| INTRODUCTION

Graduate-level attrition has historically been underexplored within higher education literature and within engineering education literature more specifically. However, in the United States, doctoral attrition has remained a critical issue for graduate programs for decades. Researchers have estimated that anywhere between 40% and 60% of doctoral students depart from their graduate programs depending on their discipline (Bair & Haworth, 1999). In engineering, 35% of women, 24% of men (Council of Graduate Schools, 2008;Sowell et al., 2015), and 57% of Black students (Sowell et al., 2015) depart from their graduate programs. It is estimated that a large majority (70%) of engineering doctoral students in the United States consider attrition at some point during their programs (Bahnson & Berdanier, 2023), while 10-year engineering PhD completion rates hover around just 60% depending on the engineering discipline (Council of Graduate Schools, 2008).

This attrition epidemic in higher education has widespread negative consequences at both the individual and structural levels. Students who depart can lose employment opportunities and experience negative mental health consequences associated with their departure (Geisinger & Raman, 2013;Johnson, 2012); faculty experience interruptions in research output that can affect tenure, promotion, and grant funding (Golde, 2005); and administrators and institutions can experience monetary losses from government grants, scholarships, and fellowships (Johnson, 2012). Having fewer doctoral students also decreases the availability of qualified individuals eligible to participate in and diversify the professoriate (Burt, 2020). Within engineering, attrition has widespread negative consequences for economic growth and development in the United States. This is due to the increasing demand for people with advanced engineering degrees to keep up with worldwide technological advancement that cannot be met with current enrollment numbers (Celeste et al., 2014;Okahana et al., 2018).

In an effort to address graduate-level attrition, the National Academies published a call to action in 2018 urging further research on graduate student experiences in graduate programs (National Academies, 2018). Engineering education literature exploring this topic has taken a student-centered approach, investigating the important psychosocial factors that students attribute to their attrition considerations. However, there remains a deficit of research exploring the influences of departmental policies, degree requirements, and organizational culture among other things, on graduate student attrition. In this paper, we explore how these structural factors may be contributing to engineering graduate attrition, which we define as any premature departure from an intended degree.

| LITERATURE REVIEW

Attrition is often examined through psychological and sociological lenses to understand the factors that influence departure or inhibit persistence in graduate programs. Literature indicates that self-efficacy (Braxton & Baird, 2001;Kelley & Salisbury-Glennon, 2016;Litalien et al., 2015), competency and academic identity development (Berdanier, 2019;Cruz et al., 2019;Hasbún et al., 2016;Jazvac-Martek, 2009;Sverdlik et al., 2018), and writing competency (Zerbe & Berdanier, 2020) can influence students' perceptions of their abilities to persist in their degrees. Chronic stress and lack of work/life balance, which are both common in graduate programs (Evans et al., 2018), also substantially impact attrition and persistence because of their detrimental effects on graduate students' mental health (Allen et al., 2020;Bekkouche et al., 2021;Schmidt & Hansson, 2018). Students with marginalized identities contend with added factors such as systemic racism, sexism, microaggressions, and other forms of discrimination (Bekkouche et al., 2021;Jones-White et al., 2021;McGee et al., 2021;Wilkins-Yel et al., 2022). These factors exacerbate negative mental health effects and perpetuate hostile working environments, which can lead students to consider attrition.

Researchers across disciplines have explored how advisor-student relationships and socialization processes affect graduate students' experiences. Foundational works by Lovitts (2001), Nelson (Lovitts & Nelson, 2000), and Nerad (2004) examined graduate students 0 decision making with regard to departure and found that negative advisor relationships greatly influenced attrition. These negative relationships can stem from a variety of reasons, including power dynamics, advisor matching processes, or mismatched expectations between advisors and students (Ampaw & Jaeger, 2012;Artiles Fonseca, 2019;Artiles & Matusovich, 2022;Barnes, 2010;Hunter & Devine, 2016;O'Meara et al., 2013). Advisor behaviors, including exploitation of these dynamics, rude or demeaning behavior, lack of research or personal support, and miscommunication, can also cause students to consider attrition (Golde, 2005;Lovitts, 2001;Maher et al., 2020;Ruud et al., 2016). These advisor relationships are integral to graduate student socialization experiences as advisors teach research domains and departmental norms. When students have challenging relationships with their advisors, they may lose opportunities

| ANALYTICAL FRAMEWORK

This study seeks to address these research questions by applying systems theory as an analytical framework to explore the attrition experiences of engineering graduate students. Popularized by von Bertalanffy in the late 1960s and early 1970s (1967,1972), systems theory investigates complex systems holistically to understand system behavior. The goal of this theory is to develop an understanding of a given system-whether that be a biological ecosystem, a business organization, or an educational program-by exploring interactions between all the components that encompass that system (Laszlo & Krippner, 1998). Holism, which is the idea that the parts of a whole are interconnected, is a core tenet of systems theory (Almaney, 1974). Through holism, systems theory posits that the components of a system cannot be understood in isolation, the system is greater than the sum of its parts, and all components of a system must be explored through systems thinking approaches to fully comprehend the system of interest. Although originally applied to biological processes, systems theory has been adopted by a wide range of disciplines including physics, psychology, social work, political science, organizational management, and education (e.g., Adams et al., 2014;Chikere & Nwoka, 2015;Clancy et al., 2008;Cox & Paley, 1997;De Bot et al., 2007;Jackson, 2000;Laszlo & Krippner, 1998;Rousseau, 2015;Schneider et al., 2017;Wulczyn et al., 2010). In this paper, we posit that graduate attrition is part of the educational system and apply systems theory as an analytical framework to further explore and understand engineering graduate attrition.

In accordance with systems theory, understanding the larger attrition system requires an exploration of all the components that can influence graduate attrition. These components can generally be categorized into one of three overarching categories: sociological, psychological, and structural factors influencing attrition. Engineering attrition literature to date has explored attrition mainly through sociological and psychological lenses. Researchers have studied how social interactions between engineering graduate students and their communities (sociological lens) and how individual motivation and mental health (psychological lens) can influence engineering graduate students' attrition. Only a handful of scholars, however, have critically analyzed graduate attrition through a structural lens. These researchers have explored how departmental culture (Gardner, 2009;Golde, 2005) or milestones including qualifying exams, proposals, and dissertations (Liera et al., 2023;McLaughlin et al., 2024) can affect graduate students' educational experiences. There are other structural components ripe for exploration in engineering contexts, though. For instance, when looking at the system of graduate education through a structural lens, the policies governing graduate students remain relatively obscured. These policies are formed over long periods of time, and students are rarely privy to their existence and rationale until they are personally affected. Faculty tenure and promotion requirements, technoeconomic priorities governing funding agencies, graduate admissions decisions, and even the influence of undergraduate class enrollment on the availability of teaching assistant (TA) positions remain mysterious to most graduate students. Departmental policies surrounding milestones may differ even within the same university, and applications of policy are often subject to interpretation by administrative individuals given their personal commitments to resolving issues. As faculty who are principal investigators are allowed nearly full autonomy in their mentorship styles, graduate student experiences are also impacted by the hidden and explicit policies and expectations of their advisors.

We posit that these policies are another aspect of the structural component of graduate education and attrition experiences that must be explored to understand the attrition system more comprehensively. Because existing literature in engineering education has explored engineering graduate education and attrition through sociological and psychological lenses but not extensively explored the structural components influencing attrition, it is our responsibility to do so. Applying systems theory as an analytical framework in this study allows us to explore another facet of engineering graduate attrition: the structural component of graduate education influencing attrition. In doing so, we add a structural perspective to existing literature on sociological and psychological perspectives of attrition, gaining a more holistic understanding of engineering graduate attrition.

| POSITIONALITY

At the time of the data collection and analysis, the first author was a graduate engineering student specializing in engineering education. The first author conducted, coded, and analyzed interviews with participants with guidance from the second author, who is a tenured engineering professor holding a PhD in engineering education, with over 10 years of experience studying graduate-level engineering education, attrition, and persistence. The first author's identity as a graduate student at the time of the interviews allowed her to develop rapport with participants. It also eliminated any perceived power dynamics between interviewer and interviewee, which encouraged participants to delve deeper into their experiences as graduate students and the effects of those experiences on their mental and emotional health.

We understand that our positionalities and identities influence our research including the research design and interpretations of findings. This is particularly evident in our choice of analytical framework and discussions of structural components that can influence attrition throughout this work. For example, we subscribe to the idea that an absence of structure or policy is still evidence of a values system and is part of the structural component of the graduate education system. This is consistent with conversations in other domains on "null curriculum" and "null policy," which state that the absence of conversation is an implicit affirmation of what is not addressed such that inaction can still be conceived as a form of action (McConnell & 't Hart, 2019).

This belief structure affords us the ability to interpret certain factors of attrition as structural when others may not, and aligns with our systems theory approach to attrition encompasses sociological, psychological, and structural components. Although we focus on the structural components of attrition in this work, we still acknowledge and adamantly believe that the sociological, psychological, and structural components of attrition do not exist in vacuums.

Instead, these components continuously affect and influence each other and engineering graduate attrition is often a result of a combination of these three components. At the same time, we also acknowledge that our positions in academia are due to our persistence and acceptance of the dominant and normative structures in engineering, which also influence our interpretations in ways we may not even be able to identify fully.

| METHODS

This qualitative work is part of a larger NSF-funded, IRB-approved, nationwide study on engineering graduate student attrition. The participants in this study are a subset of participants from a longitudinal, quantitative, SMS-based study on graduate student experiences. Because we were specifically interested in exploring the narratives of students who have chosen to depart from their PhD programs before receiving their intended degrees, only participants who met this criterion were selected for the current study. For more information on the larger longitudinal study, refer to our team's prior and ongoing work (e.g., Jwa & Berdanier, 2022).

| Participant recruitment

Participants for the current study were recruited as a subphase of the quantitative, longitudinal study mentioned above. To recruit participants for the original larger scale study, we emailed the graduate student coordinators and department chairs in every engineering department at the top 50 engineering PhD-granting universities in the United States as per ASEE's 2020 Engineering by the Numbers report (Roy, 2020). In these emails, administrators and faculty were asked to forward the study description and recruitment survey to their engineering graduate student listservs. The study description informed students of our interest in exploring their graduate school experiences, and the recruitment survey asked students to share demographic information including gender, race/ethnicity, citizenship status, number of years in graduate school, engineering discipline, and degree program. Participants were selected for maximum variation (Creswell & Poth, 2016) in gender, race, engineering discipline, year in program, and attrition consideration to improve participant representation and increase communication of the original study. In the recruitment survey, students also indicated how often they considered leaving their graduate programs. They provided their contact information depending on their interest in participating in the original longitudinal SMS-based study.

Students selected to participate in the original study completed either 1 or 2 years of longitudinal surveys three times per week and periodically informed the researchers about whether they intended to continue persisting in their engineering graduate programs. During the course of the longitudinal study, several students indicated that they were leaving or had left their PhD programs. As it is generally difficult to recruit participants who are leaving or have left their programs for research, and because of our research team's general interest in understanding the experiences of students who depart from their engineering graduate programs, we chose to follow up with the students who were no longer continuing in their PhD program to gauge their interest in participating in research on attrition. Of the 25 students who indicated they were departing, 8 agreed to be interviewed for this study. During the interviews, however, one participant indicated that they had always intended to obtain no more than a master's degree even though they were enrolled in a PhD program. Their experiences did not align with our definition of attrition, which we defined as a student departing from their graduate program before receiving the degree they had been working toward, and they were removed from the study. Table 1 provides the race/ethnicity and gender identity for each of the seven participants of this study. Although participants spanned the engineering disciplines of mechanical, civil/environmental, electrical, chemical, and bioengineering, their individual disciplines are not included in Table 1 to maintain their anonymity where possible and because their engineering disciplines were not salient to their attrition experiences. All participants were US citizens or permanent residents, aligning with the goals of the larger study.

| Data collection

Data were collected through participant interviews conducted by the first author. These interviews followed a semistructured interview protocol to allow for follow-up questions about salient participant experiences (Rubin & Rubin, 2011). This protocol was adapted from previous research conducted by the research team related to graduate student attrition (Sallai et al., 2023;Shanachilubwa et al., 2023). In the interviews, participants were asked about their journeys into engineering and graduate school and to describe their experiences throughout graduate school, including the level of support they received. Participants were also asked about the milestones they attempted and/or completed for their PhD programs and why they may not have been successfully completed. Because all participants had left or were in the process of leaving their PhD programs, they were asked about their experiences with departure such as when they began considering leaving, the experiences and situations that contributed to their departure, and how they felt once they knew they would be leaving their programs. Participant interviews lasted between 90 and 120 min depending on how much the participants wanted to share. Interviews were conducted online via Zoom because participants were located across the United States Participants were compensated with $10 Amazon gift cards at the end of the interviews. Audio recordings of interviews were sent to a secure professional transcription service and all identifying information was removed during the cleaning and validation of transcripts. All participants selected their own pseudonyms reflected in this paper.

| Data analysis

To increase trustworthiness in the data analysis process, we outline the rationale for choosing a qualitative method infused with a root cause analysis (RCA). During the interview and transcription validation processes, the authors began noticing a pattern among participants' attrition stories. When asked directly about why they departed from their PhD programs, participants provided straightforward, surface-level reasons that engineering graduate programs could readily attribute to students' preferences or lack of preparation such as failing qualifying exams or leaving graduate school for an industry job. However, when probed further about their experiences in graduate school, and with departure specifically, participants described structural factors, such as departmental and management policies, that hindered their ability to succeed as graduate students. These seemed to be the underlying causes of their attrition, even when attrition was attributed to something else. This observation suggested a possible misalignment in dominant attrition narratives between administrators and students in the formal reasons for why students leave and root causes of their departure. Guided by this observation, we conducted a qualitative RCA for participants' attrition from their programs. This analysis was conducted through a constructivist paradigm (Charmaz, 2017) to reiterate the importance of each participant's voice and the validity of their individual experiences.

RCA is a systems engineering approach to problem-solving that aligns well with our application of systems theory as an analytical framework. In RCA, the researcher examines the core reasons for undesirable outcomes that might not be visible at first glance in order to develop solutions that effectively target the problem instead of the symptoms (Latino et al., 2019;Root Cause Analysis, n.d.). This data analysis method allowed the researchers to "pull back the curtain" on attrition, so to speak, by exploring the core structural components that influenced participants' attrition from their engineering graduate programs. Because of the exploratory nature of this study, participants' RCAs were conducted in two separate rounds of open and axial thematic analysis to disentangle participants' perfunctory reasons for attrition from the underlying causes of attrition. First, we characterized what we describe as the participants' "official reason" for attrition. This was defined as the formal and easily documentable justification by graduate programs, and students themselves, for why participants were no longer enrolled in their intended degrees. Usually concise, these reasons tended to align well with reasons for attrition previously documented in the literature T A B L E 1 Demographic information, including race/ethnicity and gender for each of n = 7 participants.

Participant pseudonym

Self-identified race/ethnicity Self-identified gender

Faith

White Woman Jasper White Man Karina White Woman Oscar White Man Rose Asian, White Woman Sally Black Woman Samwise White Man

(e.g., Berdanier et al., 2020), including narratives around "fit" (Austin, 2002;Austin et al., 2009;Gardner, 2008). These reasons were coded using open coding procedures and grouped into overarching themes in the axial coding process. As presented in Section 6, four formal reasons for attrition emerged: change of career goals, perception of fit, not returning from leave of absence, and unfulfilled degree requirements.

The second round of coding was exploratory in nature, as few researchers have investigated the structural factors within graduate programs that affect student attrition. In this round, the first author began by open-coding the interviews for structural factors that contributed to each participant's negative graduate school experiences and their eventual attrition. These factors were defined as situations and policies beyond participants' control and were situated within a graduate education context. For example, a participant could report working in a toxic lab environment that was, from their perspective, unmediated by their advisor, unconducive to proper research training, and affected their work progress. This situation would be coded as poor management by advisor, toxic work environment, and lack of training. While the open coding process was informal to allow participant interviews to guide exploration of salient structural or organizational causes of attrition, memoing occurred for each participant interview following the code development process (Saldaña, 2011). Situations and policies that were identified during open coding were discussed with the research team to ensure that introductory codes accurately captured the experiences that participants described. The axial coding phase generated three structural factors that were root causes for participants' attrition from their graduate programs. These causes resulted from examinations of the common themes between codes developed during the open coding phase and in relation to the authors' stance that inaction can be a form of action (McConnell & 't Hart, 2019). These themes were discussed among the researchers before finalization to check for quality and understanding of participants' experiences and ensure pragmatic and procedural validity. The research team maintained a commitment to quality and rigor in the data analysis process through memoing and peer debriefing in the analysis and interpretation stages.

| Limitations

As with all research, this study inherently has limitations that must be acknowledged. Participants of this study were recruited through voluntary recruitment measures to discuss their experiences with attrition, which can be a very stressful and emotive experience. This form of recruitment could lead to self-selection bias within the participant sample, as only those who were compelled to share their experiences are represented. These participants were also selected from the 50 universities who confer the most engineering PhDs in the United States. While this was deliberate, as most people who want engineering PhDs apply to these institutions, there is opportunity in future work to explore engineering programs in universities not consistently on this list. As this work is part of a larger study focused on domestic US students' experiences, there is no international student representation. While this limits the attrition narratives and structural root causes that may influence students' attrition, international students experience power dynamics and academic pressure differently in graduate school due to visa and immigration requirements (Bluestein et al., 2018) in ways that we cannot do justice to in this paper. Therefore, the findings of this work may not be applicable for international students, as we do not account for these different contextual features. The small sample size and lack of racial/ethnic diversity in participant representation may also contribute to an incomplete representation of structural root causes that affect students' attrition. However, it is extremely difficult to recruit students who are or have departed from their engineering PhD programs, and we explore the attrition experiences of racially or ethnically minoritized individuals where possible. Many of the participants attended graduate school during the COVID-19 pandemic. During this time, they and the administrators and faculty of their programs were navigating moment by moment how to stay safe and preserve a semblance of well-being amid unprecedented pressures in work-life balance and bereavement and in the "dual pandemic" of COVID-19 and violent racism in the United States (Coley & Thomas, 2023). While participants acknowledged the pandemic's effect on support structures, mental health, and motivation, these participants were also clear that their experiences with administrative components of their departments and their relationships with their advisors were, in their perspectives, not affected by the pandemic. Last, we acknowledge that this study presents the perspectives of graduate students who decided to leave their program of study: It is likely that each of these participants' doctoral advisors (or other university administrators) would likely have a different perspective and account of events from their vantage point. In approaching this work, we both validate the lived experiences and perspectives of the graduate student participants, but also honor that all stakeholders in the educational system, including faculty advisors, are experiencing tensions, including a system of overwork and competing priorities in education and research.

In their interviews, participants described four formal reasons for their attrition from their programs. These reasons were considered as "voluntary" and "involuntary" processes. Changes in career goals, perceptions of fit, and not returning from leaves of absence produced voluntary attrition, while unfulfilled degree requirements produced involuntary attrition. Four of the participants departed from their graduate programs voluntarily. Jasper received a job offer from a company he was interning at during the summer and left his PhD after accepting the offer. Karina departed because she had experienced a difficult working environment that did not align with her work preferences. Rose and Samwise both opted to take leaves of absence due to the stress and anxiety associated with preparing for and taking their qualifying and preliminary exams, respectively, after which they decided not to return to their programs. The other three participants, however, left their programs involuntarily because they were unable to meet or pass the necessary requirements to continue in their PhDs. Faith's program required all students to maintain a minimum coreclass GPA, which she was unable to maintain over multiple semesters. After two attempts, Oscar was unable to pass his qualifying exam. Sally had received a conditional pass for her qualifying exam but was unable to satisfy the condition. All three participants were informed by their departments that they would no longer be able to continue in their PhDs and subsequently left their graduate programs.

While these formal reasons for leaving graduate school appear straightforward, RCAs of interviews with participants yielded a more nuanced understanding of the underlying structural causes of participants' departure. Three main structural causes were found to contribute to participants' attrition: working environment, advisor policies and practices, and department policies and practices. These themes were considered structural graduate school issues because they were described by participants as things beyond their control. Here, we subscribe to the critical sociological interpretation of the word "structural" to apply to practices and norms that govern the participants' decision-making processes and degree progress (Godwyn & Gittell, 2012). While sometimes structural causes are literal (e.g., written policy), they also can be unwritten practices, including habitual recurrent inaction or inattention that, in its absence, promotes a policy of neglect and serves as a barrier to PhD persistence and completion. In Table 2, we provide the thematic synthesis of the formal reasons for and root causes of departure, along with definitions of each theme.

While many of these root cause themes may seem similar to factors that have been well established in higher education and graduate attrition literature, we highlight that the way that the structural causes for attrition were outlined by the participants showed deeper structural and systemic issues at play, such that our naming of the themes seeks to reflect those differences. For example, "advisor relationship" is a common reason for attrition. However, what we found was that for our participants in this study, it was not necessarily the relationship with the advisor that was unsuccessful, but it was the habits, behaviors, and working practices of the advisor that served as their policy of group governance that was unsuccessful. Sometimes these were active policies and behaviors, but sometimes these

T A B L E 2 Official reasons for attrition and root cause structural factors contributing to student attrition. Formal reason for attrition Definition Change of career goals Receiving an immediate job offer Perception of fit Mismatch in student's expectations or preferences and their experiences, either within their department, research lab, or with their advisor Not returning from leave of absence Choosing not to return to a graduate program after taking an authorized extended time away (usually one or more semesters) Unfulfilled degree requirement Inability to successfully complete degree requirements (i.e., GPA minimums, course completion, qualifying/preliminary exams, dissertation defense) Structural root causes of attrition Definition Structure of working environment Structures dominating research and learning spaces including preexisting lab cultures, colleague interactions, and research training and collaboration opportunities Advisor policies and practices Advisors' handling of research and lab responsibilities, managerial role, and advising style with their students Department policies and practices Departmental culture including graduate program policies and requirements, expectations, organization, and communication with students

were "null policies" (the lack of policy) and behaviors associated with advisor inaction. Therefore, we named the advisor-related structural cause of attrition Advisor policies and practices to emphasize this distinction between sociological and structural factors of attrition. In the following subsections, we work to demonstrate how student attrition decisions are often governed by undercurrents of structural policies and null policy or patterns of inattention.

| Working environment culture as evidence of structural issues

Karina and Samwise's working environments significantly impacted their attrition from their graduate programs. Working environments were defined by interactions with colleagues, pre-existing lab culture, and training and collaboration opportunities. Neither participant experienced positive working environments during their time in graduate school. They described their research labs as toxic working spaces that hindered training, research progress, collaboration, and connection between labmates. One of the common recommendations in the literature is for incoming students to speak with current students, specifically asking about the culture and environment of the research laboratory (Gardner, 2007;Maher et al., 2019;Wofford & Blaney, 2021). However, despite following this advice, students can experience a "bait and switch": Karina, for example, experienced a very different research environment than the one advertised during recruitment. Older students had described the lab culture and working environment as a positive, collaborative community when she was deciding whether to join their group. After she began working in the lab, however, Karina realized she had been misled, as the lab culture was completely different to previous descriptions. Instead of a collaborative research environment that fostered camaraderie, she dealt with lab toxicity. In one instance, she described how the research faculty member hired by the advisor to manage the lab was known to purposefully withhold vital lab equipment, noting this individual "actively does not share information and takes reagents and is hindering some progression." She further described the challenging relationships with older graduate students that eliminated the potential for collaboration and mentorship, saying:

The fifth and fourth year [students] when I joined the lab were incredibly rude and there was definitely, I learned later, a toxic culture in the lab in previous years … there were a lot of comments [from the older students in the lab] about "why would we [the older, experienced students in the lab] even train you … like, you have a protocol," or "that's not my responsibility," etc. There was never any benefit from having the older students around to train us. It's like us [the first and second year students] floundering and getting everything done from ground zero. So, it was reinventing the wheel all over again and voicing that to our PI.

Karina tried to agentically manage this toxic lab culture by requesting interventions and support from her advisor and building a positive counterculture with the other first-and second-year students in the lab. However, the toxic culture and lack of collaboration remained detrimental to her experiences, as she increasingly felt she was wasting her time in her graduate program and lab because she was not receiving the training and gaining the skills she expected from graduate school. Samwise also experienced a toxic working environment, stunting his ability to receive proper research training. In his case, only one senior graduate student in the lab was willing to train incoming students. This student, however, had poor mentoring skills and purposefully withheld protocols. Samwise recounted the senior student saying things like:

[my labmate would say things like] "I'm not gonna share my protocols with you. You need to develop them on your own because it's a good scientific exercise …" And learning these experimental techniques from her, I would watch her do them and I would have to sort of "please give me your protocols," beg a bit. And I would then write my own protocols and say, "Hey, can you review these for me, then? Am I missing anything? Would you say they're complete?" […] And then she would just go on a tirade of Microsoft Word comments, with just so many mean comments and "I don't know why you would say this here" kind of a thing.

Samwise described this constant pattern of derogatory behavior, noting that the "feedback was really just unhelpful and mean […] just constant criticism, but not even helpful criticism. And yeah, that was kind of, I guess her M.O. and other students, new students in the lab, we kind of felt the same in terms of how she was treating us."

Although he and other first-year students tried to address their issues with working with this labmate, nothing changed, even after reporting these interactions to his advisor. Observing his advisor's ongoing failure to intercede, Samwise made the decision to distance himself from this labmate to attempt to reduce the toll that trying, and failing, to please her and still not receiving adequate training was taking on his mental well-being. This resulted in his decision to leave, noting "it's not that I'm not interested [in my research], it's that I just like, I don't know how [to do my research]. It's like nobody's helping me [learn] and I'm trying to figure this out, but it was incredibly difficult."

Prior literature has discussed laboratory dynamics and interpersonal relationships as being essential to thriving in graduate school. But in this work, we take the critical perspective that these patterns of behaviors and climate issues are not just one-off problems relating to interpersonal issues; they are established patterns of toxic behavior that are evidence of systemic problems. While, yes, there are individuals and personal interactions that are occurring, the presentation of issues here-as well as the wide breadth of climate issues documented in the literature across institutional types and geographic locations-demonstrates a deeper structural problem within engineering. Both Karina and Samwise agentically tried to navigate their situations, but the structural climate of the lab made it impossible for them to make progress and succeed in their environment.

| Advisor practices and informal policies can be structural barriers

While "advisor relationship" is commonly addressed in the literature as an interpersonal issue and main contributor to attrition, our participants highlight in their experiences that sometimes the actual one-to-one relationship with advisors was not an issue as much as the patterns of behavior and policies of practice. The enacted practices, or lack thereof, indicated to our participants how their advisors valued student learning. Here, we defined advisor policies and practices as the (in)actions taken by an advisor with regard to lab/research management and the advising styles and policies implemented during advising. This theme related to how advisors' interpretations of their responsibilities to advisees manifested in their interactions with them. Continuing to explore Karina and Samwise's struggles: each participant discussed toxicity issues with their advisors in an attempt to improve their lab culture but was met with inaction that perpetuated lab toxicity. In one instance, Karina described a situation in which her advisor failed to provide any conflict management or reiterate the importance of a positive, collaborative lab environment: Behind my back […] I had overheard [labmate] said, "Why is [Karina] even here? She can't even figure out a known protocol or know what we do." And then had already at this point talked about other first years as calling their questions ridiculous and that they should just 'Shut up and read' or why were they even here. [I] brought up this concern to the PI. I was like, "Look, this fifth year is saying negative things about the first years. It's not okay." [But it] never got addressed. I'm not okay with that. [My PI is] so nonconfrontational and distant that it's ridiculous.

Karina experienced advisor inaction with regard to lab management that led to the perpetuation of toxicity in her lab and to the perception of lack of fit within the research group and graduate program. This ultimately caused her attrition from her program.

In addition to not feeling like he was getting the mentorship to develop technical research skills in the laboratory, Samwise recounted feeling powerless in preparing for his preliminary exams, in which he was expected to write and present a research proposal to his committee which included his advisor. Although there was no official departmental policy against it, Samwise's advisor refused to provide guidance: And then finally, prelims were coming up [… My advisor then said] "Okay, this is an exam. So, I'm gonna provide minimal input, 'cause this is … me and the committee trying to examine whether you're a worthy [PhD] candidate. And so, I'm gonna provide minimal input on things like if you're writing your dossier or you're working on your presentation." And so, it was just like, 'I'm still formulating my project ideas and you're telling me now that … you're gonna provide minimal input on the direction of my thesis!' And it was just, oh, it was a lot of stress on top of already not being able to perform my experiments [because of lack of training]. And so, I just felt like I could not complete my prelims.

This sudden change of expectations reinforced the lack of technical support that Samwise had already felt, especially given that this advisor policy related to guidance on the dissertation proposal development was unexpected. Samwise later explained that this made him feel extremely overwhelmed about his preliminary exam, inspiring him to take a leave of absence and eventually leave his PhD program. Here, we see advisor policy coupling with inaction around managing laboratory working culture that caused unfulfilled program requirements (the preliminary exam), inspired him to take a leave of absence and not return, and ultimately caused his departure.

Another participant, Jasper, also experienced issues with his advisor's advising policies and style that contributed to his decision to leave his graduate program. He described his advisor as "extremely hands-off" due to his prominent administrative position within the department. Jasper would spend months "going down rabbit holes" of research before having conversations with his advisor to interpret or trouble shoot his research. He began feeling frustrated that he was not making enough research progress toward his degree, which then led to feelings of stuckness and questioning whether to continue in his PhD. These feelings of stuckness became more prevalent due to his advisor's ambiguous policies regarding timing of the dissertation proposal (in their department, called "candidacy"):

[I felt] a little bit frustrated because [not being able to take candidacy] both made me feel like I wasn't making as much progress because it wasn't clear when I would be ready to do it. With everybody else [in my program] it was sort of once you finish your classes by definition, then you are ready [to take candidacy]. In my group, it was sort of, you are ready to take this whenever your professor says to you that, "Hey, I think you're ready to take it" and mine never did, and was never quite clear about what criteria that would be beyond just "We have more work to do."

As a result, Jasper was unable to attempt this milestone and lost opportunities for research funding that could help him progress in his degree and feel less stuck:

There are some other benefits if I had completed that milestone [candidacy], basically would have automatically given me a master's degree if I had completed that … There are certain fellowships that you're only eligible for if you've completed your candidacy … There were a lot of fellowships that I was not able to apply for because I didn't have that and if I had applied for them, then I might have felt a little bit more independent.

Jasper continued to feel stuck in his PhD and worried about how much longer he would have to spend in graduate school to actually complete his degree. Jasper eventually left his program because of his advisor's policies that created too much uncertainty about his graduation timeline, and he had received a job opportunity that provided more guidance and gave him more certainty of his work timelines. To that end, Jasper's experience is another indication that the reason for departure (changing career goals when presented with a job opportunity) masks underlying structural issues including unclear and seemingly arbitrary policies about readiness for academic milestones that could have helped Jasper to persist.

In sum, the experiences of these three participants highlighted in this section show further how official reasons for departure differ from the root causes of departure, and how these causes are, at their core, related to structure and policy. While faculty autonomy and decision making around how the research group operates is fully appreciated, the lack of clarity and inattention to student issues is at best a form of "benign" neglect. Here, we operationalize "benign" neglect as (perceived) unwillingness to engage. We propose the modifier "benign" to capture an optimistic assumption that there is likely not malicious intent behind advisor actions-but, as recent work by Fleming et al. (2024) noted, advisors' working styles or skills in handling situations can have a multitude of downstream effects on student well-being and career decisions. While perhaps not of malicious intent, avoiding sensitive issues or confrontation causes inadvertent harm to the students. At worst, even unintentional (benign) neglect in the discomfort or unwillingness to intercede in laboratory environments promote workplace hostility and bullying from other students and even research faculty (recalling Karina's experience). These issues go far deeper than considering just differences in "advising style" or "interpersonal traits"-these issues as presented are ultimately structural in nature, and we argue that if these factors of attrition continue to be conceptualized as individual issues, they can never be fixed.

| Department policies and practices can unintentionally impede persistence

Departmental policies and practices impacted many participants' attrition. These policies and practices shaped the departmental culture and were defined by graduate program policies regarding degree requirements and department organization and communication with students. While departmental culture contributed to many participants' attrition, it considerably influenced the attrition of students who left involuntarily.

Sally, for example, had taken her qualifying exam and received a conditional pass. In her department, the qualifying exam was course-based, meaning it tested students' knowledge of four key subject areas that students were supposed to have completed classes in before taking the exam. Requirements for the exam itself were very clear, but requirements of what conditions could be applied for a conditional pass were vague and left to the discretion of the committee of professors grading the exam. Sally's condition to officially pass her exam was that she had to write a publishing-quality research article within a year. Sally attempted and eventually failed to meet this condition. She expressed frustration with the departmental policy of allowing this type of condition to determine whether a student was good enough to be in the PhD program, given differences in research domains, variations in project maturity, previous experiences writing papers, and the inherent nonlinearity of real laboratory research. Some of her frustration stemmed from a lack of research clarity and the shift in research focus that occurred during that year she was writing her paper. Sally explained, "'Cause I don't feel like my work was that focused. And that didn't really come … As I was writing, I was like, oh, okay [this research doesn't make that much sense]. But I had to finish the paper because I had a deadline, and so was just kind of committing to [the original research idea]." Upon reflection, Sally described frustration at being forced by the departmental conditional pass policy to write a paper within a year without consideration for the possibility and reality that research is not linear, can find new directions, and progress can stall: I think it was difficult because, I don't know, I don't know if it's fair to give a one-year deadline to write a paper. Maybe, I don't know. 'Cause things can happen and then you never know what will potentially pop up. Research isn't very linear. I mean, ultimately, we had to change the direction of the paper because there wasn't enough data for the initial idea. And so, it was kinda frustrating that [not passing] was ultimately the ending and this was where it was going on.

Faith experienced similar frustrations with the policies and practices of her department that ultimately forced her to leave her PhD. Faith's department had a minimum GPA requirement for core classes. While this requirement was in the graduate student handbook, it was not widely discussed by the department, faculty, or graduate students. Faith explained that at orientation they did not even discuss the GPA requirement: "They just told us we had to take classes and we had to take a qualifier exam and then a prelim and then defense at the end. They didn't talk a lot about [the GPA requirement]. We kind of just got a really thick … a grad guidebook and it's like, here are the requirements." Throughout her time in school, Faith also received consistent messaging from her advisor, faculty, and peers that classes were not very important and did not predict student success. She did not have the minimum core-class GPA and was eventually informed by her department that she needed to raise her GPA by taking more classes. Her department communicated this after class registration deadlines and informed her again the next semester so late that she believed she had been cleared of the GPA requirements:

It was like halfway through the semester … [the department was] like "Your GPA is not where it should be" and it was kind of like, "Hey, you should take another core class to raise your GPA," but it was past the add-drop deadline. So, I couldn't do anything that semester, so I waited till the spring … That [class] I also got a B-… They didn't say anything at the end of spring … I thought that if they didn't say anything, I was like, "Okay, I guess it's fine. I don't know." But halfway through the summer they're like, "Okay, no, your GPA is still not where it should be. You should sign up for another core class." So that's when I retook the same one I took back in [my first semester].

Faith explained that even though she retook multiple core classes, her GPA was not meeting the department requirements and she was eventually encouraged to leave her program. However, she also explained that her undergraduate university had not graded on a plus/minus system, but her graduate school did and because her department never mentioned this, she did not realize until much later into taking classes that there were such significant differences in the GPA weighting.

Oscar's experience with formal policy was perhaps the most egregious of all participants. Like Sally, his qualifying exam was meant to be a course-based exam. The departmental policy in his program was that students could be tested on anything related to the courses they had taken in graduate school. On his first attempt at the qualifying exam, Oscar had never attempted an oral exam before and was so nervous he knew he had done poorly and failed his qualifying exam. When Oscar went in for his second attempt, however, he was asked questions about his specific research and his general research area, neither of which was taught in any courses in his department:

The limitations that hypothetically are put on the qualifying exam are that it's only supposed to be about your coursework, courses that you have completed at my graduate school. My advisor asked me a bunch of questions about my research, although not even really about my research, it's not stuff that I was implementing in my research, so I didn't answer those. He wasn't very thrilled about that. And I was like, "I've literally never touched on this." Some stuff that was [asked on the exam] … I've done in coursework … in undergrad. I haven't done it in grad school … Oscar failed his second attempt at his qualifying exam, which meant he was forced to leave his PhD program even though he still wanted a PhD. He felt he had been unfairly tested and misled by the department's official policy on what could be included in the qualifying exam. The only participant to raise concerns with departmental administration, Oscar arranged to speak with the graduate program department chair to bring this issue forward and reconcile these differences in policy and practice for future students. However, the discussion with the department chair surprised him: And so, I met eventually with the chair of [my] graduate studies department after this [second] attempt. I was like, "Look, I'm fine that I failed. I'll just leave … I'm not enjoying myself enough here, anyways, this is just a convenient path out, I suppose. But like the website says this, the emails say this, this was my experience." And [the department chair's] response was, "Yeah. We say that, but the thing is the professors can ask whatever they want." And then I said, "Okay. Well, say that. I need to be able to … The exam already covers basically whatever the professors want, but I need to know that it actually just covers literally anything that they want and there's no reason for me to try and prepare." 'Cause I would've, I think, done just as well if I spent zero time studying for the exam, if I hadn't put any time in, because so many of the questions came from outside of the scope of the already very broad scope of what I was trying to prepare for.

From that conversation, Oscar learned that the department was aware of this discrepancy in policy and practice but continued to falsely advertise the expectations of the qualifying exam to students. The department chair also lacked remorse and seemed unwilling to change departmental practices or update the handbooks to better reflect departmental realities.

Sally, Faith, and Oscar all left their graduate programs because they were unable to fulfill their degree requirements. But on closer inspection we see that the effect of departmental policies and practices was a root cause of their inability to complete these requirements and led to their attrition. In total, the findings related to this theme are most connected to literal interpretations of "structural" issues, specifically to enacted, written, and formal policy language which is how the term "structural issues" is typically envisioned. Detrimental department policies and practices tended to manifest as participants were attempting degree milestones or meeting degree requirements. There was also minimal discussion of departmental interest in reflecting on how their policies or practices were negatively impacting student experiences and could contribute to attrition.

| DISCUSSION AND IMPLICATIONS

The purpose of this study was to determine how structural factors impact engineering doctoral students' attrition decisions, interrogating the differences between official reasons for and root causes of attrition from a critical structural systems lens. Approaching this work by applying systems theory as an analytical framework allowed us to investigate attrition from a structural perspective, expanding our understanding of the components that affect and contribute to attrition. While previous literature has extensively explored the sociological and psychological components of attrition, this work explored the structural components to enhance the holistic examination of graduate attrition that is encouraged by the systems theory framework. From our investigation into the structural components, we see how subsystems-a lab group comprising a PI, multiple graduate students, and research staff within a department, within a university-can be governed by both formal and informal policies, some of which are written and some of which are policy via inaction. We also find there may be deeper structural roots to what are often considered interpersonal student issues. The holistic approach central to systems theory helps us to connect themes in ways that are typically held separately in prior graduate education literature. For example, most literature holds the advisor relationship separate from departmental policies, but this approach finds that these aspects are innately entangled, calling to mind other recent literature on advisor matching processes (Artiles et al., 2023;Artiles & Matusovich, 2022). This approach also helps us to explore how a lack of policy can be a structural component that impacts students such as the lack of enacted policy in Oscar's case, or the lack of policy surrounding interventions on hostility for Karina and Samwise that ends up representing a commitment to avoiding difficult conversations at the expense of student well-being.

To implement systems theory as the analytical framework, we conducted a qualitative RCA of structural factors that contribute to attrition. Through this analysis, we identified three main structural factors contributing to attrition: working environment, advisor policies and practices, and department policies and practices. This work adds value to the engineering education literature related to doctoral student attrition in multiple ways. First, this research adopts an RCA methodological approach to understand graduate student attrition, a method of analysis that is sparsely used in qualitative and engineering education research. This analysis shows that the reasons given for attrition are often not the same as the underlying causes of attrition. Second, this is one of the first studies to specifically center the role of academic structures in engineering doctoral attrition, both formal and informal, present and "null." This is one of the first works to highlight how many of the traditionally "interpersonal" factors that are documented in the literature may actually have structural underpinnings, which corroborate widespread documentation of hostile work environments and "chilly climates" in engineering (Malicky, 2003;Morris & Daniel, 2008;Rinc on & George-Jackson, 2016). To that end, we promote that if we can consider how the factors of attrition that seem to be individualized may actually be systemic, we can better envision how policy can be created and enacted to improve climate and culture, and, therefore, persistence. The findings from this study indicate that structural components within higher education can be integral to students' attrition experiences, suggesting the necessity for critical reflection from within academic structures to better address attrition.

7.1 | The story is not the story: How conventional characterizations of attrition may ignore or obscure underlying structural issues Each participant indicated an official reason for attrition that aligns well with conventional faculty narratives on attrition, including "lack of fit" or "inability to 'cut it' in graduate school," which is the anecdotal version of the theme "inability to complete departmental milestones." However, in-depth conversations throughout this study challenged these narratives as participants revealed the underlying working environments, advisor policies and practices, and department policies and practices that drove their attrition. Administrators or faculty listening to Jasper's story, for example, might assume he left his PhD because his career goals changed. While there is some truth in that, Jasper revealed that his advisor's policies on milestone completion, lack of clear expectations, and extremely hands-off approach pushed him to accept a job when the opportunity arose. Karina's narrative may suggest a "lack of fit" problem where she may not have been suited to work in her research environment. However, the toxic working environment and her advisor's lack of action to mitigate it pushed her to leave. Faculty may be quick to attribute Rose and Samwise's not returning from leaves of absence to the idea that they "couldn't 'cut it'" in a graduate environment, but Rose's challenges with her department's policy on qualifying exams and Samwise's hostile labmate and advisor's advising policies with regard to milestones were the root causes of their leaves of absence and subsequent attrition. Finally, Sally, Faith, and Oscar were all required to leave their PhD programs because they did not successfully complete a degree requirement. However, departmental policies and practices related to these requirements negatively affected their ability to successfully complete these requirements.

RCA of this study supports existing literature that disproves commonly held faculty assumptions about attrition and suggests that attrition is a complex process (Berdanier et al., 2020;Shanachilubwa et al., 2023;Zerbe et al., 2022). Previous studies have explored this complexity through analyses of sociological and psychological factors that affect attrition. In considering the system of graduate education and attempting to view attrition more holistically through the systems theory (Almaney, 1974;Laszlo & Krippner, 1998), we find that structural components of the graduate education system can be misconstrued as interpersonal issues but play an important role in students' attrition experiences. While sociological and psychological attrition literature highlights the agentic nature of graduate attrition, we also want to acknowledge tension in students' agency found when discussing structural components. We agree that attrition is often an agentic process in which students have decision-making power to determine whether they should persist or depart. However, because all parts of a system are interconnected and can influence one another (Laszlo & Krippner, 1998), we understand that structural components are beyond students' control and can hinder students' agency in attrition decisions. From this work, we posit that it is essential to holistically explore systems to define and mitigate any tensions that might arise and provide viable solutions to complex problems.

7.2 | Engineering literature to date has not considered "benign" neglect as evidence of structural issues causing doctoral attrition

The ways in which advisor administrative policies affect attrition is particularly salient in this work. While graduate education literature across disciplines has extensively explored the advisor-student relationship (Ampaw & Jaeger, 2012;Artiles & Matusovich, 2022;Golde, 2005;Hunter & Devine, 2016;Lovitts & Nelson, 2000), it has mainly been addressed through a socialization lens, where advisors' mentorship competency and success in teaching research and departmental norms and the quality of advisors' personal interactions with students have been found to significantly influence graduate student experiences and attrition. Advisors, however, also have managerial responsibilities because graduate students are often in liminal spaces between being students and employees of their universities. This study found that advisors' managerial roles significantly influenced students' attrition. These roles are considered a structural factor because the advisor is akin to being a manager in workplace settings, holding responsibilities to manage workplace conflict.

Structural issues related to an advisor's management policies manifested in a variety of ways. One significant way was advisors' "benign" neglect to their labs, that is, a perceived unwillingness to mitigate conflict between students in their research group (from the students' perspective). Karina and Samwise both experienced toxic and hostile working environments that adversely affected their research training. Karina's advisor was averse to conflict and seemed to refuse to manage the internal conflicts within his research group, resulting in a pitting older graduate student against younger ones which perpetuated the toxic working environment. Samwise's advisor disregarded students' concerns for working with their hostile labmate and gave them no alternatives to learning research techniques. This caused Samwise to flounder in preparation for the dissertation proposal milestone. It could be argued that neither advisor fulfilled their managerial responsibilities to their graduate students, perpetuating unfavorable working conditions and disenchantment with academic systems. Jasper's advisor had a strict policy on when students were allowed to complete the second of their three milestones for their PhD programs. He made it very clear that students could complete this milestone only when he deemed they were ready, but he failed to define what constituted being "ready." Participant interviews show how advisors' mishandling of student issues, often through non-action or null policy, considerably impacts attrition.

The departmental effect on attrition is also notable in this work, as participants were negatively affected by departmental policies and practices. Interestingly, these issues generally related to degree requirements including the completion of milestones. Faith, like many graduate students, believed that courses were less important in graduate school. Although this narrative is perpetuated by faculty, including Faith's advisor, her department still required students to maintain a relatively high minimum GPA. While it is true that all graduate students had access to the policies in a handbook, the mixed formal, informal, and-most importantly, delinquent-messaging meant that she was not able to remedy her GPA in time, resulting in an involuntary departure. Other participants, including Sally and Oscar, struggled with their qualifying exams. In both cases, there was a misalignment between the policy surrounding the exam objectives and the reality of what was tested. Qualifying exams are an area ripe with criticism in graduate education literature. Researchers advocate for critical reflection on behalf of departments to understand what structures qualifying exams are upholding, including inequitable and potentially biased practices (Posselt, 2020). They also argue that qualifying exams in their current state are poorly designed because they do not test necessary professional skills for doctoral programs including research and communication skills (Cassuto, 2015;Riviere, 2016).

Many of the advisor and departmental attrition factors connect with larger conversations in graduate educational literature surrounding scaffolding within graduate programming (Ferdinand-James & Medina-Charles, 2022;Pittaway et al., 2023;Riviere, 2016), especially with respect to considering how milestones can and should be scaffolded. In our interpretation of the body of literature addressing milestones within doctoral attrition and completion across disciplines, advisors and departments sometimes forget that the goal of a doctoral program is to develop a student into an independent researcher by the time they graduate and not to have enrolling students as already independent researchers. This may lead them to provide inadequate advising, mentorship, support, and examinations to help students achieve this objective. In combination with ambiguous policy, policy not upheld, or null/absent policy signaling values within a research group or a department, we show here that the lack of attention to structural graduate student issues is perhaps not malicious but is certainly not without consequence.

| Implications for research advisors, graduate chairs, and policymakers

This research is particularly relevant to faculty and administrators in graduate engineering programs. While relevant recommendations are provided for each stakeholder, we do not take the stance that graduate students should (or can) be responsible for changing structural issues that affect them. First, faculty who advise graduate students must consider that they have a responsibility to manage their research labs (or hire an appropriate person such as a postdoc or research faculty to manage their labs if they are too busy). While graduate students can guide lab culture, it is the faculty member's responsibility to set and maintain appropriate lab culture. This includes promoting and rewarding collaboration between students and teaching older graduate students how to adequately mentor younger graduate students. Because we take the view that inaction is a form of action (that is often considered support of systemic misbehavior) (McConnell & 't Hart, 2019), faculty advisors must be wary of deliberate or inadvertent inaction in relation to their lab environments. This is because inactions can be interpreted as approval of toxicity in lab environments or lack of regard for graduate students.

Maintaining a positive working environment for students not only improves their personal graduate school experiences but also fosters productivity. Because learning how to be an adequate manager is not something advisors are taught before they establish their labs, administrators and departments can step in to fill this knowledge gap through workplace training opportunities. Considering parallel conversations on managing workplace bullying and hostility, writing and enforcing clear consequences, and reporting structures in laboratory expectations documents would be an important first step. While it is not the goal to force all lab members to be best friends, it is an essential professional skill that all graduate students learn to be collegial and respectful with colleagues, even if they do not personally get along.

It is also important for advisors to keep in mind the stage that their graduate students are in as they consider advising and mentoring styles and the personal policies that any stylistic choices might convey. For example, a senior graduate student is better able to work independently, and a more hands-off advising style may be suitable for them. However, a first-or second-year graduate student is still being socialized into graduate school, research, and their specific research domain. They may require a more hands-on approach to learn how to properly find, read, and synthesize literature, how to develop research plans, and how to write for academic publications. We are not advocating for coddling or micromanaging of graduate students. We are, however, advocating for advisors to consider how research can be appropriately scaffolded and to remember that the independent researcher is meant to be developed by the end of the doctoral degree, not already present fully in the beginning.

There are increasing discussions related to the structures of current milestones and the accuracy with which they test and prepare graduate students for their research and careers (Cassuto, 2015;Liera et al., 2023;McLaughlin et al., 2024;Riviere, 2016). Administrators and faculty who lead departments could benefit from internal reviews of their degree requirements to determine whether those requirements align with their overall goals for students by the time they complete their PhDs. One such example is the option to alter qualifying exams to better reflect research learning and knowledge acquisition. Departments must also provide clear guidelines about what is expected of students to remain in their graduate programs and to successfully complete their milestones. Supplying students with rubrics that clearly define how qualifying exams are graded before the actual exam is one way for departments to mitigate uncertainty and potential biases that can arise in scoring. Departments must also hold the administration and faculty accountable to adhere to these guidelines to remain consistent, minimizing intentional or unintentional discrimination and biases.

Most importantly, departments, administrators, and faculty must acknowledge that they can also be responsible for graduate student attrition, even when it seems that the graduate student "simply" did not meet the requirements set forth by their department. Administering exit surveys for Master's students similar to the exit interviews conducted with PhD graduates is a vital opportunity for departments to learn about their departmental culture and develop actionable items to improve graduate student experiences. It is important to conduct these surveys with Master's students as well because many departing engineering PhD students depart with Master's degrees, thereby documented as successes in departmental statistics. Departmental self-reflection on the roles that structure and policy can play in attrition will improve the quality of education and personal and professional experiences for graduate students and has the potential to decrease engineering graduate attrition. Future work could explore the implications of altering qualifying exams to be more consistent with the goal of developing independent researchers or investigate the viability of improved working environments and increased satisfaction with advising through manager training for advisors.

| CONCLUSION

This study applied systems theory as an analytical framework to investigate structural factors that can contribute to engineering graduate students. In doing so, we are developing a more holistic understanding of attrition within the graduate education system. In this work, we explored the experiences of seven engineering graduate students who were leaving or had already left their PhD programs. The methodological approach used, namely the RCA approach, provides a valuable methodological contribution to the engineering education literature, as this is one of the first studies to analyze qualitative interviews in this way. This analysis revealed three structural factors that considerably impacted engineering graduate students' attrition. Although these students had officially left their programs because of common attrition narratives related to "lack of fit" or "inability to cut it in graduate school," there were underlying structural issues that influenced their attrition: working environment, advisor policies and practices, and department policies and practices. This work emphasizes the fact that educational structures such as advisor and departmental policies can and do influence attrition. Advisors' failures to meet managerial requirements related to mitigating hostile working environments and unwillingness to provide appropriate scaffolding to teach research independence can significantly negatively impact students' experiences. The enactment or lack thereof of departmental policies regarding qualifying exams and other degree requirements is also capable of impacting attrition. Ultimately, this work adds value to the engineering graduate attrition literature in multiple ways. It critically analyzes the role of advisors and departments from a structural perspective, supplementing the sociological and psychological perspectives commonly explored in attrition literature. This work also exemplifies the complexity of attrition, asserting that attrition is more complex than faculty or administrators may want to admit. This work is especially important for faculty and administrators who interact with and have the power to influence graduate programs and policies. These individuals can engage with this work to critically reflect on enactment of their graduate policies and practices to ensure that they are developing graduate school environments that set students up for success instead of failure. Doing so can help mitigate unnecessary and often unwanted attrition by both students and faculty.