Das ist kein Ausreißer. Das ist die Norm. Stanford verlangt drei. Harvard GSAS verlangt drei. Chicago MAPSS verlangt mindestens drei. NYU GSAS und ASU — drei. Im US-Forschungsuniversitätssystem ist drei die Erwartung des Bewerbungsformulars.

Halten wir hier kurz inne. Ein Masterabschluss ist, im offiziellen Selbstverständnis der Universität, das Zeugnis derer, die die Arbeit getan haben — Klausuren geschrieben, Hausarbeiten verfasst, Noten gesammelt. Das Bachelorzeugnis ist die Quittung. Das Diplom ist die Quittung der Quittung.

Und obendrein müssen drei ältere Personen bürgen.

Warum?

Die offizielle Antwort: Schreiben erfassen, was Zeugnisse nicht erfassen. Durchhaltevermögen. Neugier. Kollegialität. Forschungspotenzial. Die weichen Eigenschaften. Das, was eine Note nicht zu fassen bekommt.

Das ist eine kohärente Theorie. Sie ist auch fast vollständig falsch. Und sobald man sieht, was das Schreiben tatsächlich tut, kann man es nicht mehr nicht sehen.

Gehen wir das Schritt für Schritt durch.

Was das Schreiben tatsächlich misst

Das Schreiben misst nicht die Studentin. Es misst den Zugang der Studentin zur schreibenden Person.

Überlegen Sie, was nötig ist, um ein starkes Schreiben zu produzieren. Eine Professorin muss eine Bachelorabsolventin gut genug kennen, um eine Stunde lang sorgfältig über sie zu schreiben. Vorlesungen leisten das nicht. Klausuren leisten das nicht. Die Beziehung entsteht woanders: in Sprechstunden, in unbezahlter Laborarbeit, in studentischen Hilfskraftstellen, in Bachelorarbeiten mit besonderer Betreuung — in den kleinen Ritualen, in denen Studentinnen lernen, Lehrende anzusprechen, und Lehrende lernen, Studentinnen ernst zu nehmen.

Diese Rituale kosten Zeit. Die Studentin, die am Wochenende kellnert, hat diese Zeit nicht. Die Studentin, die zwei Stunden aus dem Elternhaus pendelt, weil eine Wohnung in Universitätsnähe unbezahlbar ist, hat diese Zeit nicht. Die Studentin aus einer Familie, in der Erwachsene mit Titel nicht angesprochen wurden, weiß noch nicht einmal, dass es diese Rituale gibt.

Die Verzerrung steckt nicht in der Sprache des Schreibens. Sie steckt in den Voraussetzungen.

Woher die Praxis kommt

Die Genealogie ist der Teil, den die meisten Kommissionen nie untersuchen. Sie ist auch der Teil, der sich, einmal bekannt, kaum verteidigen lässt.

Der moderne amerikanische Zulassungsessay — und sein Begleitstück, das Empfehlungsschreiben — wurde nicht erfunden, um verborgene Talente zu entdecken. Er wurde in den 1920er-Jahren in Harvard, Yale und Princeton erfunden. Er wurde erfunden, weil diese Institutionen ein Problem hatten.

Das Problem war: reine akademische Leistung ließ zu viele Juden zu.

Jerome Karabel, The Chosen, dokumentiert das im Detail. Die Verlagerung vom quantifizierbaren akademischen Zeugnis zur “Charakterbeurteilung” — Schreiben, Interviews, “well-roundedness” — war eine bewusste institutionelle Antwort. Die Kuratorien brauchten ein Werkzeug, das es erlaubte, akademisch brillante Bewerber abzulehnen, ohne den Grund zu nennen. Empfehlungsschreiben waren dieses Werkzeug. Sie waren nicht quantifizierbar. Sie waren nicht überprüfbar. Sie funktionierten.

Ein Jahrhundert ist vergangen. Das Werkzeug wurde nicht aussortiert. Es wurde verfeinert.

Noch etwas ist beachtenswert. Die Drei-Briefe-Norm ist amerikanisch. Deutsche Universitäten lassen Studierende in konsekutive Masterprogramme typischerweise ganz ohne Schreiben zu — Zeugnisse und strukturierte Kriterien leisten die Arbeit. Britische Programme verlangen typischerweise zwei. Die kontinentalen Systeme führen Masterzulassungen über Zeugnisse und strukturierte Verfahren — und der Himmel ist nicht eingestürzt. Das ist wichtig, weil es zeigt: Die Anforderung ist keine logistische Notwendigkeit. Sie ist ein kulturelles Artefakt — spezifisch für eine bestimmte Zulassungstradition, exportierbar oder zurücknehmbar nach Belieben. Das amerikanische Drei-Briefe-Regime ist nicht, wie Universitäten Menschen zulassen. Es ist, wie diese Universitäten Menschen zulassen. Sie könnten morgen damit aufhören.

Die Kommission, von innen

Ich habe in Masterzulassungskommissionen gesessen. Ich möchte nüchtern beschreiben, was dort passiert.

Ein Schreiben mit einem Namen, den die Kommission kennt — eine ältere Person an einer Eliteuniversität, eine bekannte Mitautorin eines Mitglieds — leistet für die Bewerberin Arbeit, derer sich die Kommission weitgehend nicht bewusst ist. Der Inhalt wird als Bestätigung dessen gelesen, was die Unterschrift bereits impliziert. Ein Schreiben einer unbekannten Person, egal wie gut formuliert, wiegt schlicht nicht so viel. Niemand diskontiert es bewusst. Niemand muss es.

Das ist keine Korruption. Das ist bayessch. Wenn das Schreiben ein Signal ist, dessen Verlässlichkeit von der Quelle abhängt, dann ist eine Gewichtung nach Quelle rational.

Hier aber die Rekursion. Die Verfügbarkeit hochglaubwürdiger Quellen ist selbst nach genau jenen sozialen Faktoren stratifiziert, an denen die Kommission eigentlich vorbeisehen sollte. Die Bewerberin von einer Regionalhochschule — kompetente Betreuer, kein Renommee — beginnt mit einem Nachteil, der nichts mit ihrem Können und alles mit der Geografie ihrer Ausbildung zu tun hat.

Jenny Posselt, Inside Graduate Admissions, 2016. Die gründlichste Ethnografie, die wir haben. Befund nach Beobachtung von Kommissionen mehrerer amerikanischer Eliteprogramme: Die Kommissionen verließen sich stark auf Signale aus prestigereichen Netzwerken und Hochschulen, und diese Verlässlichkeit operierte unterhalb der Schwelle bewusster Abwägung. Die Mitglieder glaubten, sie würden Leistung beurteilen. Sie beurteilten Leistung gefiltert durch ein Netzwerk, dessen Zusammensetzung selbst eine Funktion dessen war, wer in vorangegangenen Generationen aufgenommen worden war.

Lassen Sie diese Schleife lange genug laufen, und Sie haben keine Meritokratie.

Sie haben eine erbliche Elite, die gelernt hat, sich in der Sprache der Leistung zu beschreiben.

Die Verteidigungen

Es gibt drei. Sie fallen alle. Schnell:

1. “Schreiben liefern Kontext.” Ja — für die Bewerberin, die welche hat. Die Bewerberin ohne schreibwillige Person profitiert nicht. Schlimmer: Das Fehlen wird als Mangel gelesen. Sie verliert doppelt. Das ist eine Verteidigung des Erlaubens von Schreiben, nicht des Verlangens.

2. “Schreiben erfassen Soft Skills.” Tun sie nicht. Sie erfassen die Investition der schreibenden Person in die Studentin — selbst eine Funktion strukturellen Zugangs. Wenn wir Durchhaltevermögen und Kreativität messen wollen, gibt es bessere Instrumente: strukturierte verhaltensbezogene Interviews, rubrisch bewertete Schreibproben, Portfolios. Wir benutzen sie nicht. Wir benutzen das Schreiben. Weil das Schreiben für die Institution billig ist und weil die Kommission das Gefühl hat, einer “ganzen Person” begegnet zu sein.

3. “Alle stehen vor derselben Regel.” Eine formal gleiche Regel, angewendet auf substantiell ungleiche Vorbereitung, produziert ungleiche Ergebnisse. Das ist das älteste Resultat der Sozialschichtungsforschung. Es hört nicht auf wahr zu sein, nur weil Zulassungskommissionen lieber nicht darüber nachdenken.

Wozu das Schreiben tatsächlich da ist

Streifen Sie die offizielle Theorie ab und sehen Sie sich die Funktion an.

Das Schreiben selektiert für Studierende, die die richtige Art von Bachelorerfahrung hatten — kleine Kohorten, ansprechbare Lehrende, eine Kultur, in der das Anklopfen an einer Bürotür normal war. Es selektiert für Studierende aus der richtigen Art von Familie — eine, in der Titel nicht einschüchterten. Es selektiert für Studierende von der richtigen Art von Institution — eine, deren Unterschriften die Kommission ohnehin schon kennt.

Es ist keine Messung von Können. Es ist eine Messung vorheriger Selektion. Es protokolliert, wer bereits zuvor in das Netzwerk aufgenommen wurde, aus dem die Kommission schöpft.

Das ist, was Selektionsmechanismen typischerweise tun. Sie reproduzieren die Population, die sie entworfen hat. Das Empfehlungsschreiben ist darin nichts Besonderes. Es ist nur besonders darin, mit welcher Selbstsicherheit es behauptet, etwas anderes zu tun.

Was stattdessen?

Der Fall gegen die Pflicht ist kein Fall gegen die Bewertung von Können. Er ist ein Fall für die Bewertung mit besseren Instrumenten.

Eine strukturierte Schreibprobe, von zwei unabhängigen Leserinnen anhand einer veröffentlichten Bewertungsmatrix beurteilt. Zeugnisse und Notenverläufe — schon vorhanden. Ein kurzes strukturiertes Interview: dreißig Minuten, dieselben Fragen für jede Bewerberin, dieselben Dimensionen. Portfolios, wo das Fach es zulässt.

Keines davon ist verzerrungsfrei. Alle sind weniger verzerrt als das Schreiben. Alle sind transparenter darüber, was sie messen.

Für die Bewerberin, die ein Schreiben einreichen möchte — bitte. Macht es optional. Behandelt sein Fehlen als null Information, nicht als negative. Gewichtet es bescheiden. Die gegenwärtige Praxis — drei Schreiben zu verlangen und ihr Fehlen als Mangel zu lesen — ist die Praxis, die das Schreiben von einem nützlichen Zusatzsignal in einen Klassenfilter verwandelt.

Schluss

Stellen Sie sich eine Frage. Wenn Sie in einer Kommission sitzen, stellen Sie sie ehrlich.

Was würde sich an Ihrer Bewertung der Bewerberinnen ändern, wenn die Schreiben morgen verschwinden würden?

Wenn Ihre Antwort lautet wenig, dann ist das Schreiben dekorativ, und Sie erheben eine Steuer auf jede Bewerberin um eines Rituals willen. Wenn Ihre Antwort lautet vieles, dann sehen Sie genau hin, was die Schreiben tatsächlich tun. Sehen Sie hin, was die Literatur sagt, dass sie tun. Entscheiden Sie, ob das ein Verfahren ist, das Sie weiter verteidigen wollen.

Die wohlwollendste Lesart des Schreibens: Es wurde für ein Problem entworfen, das wir heute nicht mehr haben — eine Zeit unzuverlässiger Zeugnisse, nicht standardisierter Hochschulen, Kommissionen ohne anderen Einblick.

Die unwohlwollendste Lesart: Es wurde für genau das Problem entworfen, das es bis heute löst — die falsche Sorte Mensch fernzuhalten, während die Institution behaupten kann, sie wähle nach Leistung aus.

Die Wahrheit liegt vermutlich dazwischen. Es ändert nicht viel. Wofür auch immer das Schreiben entworfen wurde — was es jetzt tut, ist das, wofür wir verantwortlich sind.

Und was es jetzt tut, ist eine Steuer für jene Bewerberinnen zu erheben, die sie am wenigsten zahlen können.

Schafft die Pflicht ab. Behaltet das Schreiben als optionalen Eingang. Baut den Rest der Beurteilung auf Instrumenten auf, die ehrlich darüber sind, was sie messen. Die Studierenden, die Sie am dringendsten zulassen wollen — jene, deren Können wirklich vorhanden, deren Zugang aber schmal war — werden ihren Weg hindurchfinden.

Genau dafür, schließlich, sollte das Tor da sein.

This is not an outlier. It is the norm. Stanford requires three. Harvard GSAS requires three. Chicago MAPSS requires a minimum of three. NYU GSAS and ASU — three. Across most of the American research-university system, three is what the application form expects.

Pause on this. A master’s is, in the official self-image of the university, the credential earned by undergraduates who did the work — sat the exams, wrote the papers, accumulated the grades. The transcript is the receipt. The diploma is the receipt for the receipt.

And yet, on top of that, three senior people must vouch.

Why?

The official answer is that letters capture what transcripts cannot. Persistence. Curiosity. Collegiality. Research potential. The soft qualities. The things a number cannot hold.

This is a coherent theory. It is also almost entirely false. And once you see what the letter actually does, you cannot unsee it.

Let me walk through this.

What the letter actually measures

The letter is not a measurement of the student. It is a measurement of the student’s access to the writer.

Think about what is required to produce a strong letter. A senior academic must know a recent graduate well enough to spend an hour writing carefully about her. Lectures will not produce this. Exams will not produce this. The relationship is built somewhere else: in office hours, in unpaid lab work, in research assistantships, in honors theses, in the small rituals where students learn to approach professors and professors learn to take students seriously.

These rituals cost time. The student who waits tables on weekends does not have that time. The student who commutes two hours from a parental household because campus rents are ruinous does not have that time. The student raised in a family where adults-with-titles were not addressed does not even know the rituals exist.

The bias is not in the letter’s language. The bias is in the prerequisites.

The empirical record

You do not have to take this on faith. The literature is now embarrassing.

A 2025 study in Research in Higher Education — Kim, Park, Lo, Baker and colleagues — ran NLP across 615,557 high-school counselor letters submitted through the Common Application. Undergraduate admissions, not graduate. But the same instrument, the same mechanics. The findings: systematic differences in length and content across nearly every demographic line. Letters about private-school students were longer. More sentences on “personal qualities.” The differences survived controls for academic record and school type.

Berkeley, undergraduate admissions, 2016–17. The university ran a pilot: introduce letters into a previously letter-free process. Roughly 30% of applicants invited; a subset scored both with and without their letters. Jesse Rothstein had the full dataset. The early Rothstein report found letters helped underrepresented applicants — boosts of up to four percentage points for moderate-likelihood candidates. Encouraging, if you stopped reading there. Subsequent work with Ben-Michael and Feller (NBER w30940) sharpened the picture and made it less encouraging: the impact of letters increases with applicant strength, and on average there is “little difference” for disadvantaged applicants, with the result “more mixed” than the headline suggested. The cleanest natural experiment we have does not vindicate the letter. It produces an ambiguous, applicant-strength-dependent picture in which the candidates with the most to recommend them benefit most. Which is, when you think about it, exactly the dynamic the practice’s critics warned about.

Then the language. Trix and Psenka, 2003. Madera et al., 2009 and 2019. Akos and Kretchmar, 2016. Newkirk-Turner and Hudson, 2022.

The pattern, replicated again and again:

• Letters about women — communal language. Warm. Helpful. Supportive.

• Letters about men — agentic language. Analytical. Ambitious. Leading.

• Letters about Black applicants — what the literature politely calls “phrases that bias readers.” Hedges. Doubt-raisers. Grindstone words. Diligent. Hardworking. In the encoded vocabulary of admissions, these mean: competent, but not exceptional.

Newkirk-Turner and Hudson — looking at 161 letters submitted to a speech-language pathology graduate program — found 202 instances of biasing phrases. The phrases were unrelated to the applicant’s GPA, undergraduate institution, or letter length. They were, however, related to admission offers. The phrases predicted outcomes independently of the academic record they accompanied.

So the letter discriminates twice. Once in who can obtain a strong one. Again in how it reads once written.

Two filters in series. Both pointed the same direction.

Where the practice came from

The genealogy is the part most committees never investigate. It is also the part that, once known, is hard to defend.

The modern American admissions essay, and its companion the recommendation letter, was not invented to find diamonds in the rough. It was invented in the 1920s at Harvard, Yale, and Princeton. It was invented because those institutions had a problem.

The problem was that pure academic merit was admitting too many Jews.

Jerome Karabel, The Chosen, documents this in detail. The pivot from quantifiable academic record to “character assessment” — letters, interviews, “well-roundedness” — was a deliberate institutional response. The trustees needed a tool to reject scholastically brilliant applicants without saying why. Letters were that tool. They were unquantifiable. They were unchallengeable. They worked.

A century has passed. The tool has not been retired. It has been refined.

One more thing worth noticing. The three-letter norm is American. German universities typically admit students into consecutive master’s programs without letters at all — transcripts and structured criteria do the work. UK programs typically ask for two. The continental systems run master’s admissions on transcripts and structured assessment, and the sky has not fallen. This matters because it tells you the requirement is not a logistical necessity. It is a cultural artifact — specific to a particular admissions tradition, exportable or retractable at will. The American three-letter regime is not how universities admit people. It is how these universities admit people. They could stop tomorrow.

The committee, from the inside

I have served on master’s admissions panels. I want to describe, plainly, what happens.

A letter from a name the committee recognizes — a senior figure at an elite institution, a known collaborator of someone in the room — does work for the candidate that the committee is largely unaware of. The content of the letter is read as confirmation of what the signature already implies. A letter from an unknown recommender, however well written, simply does not weigh as much. Nobody discounts it consciously. Nobody has to.

This is not corruption. It is Bayesian. If the letter is a signal whose reliability depends on the source, then weighting by source is rational.

But here is the recursion. The availability of high-credibility sources is itself stratified by exactly the social factors the committee is supposed to look past. The candidate from a regional public university — competent advisors, no fame — begins with a disadvantage that has nothing to do with her ability and everything to do with the geography of her education.

Jenny Posselt, Inside Graduate Admissions, 2016. The most thorough ethnography we have. Her finding, after observing committees at multiple elite American doctoral programs: committees relied heavily on signals from prestigious networks and institutions, and that reliance ran below the threshold of conscious deliberation. The members believed they were judging merit. They were judging merit as filtered through a network whose composition was itself a function of who had been admitted in previous generations.

Run that loop long enough and you do not have a meritocracy.

You have a hereditary elite that has learned to describe itself in the language of merit.

The defenses

There are three. They all fail. Quickly:

1. “Letters give context.” Yes — for the candidate who has them. The candidate without a willing recommender cannot benefit from contextual nuance. Worse: her absence of a letter is read as a defect. She loses twice. This is a defense of allowing letters, not of requiring them.

2. “Letters capture non-cognitive skills.” They do not. They capture the recommender’s investment in the student — itself a function of structural access. If we wanted to measure persistence and creativity, we have better instruments: structured behavioral interviews, rubric-scored writing samples, portfolios. We do not use them. We use the letter. Because the letter is cheap for the institution and lets the committee feel it has met a “whole person.”

3. “Everyone faces the same rule.” A formally equal rule applied to substantively unequal preparation produces unequal outcomes. This is the oldest result in social-stratification research. It does not stop being true because admissions committees would prefer not to think about it.

What the letter is actually for

Strip away the official theory and look at the function.

The letter selects for students who had the right kind of undergraduate experience — small classes, accessible faculty, a culture in which approaching a professor was normal. It selects for students who came from the right kind of family — one in which titles did not intimidate. It selects for students who attended the right kind of institution — one whose faculty’s signatures the committee already trusts.

It is not a measurement of ability. It is a measurement of prior selection. It records who has already been admitted, before, into the network that the committee draws from.

This is what selection mechanisms tend to do. They reproduce the population that designed them. The recommendation letter is not unusual in this respect. It is only unusual in the confidence with which it claims to do something else.

What to do instead

The case against requiring letters is not a case against assessing ability. It is a case for assessing it with better instruments.

A structured writing sample, evaluated against a published rubric by two independent readers. Transcripts and grade trajectories — already in hand. A short structured interview: thirty minutes, the same questions for every candidate, scored on the same dimensions. Portfolios where the discipline allows.

None of these is bias-free. All of them are less biased than the letter. All of them are more transparent about what they measure.

For the candidate who wants to submit a letter — fine. Make it optional. Treat its absence as zero information, not as negative information. Weight it modestly. The current practice — requiring three letters and reading their absence as a defect — is the practice that turns the letter from a useful supplementary signal into a class filter.

Closing

Ask yourself one question. If you sit on a committee, ask it honestly.

What would change in your evaluation of candidates if the letters disappeared tomorrow?

If your answer is very little, then the letter is decorative and you are imposing a tax on every applicant for the sake of a ritual. If your answer is a great deal, then look hard at what the letters are doing. Look at what the literature says they are doing. Decide whether that is a process you want to continue defending.

The most generous reading of the letter is that it was designed for a problem we no longer have — an era of unreliable transcripts, unstandardized institutions, and committees with no other window into the candidate.

The least generous reading is that it was designed for exactly the problem it still solves: keeping the wrong sort of person out, while letting the institution maintain that it admits on merit.

I suspect the truth is between these. I also think it does not matter very much. Whatever the letter was designed for, what it does now is what we are responsible for.

And what it does now is tax the candidates least able to pay.

Abolish the requirement. Keep the letter as an optional input. Build the rest of the assessment on instruments that are honest about what they measure. The students you most want to admit — the ones whose ability is real and whose access has been narrow — will find their way through.

That, after all, is what the gate is supposed to be for.

The student knows what is happening. The student also knows that this person controls the recommendation letters that decide their next position, holds nominal authority over the project’s data, sits on the committees that will evaluate the dissertation, and has a thirty-year head start on every alliance inside the institution.

This article is for that student.

It is not a guide to feeling better. It is a guide to the legal and procedural mechanisms that exist, in most European jurisdictions, to constrain what a supervisor can do to a doctoral candidate. The architecture is uneven across countries. The underlying truth is consistent.

Doctoral candidates have substantially more rights than their supervisors generally let them know they have.

That sentence is the whole article, basically. The rest of it is a map.

A standard caveat first, because the rest of the article does not contain caveats and I want to get them out of the way. This is not legal advice. Every situation has specifics that matter. Every European jurisdiction has its own statutes and ministerial orders and institutional regulations. What follows is comparative-legal journalism. Its job is to make sure you know the regulations exist and where to look. If you are in a real situation, talk to your doctoral student union, to your national PhD council, or to a lawyer with employment-law expertise. Talk to them before, not after.

Now to the point.

The first thing to understand is that most “rules” in academia are not rules.

Doctoral candidates routinely encounter sentences like these. This is how we do things in this group. You cannot publish without my approval. You cannot change supervisors. You cannot defend until I am ready. This is normal in this field.

Almost none of these are statements about the law. They are statements about the supervisor’s preferred operation of the supervisor’s group. The legal regime that governs your doctorate is set by national statute and by university regulation. It is not set by your supervisor’s preferences.

This sounds obvious. It is not obvious. The single most useful thing a doctoral candidate in distress can do is to learn the difference between what the supervisor wants and what the regulation says. Almost nobody does this until they are forced to. Almost everyone who does it is surprised by what they find.

The regulations come from three layers, roughly. The national statute. The Finnish Universities Act 558/2009. The Dutch Higher Education and Research Act, the WHW. The French Code de l’éducation, plus the Arrêté du 25 mai 2016 fixing the national framework for the doctorate. The German Landeshochschulgesetze. The Swedish Higher Education Ordinance, 1993:100. These are public documents. They are not long. The chapters relevant to doctoral candidates are usually under fifty pages.

Then the institutional layer. Each university publishes its own Promotionsordnung, promotiereglement, doctoral regulation, faculty rules. These bind the institution. They cannot be overridden by departmental practice, and they cannot be overridden by your supervisor.

And then the third layer, which is the one almost nobody uses. Employment law.

In most of the Nordic countries, in the Netherlands, in parts of Germany, and in an increasing number of European systems, doctoral candidates are not students. They are employees. They have employment contracts, they pay social-insurance contributions, they accrue pension rights, and they are protected by the same body of labour law that protects everyone else who works for a wage in their country.

This is the layer where the strongest protections are. And it is the layer that doctoral candidates use the least, because they think of themselves as students rather than as workers. The supervisor is happy to encourage this confusion.

A doctoral candidate facing supervisor problems who has not read all three layers is fighting with one hand tied behind their back. Reading them takes a weekend. It is the most valuable weekend that most doctoral candidates will spend during their doctorate.

If you are an employee, the supervisor relationship is, legally, a workplace hierarchy. It is bound by the same laws that govern hierarchies in every other workplace.

This is the lever that almost nobody pulls.

Most European jurisdictions have explicit statutory provisions against workplace harassment. In Finland, the Occupational Safety and Health Act imposes an obligation on the employer to address harassment that endangers the employee’s health. In Austria, the Arbeitsverfassungsgesetz and the Gleichbehandlungsgesetz do similar work. In Sweden, the Discrimination Act and the Work Environment Act. In the Netherlands, the Working Conditions Act and the general “good employer” obligation in the Civil Code.

The pattern is the same across these jurisdictions. The employer is liable for failing to address harassment. Not the supervisor personally. The employer. The institution.

This changes what your complaint is about.

Your complaint is not against your supervisor. It is against the institution, which has a statutory duty to act and which, by failing to act, is itself in breach.

That is a different complaint. It points at a different actor. It triggers different procedures. And the institution knows it.

There is also discrimination law. If your supervisor’s hostile conduct correlates with gender, ethnicity, age, religion, sexual orientation, disability, or another protected characteristic — and it often does, in patterns that the candidate is not the first to notice — there is a parallel claim that is procedurally distinct from a harassment claim and frequently more powerful.

There is the right to a healthy work environment. In the Nordic and Dutch systems particularly, the employer has an active duty to maintain a working environment that does not damage the employee’s health. Including mental health. Documented stress, sleep disturbance, anxiety, or depression caused by workplace conditions creates an enforceable obligation on the employer to act.

And there is occupational health. In Finland every employer is required to provide occupational health services. They are confidential. They are separate from the employer. They can document workplace-related health impacts in ways that become evidence later. Most candidates do not know they are entitled to them. Most candidates assume that using them will be visible to the supervisor. It will not.

Reframe your situation. Stop saying my supervisor is treating me badly. Start saying my employer is failing to discharge a statutory duty. The first sentence is true. The second sentence is also true, and it is the one that wins.

The second layer is institutional procedure.

Almost every European university has, on paper, procedures for doctoral candidates in supervision difficulty. They are typically underused. Partly because candidates do not know about them. Partly because supervisors prefer it that way.

The menu is roughly the same across countries.

There is the doctoral committee, or the doctoral school. Your supervisor is rarely the only academic body with authority over your doctorate. In most European systems, a doctoral committee, a doctoral school director, or a faculty doctoral office holds the procedurally significant decisions. Admission to defense. Examination committee composition. Evaluation of progress. These bodies can be approached directly by the candidate. The supervisor has no formal authority to prevent this. The supervisor would prefer that you not realise this.

There is the ombudsperson. Most European universities have one. The role is confidential. It is not management. It cannot impose solutions. But it can open conversations the candidate cannot open alone, and its involvement is often noticed by deans and rectors in ways that single-candidate complaints are not. The German DFG Code of Conduct, binding on every DFG-funded institution, requires universities to maintain ombudsperson structures specifically for these conflicts. The Dutch national ombudsperson for academic integrity (LOWI) covers a different and overlapping function.

There is the supervisor change procedure. Almost every doctoral regulation in Europe contains one. They are often buried in the document, rarely advertised, and usually structured to require the cooperation of the current supervisor. But the cooperation requirement is procedural, not absolute. In Sweden, the right to request a change of supervisor when the relationship has broken down is statutory under the Higher Education Ordinance. In the Nordic systems generally, supervisor change is a recognised institutional response to documented difficulty, and faculties have a duty to facilitate it.

There is the individual study plan. Most modern European doctoral systems require one, typically reviewed annually. This document is binding on the institution, not just on the candidate. If your supervisor has begun to claim you are “behind” or “underperforming”, the binding instrument is the individual study plan, not your supervisor’s email. A supervisor cannot unilaterally redefine what counts as adequate progress. The institution can, but only by following the procedure laid out in the regulation.

And there is the multi-supervisor structure. A growing number of European universities — Sweden across the board, the Netherlands across all institutions, Switzerland especially since the ETH reforms of 2020, and increasingly German and Finnish faculties — require every doctoral candidate to have at least two formal supervisors.

This is not pastoral kindness. This is structural. When the primary supervisor is the problem, the second supervisor is the formal channel through which the situation can be escalated without the candidate having to navigate the conflict alone. This is one of the few unambiguously good reforms that European doctoral education has produced in the last decade. If your university already has it, use it. If your university does not yet have it, this is the reform to push for.

The third layer is the defense itself.

The single most common form of supervisor abuse against late-stage doctoral candidates is blocking the defense. It is also the form most amenable to legal remedy. Because the authority to admit a dissertation to defense almost always sits with a body other than the supervisor.

Germany. The doctoral committee, the Promotionsausschuss, admits the dissertation. The supervisor is one of two or three referees. A negative supervisor report does not end the procedure. It triggers committee review.

Netherlands. The PhD Board and the Doctorate Board hold the legal authority over the defense procedure. Not the supervisor. Article 7.18 of the WHW vests the conferral of the doctorate in the institution. The disputes regulation provides an administrative-law remedy if the candidate believes admission has been unjustly refused, with the General Administrative Law Act, the Awb, governing the objection procedure. Article 7:13 of the Awb governs the advisory committee that prepares the institution’s decision on the objection. The candidate has six weeks from the date of the contested decision to file the objection. This is not theoretical law. Dutch universities run these procedures regularly.

France. Under the Arrêté du 25 mai 2016, the head of the institution authorises the defense, on the proposal of the doctoral school director, after the rapporteurs submit favourable reports. The thesis director is procedurally constrained from blocking unilaterally. The arrêté itself is the relevant text, and the comité de suivi individuel, mandated by the same instrument, is the formal first step when the supervision relationship has begun to fail.

Sweden, Finland, Denmark, Norway. The faculty grants permission to defend. In Finland the väittelylupa. The supervisor is consulted. The supervisor does not decide. Refusal is an administrative decision subject to administrative-law review under, in Finland, the Hallintolaki 434/2003.

A supervisor who tells you that they personally can prevent your defense is, in most European jurisdictions, overstating their legal authority.

They can complicate the path. They can delay it. They can damage you politically. They cannot, in most modern doctoral regulations, definitively block it.

This is worth saying twice. Most candidates do not believe it on the first reading.

So what do you actually do.

What follows is procedural advice, not legal advice. It applies whether your situation is just beginning to feel wrong or already in deep crisis. Roughly in order.

First, document everything. Privately and securely. Start a timeline. Date every interaction that troubles you. Save email threads to a private storage account that your employer cannot access. Not your university Google account. Not OneDrive provisioned through your faculty. Not Dropbox under your work email. Use a personal account on a separate device if you can.

Be precise. Write down what was said, by whom, when, in front of whom. Note the tone, but distinguish between what you observed and what you inferred. A timeline assembled in real time is overwhelmingly more credible to any institutional or legal actor than a reconstruction created after the relationship has broken down.

Be careful with screenshots. The personal data of your supervisor and your colleagues is their data under the GDPR, and publishing or sharing it inappropriately can create liability for you, independent of the underlying conflict. Save evidence. Do not broadcast it.

Do not post about your case on social media. The institutional legal team, if it comes to that, will read every word. I have written about this elsewhere. The short version is: anything you write publicly about the case becomes the case.

Second, read the regulation. The doctoral regulation of your university. The relevant national statute. Your employment contract, including any references to collective agreements. The institution’s anti-harassment and equality policies.

Identify, on paper, who has the formal authority to admit your dissertation to defense. What procedure exists for supervisor change. Who the ombudsperson is, and how they are contacted. What the formal complaint procedure for harassment is. What the institutional disputes-resolution procedure is. What the protections against retaliation are.

Most candidates discover, in this exercise, that the institution is procedurally on their side in ways they had not realised.

Third, talk to your union and your PhD council. Almost every European country has a doctoral student association or union with experience in exactly the situation you are facing. Tieteentekijät in Finland. SFS-DK in Sweden. PNN in the Netherlands. The German Doktorandenverband and its disciplinary equivalents. The CJC in France. Their advice is free, confidential, and informed by repeated experience of patterns you are encountering for the first time.

In jurisdictions where the doctoral candidate is an employee, your trade union may also offer free legal aid as part of the membership package. Most academic unions in the Nordic countries, the Netherlands, and Germany do. If you are not yet a member, joining is the most cost-effective insurance available to you. Most unions accept new members and provide consultation immediately, with the legal-aid right activating after a short waiting period. The consultation alone is often enough to clarify the path.

Fourth, find the sympathetic institutional actor. In every European university there is, somewhere, an institutional actor willing to read the regulation in your favour. The doctoral school director. The dean. A senior professor in the department who is not your supervisor’s ally. The equality officer. The ombudsperson.

Your task is to find this person and to give them a documented, legally framed case that makes acting easy. The work is procedural, not emotional. The argument that wins is the regulation says X, and the institution is failing to apply X. Not my supervisor has been unfair to me. Both may be true. Only the first creates an obligation to act.

Fifth, use the formal procedures, in the right order.

Informal mediation through the ombudsperson or the second supervisor when the relationship has not yet broken irrecoverably and the candidate would prefer to continue with the same supervisor.

Supervisor change when the relationship has broken and the candidate wishes to continue the doctorate with a different supervisor.

A formal harassment complaint when the conduct meets the institutional definition of harassment and the candidate wishes the institution to investigate. This is procedurally distinct from supervisor change. It triggers different obligations on the institution.

An administrative-law objection when an institutional decision is itself the problem and the candidate wishes to challenge it. In the Netherlands, this is the bezwaar procedure under the Awb. In Finland, the path runs through the Hallintolaki. In Germany, it depends on the Landeshochschulgesetz.

An external complaint to a regulator when internal procedures have failed. Equality ombudspersons. Occupational safety authorities. National research councils, where research-funding terms are involved.

And, last, litigation. Labour court. Administrative court. Civil court. Litigation is expensive, slow, public, and sometimes the only path that creates real consequences. Talk to your union before going there. Almost everyone who ends up in litigation could have prevailed through earlier procedures, given better information earlier.

Sixth, protect your career path. Throughout this process, continue to publish. Continue to attend conferences. Continue to develop relationships with senior researchers in your field outside your group. The single most effective response to a supervisor who controls your career through letters of recommendation is to develop a network that does not require those letters. This is also the single most underrated piece of advice in the doctoral handbook.

If your supervisor has begun to spread negative narratives about you within the field, the counterweight is a track record visible to people who do not need to ask your supervisor what they think. Talks at international workshops. Co-authorships outside your group. Visits abroad. An engaged presence in your subfield. All of these are insurance against a supervisor who would prefer you to be silent.

The hardest thing about supervisor bullying is that it is, at its core, an attempt to convince the candidate that the regulation does not apply to them. That their case is special. That the procedures exist for other people. That they should not make trouble. That they should wait. That they should be patient. That they should consider their career.

Every one of these claims is structurally false.

The regulation applies to every candidate. The procedures exist for exactly your situation. The institution has obligations toward you that it cannot dissolve through informal pressure. The supervisor’s authority is significant but bounded. By employment law. By university statute. By the terms of your individual study plan. By the doctoral regulation. By the national statute. By the constitutional right to administrative due process. By the GDPR. By the equality directive.

The system is designed, on paper, to constrain exactly the behaviour you are encountering. Most of the time it does not. Because nobody invokes it.

The act of invocation is what activates the protections.

Read the regulation. Identify the actor with the legal authority. Build the documented case. Engage the procedure. The path is harder than it should be. It is also, in most European systems, more available than your supervisor wants you to know.

You are not alone. The pattern you are encountering has been encountered before, documented before, litigated before. Many candidates prevailed. The legal architecture exists. Use it.

Sources and primary references

Finnish Universities Act, Yliopistolaki 558/2009, at finlex.fi. Finnish Administrative Procedure Act, Hallintolaki 434/2003, at finlex.fi. Finnish Occupational Safety and Health Act, Työturvallisuuslaki, at finlex.fi. Dutch Higher Education and Research Act, WHW, at wetten.overheid.nl. Dutch General Administrative Law Act, Awb, including Article 7:13 governing advisory committees in objection procedures, at wetten.overheid.nl. French Arrêté du 25 mai 2016 fixing the national framework for the doctorate, at legifrance.gouv.fr. Swedish Higher Education Ordinance, Högskoleförordningen 1993:100, at riksdagen.se. Austrian Arbeitsverfassungsgesetz, at ris.bka.gv.at. DFG Code of Conduct, Guidelines for Safeguarding Good Research Practice, at dfg.de.

National doctoral student organisations: Tieteentekijät (Finland). SFS-DK (Sweden). PNN (Netherlands). Doktorandenverband (Germany). CJC (France).