Abstract
The media is replete with stories about the impact on families when they drastically alter their lives to provide care to a service member who is seriously wounded, ill, or injured (WII), though few studies have systematically examined these impacts. Using a 2008 survey of seriously WII service members, the authors found that 27% of caregivers provided an average of more than 40 hr of care per week, 64% of caregivers have provided care for more than a year, and 50% expected they may need to provide care over the long term. The probit and ordered probit results show that significant others (such as a spouse, fiancée, or girlfriend) bear a greater caregiving burden in both intensity and duration compared to other caregivers. The results also show that duration of caregiving for Reserve Component service members is greater than for Active Component service members.
Introduction
As a nation, the United States is grappling with the challenges of the care and treatment of its wounded, ill, and injured (WII) service members and veterans of Operation Enduring Freedom (OEF) and/or Operation Iraqi Freedom (OIF). This is a more common phenomenon now, as many casualties of these wars would not have survived in previous wars (Department of Veterans Affairs, 2007), due to air medical evacuation, improved body armor, and improvements in battlefield medicine. With so many casualties surviving what were previously fatal injuries, there is concern and ongoing dialog as to the long-term financial and emotional effects these injuries will have on service members, their families, and their communities (Kreutzer, Gervasio, & Camplair, 1994b; Schwarz et al., 2007).
Various commissions and committees have been established in recent years to deal with the issues faced by seriously WII service members and veterans. These include (a) The President’s Commission on the Care for America’s Returning Wounded Warriors (Dole–Shalala Commission); (b) The Veterans’ Disability Benefits Commission; and (c) The Department of Defense (DoD) and Department of Veterans Affairs (VA) Wounded, Ill, and Injured Senior Oversight Committee. In particular, these commissions and committees recognize the importance of caregivers as well as the substantial responsibilities that these caregivers have in helping to take care of seriously WII service members.
For example, the Dole–Shalala Commission explicitly recognized that “the most seriously injured service members and their families are embarking on a long journey together.” During the road to recovery, family members of the most seriously injured may be required to temporarily relocate to a different part of the country to be near the facility where their loved one is being treated. Relocation may require them to give up the lives they know—jobs, school, homes—and live for an uncertain period far from their existing network of friends and family. (Dole & Shalala, 2007, p. 19 )
This relocation is common as the military has established centers of excellence at its major medical centers for the treatment of the most seriously injured. Family members of seriously WII service members may also experience economic and/or emotional stress (Anderson, Parmenter, & Mok, 2002; Coe & Van Houtven, 2009; Dole & Shalala, 2007; Kreutzer, Gervasio, & Camplair, 1994a; Murray-Swank et al., 2007). Nevertheless, family support is critical to the service members’ successful rehabilitation (Sander et al., 2002), and assistance from family and friends is more satisfying to disabled individuals than assistance from others (Matthias & Benjamin, 2008).
Given the challenges that families of seriously WII service members face, the purpose of this study is to examine the impact on family and friends—the “caregivers”—who provide nonmedical care, support, and assistance for wounded warriors. Note that in their landmark book, The Three Trillion Dollar War, Stiglitz and Bilmes (2008) made high-order estimates of caregivers’ lost earnings. Christensen et al. (2009) went further and provided more specific estimates with additional details of the types of economic impacts these caregivers face. Specifically, they found that 77% of caregivers were working prior to becoming a caregiver and that 87% of these caregivers either quit or took time off from work to be a caregiver. There were similar impacts for those in school prior to becoming a caregiver. They estimated that the average loss of earnings and benefits for caregivers of seriously WII service members was US$3,200 per month. Given that on average these service members had a caregiver for more than 1.5 years, the average earnings and benefits losses are about US$60,000.
The purpose of this study is to add to our understanding of the impacts on caregivers of the seriously WII service members by using the survey data collected by Christensen et al. (2009) to explore the relationships of various caregiver and service member demographics with the intensity and duration of caregiving. Understanding these relationships is necessary to identify differences in burden so that policy makers can consider them when addressing support for caregivers.
Background
When a service member is seriously WII, DoD will provide financial assistance in the form of “Invitational Travel Authorization” orders to eligible immediate family members to be at the bedside of service members who are categorized as very seriously injured (VSI) or seriously injured (SI). (Those categorized as VSI are medically deemed to be in danger of imminent death. The SI classification denotes a level of severity that is cause for immediate concern but with no danger of imminent death. During the 2003–2008 period, there were 4,575 service members classified as VSI or SI with a ratio of 3.7 SI to 1 VSI; see Christensen et al., 2009.) Travel to and from the hospital, lodging, meals, and incidental expenses are paid for by the government, but this per diem is not meant to be an income replacement program for family members—it’s to cover travel expenses.
Following hospitalization, there is often an extended period of treatment and rehabilitation that can continue at a DoD, VA polytrauma, or civilian hospital. During this time, the service member remains at or near the facility in an outpatient status, and a medical provider evaluates the need for a nonmedical attendant who may receive per diem to cover travel expenses. If eligible, the service members designate a nonmedical attendant (usually a spouse or parent) to help with activities of daily living that may be difficult for them to perform independently.
Nonmedical attendants or caregivers often take on tremendous responsibilities to promote the well-being of the service member. These responsibilities frequently include driving service members to therapy appointments, picking up medications, providing a wide range of personal care, taking care of children, paying the bills, and providing emotional support, and “in short, find they have a full-time job—or more—for which they never prepared” (Dole & Shalala, 2007, p. 19). On average it takes these VSI/SI soldiers about 2.5 years after injury to process out of the military to veteran status (Christensen et al., 2009). As a result of the service member’s injury, caregivers may experience both economic and emotional stress as there is a spillover effect for caregivers (Bobinac, van Exel, Rutten, & Brouwer, 2010).
Most of the literature investigating the employment effects of caregiving focuses on caregivers of elderly or disabled parents or children. While not focused on seriously WII service members, this literature has some application to their caregivers. The literature shows that the labor force participation rate and hours worked are less for caregivers compared to noncaregivers (Leigh, 2010; Lilly, Laporte, & Coyte, 2007). Furthermore, female caregivers are more likely than male caregivers to reduce their work hours or leave the labor force (Ettner, 1995, 1996; Lukemeyer, Meyers, & Smeeding, 2000; McLanahan & Monson, 1990; Muurinen, 1986; Pagani & Marenzi, 2008; Stone & Short, 1990; Wakabayashi & Donato, 2005). In addition, Choi (2010, p. 493) found that “a father’s poor heath status is a significant predictor of lowering a daughter’s educational attainment and working probability during her subsequent adulthood years.” This finding suggests that daughters in the role of the caretaker may negatively impact their education and future earnings. We also expect that the burden of caring for the seriously WII service members will fall on women—the wives and mothers of the service members. So by inference, our presumption is that caretaking duties may also affect both education attainment and/or work status for wives caring for seriously WII spouses or mothers caring for their adult sons.
Beyond the employment impact, research has shown that while physical health of caregivers was only slightly worse, psychological distress including anxiety, depression, and hostility were prevalent (Caldwell, 2008; Hanks, Rapport, & Vangel, 2007; Sander et al., 2007; Williams, 1993). The emotional impacts that caregivers face include depressive symptoms, high levels of emotional distress, and family dysfunction (Anderson et al., 2002; Coe & Van Houtven, 2009; Kreutzer et al., 1994a; Marsh, Kersel, Havill, & Sleigh, 2002) as well as anxiety and clinical depression (Murray-Swank et al., 2007). For example, Marsh et al. (2002) found that one third of caregivers had “clinically significant symptoms of anxiety and depression” at both 6 months and 1 year after injury. Furthermore, the emotional and employment impacts are related as caregiver depression is associated with a higher likelihood of time missed at work (Wilson, Van Houtven, Stearns, & Clipp, 2007). In addition, Ergh, Hanks, Rapport, and Coleman (2003) found that about a third of caregivers of individuals with traumatic brain injury who reported low social support also had significant life dissatisfaction (also see Arzi, Solomon, & Dekel, 2000). On the positive side, they found that the reverse was true—high social support is a moderating factor for life dissatisfaction. Similarly, Rubin and White-Means (2009) found lower quality of life among “sandwich caregivers”—those individuals who are caring for both their minor children and their elderly parent.
Caregivers of those with a mental disability were more likely to report unmet needs compared with those with a physical disability (Vecchio, Cybinski, & Stevens, 2009). This is concerning for caregivers of seriously WII service members who often have both physical and mental disabilities. The literature shows that the caregiving relationship affects not only the caregiver and the care recipient but also the whole family equilibrium (Marsh et al., 2002).
Although the United States government has made efforts to provide assistance, some seriously WII service members and families still struggle. (This assistance includes Traumatic Service member’s Group Life Insurance [TSGLI], which pays US$25,000–US$100,000 to living service members depending on the specific injury; Family Medical Leave Act [FMLA], which has been modified for the families of WII service members to have 26 weeks of FMLA leave in a 12-month period compared to the normal 13 weeks; and the Caregivers and Veterans Omnibus Health Services Act of 2010, which provides eligible caregivers a monthly stipend and access to health care insurance.) The question we explored is to what degree do demographic factors of both the caregiver and the service member affect how the family responds in terms of how much care they provide and for how long. These are relevant questions as Kolodinsky and Shirey (2000) found that labor force participation and work hours of daughters were affected by their elderly parent’s characteristics. Furthermore, the demographics of daughters residing with their elderly parent affected their labor supply compared to those not coresiding. Similarly, Lai and Leonenko (2007) studied the effects of caregiving for Chinese Canadian caregivers. They found that demographic factors such as age, employment, and caring for more than one person were correlated with the perceived economic costs.
With this as context, this study examines the impact that the relationship to the care recipient, employment, schooling, and education level has on the intensity and duration of caregiving for seriously WII service members. We also examine the impact that demographic factors of the care recipient have on the intensity and duration of caregiving. It is important to understand these relationships for the seriously WII service members as the types of conditions they face are very different from the chronic illnesses for which the elderly typically require caregivers. The types of primary conditions that seriously WII service members commonly face are extremity condition/amputation (38%); neurological condition, most commonly traumatic brain injury (31%); burns/disfigurement (10%); vision (10%); and mental condition, most commonly posttraumatic stress disorder (8%). (The distribution of injuries across these groups is consistent with injuries stemming from the blast injuries that service members are frequently exposed to.) Furthermore, these are the primary conditions. Many service members and veterans have secondary and tertiary conditions (Christensen, McMahon, Schaefer, Jaditz, & Harris, 2007).
With this background, the next section discusses the methodology behind the survey of seriously WII service members regarding their caregivers. The section that follows discusses the probit and ordered probit model results of the relationship between caregiver and service member demographics, and the intensity (weekly hours) and duration (past and future months) of caregiving. The final section discusses the implications of these results.
Data and Method
To understand the relationship between demographics and the caregiving burden, we rely on a 2008 survey of active duty seriously WII service members that asked them about the support (i.e., assistance with treatment, rehabilitation, and recovery transition) they received from their caregiver. The survey targeted VSI/SI service members as a proxy for seriously WII service members as there is no standard definition of what is seriously WII. The VSI/SI designation is a reasonable proxy for seriously WII as service members with this designation have serious-enough injures to warrant immediate travel of family members to their bedside. The survey was sent to 1,100 service members, and there were 263 survey respondents of which 248 completed the survey. Given our estimate of the size of the VSI/SI population still on active duty, the 248 completed surveys represented about 11% of the VSI/SI population at the time of the survey. 1 Note that because 95% of respondents were male, we limited our analysis to male respondents to limit heterogeneity due to gender. Further 90% of respondents indicated that they had a caregiver. This combined with missing observations results in a sample with about 200 male service members designating a primary caregiver.
The survey instrument was developed in consultation with each of the military services’ wounded warrior programs. The objective of the questionnaire was to understand (a) the relationship of the caregiver to the service member; (b) demographics of the caregiver; (c) economic impacts on the caregiver to include employment, education, child care needs, and financial obligations; and (d) the duration and intensity of caregiving. The various wounded warrior programs provided feedback on the survey instrument both in terms of wording and content as well as valuable feedback on the ability to effectively administer the survey.
Table 1 presents the distribution of responses to the survey questions that indicate the intensity and duration of caregiving—average weekly hours of caregiving, past months of caregiving, and expected future months of caregiving. Note that these questions have categorical responses, which drives the need for our probit and ordered probit models. The descriptive data show that 27% of caregivers provided, on average, more than 40 hr of care per week. Furthermore, the majority (64%) of these caregivers had provided care for over a year, and 50% expected that they will require care for the long term.
Descriptive Statistics for Caregiving Time
Note: Figures represent male service members with a primary caregiver.
Table 2 shows the characteristics or demographics of the service members and their caregivers. Note that we included some of the demographics of the service member as controls in our analysis. For example, we used age, rank, and education level as instruments for type of assignment. These are correlated with the types of risks and injuries to which a service member is exposed. Ideally, we would have controlled for the nature of the disability or disabilities, but that information was not available in the survey. Other service member demographic factors such as number of dependents and whether they were in the Reserve Component (RC) reflect different family obligations and support structures and these may influence the demands on the caregiver. Our sample of service members is relatively young with almost half of the sample under the age of 30 and more than 80% of the sample under the age of 40.
Descriptive Statistics: Seriously WII Service Members and Their Primary Caregivers (N = 202)
Note: WII = wounded, ill, or injured. Figures represent male service members with a primary caregiver.
Table 2 also shows the demographics of the caregivers. Note that the relationship of the caregiver to the service member is highly collinear with marital status. Therefore, we have not included service member marital status as a demographic control variable. For married service members, the spouse is the caregiver in almost all cases. For other service members, their mothers are the most common caregiver (nearly half). Note that in most cases (82%), the primary caregiver had help from other family members or friends.
We used probit and ordered probit models to estimate the impact of caregiver and service member demographics on three dependent variables: (a) the average hours of weekly caregiving, (b) the number of past months of caregiving, and (c) the expected future months of caregiving. In all cases, our independent variables are the caregiver and service member demographic variables as shown in Table 2. We used discrete choice models—probit and ordered probit—because the survey data are categorical not continuous variables.
Average weekly hours of caregiving, for example, has five categories: 10 hr or less, 11 to 20 hr, 21 to 30 hr, 31 to 40 hr, and more than 40 hr. The ordered probit includes all of these categorical responses. The same holds for our two other dependent variables in those ordered probit models. For the binary or probit model, we examined the probability of providing a weekly average of more than 40 hr of care. For the past months of caregiving, our probit model examines the probability of having provided more than 12 months of care. Similarly, for the expected future months of caregiving, our probit model examines the probability of needing to provide care for the long term.
Results
Average Hours per Week of Care
We first present the results of our probit and ordered probit regressions for the impact of demographic characteristics on the average weekly hours of caregiving (see Table 3). Note that Table 3 shows these results for a version of the model that includes the age of the caregiver but not the service member. We did not include age variables for both caregivers and service members in the same regression due to multicollinearity. This arises as the vast majority of caregivers are significant others and their ages are highly correlated with the service member’s age. For completeness, we also estimated the model two additional ways: without age variables for the caregiver and the service member and with age variables for the service member only.
Probit and Ordered Probit Regressions on Average Caregiving Hours per Week
Note: SM = service member
p < .1. **p < .05. ***p < .01.
While we report the results of only the model with caregiver age variables, we note that for the most part, the marginal effects and coefficients of the model are stable and that statistical significance for the other variables is common across the three models.
Looking at the specific results for the average hours of weekly caregiving, the probit results show that significant others are much more likely to provide care more than 40 hr per week compared with other caregivers. In fact the marginal effect is 15%. This is substantial given that the data mean is 27%. Why would this be the case? The data can’t answer that question, but it is perhaps indicative of social norms and expectations to which caregivers are subject such as wives caring for their husbands. Caputo (2002) raised the idea of caregiving being influenced by social norms. Caputo looked at whether inheritance-related factors influenced the probability of adult daughters providing care for aging parents and found no effect, which led to his conclusion that this finding suggests conformance to social norms. The same may apply with significant others caring for their WII service member.
We do not know the long-term effects caregiving has on these young spouses. However, the potential long-term effects are concerning in light of research showing that wives caring for their husbands had more negative outcomes (such as health status, satisfaction with themselves, and depression) than other caregivers or noncaregivers (Hoyert & Seltzer, 1992). In addition, Hoyert and Seltzer (1992) found that “those who provide care for a longer period of time exhibit more negative outcomes relative to those who provide care for a shorter period of time” suggesting “that accumulation of stress over the course of caregiving may occur.” (p. 79) Similarly, Lee, Walker, and Shoup (2001) found that among employed caregivers, those who provided more caregiving hours reported poorer emotional health. This is concerning over the long term for spouses caring for seriously WII service members who face enormous physical and mental health issues.
The probit results also show that primary caregivers who have the support of one other caregiver are 30% more likely to provide more than 40 hr of care per week. This result is somewhat counterintuitive as one could expect that having additional caregivers would mean a reduced burden on the primary caregiver. One possible explanation, however, is that the additional caregiver is providing less direct caregiving support to the service member, but may facilitate the primary caregiver focusing on the caregiving role and free him or her from some of his or her noncaregiving responsibilities. While this may be one explanation, it may also be the case that this result may be just picking up for the severity of the injury.
Another finding of interest is that caregivers of RC service members are less likely to provide more than 40 hr of care per week compared with Active Component (AC) service members. The marginal effect is 16%. A possible explanation for this is that a RC service member often has less access to the DoD services and support than AC service members do. Generally, AC service members live near a military base with all of its associated medical and nonmedical services and support functions. This is often not the case for RC service members. Furthermore, for RC service members, their families are generally not as steeped in the military culture and the DoD system and may not be aware of or know-how to access available services. Hence, it may be the case that caregivers of RC service members may have to deal with more noncaregiving issues and therefore cannot devote themselves to the caregiving role to the same degree as caregivers of AC service members.
Looking at the correlation of caregiver education level with the probability of providing care for more than 40 hr per week, the probit results show that caregivers with an associate’s degree or some college are less likely to provide more than 40 hr of care compared with those with a high school education or less. This result is not inconsistent with Mentzakis, McNamee, and Ryan (2009) who found that “wealthier individuals are less likely to be caregivers.” Although education level and wealth are not the same, they are positively correlated.
Other variables of interest were not significant. For example, working or being in school prior to caregiving is not correlated with the number of hours of care per week. We expect that these variables are correlated with the decision to be a caregiver, but we do not observe that decision. In addition, we would expect that the distance the caregivers have to travel affect their decision to be a caregiver. Again, we do not observe that decision, but for those who have chosen to be caregivers, distance from home is not correlated with the hours per week they provide care.
While we note these correlations from the probit analysis for the probability of providing an average of more than 40 hr of care per week, we did not find statistical significance for these same variables in the ordered probit regression, which looks at several levels of caregiving hours per week.
Past Months of Caregiving
Table 4 shows the probit and ordered probit results for the impact that various demographic factors have on how many months caregivers have provided care. As was important with average weekly hours of care, the relationship of the caregiver to the service member is correlated with the number of months they have provided care. The marginal effect is larger for significant others (63% compared with other caregivers) than it is for mothers (44%). Both effects are large and statistically significant. The effects are also highly significant in the ordered probit model.
Probit and Ordered Probit Regressions for Past Months of Caregiving
p < .1.**p < .05.***p < .01.
The results show that caregivers of RC service members were more likely to provide more past months of care than AC service members. These results are significant in both the probit and ordered probit models. This is an interesting finding given that caregivers of RC service members were less likely to provide more than 40 hr of care per week. That said, the same reasoning of AC service members and their families having better access to and knowledge of DoD services than RC service members and their families applies. While this lack of access to or knowledge of services may limit caregivers of RC service members to focus on the caregiving role to the same degree, it appears to be associated with an increase in the period of time that they need to provide care. This seems logical given that RC compared to AC service members take an average of 2 to 3 months longer to process out of the military to veteran status following injury (Christensen et al., 2009).
The results show that working prior to caregiving is positively correlated with the number of months the caregiver provides care. While the survey data do not answer why we observe this, one possible explanation is that those who were not working prior to being a caregiver may face an increased obligation to provide financially for the service member. This would be particularly true for spouses who make up the majority of caregivers. These spouses would likely have less time available for caregiving. This logic is consistent with Mentzakis et al. (2009) who found that coresidential caregiving competes with employment. This could explain the result showing that those in school prior to becoming caregivers provided care for a shorter period. Prior to caregiving, they were already pursuing economic advantage through education. Now with the service member’s injury, the new long-term realities may increase the need to pursue breadwinner roles through work and/or education leaving less time for caregiving.
Although we did not generally observe a significant relationship between the number of children the service member had and the average hours of care per week, the results show a statistically significant relationship for the duration of past caregiving. The marginal effect is sizable—31% for service members with two children, indicating that caregivers providing care for service members with two children were more likely to provide care for more than 12 months. This finding is consistent with Fredriksen and Scharlach (1999) who found that caregivers with child care responsibility had higher levels of strain.
Future Months of Caregiving
Table 5 shows the probit and ordered probit results for the impact various demographic factors have on the expected future period of caregiving. As with the average hours of care per week and past months of caregiving, the results show that significant others are expected to provide more care over long term than other caregivers. (The marginal effect is 38%.) This result is also highly significant in the ordered probit model. There is no reason to believe that service members with a significant other have more debilitating injuries. Rather it is logical that social norms and expectations cement this caregiving role creating long-term expectation.
Probit and Ordered Probit Regressions for Expected Future Months of Caregiving
p < .1. **p < .05. ***p < .01.
The results do not show any difference in the duration of expected future caregiving between the caregivers of AC and RC service members although this had a strong and significant correlation with average hours of care per week and duration of past caregiving. For the reasons of RC and AC differentials in access to or knowledge of services offered by DoD, those findings are logical. Looking forward at expected future caregiving, it is also logical that we would not see a difference between the caregivers of AC and RC service members as almost all seriously WII service members—AC or RC—are in the process of transitioning to veteran status. This means that in the future, the services available to all are from the Department Veterans Affairs that has a more geographically spread distribution of services across the nation rather than the services generally being in close proximity to military bases.
Finally, we note that the expected duration of future caregiving is less for caregivers of officers than enlisted service members. As we stated previously, we view rank as an instrument for the types of risk a service member would face and by extension the types and severity of injuries they are likely to have. This result is consistent with our expectation that on average the injuries of officers are less severe.
Conclusions
It is clear from the literature that caregivers face many economic and noneconomic burdens. These include reduced labor force participation, reduced work hours, lower educational attainment, poorer health, anxiety, depression, emotional distress, and family dysfunction. Now with OEF since 2001 and OIF since 2003, the United States has a growing number of new caregivers who in many respects are different from caregivers who typically care for aging or elderly spouses or parents. In most cases, these are circumstances where the family could see the likelihood of needing to provide care growing overtime giving them more time to prepare.
In contrast, the caregivers of seriously WII service members were faced with these challenges without warning. These families carry a tremendous burden as a result of service to the nation. We explored the issue of understanding the relationship between the caregiving burden and the demographics of both the caregiver and the service member. Given that these burdens come suddenly, understanding the relationship of demographics and the caregiving burden is important so policy makers can understand which caregivers are most at risk for poor outcomes and target policies to mitigate the burdens when possible.
We found that significant others compared to other caregivers bore the largest burden in terms of intensity (average weekly hours) and duration (past months) of caregiving. Clearly this is an area of concern. Policy makers should explore ways to mitigate these burdens. In doing so, policy makers should consider the fact that for spouse caregivers of elderly veterans, support from the veteran is a strong predictor of caregiver satisfaction (Dorfman, Holmes, & Berlin, 1996). Conversely, lack of support is a strong predictor of caregiver strain.
Caregivers of RC service members are another area of concern. By nature of the RC commitment, these service members and their families are less connected with the military culture and way of life. When they have periods of activation, it is more disruptive to these families as they have to put other jobs on hold. Furthermore, these families may have less access to (proximity) and knowledge of the services available to them as they are not as adept at navigating the DoD system. Steps to reach out to these families to increase awareness of what is available to them may help to mitigate poor outcomes. In addition, more integration of DoD and VA services may benefit these families as VA services are more widely distributed throughout the country than DoD services.
It seems clear that some of the correlation between average hours of care per week and duration of past caregiving is influenced by social norms that wives provide care for their husbands. Although we are not making a case for or against this social norm, it does raise the question of the degree to which caregiving persists in some cases as a function of this social norm rather than an actual need. Our point is that with all of the potential poor outcomes associated with caregiving, there are real economic and health benefits to not have caregiver roles persist longer than actually needed. Accordingly, future research should explore ways to understand the degree to which social norms increase the intensity and/or duration of caregiving.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was funded through a contract with the Office of Naval Research, Department of the Navy.