For most of human history, learning a demanding craft meant putting one novice under one master, for years, with the novice’s entire future in that one person’s hands. It is the oldest high-stakes relationship in working life, and it is the very structure of a PhD: one expert, one dependent newcomer, a body of tacit knowledge that passes only through close contact, and a power asymmetry total enough that the master’s signature decides whether the years counted for anything. The trades noticed the danger in this arrangement a very long time ago. A gifted carpenter is not automatically a good teacher, and a bad master could waste, or break, a young person’s decade. So the crafts did something the modern doctorate, remarkably, did not. They engineered the relationship, instead of leaving it to the master’s temperament.
A traditional apprenticeship runs on written articles of agreement that state what each side owes the other, a logbook that records what was actually taught and done, defined milestones that mark progress independently of the master’s mood, and, crucially, an external assessor, a guild, a board, a journeyman’s examination, so that no apprentice is ever wholly captive to one person’s opinion of them. Strip all of that out and keep only the power asymmetry, and you have the modern PhD supervision relationship: the oldest apprenticeship structure in the world, running with almost none of the safeguards the trades built to make it survivable. This article is about putting the safeguards back, from below, by the one person with the strongest incentive to.
Table of Contents
- Simon, whose PhD is really about one person
- The articles of agreement nobody wrote: five mismatches
- Keeping the logbook your supervisor never will
- The conversation Simon kept not having
- When it is no longer a logbook problem
- A word for the graduate school that owns the vacuum
Simon, whose PhD is really about one person
Simon is in the fourth year of his PhD in materials science, and if you ask him how it is going, he will talk about his supervisor. Not his experiments, which are progressing, and not his funding, which is secure. His supervisor, who takes three weeks to answer an email, returns manuscripts with the single comment “needs work”, and in their last meeting proposed a new research direction as if the previous eighteen months had been a warm-up. Simon has read the productivity blogs. He has a reference manager and a colour-coded calendar. None of it touches the actual constraint on his doctorate, which is not a tooling problem. It is a master-apprentice relationship with an unmanaged power asymmetry at the centre of his working life, and no logbook.
Simon’s instinct that this is what matters is correct. I keep returning to the numbers on this because they deserve the attention. Levecque and colleagues [1], studying more than 3,600 PhD students, found supervision climate and leadership style among the strongest predictors of doctoral mental health: an engaged, inspirational style predicted better outcomes; a hands-off, laissez-faire style predicted worse. Cornér, Löfström and Pyhältö [2] found the same from the other direction: how doctoral students perceived their supervision related directly to their burnout. And Nature’s global PhD survey [3] found satisfaction with the supervisory relationship tracking overall PhD satisfaction more closely than almost anything else students report. For most doctoral researchers, this relationship is not one variable among many. It is the working environment itself.
And almost nobody is trained for it, on either side. Supervisors are hired for their research, not their teaching, just as a master craftsman is chosen for his craft and not his pedagogy, and most have never had a day’s instruction in giving feedback, setting expectations, or developing a person. The student, meanwhile, is told to “manage upward” as though the phrase came with a manual. Two people acting in good faith, a structural vacuum between them where the articles of agreement should be, and a student who concludes the problem is personal. It usually is not. A difficult supervisor relationship is rarely a character problem. It is usually a structure problem wearing a personality.
The articles of agreement nobody wrote: five mismatches
The most common mistake is treating “my supervisor is difficult” as one problem. It is at least five different problems, each of them a clause the apprenticeship agreement would have covered and the PhD left blank, and they have five different fixes. Diagnose before you strategise. Which of these is actually yours?
An availability mismatch. Your supervisor is competent and benign but chronically absent: slow replies, cancelled meetings, no time. The most common pattern, and the most fixable, because what is missing is structure, not goodwill.
A feedback mismatch. You get responses, but they do not help: vague (“needs work”), purely negative, or arriving too late to act on. This is a specification problem, and most students send vague requests, which reliably produce vague feedback.
A communication mismatch. You do send updates, but they read like a shopping list: ten items, no shape, no request. Put yourself in the world of a PI who is toggling between five projects, two grant proposals and a paper revision. A list raises no interest and asks for nothing, so it gets nothing. The fix is to tell the story instead: where the project stands, what moved, what is blocked, and the one clear request you need answered. Interest follows shape, and decisions follow requests.
An expectations mismatch. You and your supervisor are running different mental models of the project, its scope, timeline, authorship, and what “done” means. Nobody wrote the model down, so each of you discovers the other’s version only on collision.
An interests mismatch. Your supervisor’s incentives and your development really do diverge: they need output for a grant, you need first-author papers and skills for your next position. This one is real and structural, and pretending it away helps nobody. It is managed by making your interests visible and enlisting other mentors, not by hoping.
Notice what is not on the list: hostility, humiliation, credit theft, discrimination. Those are not mismatches, and the penultimate section treats them separately, because everything between here and there assumes a flawed relationship between people of reasonable goodwill. Structure fixes mismatches. It does not fix mistreatment.
Keeping the logbook your supervisor never will
Here is the reframe that changes what is possible: you cannot change your supervisor’s personality, but most of what frustrates you is not personality. It is a missing logbook, and a logbook can be kept from below. Supervisors almost never resist structure that costs them nothing and saves them time. Five pieces do most of the work.
An agenda you send the day before. One page, three sections: what I did, what I am deciding, what I need from you. Send it 24 hours ahead. It converts your meeting from a status ramble into a decision session, and it quietly trains a disorganised supervisor to prepare, because arriving unprepared to a prepared agenda is visibly awkward.
A written summary the same day. Five lines after every meeting: what we agreed, what I will do by when, what I understood you would do. This is not bureaucracy. It is the cheapest insurance in academia, the logbook entry that surfaces an expectations mismatch within a day instead of within a year, and the record you will be very glad of when a research direction “changes” and two memories disagree.
Feedback requests specific enough to answer. Replace “any comments welcome” with “I need a decision on whether the argument in section 2 holds; the analysis in section 4 is stable, please skip it; I need this by the 15th to keep the submission date.” Vague requests hand the job of prioritising to the busiest person in the relationship, so they defer it. Specific requests do that work for them, and specific requests get answered first.
Questions that arrive worked, and in batches. The counterpart of the specific feedback request is the well-built question. Before you ask, outline the situation, name the problem you are trying to solve, list at least two solutions you considered, and say which one you lean towards and why. This is the format we teach in Fast Forward, and it pays both sides at once. You learn to work the problem before handing it over, which is the skill your doctorate exists to build. Your supervisor gets a specific problem instead of a generic question, and specific problems get actionable answers. Then batch: collect the non-urgent questions for the meeting or the office hour instead of firing them one at a time. And do not read a slower reply as neglect; a supervisor who answers at set times rather than instantly is running the same machinery from the other end, which I show in my post on protecting your research time as a PI.
A default-forward rule for silence. Agree it once, out loud: “If I have not heard back on a decision within two weeks, I will proceed with the option I flagged, and you can correct me at our next meeting.” This single sentence dissolves the most corrosive pattern in doctoral work, the weeks lost waiting for permission that was never actually required. It turns silence from a blocker into a decision.
Simon built all five, starting with the agenda, and within two months the same supervisor, same personality, same overloaded calendar, was returning decisions inside a week. But the effect he had not predicted was the one worth dwelling on. To fill the agenda each week he had to break the project into named work packages, which meant he always knew his own next three steps, and therefore so did his supervisor. He had discovered from below what our Fast Forward participants report from inside the same experience. As one of them put it: “It also improved my communication with my supervisor and co-authors because I knew exactly what the next important steps were.” Read that carefully, because it inverts the usual assumption. We think communication with a difficult senior person is about managing them. Often it is about making yourself legible: structure travels up a relationship as reliably as it travels down, and the person who builds it decides which way it flows. Nothing about Simon’s supervisor changed. The interface did.
The conversation Simon kept not having
Structure handles most mismatches silently, which is its great virtue: nobody has to be confronted. But sometimes the pattern itself needs naming, and this is the conversation students defer for semesters, because the power asymmetry makes it feel dangerous. It is less dangerous than the alternative, which is resentment quietly compounding until it leaks out at the worst possible moment.
The format that works is behavioural, forward-looking, and free of diagnosis. Three sentences: a specific observable pattern, its concrete effect on the work, a proposed mechanism. “The last three manuscripts came back after four to six weeks, which put each submission behind the conference deadlines. Could we agree a two-week window for decision-level feedback, and I will mark which sections need it?” No accusation of neglect, no theory about their priorities, no “I feel undervalued”. A supervisor can argue with a characterisation of their motives all day. A dated list of turnaround times and a cheap proposed fix is very hard to argue with, and most will simply say yes. If saying it aloud feels unmanageable, put the same three sentences in writing. What matters is that the pattern gets named while it is still a logistics problem, because unnamed logistics problems reliably mature into relationship problems.
When it is no longer a logbook problem
Everything above assumes good faith on both sides, and that assumption has a hard limit. If you are being bullied, humiliated in front of others, having your work appropriated, facing discrimination, or being retaliated against, you do not have a communication problem, and no amount of agenda-sending will fix it. Those situations call for a different toolkit entirely: documentation, your thesis committee or graduate school, the ombudsperson, and the formal procedures your institution maintains for situations like this one. This is the apprentice appealing past the master to the guild, and it is not disloyalty, it is using the system as designed. If the situation is affecting your health, a doctor or counsellor belongs in the picture too; this is a sensitive area, and if any of it touches your own mental health, finding the right support matters more than any strategy in this article.
One researcher in a recent cohort completed a training while, in their own words, navigating “a difficult situation with my supervisor regarding my project and contract extension”, and still held their daily structure through it. That is the real picture: structure is what you can control while the larger thing is being resolved, not a substitute for resolving it.
A word for the graduate school that owns the vacuum
And if you run a graduate school, or organise the trainings a graduate school buys, the apprenticeship comparison lands on you, because the missing articles of agreement are institutional, not personal. You hire brilliant researchers to supervise and you train almost none of them to do it (the craft itself is the subject of my article on how to supervise PhD students), then you are surprised when the relationship is where your students founder. You ARE the guild in this analogy: the external structure that decides whether an apprentice is ever wholly captive to one person’s mood. Two things follow. Every doctoral researcher needs a genuine second mentor, a committee member, a senior postdoc, a collaborator elsewhere, as basic risk management, because a PhD with a single point of failure at its centre is a fragile system and everyone experienced knows it. And the structural communication skills in this article, the agenda, the summary, the specific request, the named boundary, are trainable, which is what the Resilience Training for Researchers and Fast Forward build, with over 1,500 individuals trained across both. The trades worked this out centuries ago: you cannot make every master a great teacher, but you can give every apprentice a logbook and a route to the guild. For the doctorate, that engineering is overdue.
FAQ
How do I deal with a difficult PhD supervisor?
Here is what nobody tells you: most “difficult supervisor” problems are structure problems wearing a personality. Diagnose first; availability, feedback, communication, expectations and interests are five different mismatches with five different fixes. Then keep the logbook from below: an agenda sent the day before, a written summary after every meeting, feedback requests specific enough to answer, and an agreed default when silence blocks you. Fix the interface before you judge the person.
My supervisor never gives useful feedback. What can I do?
Look at your requests before their replies. “Any comments welcome” hands the busiest person in the relationship the job of deciding what matters, so they defer it. Ask for a decision, not comments: which sections need judgement, which to skip, by what date, tied to a real deadline. Specific requests get answered first. Vague ones get answered eventually, which in academia means late.
Can I change my PhD supervisor?
Usually yes, and it is less rare than the silence around it suggests; most institutions have a procedure through the graduate school or thesis committee. It is a heavyweight move, so exhaust the structural fixes first for real mismatches. But for mistreatment, or a relationship that structure demonstrably cannot repair, changing supervisors is not failure. Staying in a doctorate that is damaging you would be the failure.
How do I disagree with my supervisor without damaging the relationship?
Disagree about the work, in writing where possible, with evidence attached. The safe format is behavioural: the specific observation, its effect on the project, and a proposal, with no commentary on motives. Supervisors argue with characterisations; they rarely argue with dated facts and a cheap fix. And disagreeing early, while it is still a logistics issue, is far safer than disagreeing late, once it has become a grievance.
References
1. Levecque, K., Anseel, F., De Beuckelaer, A., Van der Heyden, J., & Gisle, L. (2017). Work organization and mental health problems in PhD students. Research Policy, 46(4), 868–879.
2. Cornér, S., Löfström, E., & Pyhältö, K. (2017). The relationships between doctoral students’ perceptions of supervision and burnout. International Journal of Doctoral Studies, 12, 91–106.
3. Woolston, C. (2019). PhDs: the tortuous truth. Nature, 575, 403–406. (Nature global PhD survey.)
Last Updated on August 5, 2026 by Dr Nadine Sinclair
Dr Nadine Sinclair
Dr. Nadine Sinclair is a molecular biologist, strategy consultant and co-founder of Mind Matters. A scientist by training and at heart, she conducted her doctoral research at the Max Planck Institute for Biophysical Chemistry. She has 30,000+ hours of hands-on project management experience, built across 18 years as a strategy consultant at McKinsey and as an independent consultant, working with research institutions, foundations, pharmaceutical and biotech companies and governments. She is the co-creator of the Personal Resilience Indicator and the author of On Track. Since 2018, over 1,500 researchers from more than 30 institutions have trained with Mind Matters.