“A small pawn in a big game”: an exploration of doctoral students’ experience of co-supervisory relationships

ABSTRACT Co-supervision of doctoral theses is on the rise worldwide, but we know little of the relational dynamics of the co-supervisory team and doctoral student. These teams often must work together for several years, during which supervisors may come and go. We interviewed 19 doctoral students about their experiences of co-supervision and found a complex team climate characterised by prior relations, merit- and network-gaining activities, and ‘cushioning’ adopted by supervisors to support the student during stressful periods. Doctoral students viewed themselves as passive and dependent – as pawns in a bigger game – suggesting that co-supervision might uphold the traditionally strong hierarchical structure of academia. We therefore suggest that all co-supervision teams discuss the roles and structure of supervision before commencing a research project.


Introduction
Co-supervision, also referred to as joint, dual or team supervision, has become increasingly common in higher education, and in doctoral supervision in particular (Guerin and Green 2015;Guerin, Green, and Bastalich 2011).This type of supervision, which involves two or more supervisors in a doctoral student's PhD progress (Burgess, Pole, and Hockey 1994), provides numerous benefits for both supervisors and doctoral students such as balancing absenteeism among supervisors, increasing the availability of specialised expertise and engagement in interdisciplinary projects, and increasing the opportunity for students to acquire broader skills and knowledge (e.g.Paul, Olson, and Gul 2014;Pole 1998).However, it also presents highly complex relational dynamics leading to new challenges.How these challenges arise and are addressed remains understudied (Bui 2014;Lahenius and Ikävalko 2014).
A positive student-supervisor relationship is crucial for numerous student-related outcomes, such as students' successful completion of the thesis, productivity and creativity, well-being and satisfaction with the programme, the writing and learning process, and formation of a research identity (e.g.Dericks et al. 2019;Devos et al. 2017;Lonka et al. 2019;Pyhältö, Vekkaila, and Keskinen 2015;Schmidt and Hansson 2018;Van Rooij, Fokkens-Bruinsma, and Jansen 2021).However, previous research has mainly focused on the dyadic relationship between one student and one supervisor, largely ignoring the complexity and challenges of the relationships involved in co-supervision.To shed light on this, situations and chose their words more carefully (Manathunga 2012).Some supervisors have expressed that co-supervision can increase their workload because of problems related to timing and communication, or because they experience personality clashes and conflicts arising from their differing research interests or professional competition (Manathunga 2012;Olmos-López and Sunderland 2013).Other problems include uncertainty among the supervisors concerning who takes overall responsibility for the project (Watts 2010) or the demand for the main supervisor to possess a certain level of experience and confidence, the lack of which can lead them to focus on self-preservation rather than the student's needs (Manathunga 2012).
Most previous studies on supervision have examined supervisors' views of the research process and relationship to the doctoral student; few studies have examined co-supervision from doctoral students' perspectives (Guerin and Green 2015;Guerin, Green, and Bastalich 2011;Lahenius and Ikävalko 2014).One challenge related to co-supervision, according to doctoral students, is how the diversity in supervisor feedback can result in ambivalence towards the project, difficulties in attaining autonomy, delays in their work schedule, or tension between the doctoral student and supervisory team (Guerin and Green 2015).In addition, students noted that diverging opinions were often not discussed among the team, causing questions to go unanswered (Guerin and Green 2015).Students also tend to see the main supervisor as the senior or final decision maker (Guerin, Green, and Bastalich 2011), which partially aligns with supervisors' views (Johansen et al. 2019).However, some students reported experiencing less hierarchical and more collaborative styles of co-supervision characterised by shared decision making (Guerin, Green, and Bastalich 2011).
Instead of seeing doctoral students as passive observers in co-supervision, Guerin, Green, and Bastalich (2011) stressed that students can be instrumental in resolving conflicts between supervisors and keeping everyone's best interests in mind.They saw the student 'as active coordinator and manager of supervision resources' (Guerin, Green, and Bastalich 2011, 152) using the 'big love' metaphor, which represents how students can manage multiple relationships skilfully.However, in a recent study by Riva, Gracia, and Limb (2022), the majority of doctoral students (52%) reported lacking guidance, advice and support and suffering from imposter syndrome (55%); many also seriously considered leaving their PhD studies (24%).
In sum, although co-supervision is premised on the notion that shared responsibility among the supervisors will benefit doctoral students by balancing absenteeism, adding further expertise, and promoting the development of the students (Spooner-Lane et al. 2007), the complexity of the relational dynamics presents new challenges, especially with consideration to the lengthy process of finishing a PhD program.Although co-supervision is generally seen as a strength by supervisors, many of them still report problems with this style of supervision, including confusion concerning roles, workload, and expectations and the overall lack of formal training.Understanding how doctoral students experience supervisory relationships in co-supervision is essential because a positive supervisory experience is the largest predictor of students' satisfaction, performance, wellbeing and learning (Dericks, Thompson, and Roberts et al. 2019;Devos et al. 2017;Lonka et al. 2019;Pyhältö, Vekkaila, and Keskinen 2015;Schmidt and Hansson 2018;Van Rooij, Fokkens-Bruinsma, and Jansen 2021).Nevertheless, it remains an understudied area of research.Many supervisors believe that supervision should be tailored to the doctoral student's needs on an individual level (Riva, Gracia, and Limb 2022).To help supervisors adjust their supervisory style in co-supervision, it is crucial to better understand doctoral students' experiences.Thus, the aim of this study is to understand doctoral students' experiences of co-supervisory relationships and their impacts on students.

Method
This study utilised an exploratory design to examine the subjective experiences of doctoral students during supervisory change, to illuminate their experiences of the relational dynamics of the supervisory team.

Participants and data collection
A purposive sampling strategy was combined with the snowball method to recruit participants.Three initial doctoral students -who all agreed to participate -who had finalised their PhD studies were purposely selected from different universities, gender, and subject areas.After each interview, the participants were asked to recommend other doctoral students they knew of, who then were contacted by the authors by email.In total, 26 doctoral students were contacted, of which 19 (15 women) agreed to participate in the study.Twelve participants were in the final stages of their doctoral studies, while the remaining seven were in different stages of their doctoral programmes.Participants ranged in age from 31 to 58 years (M age = 41.1).Participants represented 11 different subjects in their theses, such as business administration, health sciences, psychology, and informatics, thus ensuring a diverse sample.The majority of the students had three or four supervisors during all of their PhD-studies, five students started off with two supervisors, but the number changed during their studies, as all of the participants went through at least one supervisory change.In Sweden, most institutions require a minimum of one main and one co-supervisor per PhD-student.The supervisors are appointed by the dean of research, or the dean of the faculty preceded by a discussion among the senior members of the faculty.
All data were collected at five universities in southern Sweden between May and October 2020 through face-to-face interviews, out of which seven were conducted over the digital communication tool Zoom due to the COVID-19 pandemic.A semi-structured interview guide was used (Brinkmann and Kvale 2014).The interview guide included open-ended and probing questions such as 'Can you describe your relationship with your supervisor(s)', 'How did/do you experience the relationship with your supervisor(s) and the one between your supervisors', or 'How did the relationships with your supervisor(s) impact on you (your work)?'The interviews, all of which were audio-recorded and transcribed verbatim, ranged from 36 to 105 minutes (M time = 57 minutes).Sixteen interviews were conducted by the second author and the remaining three by the first author.Five were conducted in English and fourteen in Swedish, as preferred by the doctoral students.
The study was conducted in accordance with Swedish law concerning research involving human subjects (SFS 2003).All participants gave informed consent before the interviews and were informed both in writing and orally that their participation was voluntary, that they had the right to withdraw at any time without providing a reason, how data were stored, that analysis would be performed at a group level, and that their confidentiality would be secured.

Analysis
A thematic analysis (Braun and Clarke 2006), a method for identifying and interpreting patterns in qualitative data, was conducted.The analysis was carried out in the following steps: 1) familiarisation with the data, 2) generation of initial codes, 3) search for themes, 4) review of themes, 5) definition and naming of themes, and 6) presentation of the results.We began by reading and re-reading the transcribed interviews to familiarise ourselves with the data.Subsequently, data related to the aim of the study was extracted and codes were systematically identified to organise the data into meaningful thematic groups.These themes were then reviewed through assessment of their internal homogeneity and external heterogeneity.At this stage, the authors engaged in joint discussion until a preliminary thematic structure was agreed on.The discussion process was circular, moving from considering the individual themes to considering the entire data set.During this process, new subthemes were formed, themes were renamed, and the structure was reconsidered based on consensus between the authors.Finally, quotations were chosen to illustrate the findings.
The analysis process was carried out by both authors, who coded separately during stages one, two, and three and thereafter worked together until a consensus was reached.This procedure enhanced the trustworthiness of the analysis.Further, because thematic analysis is a reflexive process, subjectivity and pre-understanding is considered a strength rather than a weakness (Braun and Clarke, 2019).Themes were not identified but generated and constructed with help of the data and the authors' analytic resources and skills, alongside reflective and thoughtful engagement.Given that all the interviewed students had experienced at least one supervisory change during their PhD-studies, we aimed to be very mindful of the potential complexity of the context where their experiences were gained.As both authors held PhDs, they were familiar with the PhD supervision process and have extensive experience with qualitative studies (MS) and behavioural science (EH).

Results
Doctoral students' experiences of co-supervisory relationships and their impacts were structured into four distinct themes: a complex and dynamic team climate, cushioning, prior relations -the catch 22 of supervision, and network and merit gaining activities.

A complex and dynamic team climate
The participants experienced the co-supervisory relationships as involving complex and dynamic interactions between supervisors, which impacted them in different ways.The students reported many examples of a negative team climate during different phases of their PhD studies.This negative climate could be created by just one of the supervisors and included behaviours such as rudeness or ignorance of one's responsibilities, as expressed by participant 9: 'The relationship with him [the co-supervisor] is not bad only with me . . .how he treats people, it's so condescending.I feel more tense, how one talks during the meetings, to me . . .it's hard for me to talk about . . .it's the atmosphere during supervision . . . it should be in a good spirit. . . .But it's nothing we discuss openly.At all. . . .It takes so much energy.It's not reasonable'.Some students experienced well-functioning co-supervisory relationships, which were described as positive and respectful -allowing students as well as supervisors to speak freely.In those constellations, the doctoral student felt that they were a natural part of the team and valued the relationships, as participant 6 experienced: 'There was an openness to new ideas.[. ..]It never happened that they kind of spoke ill of each other or ran each other down [. ..]There was such a positive spirit.They worked on other projects together as well.The relationship seemed very important.They went to conferences together a lot and I joined them.So, it became sort of a very social thing as well'.
However, these positive relationships often focused on one or two supervisors and rarely on the whole team.When the students experienced the team climate as positive, differences between the supervisors in terms of knowledge of methods or subjects were considered an advantage, as the supervisors were seen as complementing each other.
Doctoral students also found that certain supervisors were considered 'troublemakers' within the institution -they were described by the students as showing eccentric behaviours, attitudes, or personalities, or having a bad reputation in previous supervisory or teaching contexts, which negatively impacted the supervisory team climate.Sometimes, students felt that the other supervisors were uncomfortable with discussing supervisory issues openly, causing them to ignore or avoid the issue entirely, trapping the doctoral student in an unhelpful status quo.Participant 17 explains: 'It's maybe been a little bit naïve from my side as well because I know that the main supervisor had some kind of issues in the organisation in general, and I know that the dean talked to him a few times as well about his behaviour to other people'.
This lack of disciplinary action and acceptance of such behaviours or attitudes of supervisors being described as lacking social skills, being 'big egos and strange persons' or 'eccentric professors', puzzled the doctoral students, causing them to question the world of academia, as they believed that such behaviours would have not been tolerated in other workplaces.
Hierarchies among the supervisors were a major theme in the interviews, experienced by the students as characterising many of the co-supervisory relationships as well as their relationships with the supervisors.The students experienced that everyone, themselves included, agreed on the unspoken supremacy of the main supervisor.Participant 15 explains: The main supervisor made himself heard.So, indirectly, it becomes a bit like his opinions counted most [. ..] my experience is that the main supervisor seemed to have more power.
Furthermore, the students often reported experiencing hierarchical structures, henpecking ruler techniques and power based on the supervisors' positions in academia.The students' dependency on the supervisor was also mentioned in most interviews.
Communication within the team was a particular concern for doctoral students and influenced by the team climate.The lack of mutuality could result in doctoral students not communicating openly within the supervisory team or feelings of insecurity, as in the situation described by participant 14: I could come to supervision meetings . . .and half an hour went by while they [supervisors] talked about their travels and family concerns and stuff.And I felt like "No, I don't want to argue, it has to be like this".Today, I would have said "Listen, come on.Now we have to focus".Back then I didn't dare, I often felt pretty diminished and insecure . . .Finally, doctoral students were aware of which supervisors were more inexperienced, perceiving them as hesitant, passive, or not contributing.In several cases, these supervisors communicated their superfluous role in the team and their insecurities to the students, which created a climate of uncertainty.

Cushioning
Doctoral students reported that some supervisors seemed to want to balance out unhelpful supervisors or their unhelpful attitudes by directly or indirectly supporting the doctoral students in their work and avoiding supervisory changes.In this way, such 'helpful' supervisors served as a buffer for ineffective supervisors.These helpful supervisors adopted a solution-focused orientation and leveraged strategic choices, which students appreciated.Although a well-functioning supervisory team was desired, sometimes just one supervisor was needed to make a positive difference in the doctoral student's progress.Students formed an alliance with this supervisor (who was not necessarily the main supervisor), who became more active behind the scenes, unbeknown to or in silent agreement with the main supervisor.This way, the supervisor(s) who was considered problematic, was mentally 'bracketed' to avoid disturbing the students' workflow.Several participants agreed on such a 'special arrangement' to ensure functioning supervision, such as participant 5: "So I told [my co-supervisor], "My condition is that I am only working with you.If you want to keep him there because he is your friend or whatever reason, I don't feel that way.I don't like him.I don't want his comments".So [my co-supervisor] was like, "Let him be there.Don't throw him out."Then I said, "Ok.I need to feel that he does not exist.Everything is going to go through you".
In some cases, when doctoral students were emotionally drained because of what they thought of as a destructive supervisor's comments, attitude or actions, the helpful supervisor would actively reassure or validate the doctoral student in private: "Without her [co-supervisor] I think I actually would have quit.She was very supportive, and she knows how she [main supervisor] works.She supported me, like "Don't take it personally, this is just her way of exercising power and to handle her complexes".She was a big help".(Participant 10) The helpful supervisor could also be experienced as a supportive mediator, balancing the negative input from another supervisor smoothly during meetings and helping the student navigate and move forward rather than becoming mired in conflict, as was expressed by Participant 19: "And this co-supervisor, he acted as a [. ..] a control rod in a nuclear power plant; he [. ..] sort of calmed down this main supervisor and was able to structure him".
Rarely did a supervisor address a problem in the team like for example lack of supervision or inappropriate behaviour.Only one student reported that the supervisory team acted when a main supervisor moved away and no longer fulfilled the supervisory expectations: "We talked about it in the supervisory group . . . .One of the co-supervisors thought it was a supervision issue.I thought it felt good because they [the supervisory team] thought it was important, "We will take care of it . . .and then we will see how it goes.Because it is important that it works for you." . . .We had a dialogue all this time.And one of them, this very experienced co-supervisor, stepped up and took over a lot of this main supervisor's work.So everything that he didn't answer, I sent on to [the co-supervisor] and he tried to solve it".(Participant 8)

Prior relations -the catch 22 of supervision?
The PhD-students did not only experience dependency towards their supervisors, but also between supervisors, implying that some supervisors directly exerted or was indirectly given more authority within the team.Power imbalances were particularly evident because of past relationships or planned career advancements.Some supervisors were not seen as helpful by the students who instead saw them as being more concerned with agreeing with the dominant supervisor or being silent and/or passive, as experienced by participant 6: "But then I had two co-supervisors as well, and one of them was kept on a tight rein by one of the main supervisors.And it did not go well, because she never stood up for me, or helped me, or noticed that I needed something; instead, she complied with the main supervisor all the time. . . .they [co-supervisors] were also very quiet when the main supervisor was around.
Many of the participating doctoral students had supervisors that had previously been in a studentsupervisor relationship with each other.This could add to the complexity of the relational dynamics of the team, as participant 14 describes: "Two of them [co-supervisors] were previous doctoral students of him [main supervisor].That's why that dynamic was still there.They didn't dare to contradict him".
When the students were asked if it was an advantage or disadvantage if the supervisors knew each other prior to entering the supervisory team, they often preferred that they did not know each other expecting this would result in friendly behaviour and no hidden agendas.Even when supervisors had good prior relations it could impact the doctoral students in a negative way, as explained by participant 5: " . . .I knew that they [the supervisors] had an extremely good relationship.Isn't he like a godfather to his children or something like that?I think it's bad.There comes a time when you'll have a conflict.And then the two are going to back each other up.And that's not good.You always need to have another one you can talk to, or at least who has an objective opinion.The two knowing each other has only been bad".
Previous relationships between the supervisors could also result in unresolved issues.Those constellations were often not chosen by the people involved, but rather the result of supervisors being selected at an institutional level.Such conflicts detracted from the supervision and drained the students' energy, as in participant 1ʹs example: "Their [main and co-supervisors'] relationship was always complex . . .every time the three of us met, it was much more about their complex relationship, instead of . . .instead of seeing me and what I needed. . . .When we had meetings they couldn't keep it together, but I stayed out of their discussions".
Sometimes power struggles emerged among the supervisors in relation to the subject of the thesis or to the universities to which the supervisors belonged.Doctoral students felt that these struggles distracted from their thesis and further described that some supervisors marked their power over the doctoral students as well as the other supervisors in such a way that even when replaced (e.g.due to retiring) they remained in a power position and could at times trump the new main supervisor.
The doctoral students also saw power issues as related to the older generation of supervisors experiencing less power abuse from younger supervisors Participant 10: "The younger supervisors are more manageable.I hope . . .that there will be a new generation in academia now, that will allow these power abuses to slowly fade away".
Furthermore, the doctoral students expressed a supposition that supervisors also could exert power by not allowing another supervisor into the team with obscure reasons for not to.This presented doctoral students with a dilemma, as they would miss the opportunity to obtain important knowledge from that additional supervisor.Participant 5 recalled: "They gave me an ultimatum, and they said that "[students' name], we are not accepting this.Either you have to choose to stay only with us.We are not getting him in.Or you have to decide to leave us altogether".
Based on the students' experiences, power within the team could also be exerted by choosing the language that supervision was held in (i.e.Swedish or English), which might exclude supervisors who did not speak the preferred language from actively participating in supervision meetings; slowing down the process of providing comments; declaring the inferiority of certain subject areas or affiliations; withholding information; or, as in the following example with participant 12, not deciding on the author order of a manuscript: "I think I mentioned something about that, or [my co-supervisor] did, "Should we have this author order despite the fact that he [the main supervisor] has not been involved for a year?".And then [my co-supervisor] said she would take care of it.But, you know, it didn't work.It was not the right time and stuff.At the end of a supervision meeting, she brought this up again anyway; "Is now the right time to talk about this?"."No.This is not the right time" he said, got up and walked out of the room, closing the door behind him.And we're like "What just happened?"It doesn't matter to me who is the last author.They can manage that between themselves".(participant 12)

Merit or network gaining activities
Doctoral students experienced co-supervisory relationships primarily as a form of networking, which they saw as intrinsic to academia.Supervisors often knew each other prior to joining the co-supervision and in many cases the doctoral student was supervised by someone who had also supervised one of the student's supervisors.This procedure implied that, in numerous cases, the choice of supervisors was made not by the students but by the supervisors based on their existing networks.The doctoral students saw these collaborations as based on already established roles; in other words, the supervisors were not necessarily chosen based on competence or skills but rather on networking or merit gaining: "I think it mostly was the main supervisor who chose her [co-supervisor].They were friends and I think it was a bit "you scratch my back, I'll scratch yours".(Participant 4) As a result, many students perceived the relationship between supervisors as being built on mutual benefit or 'back scratching'.At times, students perceived a lack of objectivity or interest in the dissertation among supervisors or that some supervisors seemed to have their own agendas.Notably, these experiences of 'network supervision' did not prevent some doctoral students from entering similar constellations when they had completed their PhD and begun to supervise.Their own prior relationships also worked in favour for the doctoral students -several students who had an opportunity to choose a supervisor picked their former master's thesis supervisor, with whom they had successfully collaborated in the past.
Another aspect of this theme concerned the privileging of junior supervisors' career advancement over that of doctoral students.Participant 17 recalled: "I felt quite clearly that he [a co-supervisor] had zero interest in me.He had interest to get contact . . . to do his research and to get a connection to the project leader [another co-supervisor]".(Participant 17) Students felt that the PhD supervision was an important prerequisite for academic career advancement for junior supervisors, as Participant 18 noted: "But the person also needed to supervise a doctoral student to receive tenure".
Such merit gaining was often reinforced by internal swops between main and second supervisors or by assigning supervisors external to the co-supervisory team.If internal swops occurred in agreement with the doctoral student and towards the end of their PhD studies, students were not too bothered; they did not feel that they were affected by this arrangement and saw this kind of merit gaining as a natural part of academia.
"They [supervisors] were best buddies, he [the co-supervisor] was thought of as a "sleeping supervisor", he needed legitimacy at [the institution] and engage in research activity.It was a win-win situation".(Participant 1) However, in several cases, a supervisor chosen based on merit gaining could impact the doctoral students' progress since their competence, knowledge or skills were not ultimate for the dissertation.Participant 9 described: "The one who became the main supervisor . . ..it wasn't the one who was most familiar with the project, but it would be someone who would somehow be rewarded for having a PhD student . . .who was taken to be meritorious".
These arrangements influenced some doctoral students' experiences of co-supervisory relationships, as they were seen not to be in the students' best interest.Participant 5 felt that the focus was not on him/her: "Yeah, but the thing is that the agreement between [the co-supervisor] and [main supervisor] was that [the cosupervisor] was supposed to act as my main supervisor and the other was supposed to be there in the background, coaching.All contact should be with [the co-supervisor], and the main supervisor should just be there basically coaching [the co-supervisor], not coaching me".Some students felt that these arrangements led to a lack of clarity in who had the main responsibility in the supervisory team, or who was responsible for establishing an overarching structure and workflow in the project.Many students experienced this incongruence in interests or goals problematic since they felt that their needs were being neglected in favour of supervisors' merit gaining.

Discussion
Co-supervision of doctoral students appears to be challenging, and the impact of this increased complexity must be considered in terms of the successful completion of the thesis and productivity of the doctoral students, who represent the future of academia.
The complex and dynamic team climate can be viewed from different perspectives.A dynamic climate can be fruitful if the discussion is open and the responsibilities clear (Pole 1998).However, the narratives of several doctoral students tell a different story: they saw supervisors 'thinking alike' as helpful, rather than the multitude of expertise suggested by Pole (1998).Some students even spoke of 'superfluous supervisors', which opens the question of how many supervisors are optimal in a team (Pole 1998).There is most likely no 'golden standard' for the optimal number of supervisors; however, unneeded supervisors are hardly optimal for successful thesis discussions.Olmos-López and Sunderland (2017) found that supervisors might not speak their minds as a way of showing respect towards other supervisors in the team, which makes their supervision redundant.While this is not the same as superfluous supervision, its effect on the thesis might be the same.However, from supervisors' perspective, being involved 'in training' while also avoiding responsibility might allow them a relatively low-stakes opportunity to understand the procedures, dynamics, and expectations of supervision.As formal education for supervisors is lacking as a formal requirement for academic positions, and PhD supervision courses often are either scarce or fully booked, many supervisors lack formal training (e.g.Riva, Gracia, and Limb 2022).By designing useful supervision courses and making them readily available to supervisors, the role of supervisors 'in training' might be discussed more in depth.These courses may benefit doctoral students as well by making supervisors more likely to vocally contribute and engage with their work, which could in turn improve team climate.Indeed, while a positive team climate had a beneficial impact on the students' productivity, a negative team climate made them feel out of focus and disempowered.
Several students described their co-supervisory relationships as difficult and dynamic in a negative way; they felt that they were often 'pawns in a bigger game' since the entry of new supervisors or internal swops were perceived to be based on others' needs as opposed to those of the student.The needs often involved back-scratching and relations prior to the project at hand; the catch 22 of supervision.The choice of supervisors was sometimes perceived by the doctoral students as having negative consequences for the team dynamics.Among other things, the results suggested that the main supervisor at times used his or her position to control the team (cf.Guerin, Green, and Bastalich 2011).Sometimes, no action was taken to manage this, even when the team was malfunctioning.The lack of action might have many causes, but Johansen et al. (2019) suggested it is the main supervisor's responsibility to address the problem(s).In this study, only a few students reported attempts by the main supervisor to repair a malfunctioning team.Instead, most students reported having to solve the conflicts themselves or were helped to endure through the support of one of the supervisors' cushioning.
Some doctoral students reported power struggles among their supervisors, of which they were not a part.Although co-supervision implies shared responsibility among all supervisors (Paul, Olson, and Gul 2014) in varying ratios and time periods, some students noted a clear dominance of the main supervisor.This perception was evidently shared by the co-supervisors, which other studies have also reported (e.g.Guerin and Green 2015;Johansen et al. 2019).Students felt that there was poor communication among the co-supervisory team concerning the division of responsibilities or role functions, and therefore they did not know what to expect from whom.This lack of clarity regarding the division of responsibilities was also reported in Johansen et al. (2019), who stressed the importance of distinct co-supervisory roles, especially when the composition of the team changes (e.g.supervisors are internally swopped).From students' perspective, ensuring a clear distribution of roles among the co-supervisory team was considered key for successful co-supervision (Paul, Olson, and Gul 2014), which was also observed in the present study.
Cushioning behaviours could be explained as individual supervisors' attempts at problem solving.As doctoral students often experience conflicts between supervisors as turbulent events (Cornwallet al. 2019), cushioning behaviours could help ensure forward movement: The 'cushion supervisor' could adopt a progress approach and focus on helping the student to finish their thesis.In this study, several doctoral students reported supervisors as engaging in balancing activities within co-supervision, which aimed to help the students in direct or indirect ways or support them emotionally.
The notion that novice (junior) supervisors learn from more experienced (senior) supervisors is not new (e.g.Paul, Olson, and Gul 2014;Watts 2010).It is based on the pedagogical model of learning from more experienced people, also known as the coaching model (Watts 2010).However, doctoral students questioned this 'training', finding that the focus of the mentoring supervisor was on the mentored supervisor, rather than on the person writing the thesis.In other words, the perception of many doctoral students in this study was that the co-supervision was mainly based on merit or network gaining activities -a team for the supervisors, so to speak, as opposed to the students.Additionally, relevant ideas and thoughts were sometimes not communicated between the supervisors or with the doctoral student.Doctoral students in other studies have made similar observations, reflecting an awareness of how supervision helps supervisors with career advancement, which makes the doctoral students feel like tools for the supervisors' own research productivity (Hazellet al. 2020;Olmos-López and Sunderland 2013).The 'structure of expertise' advocated for by Pole (1998), among others, might therefore not be a principal goal in putting together a cosupervisory team as the team does not end up being for the doctoral student's benefit.Despite this, we found that having the experience of back-scratching supervisors did not prevent some former doctoral students from entering similar constellations (i.e.co-supervising doctoral students with their former supervisor for mutual gain), again, the catch 22 of supervision.

Limitations
The study is not without limitations.First, the sample size can be considered relatively small, with only 19 doctoral students, which threatens the study's credibility.Further, sufficient sample variation may not have been fully achieved in terms of gender since most of the participants were women.Although many different subject areas were included, men were more difficult to recruit.This may be due to our use of the snowball method, whereby participants often recommend people similar to themselves.Of the seven doctoral students that were contacted but declined or did not respond, five were men.Thus, gender could be an interesting topic to explore as several studies focus on female doctoral students (e.g. Brown and Watson 2010;Schmidt and Umans 2014).Furthermore, most supervisors are male, and gender is an important selection criterion for the doctoral students (Robati, Bagheri, and Tonkaboni 2015).Previous studies have challenged the feminist ideal of female supervisors being more sympathetic and patient (Brown and Watson 2010) and the dynamics of the gender constellations in the supervisory teams is an interesting venue to explore further.
Second, the transferability of the results may also be limited as the study was performed only in a Swedish context (although several universities were included).PhD supervision and PhD studies differ in various countries, cultures, and contexts.Comparative studies could be one way forward to explore whether the experiences of the students vary.
Third, the study sample comprised doctoral students who had made at least one supervisory change.We chose this inclusion criterion because it secured that the doctoral student had sufficient experience with different supervisor constellations and a variety of co-supervisory relationships, but it may present biases because the participants had to provide an account of past experiences related to supervisory change.However, studies of supervision experiences report supervisory changes as a naturally occurring component of the supervision process (Lahenius and Ikävalko 2014;Wisker and Robinson 2013).Finally, while 12 participants had finalised their studies, others were still undergoing supervision.We might therefore have missed more experiences that were not voiced in our interviews due to matters of dependency.It would be of interest to include students who have not completed their PhD studies or who have left academia after completion to gain an even more honest picture of their experiences.It would also be of interest to explore the experiences of the supervisors, especially regarding the relationships within the supervisory teams.

Conclusions
The view of doctoral students as proactive and skilfully managing the 'polygamous marriage' of co-supervision (Guerin, Green, and Bastalich 2011) could not fully be confirmed in this study.Instead, doctoral students often viewed themselves as passive and dependent, as pawns in a bigger game.These findings are in line with Guerin and Green's (2015) later work describing supervisors as 'superior' to students.Hence, co-supervision may reflect hierarchical structures that are deeply rooted within academia and instil feelings of dependency in students.How cosupervision is experienced by the doctoral students is essential for students' productivity and the successful completion of their thesis.Negative experiences of co-supervision may lead to supervisory changes and doctoral students feeling passive, whereas positive experiences might make the students feel on equal terms and part of the team.It may take several generational shifts before existing power structures and hierarchies, which are visible throughout all our themes, are levelled out into a more horizontal mode of working.This is supported by Robertson (2017), who discuss different forms of power and the importance of the main supervisor to exert power effectively and productively.One way forward could be -as suggested by Riva, Gracia, and Limb (2022), -a co-production approach to supervision between doctoral students and supervisors.Such an approach might help increase awareness of each other's perspectives and improve supervisory relationships within the team by enhancing a collaborative spirit.How to establish a 'good match' between a doctoral student and supervisor is often discussed (Gube et al. 2017;Riva, Gracia, and Limb 2022;Van Rooij, Fokkens-Bruinsma, and Jansen 2021), but might need to be extended to the entire supervisory team to benefit the doctoral student.