Chapter 3. Methodological Choices and Considerations
In this chapter, I explain the methodological considerations and choices of this study.
I discuss, first, the epistemological underpinnings of the methods; second, the overall research strategy; third, the missions chosen as case studies; fourth, choices made regarding sampling and data collection; fifth, the process of data analysis; sixth, issues of credibility and generalizability; seventh, reflections on researcher role and objectivity, and finally, ethical considerations.
Epistemological Underpinnings
As explained, this study aims to contribute to the concept of moral injury by seeking holistic, in-depth insight into mechanisms of moral distress, examining veterans’ personal experiences of distress in relation to complexities of moral beliefs and socio-political processes. To achieve this objective, it deviates from the positivist traditions in which most current research on ‘moral injury’ can be situated. Current ‘moral injury’ research predominantly focuses on identifying correlations between different variables, attempting to explain moral distress in terms of isolable and one-directional causal relations between events and symptoms (see e.g. Maguen et al. 2009, Nash et al. 2013, Currier et al. 2015, Bryan et al. 2016). While such an approach helps to systematically identify empirical regularities, it inevitably overlooks and even distorts the multidimensional and multidirectional nature of the human world (cf. Kleinman 1988, Good 1994, Bryman 2012). Also, it tends to reify the concept of moral injury as a clearly demarcated entity. Consequently, current research on
‘moral injury’ has been unable to adequately capture and conceptualize complexities of both moral and contextual dimensions of moral distress.
This study approaches social reality not as a closed system of isolable facts, but as an open world shaped by countless factors and always embedded in a wider social context (cf.
Kleinman 1988, Good 1994, Bryman 2012). Accordingly, it examines moral beliefs and socio-political processes, not as independent variables of which moral distress may be a function, but as complex webs of meanings and practices in which veterans’ experiences of moral distress can take the shape they do (cf. Kleinman 1988, Good 1994, Bryman 2012). It does not examine questions such as ‘how often can moral distress be identified among a given population?’ and ‘can a correlation be found between moral distress and a particular political act?’, but such questions as ‘if veterans experience distress due to moral challenges, what does their distress entail?’ and ‘if political practices surrounding veterans’ missions play a role in experiences of moral distress, what does this entail?’. In other words, this study does not cut up moral distress into variables in order to find correlations between these variables, but attempts to better understand under what structural and contextual conditions, and in what shapes, moral distress may come into being.
44
Part I Setting the Stage Chapter 3 Methodological Choices and Considerations
The overarching objective of this approach is to contribute to the concept of moral injury.
To be able to do so, this study consistently extends its critical epistemological approach to this concept. As noted in the previous chapter, a scientific concept is always an explanation, an interpretation, and, inevitably, a judgment (Lock and Nguyen 2010, Withuis 2010, Dehue 2011). ‘Moral injury’, likewise, is not a tangible and objectively identifiable entity, but rather a useful analytical lens to understand moral dimensions of deployment-related distress.
It is from this line of thought that this study distinguishes ‘moral distress’ (an open range of experiences) from ‘moral injury’ (a particular demarcation and conceptualization of these experiences). This approach echoes an established anthropological approach, which distinguishes people’s experience of illness, on the one hand, and on the other, biomedical concepts of disease and disorder (Kleinman 1988, Good 1994, Lock and Nguyen 2010). Starting from an unspecified, open notion of moral distress, this study explores experiences that overspill the boundaries of the current ‘lens’ of moral injury, ultimately to advance this lens.
Research Strategy
The aims of this study require a research strategy that has a strong inductive component (to gain novel insight into moral, political and societal dimensions of moral distress) while also making theory development (of the concept of moral injury) possible. This led me to adopt grounded theory, both a rigorous and flexible methodology to construct or advance theory, as my overarching research strategy. Rather than testing predetermined hypotheses, this method involves theory development through a relatively open analysis of relevant data, in which data and theory continuously inform one another (Glaser and Strauss 1967, Charmaz 2006). In accordance with revised versions of grounded theory (Evers and Wu 2006, Richardson and Kramer 2006), I did start from relevant existing theories and concepts, as is also the case in more traditional research, but I let the subsequent research process be guided by a continuous interplay between data and theory.
Specifically, this study involved three alternating activities. In anticipation of the next, detailed sections on these activities, let me briefly explain what they entailed. First, I developed a preliminary theoretical framework of relevant knowledge by gathering and integrating existing insights from the fields of psychology, philosophy and social sciences, which I presented in the previous chapter. Developing this framework allowed me to move from the predominantly psychological approach of the current concept of moral injury to a more comprehensive perspective.
Second, since existing knowledge appeared insufficient to gain insight into moral, political and societal dimensions of deployment-related moral distress, I conducted a qualitative empirical study, namely a multiple case study focused on two Dutch missions. The qualitative character of the study made it possible to gain in-depth insight into the particular mechanisms of moral distress. For reasons to be discussed in the next section, as my cases I chose the Dutchbat mission to Srebrenica, the former Yugoslavia, and the TFU mission to
45
Chapter 3 Methodological Choices and ConsiderationsPart I Setting the Stage
Uruzgan, Afghanistan. The main part of the case study data consisted of interviews with Dutchbat and TFU veterans.
Third, I analyzed these interviews as narratives, which allowed me to inductively and holistically explore veterans’ accounts. In contrast to, for instance, the investigation of pre-defined symptoms through structured questionnaires, this approach made openness to new observations possible while allowing these observations to be appreciated in their wider context (cf. Mishler 1986, Kleinman 1988, Finley 2011). Ultimately, the narrative approach meant that I gained insight into the personal experience of veterans while understanding their experience in the context of political practices and public perceptions. To do so adequately, I drew on existing information on the political practices, public perceptions and operational realities that characterized veterans’ missions, thus combining and triangulating the stories of veterans with other perspectives (cf. Morey and Luthans 1984, Eriksen 2001).
The inductive aspect of the grounded theory approach allowed me to draw novel insights from the empirical material. At the same time, the continuous use of theory assured that I linked and integrated empirically grounded insights with existing theoretical insights.
Together, these two aspects made it possible to achieve the research objective of advancing the understanding of moral, political and societal dimensions of moral distress, and to further develop the concept of moral injury.
Case Selection: Dutchbat and TFU
The empirical part of this research consisted of a multiple case study focused on two missions.
The focus on just two missions, instead of a wider range, made it possible to conduct detailed analysis of the specific political practices and public debates that characterized those missions, and of the ways in which veterans’ experiences are embedded in these practices and debates. At the same time, although comparison between cases was not the main aim of the empirical research, the choice of two missions as opposed to only one mission made it possible to distinguish mission-specific observations from general patterns and mechanisms.
Mission selection was motivated by two considerations. First, the two had to be partially dissimilar, both in order to approximate the diversity of present-day missions and to enable comparison (cf. Gerring 2007). Specifically, the missions had to be dissimilar with regard to the focal points of this study, namely moral dimensions of deployment-related distress at the individual level, the role of political practices in moral distress, and the role of public perceptions in moral distress. Yet both missions had to be sufficiently representative of the variety of present-day missions. While current research on ‘moral injury’ focuses on traditional situations of war and combat (see e.g. Litz et al. 2009, Vargas et al. 2013, Shay 2014), most recent missions are actually a variation of what has been called peace support missions (Pugh 2018). The second, practical consideration was to ensure that sufficient literature on the missions, and the decision-making processes and public debates was available. Eventually, I decided to focus on (a) Dutchbat veterans deployed to the former Yugoslavia (in what is now
46
Part I Setting the Stage Chapter 3 Methodological Choices and Considerations
called Bosnia and Herzegovina), as part of the ‘blue helmet’ UNPROFOR mission, and (b) TFU veterans deployed to Uruzgan, Afghanistan, as part of the ‘green helmet’ ISAF mission.
In anticipation of detailed discussions of both missions throughout this dissertation, let me sketch three relevant characteristics. The first has to do with moral distress at the individual level. In both missions many soldiers were exposed to violence and human suffering. However, Dutchbat troops were mainly bystanders of violence, though unwillingly, whereas TFU soldiers were often used force themselves. It seemed worthwhile to examine whether and how these different positions had different psychological effects, in particular for the investigation of moral dimensions of deployment-related stress.
Second, both missions had different political characteristics which seemed related to the abovementioned differences. While neither mission was a regular war operation, both differed significantly in their approach. The Dutchbat mission was a ‘blue helmet’
peacekeeping mission, while TFU’s mission was a ‘green helmet’ counterinsurgency mission involving combat. That is, Dutchbat soldiers were deployed as UN peacekeepers with light weaponry and few troops. At least partially due to their limited possibilities and resources, they were unable to perform their tasks. This incapacity became most clear in July 1995, when the third rotation of Dutchbat could not prevent the now infamous fall of Srebrenica and subsequent mass slaughter of thousands of boys and men (NIOD 2002, Van der Meulen and Soeters 2005, Van de Bildt 2015). In contrast to Dutchbat’s mission, TFU’s mission was part of a NATO-led mission which both anticipated and authorized combat. Although the Dutch government attempted to downplay the mission’s ‘green’ element, Dutch soldiers were mandated to engage in combat when deemed necessary, which turned out to regularly be the case (Grandia 2015, Dimitriu and Graaf 2016). The selection of these two missions thus made it possible to identify and compare potential moral quandaries ‘on the ground’ related to the political practices surrounding each mission.
Third, with respect to societal dimensions, extensive public debates surrounded and shaped both missions, yet in different ways. The Dutchbat mission soon became overshadowed by allegations of cowardice in the soldiers, who would have lacked the courage to fight (Van der Meulen and Soeters 2005, Van de Bildt 2015). In the TFU mission, Dutch troops did fight, which evoked mixed reactions, however. Although it bolstered the damaged reputation of Dutch soldiers (Klep 2011), it fueled criticism in the Dutch public and in parliament that the mission was simply combat in the guise of ‘reconstruction’ (Van der Meulen and Grandia 2012, Dimitriu and Graaf 2016). Whereas Dutchbat soldiers thus were accused of having been unwilling to fight, TFU soldiers received both admiration and criticism for the fact that they did fight, and it seemed insightful to examine the potentially different ways in which these divergent public responses affected the soldiers in question.
In short, both missions are characterized by individual, political and societal challenges, in interestingly different ways, which made them suitable cases for this study. This does not mean, however, that their characteristics became predetermined themes. In line with this study’s grounded theory approach, I took the accounts of the veterans interviewed as the starting point from which to examine relevant contextual factors, not the other way around.
As will become clear, this approach resulted in the finding that Dutchbat’s and TFU’s missions had far more in common than anticipated.
47
Chapter 3 Methodological Choices and ConsiderationsPart I Setting the Stage
Sampling and Data Collection
Sample Selection
Soldiers constitute a heterogeneous population, with different ranks, tasks and experiences.
Therefore, in addition to choosing the military missions, I had to choose a sample selection.
To do so, I employed the theoretical sampling method, which is typical in grounded theory research (Glaser and Strauss 1967, Charmaz 2006, Bryman 2012). While, for instance, random sampling is driven by the aim to collect data representative of a given population, theoretical sampling serves the aim of collecting theoretically valuable data. It involves the collection of material that makes it possible to fill determined gaps in extant theory and to refine theoretical ideas that develop during research (Glaser and Strauss 1967, Charmaz 2006, Bryman 2012). Put differently, theoretical sampling implies that the search for data is led by the study’s research objectives and questions.
My sample criteria were the result of several considerations. Importantly, the first concerned the goal of gaining a better understanding of moral distress; that is, not to collect statistical data about the prevalence of moral distress, but data that helped gain an in-depth understanding in order to advance theory about the phenomenon. While this goal did not require all research participants to experience deployment-related distress, it did seem valuable that it was the case for a substantial number of the participants, either at the time or in the past. However, since only a minority of veterans appears to develop mental health problems (Engelhard et al. 2007, Richardson et al. 2010), random sampling would likely have resulted in a handful of participants with deployment-related distress. Therefore, I chose to ensure that I would reach a substantial number of participants with deployment-related distress, while also including participants without deployment-related distress for the purpose of comparison. In later sections I explain how the data collection methods I used made this possible.
A second important issue concerned the comparability of the Dutchbat and TFU missions. To be able to juxtapose the two missions with respect to their political and societal characteristics, it seemed desirable that the selected veterans were as similar as possible with respect to other characteristics. Hence, for purposes of comparison, I chose to confine the research to veterans from similar types of units. In both Dutchbat and TFU, the infantry units of the Airmobile Brigade played a major role. Moreover, in both cases these units frequently encountered local civilians as well as combatants, and were exposed to violence and human suffering. Therefore, I selected the infantry units of the Airmobile Brigade, including attached troops such as medical units. With respect to rank, I chose not to further limit my focus, because it seemed insightful to explore similarities and differences across ranks.
Data Collection
The empirical research was carried out between July 2016 and March 2017. The research was divided into one short orientation phase and two additional phases. This division was made
48
Part I Setting the Stage Chapter 3 Methodological Choices and Considerations
so that data collection could alternate with data analysis, which made it possible to examine insights gained in one phase through additional questions in the next phase. Several sources and methods were used in the empirical research phase, including the literature, interviews, participant observation and other sources such as media documents. Below, I discuss these sources and methods.
Existing Information in Literature
I drew on three kinds of empirical material in existing literature. First, I made use of studies on the political practices, public debates and operational realities that characterized the Dutchbat and TFU mission, respectively, which provided important context for the primary data (e.g. Honig and Both 1996, NIOD 2002, Klep 2008, 2011, Beeres et al. 2012, Grandia 2015). Second, I drew on large-scale qualitative and quantitative studies on deployment-related guilt and ‘moral injury’, which provided useful ‘facts and figures’ to substantiate observations that would otherwise be more impressionistic (e.g. Rietveld 2009, Nash et al.
2013, Vargas et al. 2013, Bryan et al. 2016). Third, I made use of existing personal accounts of Dutchbat and TFU veterans, dating from shortly before these veterans’ deployment to long after their deployment, which aided in triangulating my own empirical material. These accounts were found in academic studies, reports, (auto)biographies, interviews, internet blogs and documentaries (e.g. Vogelaar et al. 1996, Kroon et al. 1997, KMAR 1999, Jongbloed 2002, NIOD 2002, Vogelaar and Kramer 2004, Praamsma et al. 2005, Hetebrij 2006, 2006, 2007, Eijsvoogel 2007, Kramer 2007, NRC TV 2009, Van Bemmel 2009, Ter Velde 2010, Van Hemert 2014, Veldhuizen 2014, Freebird69 2016, Veteraneninstituut.nl 2017).
Interviews
The main part of the empirical research comprised interviews. In total, I collected and analyzed 80 interviews, of which 40 were semi-structured interviews I conducted myself. The other 40 interviews were audio-recorded ‘life story’ interviews selected from the collection of the Netherlands Veterans Institute (Veteraneninstituut).
Self-Conducted Interviews
No contact details are readily available for veterans who have left service. Therefore, to recruit veterans, I used the following routes.
- Announcements on formal and informal websites for Dutch (ex-)soldiers - Networks of service members of different ranks, established in prior research
- In several cases, chance encounters at events such as military training and Veterans Day In addition, I employed snowball sampling, identifying research participants who could then refer me to colleagues, acquaintances or friends. Snowball sampling has been identified as a method that considerably facilitates research into sensitive topics which involve hard-to-reach populations (cf. Faugier and Sargeant 1997, Atkinson and Flint 2001, Penrod et al.
2003). It makes it possible to gain access to otherwise distrustful respondents who would
49
Chapter 3 Methodological Choices and ConsiderationsPart I Setting the Stage
not have participated had they not been referred by someone they knew, and it increases the likelihood that respondents are willing to speak openly about sensitive topics (cf. Atkinson and Flint 2001, Schulman-Green et al. 2009). This advantage seems to have been especially valuable in the case of Dutchbat veterans, whose negative experiences with media had made many distrustful of interviewers. Aware of the limitations of conventional snowball sampling regarding representativeness, I drew on the above-outlined variety of starting points. Doing so resulted in ‘multiple snowballs’, which extended the sample scope beyond a single social network, thus decreasing bias (Penrod et al. 2003).
Before interviewing a veteran, I always explained my study in general terms, by which I abided with the obligation of informed consent. In the announcements posted on websites, for instance, I stated that my study focuses ‘on how soldiers and veterans deal with challenges during deployment and after homecoming’. Having noted that the words ‘moral’ and ‘ethical’
often evoke specific associations in which most veterans do not recognize themselves, I decided to avoid these terms. Instead I gave the following examples, formulated such that they were open to interpretation. ‘Seeing, doing or experiencing difficult things. Having to make hard choices. Seeing injustice. Doing something that seems not right (or simply wrong). Being treated unjustly. Or, any other challenge soldiers and veterans may face’. I made explicit in my announcement that ‘it is not necessary that you recognize yourself in this description, because it is also an interesting fact if soldiers and veterans don’t. First and foremost, I am interested in the experiences of deployed soldiers’. Also, I clarified that all
often evoke specific associations in which most veterans do not recognize themselves, I decided to avoid these terms. Instead I gave the following examples, formulated such that they were open to interpretation. ‘Seeing, doing or experiencing difficult things. Having to make hard choices. Seeing injustice. Doing something that seems not right (or simply wrong). Being treated unjustly. Or, any other challenge soldiers and veterans may face’. I made explicit in my announcement that ‘it is not necessary that you recognize yourself in this description, because it is also an interesting fact if soldiers and veterans don’t. First and foremost, I am interested in the experiences of deployed soldiers’. Also, I clarified that all