
Parameter	Resilience	Grit	Hardiness	Cortisol	DHEA	Ratio	BDNF	NPY
Fixed Effects
Intercept	52.1(3.6)***	40.20(2.7)***	32.50(2.3)***	12.37(3.4)***	48.58(6.7)***	3.56(.67)***	8.90(18.2)	19.53(10.8)
Slope	.60(.24)*	.10(.17)	.48(.15)***	−.52(.41)	3.16(.90)***	.43(.12)***	3.98(2.9)	.51(.46)
Phase=0	1.94(.47)***	.74(.37)*	1.64(.32)***	.32(.76)	−6.78(1.78)***	−.69(.23)**	14.5(5.7)*	.80(.97)
Age	.17(.15)	.08(.11)	.03(.10)	.09(.14)	−.40(.27)	−.02(.03)	.39(.72)	.19(.45)
Random Effects
SD (intercept)	4.83(.35)	3.68(.28)	3.02(.23)	5.77(.50)	7.41(1.28)	.005(.14)	5.72(2.66)	14.4(.99)
SD (slope)	1.10(.19)	.32(.29)	.46(.16)	2.29(.29)	3.81(1.04)	.53(.10)	12.0(1.81)	1.31(.48)
Corr (I,S)	.12(.18)	.06(.45)	.31(.31)	−.76(.06)	−.36(.25)	−1.0(.02)	1.0(.00)	.30(.28)
SD (residual)	2.61(.13)	2.11(.10)	1.80(.09)	4.04(.22)	9.46(.57)	1.24(.05)	29.80(1.25)	4.89(.27)
Model Wald (df)	18.68(3)***	6.35(3)	27.67(3)***	5.34(3)	102.71(3)***	79.64(3)***	7.32(3)	1.41(3)
Records (n)	437(145)	437(145)	437(145)	395(137)	391(137)	383(135)	377(134)	369(135)

Notes: Abbreviations: BDNF = brain-derived neurotrophic factor. DHEA = dehydroepiandrosterone. NPY = neuropeptide Y. Ratio = DHEA-to-cortisol ratio.

Indicators: Significant difference at p-value of less than ***.001, **.01, *.05. ; “.” Indicates model did not produce standard errors because it did not converge.

In order to keep units across measures, including the psychological indices, in a somewhat similar scale, we divided cortisol concentration (pg/mL) by 10,000, divided DHEA, BDNF, and NPY (all pg/mL) each by 100, and multiplied the original DHEA/cortisol ratio by 100.

For all outcomes, the top four rows show the “fixed” effects, including the intercept, a linear slope representing growth (or decline) across phases, and an indicator to capture the overestimation of psychological factors prior to the start of BUD/S. Given the first measurement occasion indicator, the intercept can be viewed as the “true” baseline value of the outcome trajectory across phases. The next four rows in the table present the random effects, that is, the standard deviations representing the extent of variation across the sample in intercepts and slopes, as well as the correlation between them and the standard deviation of the residual (unexplained variation in the outcome at each measurement occasion). In order to keep units across measures, including the psychological indices, on a somewhat similar scale, we divided cortisol concentration (pg/mL) by 10,000, divided DHEA, BDNF, and NPY (all pg/mL) each by 100, and multiplied the original DHEA/cortisol ratio by 100. Figure 1 illustrates the trajectories for each variable discussed in the subsequent sections.

Figure 1.

Growth models of psychological measures and physiological indicators

3.1. Psychological Growth Models

The average SEAL candidate began BUD/S with a resilience value of 53.7 and grew 0.59 units per phase after initially declining 1.78 units after the start of training. The effect of the indicator for pre-training measurement and the growth rate were statistically significant. Figure 1A illustrates the average trajectory based on the results. The solid line in the figure reflects the “true” underlying average trajectory. The trajectory began at the beginning of training at under 54 units and increased across phases to just under 56 units by the end of BUD/S. However, the coefficient for the pre-training indicator is 1.78, indicating that average resilience at the beginning of training was just under 56 units. While Figure 1A shows the average trajectory across the sample, not all candidates experienced growth in resilience. The standard deviation for the slope is 1.09, meaning that 95% of candidates had growth rates that fall between −1.54 and 2.73 (.59–1.96*1.09 and .59+1.96*1.09) units per phase. Thus, some candidates grew considerably across training, while others declined.

Results follow a similar pattern for the other two psychological outcomes but with some variation. For hardiness, the average SEAL candidate began BUD/S with a value of 34.5 plus 1.38 and grew 0.44 units per phase; however, there was substantial variation in growth (sd = 0.42). Figure 1B illustrates the average trajectory and overestimation at the start of training based on the results. This outcome behaves in much the same way as resilience, indicating that some candidates grew considerably across training while others declined considerably. For grit, the average candidate experienced no growth; while the slope estimate was .12 units, it was not statistically significant. As with resilience and hardiness, there was variability around the growth rate, but the variability was considerably less (sd = 0.30).

3.2. Biomarker Growth Models

The results for the biomarker outcomes were less consistent than for the psychological outcomes (Figure 1). The average trajectories for cortisol, BDNF, and NPY were flat: the coefficients for the slope were not significant. The only exceptions were for DHEA and the ratio of DHEA-to-cortisol. The average candidate experienced a significant growth in DHEA across training (coefficient = 2.32, p < .05), and the ratio of DHEA-to-cortisol also increased significantly across phases (coefficient = 0.40, p < .001). All biomarkers had evidence of residual variability—the standard deviations of their time-specific residuals were large—and the variability in the intercepts and slopes were also large, indicating considerable variation in the biomarkers’ patterns over time.

4. Discussion

In this study, we examined growth patterns in self-report psychological measures and physiological indicators during BUD/S selection training. The results of the full sample were almost identical to those of the successful candidates, suggesting issues with predictability. A key finding from this study was that, on average, self-reported resilience and hardiness grew throughout BUD/S despite initial declines after the start of training. In support of our hypothesis, the average candidate experienced a steady growth in DHEA concentrations and in the DHEA-to-cortisol ratio, although considerable between-subject variability was observed. Contrary to our hypotheses, no growth was observed, on average, in self-reported grit or concentrations of cortisol, BDNF, and NPY during BUD/S. Collectively, these results suggest the hardships experienced by candidates during BUD/S can elicit modest growth in self-report resilience and hardiness, as well as some physiological adaptations to stress.

4.1. Resilience and hardiness grow during adversity, though initial self-reports are likely overestimated

The results indicated that SEAL candidates undergo some growth in both resilience and hardiness throughout BUD/S but only after experiencing an initial overestimation followed by a dip in those characteristics. The overestimation appeared in the initial data collection at the beginning of training. The effect of the indicator for the start of training was statistically significant, suggesting one possibility that the average candidate overestimated his “true” baseline resilience and hardiness. Another possibility is that their psychological characteristics were actually that high, and the individual was worn down throughout Phase 1. The resilience gradually increased throughout Phase 2 and Phase 3. Additional possibilities for this effect is that the psychological traits follow a non-linear, u-shaped pattern over time. The growth rate was statistically significant, indicating that the candidates, on average, experienced a positive increase in resilience and hardiness after “recovering” from their initial overestimation. This pattern for both is reflected in Figures 1A and 1B: the solid line shows a decline in each characteristic from the beginning of training to the end of Phase 1. This pattern reflects the interpretation that these characteristics are overestimated by candidates prior to the intense stress of Phase 1. The candidates then experience a decline in the characteristics as a result of the difficulties experienced in Phase 1 and then generally an increase as the candidate successfully completes training challenges throughout the remainder of the course.

We believe there are two plausible reasons why this pattern is observed. First, the pattern that emerged aligns with a potential response to adversity: self-destructive response, increased fragility/less resilience, return to normal, growth/more resilience (Bonanno, 2004, 2005; Connor & Davidson, 2003; Harms et al., 2018; Masten et al., 1990). Specifically, there is alignment with the return to normal response or potentially the growth/more resilience response given a longer data collection period. In Figure 1A, the trajectory observed indicates an initial starting point for resilience, which we classified as an overestimation. However, another interpretation is that this initial starting point indicates a baseline measure for resilience. Then, the dip indicates a response to adversity, which in this case would be the challenges imposed during Phase 1 of BUD/S. The initial data collection at the start of training took place at the end of BO immediately prior to Phase 1, thus, justifying the interpretation that the adversity was during the first phase. Phase 1 is designed to challenge SEAL candidates and break them down mentally and physically to assess who is capable of performing the duties of a SEAL (Ledford et al., 2020). The growth is a steady return to normal or baseline, which can be seen by the final observation returning to a similar level of resilience as the initial observation.

One challenge with this interpretation is that it is difficult to determine a true baseline and endpoint based on the data collection in this study. Finding a baseline for resilience or hardiness would require multiple data collections prior to BUD/S and the initiation of training. Further, determining whether the BUD/S candidates’ endpoint in resilience was a return to baseline or a growth pattern would require several data collections following BUD/S. Observing a plateau would indicate it was a return to baseline, and an increase in resilience would indicate growth. However, there is limited research indicating how hardiness develops over time and whether it is stable or grows. Thus, this could indicate a developmental pattern of hardiness much like that of resilience.

A second explanation for the overestimation effect and subsequent growth is that the overestimation is a result of the Dunning Kruger effect (Kruger & Dunning, 1999). The Dunning Kruger effect (1999) refers to the observed pattern that individuals with little knowledge or experience in an area tend to overestimate their competence in it, then experience a significant decline in their confidence in their competency, then gradually grow in confidence and competence over time at a positive rate. In our setting, the mental preparation required for BUD/S is far less apparent than the physical preparation required. The standards required for the entrance SEAL physical screening test (PST) are well known, and the physical training to prepare for BUD/S is documented on the Naval Special Warfare website (https://www.sealswcc.com/), other sites (https://navyseals.com/buds/nsw-pt-guide/), and in a myriad of SEAL books.

However, the mental training needed to prepare for BUD/S is not as clear. There are no tangible standards or exercises that develop the psychological attributes in these sources to demonstrate what is required to start or successfully complete this training. Thus, it may be the case that SEAL candidates come to training lacking resilience and hardiness but are unaware that they lack them. This is considered meta-ignorance or “unknown unknowns” (Dunning, 2011). Meta-ignorance occurs when there is a lack of expertise or knowledge, but the person is unaware that the gap even exists. This may have been the case in this study. Due to a lack of mental training and specific skills, SEAL candidates did not know they lacked resilience or hardiness. Thus, they initially overestimate their capabilities when they answer questions to assess these characteristics. Then, when they were introduced to the hardship of BUD/S, participants were able to evaluate their resilience, hardiness, and grit more accurately during the subsequent data collection.

4.2. Prolonged exposure to intense military stress induces an increase in DHEA and DHEA-to-cortisol ratio

The average candidate experienced an increase in DHEA concentration and DHEA-to-cortisol across the six-month training. Though regular physical training does not appear to significantly alter DHEA concentrations (Collomp et al., 2015), high-intensity exercise can induce increases in DHEA among athletes (Sato et al., 2016). Block periodized training combined with operational training elicited significant increases in DHEA-S, the sulfate metabolite of DHEA, among Naval Special Warfare Operators during training blocks focused on endurance, power, and muscular strength (Oliver et al., 2015). The increases in DHEA-S coincided with higher DHEA-S-to-cortisol ratios, an indication of anabolic response (Oliver et al., 2015). DHEA contributes to the enhancement of muscle glucose metabolism and protein synthesis, two physiological adaptations that are important to improve physical performance (Sato et al., 2016). In the present study, BUD/S candidates were exposed to highly intense exercise throughout training and were expected to complete physical standards in faster times as the training progressed. Therefore, the progressive, intense physical training during BUD/S may have contributed to an anabolic response, as evidenced by small, but steady increases in DHEA and DHEA-to-cortisol across the six months.

The increase in DHEA may also be tied to its role as an immunomodulator and an anti-glucocorticoid (Prall & Muehlenbein, 2018), as there were no changes in cortisol concentrations throughout BUD/S. DHEA indirectly affects the immune system through downstream modulation of metabolites (e.g., hormones) and directly by interacting with NF-κB, T cells, and monocytes (Prall & Muehlenbein, 2018). Extended periods of intense military training have been demonstrated to cause a significant decline in immune function (Shippee et al., 1994). Likewise, athletes’ physical fatigue during prolonged periods of heavy training and competition produces a similar increase in susceptibility to illness (Walsh, 2018). In both cases—intense military training and sport—exposure to physical work, sleep disruption, low energy availability, and psychological stress are present and are risk factors for illness. Similarly, BUD/S candidates are exposed to several months of extreme physical and mental stress. The increase in DHEA observed in this study may be part of an adaptive response to the repeated stress exposure experienced during BUD/S training. Specifically, the longer candidates remain in BUD/S, the more exposure they have to physical and psychological stress causing activation of the immune response, which is counteracted through DHEA growth throughout the course.

Contrary to our hypothesis, no changes were observed in cortisol, BDNF, or NPY during BUD/S training. BDNF expression is highly regulated during transcription, translation, and post-translation to elicit various functions, resulting in substantial variability in concentrations among healthy humans (Miranda et al., 2019). Likewise, NPY release is largely dependent on the type and duration of stress (Reichmann & Holzer, 2016; Zhang et al., 2021). Provided that SEAL candidates are frequently exposed to a variety of stressors known to influence BDNF and NPY concentrations, including prolonged physical exercise and periods of limited sleep (Miranda et al., 2019; Morgan et al., 2000b; Szivak et al., 2018), these factors likely contributed to the high degree of variability in BDNF and NPY, which may have precluded the detection of significant changes in the present study.

4.3. Comparison of Psychological and Physiological Indicators

While few studies have examined psychological and physiological adaptations in strenuous environments over long durations, findings from the present study suggest that the psychological and physiological adaptations during BUD/S training are asynchronous. It was not possible to combine the three psychological and five physiological variables into one larger model due to convergence issues. Nonetheless, it was possible to examine the average growth of psychological variables alongside the growth of physiological variables during the same time periods (start of training to 1, 1 to 2, 2 to 3, and 3 to graduation). This comparison showed that resilience and hardiness dipped from their initial start points at the beginning of BUD/S as described in Section 4.1, which was followed by consistent positive growth after Phase 1. The physiological variables, such as DHEA and DHEA-to-cortisol, did not follow the initial dip but had consistent positive growth from the initial start point all the way through training.

There are several possible explanations for the disparity of patterns between the psychological and physiological adaptations. As discussed earlier in section 4.1, there is a lack of mental training material to prepare students for BUD/S. While there are countless books on the physical preparation required, preparation materials on mental exercises for this training that would develop resilience and hardiness are minimal. This could explain why there would be an overestimation and dip in the resilience and hardiness scores that are not present in their DHEA and DHEA-to-cortisol levels, which could feasibly be more stable in growth. The months, and in many cases years, of physical preparation prior to this training could present less physiological shock to the candidates in comparison to the minimal amount of psychological preparation prior to the incredibly high levels of stress experienced during training. It has been noted that BUD/S is more a test of mental toughness than physicality (Couch, 2003). The students believe they are prepared for BUD/S until faced with the harsh reality of Phase 1. The authors suggest that this is illustrated in the psychological assessments, while the physical assessments are more stable. Given this and the lack of psychological preparation, it would be unsurprising for a disparity between the two adaptations.

Another possible explanation for the disparity in psychological and physiological patterns is that there is more of an effort by the instructors during the first days and weeks of Phase 1 to present immense mental stress to remove the less committed candidates from training early. In contrast, the physical standards, although incredibly high to begin with, are increased gradually throughout training. For example, students must complete a timed four-mile beach run, two-mile ocean swim, and timed obstacle course each week (Couch, 2003). The threshold for passing the weekly evolutions is lowered over the six months so that by the end of the training, students are running, swimming, and completing the obstacle course significantly faster than at the start. The physiological indicators of the study replicate that pattern. While the psychological overestimation and dip could be explained by the very high stress put on the students, especially at the beginning of training, creating an intense shock, the physiological indicators follow the same constituent growth of physical intensity in the course.

4.4. Considerations

This study is unique because of the longitudinal collection of both the psychological and physiological information from the sample of the BUD/S students; however, it has several limitations that should be examined. Concern with self-report is a normal limitation with surveys since individuals tend to want to portray themselves in the most desirable light (King & Bruner, 2000). Another limitation was in the timing of how our data were collected. Due to the BUD/S evolution schedule, we collected the data around their training schedule, so not all the data were collected simultaneously for each class. For example, in one class, data were collected later in the afternoon on the fourth day of training, instead of the first day like the other classes. This could mean the mindset of those candidates was in a slightly different place than the other classes. Likewise, variations in collection time between classes, although all primarily at the beginning of the day between 7:00 a.m. and 1:00 p.m., may have impacted the physiological results, as circadian factors, physical exertion, and metabolic state are known to influence hormone release (Friedl et al., 2000; Lieberman et al., 2016). However, data collection timing was consistent within each class. Blood biomarkers alone are likely limited in their ability to holistically convey information on physiological status and should be interpreted in the context of other psychological and physiological domains to maximize their usefulness. Finally, the homogeneity of the classes helped us in the research development, but it may not be representative of other, more diverse groups.

5. Conclusion

This study examined the growth patterns in psychological indicators of performance and physiological indicators of stress in individuals progressing through an intense selection and training environment, in this case, BUD/S. We expected that there would be growth in these attributes throughout the six-month selection and training course. There appears to be growth in the psychological indicators, resilience and hardiness, after an initial overestimation. There also appears to be growth in two of the physiological indicators of stress, DHEA and DHEA-to-cortisol, which appears to be part of an adaptive response to the repeated stress exposure experienced during BUD/S training. However, the study did not indicate significant growth in all of our psychological (grit) or physiological (BDNF, cortisol, NPY) measures, which could indicate that those indicators do not grow throughout the six months at BUD/S or be a result of limitations in our study. While we now understand more about the growth of these adaptations, due to the study’s limitations, this study falls short of a level of predictability that could benefit the SEAL community. It is possible that including performance data is one area for improvement to determine what psychological and physiological indicators predict success through BUD/S. Additionally, more consistency in when the collection of the physiological data took place would reduce noise in the biomarkers.

Despite these limitations, there is a possibility of utilizing the growth of these attributes as a monitoring tool for positive psychological and physiological growth during training. An understanding of positive growth could be used as one of many data points to determine the potential for struggling students to either continue training, get a second chance at training by rolling to subsequent classes, or find a more suitable occupation for the student in the Navy. This research also shows a greater need for psychological training prior to the course. More psychological preparation earlier in the process could help mitigate the overestimation of these attributes and increase the student’s ability to adapt positively to the stressors of BUD/S. This study certainly suggests that more research is needed not only for predicting success but also a greater understanding of the growth of both psychological and physiological adaptations in high stress environments.

Data Availability:

The de-identified data can be made available to verified researchers by request with the corresponding author. As a stipulation of the Naval Special Warfare community, it cannot be made publicly available.