‘Treat GTAs as colleagues, rather than spare parts’: the identity, agency, and wellbeing of graduate teaching assistants

ABSTRACT In recent years, the wellbeing of students and staff in Higher Education has received increased attention in pedagogical research. However, the experiences of Graduate Teaching Assistants (GTAs), or PhD researchers who teach alongside their doctoral studies, are notably absent from this literature. In this mixed-methods study, we examined predictors of GTA’s wellbeing in UK institutions. This demonstrated that greater perceived agency and lower levels of over-commitment predicted higher wellbeing. Other variables, including social identity, role pride, work/life balance, effort-reward imbalance, and perceived effectiveness of teaching, did not predict wellbeing. We also qualitatively examined the identity management practices of GTAs, using a brief story completion task. A reflexive thematic analysis generated two dominant themes (1), the ‘paradox of credibility’, whereby GTAs reported grappling with a concern to be perceived as credible and worthy of teaching whilst also managing student expectations about their own knowledge, and (2) GTAs’ desire to use their staff-student role to engage in ‘approachability and advocacy’ with and for students. We end with a set of recommendations for Higher Education Institutions that employ GTAs, informed by first-hand accounts of lived experiences.

In pedagogical research, the wellbeing of students and academic staff has been well-explored, and wellbeing is firmly on the research agenda in Higher Education.This has involved a thorough consideration of the factors affecting undergraduate student wellbeing, academic staff wellbeing (e.g.see Guidetti, Viotti, and Converso 2020), and, more recently, postgraduate researcher wellbeing (Moss et al. 2022;Casey et al. 2022).While research in this area has made promising progress, there is a notable and crucial gap in the literature which directly addresses factors contributing to the wellbeing of academics who straddle the border between staff and student roles in Higher Education: Graduate Teaching Assistants (GTAs).GTAs, often referred to as doctoral demonstrators, or Postgraduates who Teach, are PhD researchers who engage in teaching during their doctoral studies.For some, teaching may be in the form of contracted hours of teaching, whereas for others this may be ad-hoc hourly paid teaching provision.GTAs are increasingly resourced to deliver teaching and learning in HE (Winstone and Moore 2017).However, while GTAs typically receive some teaching training, provision of training varies considerably by institution and by subject area (e.g.Nasser-Abu Alhija and Fresko 2020).The role of GTAs have been subject to some pedagogical research, but this has mainly been in relation to GTA's teaching effectiveness (Santhanam and Codner 2012), training needs (Barr and Wright 2019), and career motivations (Nasser-Abu Alhija and Fresko 2020).Importantly, GTAs do not fit neatly into the category of either staff or student; therefore, it is crucial to understand predictors of wellbeing among this unique group in Higher Education.
GTAs have unique experiences because they sit at the intersection of staff/student and, therefore, must navigate these identity tensions in both their doctoral research and their teaching practice (see Raaper 2018).As Winstone and Moore (2017) demonstrated, GTAs must engage in 'identity work' throughout the process of their studies, owing to the transitory space that they occupy in the university.In their qualitative focus group study, Winstone and Moore (2017) describe how GTAs benefit from this identity malleability, by allowing GTAs to occupy multiple spaces as 'student' and 'teacher' within academia.However, because there are inherent conflicts in GTA's identity, this can often create challenge and discomfort, particularly for less established GTA roles, such as hourly paid demonstrators.Importantly, in the UK, the terms used to describe the role of PhD researchers are not consistent across institutions or faculties.For example, some departments adopt the term 'postgraduate researchers' (PGRs), whereas others use 'PhD students', and others refer to 'doctoral trainees'.Similarly, there are inconsistencies across institutions and contexts about whether PhD researchers are treated as staff or student, which highlights how GTAs occupy, as Muzaka (2009) notes, an 'ambiguous niche' (p.1), due to occupying spaces of teacher, researcher, student, and employee simultaneously.
Although there are no published statistics which comment directly on the wellbeing of GTAs, scholars have raised concerns about increasing levels of depression, anxiety, and other wellbeing issues among graduate students more generally (see Casey et al. 2022).Given the complexity and ambiguity surrounding the role of GTAs, we theorise that such wellbeing concerns may be heightened for PhD researchers who also teach alongside their studies.Indeed, recent research has suggested that support, training, and mentorship of GTAs is generally low (Shum, Lau, and Fryer 2021), which can bring about challenges in working conditions for this group of educators (Nasser-Abu Alhija and Fresko 2020).Research has also shown how the contributions of GTAs to teaching provision are often undervalued and experience disproportionately high workloads, limited agency, and excessive responsibility (Park and Ramos 2002).This means that GTAs report a need for clearer guidelines and training, in regards to their pedagogic practice (Green 2010) and professional development (Reeves et al. 2018).Despite the useful evidence which explores the teaching provision and training needs of GTAs, investigations into GTA wellbeing, including a consideration of predictors of wellbeing and identity management practices, are notably lacking.

Predictors of wellbeing
In order to establish which factors contribute to GTA's wellbeing, the present study is informed by the literature which assesses predictors of other groups of teachers', students', and academics' wellbeing.Our variables of interest include perceived agency over teaching, which has been related to educator and student wellbeing in tertiary education (Averill and Major 2020) and perceptions of work/life balance, given how negative work/life balance has been associated with negative wellbeing in academics (Bell et al., 2012).Research also demonstrates that more positive work/life balance can buffer exhaustion in PhD students (McAlpine, Skakni, and Pyhältö 2022).We will also investigate perceived effort-reward imbalances; research demonstrates how high effort, low rewards can lead to strain outcomes amongst academic employees, including psychological and physical symptoms, job satisfaction, and leaving intentions (Kinman 2019).
We also examined factors which could potentially provide a protective, positive influence over GTAs' wellbeing, including the extent to which being a teacher contributes to their overall sense of identity, as this has been thought to positively impact teacher's wellbeing (Skinner, Leavey, and Rothi 2021) and the extent to which pride in one's role as an educator can bolster mental wellbeing (Beltman, Mansfield, and Price 2011).This may include all facets of GTA's social identity, including role identity, professional identity, and identities outside of work.Finally, given (a) how researchers have identified a link between perceived effectiveness of teaching and educator's wellbeing (Mehdinezhad, 2012) and (b) the research which highlights the lack of teaching training that GTAs typically receive (Shum, Lau, and Fryer 2021), we investigated whether perceived effectiveness of one's own teaching can impact GTA's wellbeing.

The present study
In this study, we assessed how the following factors may impact GTA's wellbeing: social identity, work/life balance, perceived agency over teaching, effort-reward imbalance, perceived teaching effectiveness, and role pride.We also investigated how GTAs manage their professional identities in teaching contexts and collected qualitative responses to form practice-based recommendations for employers of GTAs.Overall, this correlational study aimed to investigate predictors of wellbeing in PhD students who engage in teaching work alongside their doctoral studies.For the quantitative items, we hypothesised that higher agency over teaching practice, teaching social identity, selfconcept, and perceived effectiveness of teaching would all positively predict higher scores of selfreported wellbeing.

Participants
Participants were 83 PhD researchers 'who do any undergraduate teaching alongside doctoral studies' in any discipline based in the UK, recruited online via social media.In the UK, a typical doctoral programme is 3-4 years, and can be completed full time or part time.We decided to recruit only UK-based GTAs, in order to be able to directly inform policy and practice in the specific UK context.There was no other exclusion criteria.The mean age of participants was 30 (SD = 7.36, range = 21-53).Participants had an average of 3.61 years teaching experience (SD = 3.01) and spent Note.In terms of the nature of participants' teaching duties, 54% of the sample reported teaching in more than one type of setting, hence the reported percentages exceed 100%.*Participants were asked to self-define their membership of any underrepresented group in academia using a free-text box.The three most frequently reported memberships included: racial minority (16%), LGBTQIA+ (13%), and gender minority (11%).
an average of 6.63 hours per week teaching (SD = 5.2).See Table 1 for the full demographic information.

Procedure
Participants were invited to complete an online study which lasted for 15-20 minutes online.The survey was built and hosted on Qualtrics.Ethical approval for this study was granted by the local School of Psychology Ethics Committee on 9 April 2021 (Reference: PSYC-248).Survey data was collected from April 2021 to June 2021 and aimed to assess participants' perceptions regarding their teaching role and their current mental well-being.

Quantitative survey measures
We used, where possible, existing, validated scales in the literature.However, we also created some scales for the purpose of this study.

Well-being
As the core outcome variable, we measured wellbeing using the Short Warwick-Edinburgh Mental Wellbeing Scale (Tennant et al. 2007), which comprises 7-items that asked participants to describe their experiences over the last two weeks (e.g.'I've been feeling useful') on a 1 (none of the time) to 5 (all of the time) scale.This scale was chosen as is it widely used in pedagogical research (e.g.Slack and Priestley 2023), has robust construct validity and sensitivity (Tennant et al. 2007), and is designed to be appropriate for a wide range of contexts and samples (Stewart-Brown et al. 2011), including PhD researchers (Casey et al. 2022).

Social identity
To measure the extent to which participants' teaching practice formed an integral part of their overall social identity, we used a short-form social identity measure adapted from Luhtanen and Crocker (1992).Participants indicated their agreement with four items (e.g.'Being a good teacher is an important part of my self-image' on a 1 (strongly disagree) to 5 (strongly agree) Likert scale.

Role pride
To measure self-reported pride over GTA's teaching role, we also included one item, 'I am proud of my teaching role'., that participants rated on a 1 (strongly disagree) to 5 (strongly agree) Likert scale.
After reviewing the literature, this item was included in order to capture the potentially protective influence of role pride on wellbeing, an effect that has previously been reported in secondary school educators (Stoloff et al. 2020).

Perceived agency
To measure the extent to which participants reported agency over their teaching role, an adapted version of the Agency of University Students (AUS; Jääskelä 2017) was used.In this measure, participants indicate their agreement with 16 statements on a 1 (strongly disagree) to 5 (strongly agree) Likert scale.Items include agency over workload (e.g.'I feel that the workload demanded of graduate teaching assistants is too excessive'), teaching effectiveness (e.g.'I believe that I can teach the course content effectively') and overall agency over course structure (e.g.'I feel that I can decide how the course content is delivered').This scale originally referred to 'university students' throughout, but we elected to use the phrase 'graduate teaching assistant' throughout the amended version of this questionnaire and explained to participants that while terminology may vary, this includes any PhD researcher who teaches undergraduates alongside their studies.

Work/life balance
To measure participants' perceived work/life balance, we created a work-life balance scale.This was designed on the basis of past literature which has sought to assess academics' beliefs regarding the extent to which they are able to effectively split their time and energy between their work and nonwork activities (e.g.Bell et al., 2012).The scale comprised of four statements measured on a 1 (Strongly disagree) to 5 (Strongly agree) Likert scale, including items such as 'I have a good work/life balance' and 'I am able to enjoy leisure activities outside of my teaching role'.

Effort-reward imbalance and overcommitment
To measure effort-reward imbalance (i.e. the extent to which participants feel that their efforts are suitably rewarded) and overcommitment (i.e. the extent to which participants believed that they often dedicate an excessive amount of time and mental resources to their role), we used the shortform Effort-Reward Imbalance Scale (Siegrist, Li, and Montano 2014).The short-form Effort-Reward Imbalance Scale consists of three subscales: effort, reward, and overcommitment.We omitted one item from the effort subscale, 'I have many interruptions and disturbances while performing my job'., and two items from the reward subscale, 'My job promotion prospects are poor'.and 'My job security is poor' due to a lack of relevance to PhD researchers.We also amended item 3 of the effort subscale by changing 'Over the past few years' to 'Over time' due to the nature of PhD teaching work.Our final amended scale consisted of 13 items measured on a 1 (strongly disagree) to 5 (strongly agree) Likert scale.The overcommitment subscale was used to form the overcommitment variable analysed in the current study.Whereas, scores from both the effort and reward subscales were used to create the effort-reward imbalance variable.Please see the data preparation section for details on how this variable was calculated.

Perceived effectiveness of teaching
To assess participants' self-perceptions of their teaching effectiveness, a perceived teaching effectiveness scale was created based on past literature where an association between low self-efficacy in regards to teaching competence and mental well-being has been reported (Shoji et al. 2016).
Participants were asked to rate the extent to which 'students learn effectively from your teaching' on a 1 (not at all effective) to 5 (entirely effective) scale.

Quantitative data preparation
To prepare the quantitative data for analysis, we first removed three participants for answering 'no' to the item 'do you currently teach?', and their data were not included in data analyses.All scales were averaged to create an index for each of the variables of interest.To form the effort-reward imbalance variable, participants' scores from the effort subscale were divided by their scores from the reward subscale to produce an effort-reward imbalance score.Quantitative data analysis was conducted in MATLAB.

Identity management
In order to explore GTA's identity management in their roles, we created a qualitative measure, inspired by story completion methodologies (Gravett 2019).In a typical story completion task, participants receive a 'stem' of a story and are asked to complete the narrative, which provides the researchers with richer, more nuanced insights compared with traditional free-text boxes.In story completion tasks, participants typically are given a fictional story to complete; however, given our interest in participants' own experiences of their identity management as GTAs, we created a brief story completion stem which involved centring participants' experiences in their own textual responses.In our measure, we provided participants with a story stem, in which they were asked to

GTA support recommendations
As a final, exploratory measure, participants were provided with free-text essay boxes to provide any more information on how universities can better support them in their role as a GTA.

Demographic characteristics and wellbeing
Given that our sample featured some atypical characteristics for a sample of students, such as a high prevalence of first-in-family students (54%), we first checked whether wellbeing scores were impacted by participants' demographic characteristics.To achieve this, three independent samples t-tests, four one-way independent samples ANOVAs and three Pearson's correlations were performed as appropriate.It was revealed that well-being scores did not vary as a function of any of the demographic characteristics listed in Table 1 (all > .05).Therefore, due to the lack of differences between groups, for the remaining analyses, we investigated predictors of wellbeing across the whole sample and did not split the sample by demographic groups.

Identity management
First, responses to the qualitative self-labelling question were analysed using content analysis.The responses were read and categorised by identity label (see Table 2).Interestingly, just under half of participants in the sample chose to introduce themselves as a 'PhD student', thus emphasising to students their role as a fellow learner, rather than as an educator or member of staff.Notably, some students took an opposing approach favouring labels which centred their identity away from their role as a student and affirming their authority as a member of staff (e.g.'teaching assistant/associate').Whereas, other students opted for labels that incorporated both their student and educator identities (e.g.'graduate/  doctoral tutor') or labels where neither identity are highlighted (e.g. using a personal introduction).

Reflexive thematic analysis
We then conducted a reflexive thematic analysis on the qualitative responses (Braun and Clarke 2019).We chose this analysis because we were interested in patterns of meaning and wanted to centre our subjectivity in the analysis, using it as an 'analytic resource' (Braun and Clarke 2021, 3), given that both authors have recent lived experience of a GTAs role in Higher Education.Both authors first read the whole dataset, marking initial codes ideas for subsequent phases of the analysis.We then regrouped to share ideas and generate initial codes.Following this, the second author (MP) coded all of the responses.After responses were coded, initial themes were generated and critically evaluated by both authors.Themes were then refined until both authors agreed on the finalised versions.Importantly, some participants in the sample did not discuss their identity management in any depth, and instead elected to simply reiterate their job title.For example, one participant reported that they refer to themselves as Study Skills Tutor because 'it's my job title".These types of responses were not analysed further, and instead we focused on the responses with richer responses.Two themes were generated from the qualitative data (1): the 'paradox of credibility' and (2) a concern for 'approachability and advocacy'.

Theme 1: paradox of credibility
The first theme that we identified across the qualitative data was the notion of a 'paradox' in how credible GTAs felt in the classroom with students.For example, some participants spoke about their concern for coming across to students as credible, knowledgeable, and 'worthy' of their teaching role.For example, one participant ['PhD student', hourly paid] referred to this as wanting 'to appear as someone who is "qualified" (e.g.know enough about the topic) to teach/facilitate the class in [students'] eyes.' and another reported that they 'try and give some sense of authority, and worthiness of being there'.['PhD student', hourly paid].Some participants in the dataset seemed to be concerned with appearing credible to students, which was in places seen to be at risk due to participants' early career status; that is, some participants felt that students use job titles as a proxy of credibility, authority, and thus suitability to teach, which was problematic for PhD students.For example: one participant noted that they use their job title to 'encourage a little respect in regards to the student/ lecturer relationship' ['Associate Lecturer', hourly paid].This desire to be respected and viewed positively by students was an identity management concern that echoed throughout the data and, at times, was discussed through a lens of wellbeing and identity safety.Another participant elaborated on this: 'I try very hard to avoid terms that may subconsciously belittle my status as a valid teacher (this is an especial issue when you are a young women in a teaching position who looks to be a similar age to the students you are teaching).So I avoid terms like 'Teaching Assistant' or 'Student Assistant' -instead I prefer to use terms that more accurately reflect my role and its level of responsibility.' ['Graduate Teaching Associate/Guest Lecturer', hourly paid] This concern for wanting to accurately and transparently reflect job titles, roles, and levels of experience, particularly to students, was echoed throughout the data.Importantly, some participants wrote about how they use terms such as 'PhD researcher' or 'PhD student' to make it clear to students where they are at with their training and to assure students that they are not fully 'qualified', i.e. they are not lecturers.For example, one participant commented that 'I believe it is important for students to know that I am not a qualified lecturer and that I am still in training'.This notion of being 'not qualified' was interesting and we conceptualised this as speaking to a broader concern that GTAs had with being perceived as credible enough to teach the class (and thus not undermining their role), while also caveating their contributions and minimising expectations of their knowledge, Similarly, for some participants, the word 'student' (i.e.'PhD student') in an introductory greeting was used as a mechanism to manage student's expectations of their pedagogic competence and subjectspecific knowledge too.For example, one participant made this explicit in their response: 'I think establishing my role as a phd student would allow me more room for accidental errors that I might make'.In this sense, participants used the label of PhD student to allow themselves space for mistakes, which was on one hand a benefit, but on the other hand served to create an expectation of mistakes, which again put GTAs sense of credibility at risk with students.This echoes previous literature on GTA's classroom management techniques (e.g.Pytlak and Houser 2014) and demonstrates the complex grappling with identity, particularly in terms of educational credibility, that GTAs must contend with.
In our analysis, we conceptualised these discussions of credibility as occupying a distinct paradox, in that participants were, on one hand, keen to appear credible and 'worthy' to students, but also felt it is important to minimise student's expectations of their knowledge and experience, in order to create space for mistakes.Therefore, in this sense, there is a narrative of establishing their 'worth' but also not over-stating their competencies in a way that may make them vulnerable to students' negative responses.We noted a distinct undertone of under-confidence in GTAs' own teaching ability and knowledge, which prompted many of these credibility issues.For example, one participant reflected that they introduce themselves as a PhD student 'probably subconsciously to manage [student] expectations of my knowledge' .We observed an inherent tension between a concern for credibility, with a concern for staying firmly within the realm of one's capabilities, in terms of subject content and skill.For the latter of these concerns, this was conceptualised as coming from a place of insecurity or anxiety about participant's ability to meet student expectations and a fear of making mistakes of 'getting things wrong' in their teaching practice.
Relatedly, despite this caveating of teaching through expectation management, many participants in our sample also discussed their concern to be 'taken seriously' by students in order to assure students that they have the necessary skills and knowledge to teach them.For example, one participant wrote that 'to my students I might downplay it a bit, to increase their confidence in my ability to teach them' ['PhD researcher', fixed term contract].This also suggests that the GTAs in our sample of participants had an inherent concern with communicating to students that they are able to teach the subject material.This may reflect the recent, perhaps problematic, emphasis on 'student as consumer' in Higher Education, whereby students are positioned to be paying consumers, and academic staff must be attentive to provide 'value for money' and meeting student expectations.The marketisation of Higher Education has been discussed in the unique context of GTAs, identifying how 'student as consumer' discourses may inhibit institutional support for GTAs (Raaper 2018).

Theme 2: approachability and advocacy
The second theme generated by our analysis was the notion that GTAs grappled with their sense of 'approachability and advocacy' in the classroom, which referred to participants who shared an overriding concern to be perceived as welcoming and relatable to students, above and beyond concerns about credibility.In some instances, for example, participants in the data wrote about their concern for students in their classroom to feel relaxed and welcomed, which informed their decision about which self-label to ascribe.For some participants, this concern for making students feeling at ease meant that participants elected to use job titles and phrases that were perceived as being understandable to students: 'Typically I would refer to myself outside of this situation as a post graduate researcher, but I feel that students at an undergraduate level are more familiar with the term PhD student.I think as well referring to myself as a student may make them more relaxed and feel that they are on the same level as me.' ['PhD student', hourly paid] For some participants, this concern for being approachable was related to an implicit desire to be able to advocate for their students in Higher Education.In many cases, this concern was particularly aligned with issues of equality, diversity, and inclusion, and participants were keen to use their platform as a GTA to advocate for students on these kinds of issues.For example, one participant described how they introduce themselves as a disabled person, in order to create an inclusive space that fosters representation in academia: 'I'm very aware of how I introduce myself.I have a passion for connecting with my students as person to person, to show a rounded view of myself including my vulnerabilities, and to show that I see my students as human beings with other interests and priorities and needs in their lives.I also deliberately model being a disabled person in academia, to let my students know that disabled people absolutely belong in and can succeed in the academy.' ['Lecturer studying for a PhD', fixed term contract] Participants were thus careful to select introductory greetings that 'set their stall out' to students and communicated the kind of teacher they hoped to be.This identity management decision was often a thoughtful, complex one, which was informed by the tension of wanting to come across as human, relatable, and welcoming, whilst also wanting to be perceived as a legitimate member of teaching staff.Generally, participants spoke about an overriding concern to be relatable to students and, in some instances, give the impression that they are 'one of them'; for example, one participant wrote: 'I also like my students to know my research area and that I am a student as well, which hopefully makes me more relatable?'['Associate Lecturer', hourly paid] and another noted that they introduce themselves as a PhD student 'to seem more relatable as a student'.Similarly, other participants noted a more general concern for their job title to be suitably 'accessible' and understandable to students that they teach; for example, one participant noted that they introduce themselves as a PhD student 'because it's a succinct summary of what I do and it's accessible to most people'.
For some participants, their choice of job title when introducing themselves to students was more broadly informed by a desire to be approachable in a way that opens up conversations with students about their own interests and positionality.For example, one participant explained that they introduce themselves as PhD researcher because it 'gives me an opportunity to introduce myself and my research.Also initiates the conversation of which areas they find interesting and their aims going forwards'.and another participant reported that they introduce themselves as a PhD student to demonstrate to students that 'we have something in common'.

Five recommendations for improving wellbeing of GTAs
At the end of our survey, we asked participants to share any recommendations for how their universities can better support them in their role as a GTA.80 participants elected to share thoughts in response to this item.A synthesis and content analysis of these textual responses revealed five core recommendations that were echoed throughout the dataset.We collate and describe each of these recommendations below (see Table 3).These recommendations reflect participants' views and are intended for Higher Education institutions who employ GTAs.This includes recommendations for individuals who linemanage or support GTAs, as well as more broader institutional-level recommendations.

Discussion
In this mixed-methods study, we investigated (1) predictors of GTAs' wellbeing (2), GTA's identity management in the classroom, in response to the unique challenges of navigating the dual identities of both doctoral student and teacher (e.g.Nasser-Abu Alhija and Fresko 2020).Similar to previous research which has aimed to investigate staff wellbeing in higher education (Bell et al., 2012;Averill and Major 2020), our findings demonstrate that wellbeing is protected when GTAs are afforded agency over their teaching practice and when perceived commitment to their teaching role is nonexcessive.The recommendations that participants provided for institutions who hire GTA revealed that one way to facilitate better wellbeing is through the creation of an open dialogue in which both staff and GTAs can collectively contribute to decisions regarding the content and structure of the modules on which they teach.Creating a space for these discussions would also provide GTAs with a means to negotiate appropriate pay, training opportunities, and more appropriate, tailored mental health support.
These findings echo the recommendations made by Ryan, Baik, and Larcombe (2022), who suggest that a cultural 'whole-of-university' approach to support wellbeing of research postgraduate students offers one solution to tackling poor wellbeing of this group.However, in order to achieve this kind of holistic approach to tackle wellbeing, there should now be more rigorous and comprehensive research into the unique experiences of GTAs.Indeed, empirical research that centres the experiences of this group is notably lacking from the pedagogical literature currently.This represents a crucial avenue for future research, and we call for researchers to empirically address the utility of the recommendations highlighted by the present study.Further, future work in this area should also investigate how implementation of these recommendations may be achieved in practice.
There are some limitations that must be acknowledged here.For example, even in a UKspecific context, there is large variability in the experiences and management of GTAs.Some institutions, for example, treat GTAs as academic employees and offer rigorous and regular training, whereas others rely heavily on hourly-paid, precariously employed PhD researchers.This means that there may be important contextual information that our study does not represent well, which may inform or influence wellbeing experiences.Similarly, there may have been differences in the experiences that GTAs have before starting their studies; some GTAs in our sample may have come into their role with previous teaching or professional experiences that bolstered their wellbeing and facilitated identity management.Some GTAs may also have diagnosed mental health conditions, which was not captured in our data, largely due to ethical constraints.Also, it is useful to recognise that the anticipated outcomes of a GTA contract may differ considerably across PhD researchers.For example, some may view their role as an important and necessary step in establishing an academic or teaching career in Higher Education whereas, for others, engagement with teaching is motivated purely by practical, financial reasons.Further, some of the themes generated from our analysis may be compounded by demographic variables, including gender and race.For example, the 'paradox of credibility' might be experienced in heightened ways for GTAs who occupy underrepresented positions in Higher Education.Therefore, it is important that research does not treat GTAs as one homogeneous group, as others have advocated for also (e.g.see Winstone and Moore 2017).

Conclusion
This study demonstrated that GTA wellbeing is associated with higher perceived agency over teaching and lower over-commitment to teaching.However, GTA wellbeing was not significantly impacted by work/life balance, perceived effectiveness, or social identity of teaching.Our qualitative results highlight the complex and nuanced identity management processes that GTAs grapple with in their role, as they straddle the border of staff and student.GTAs also provided recommendations, in light of these experiences, for institutions.Hence, important changes must be made to higher education policy in order to ensure the longevity of effective teaching practice in higher education.For example, training and CPD opportunities offered to GTAs should be offered by institutions and must provide sufficient preparation for the GTA's future teaching career.In addition, GTAs need to be given space to negotiate both the size of their workload and the pay that they will receive.Ultimately, GTAs occupy a unique space within Higher Education and it is important that their experiences are not excluded from the wellbeing conversation.
This project contains the following data: • GTA wellbeing raw data file.csv(Contains the raw data obtained from Qualtrics).
• GTA wellbeing processed data.csv(Contains the processed data on which all statistical tests were performed).
• Supplementary Analyses.docx(Contains details of analyses performed to investigate whether wellbeing scores were impacted by participants' demographic characteristics) • Correlations.docx(Contains the results of inter-variable correlations examining the relationships between wellbeing and each variable of interest) Data are available under the terms of the Creative Commons Attribution 4.0 International license (CC-By Attribution 4.0 International).

Figure 1 .
Figure 1.Wellbeing scores plotted against a model of perceived agency, work/life balance, effort-reward imbalance, overcommitment, social identity, pride, and perceived effectiveness of teaching.Note.A figure showing wellbeing scores plotted against a model of perceived agency, work/life balance, effort-reward imbalance, over-commitment, social identity, pride, and perceived effectiveness of teaching.Error bars represent ± 1 confidence interval.
'Imagine that you are entering a classroom and you are teaching a new group of students.What word would you use to fill in the gap in the following sentence?'Participants were presented with a text box that followed the text 'Hello, my name is ______ and I am a . . .'.Following this, participants were prompted to explain their answers using a separate essay box, which was coupled with 'Why do you introduce yourself in this way?And 'Is this something you are consciously aware of in your teaching practice?'.

Table 2 .
Frequency of identity labels.

Table 3 .
Recommendations for GTA support with illustrative quotes from participants.Finally, participants spoke regularly about the desire to be treated like 'academics in training', rather than as resources, or even 'cheap labour', who are not afforded the same rights and privileges as academic staff.Participants reported that departments can support GTAs by inviting them to have a 'seat at the table', e.g. through staff meetings, CPD opportunities, and social events.Note.The notation n indicates the number of participants who endorsed each recommendation.Note, some participants endorsed multiple recommendations.