When mentorship goes wrong: from personal vignettes to a system-level playbook for safer, fairer academic training

Mentorship is supposed to be ballast, a steady hand on the rudder while someone learns to sail in heavy weather. In medicine and science, we romanticize this role. We picture the senior clinician who opens doors, the principal investigator who fights for a stipend, the professor who notices a spark and fans it. We tell trainees to "find a mentor." We rarely teach them what to do when the relationship curdles.On paper, mentorship works. A systematic review in JAMA linked mentoring to career guidance, research productivity, and grant success, even while highlighting heterogeneity and methodological limits across studies [1]. Qualitative work across academic health centers has mapped the features of both successful and failed pairings: clarity of expectations, aligned values, reciprocal commitment, and the mentor's willingness to spend social capital distinguish the former; opacity, misalignment, and one-sided commitment mark the latter [2]. Alongside these encouraging findings, a parallel literature documents negative mentoring experiences. Taxonomies of "harmful mentorship" describe neglect, manipulation, lack of expertise, and broader dysfunction [3]. Team-level research on psychological safety shows how fear of embarrassment or retaliation silences questions and concerns, undermining learning precisely where stakes are high [4]. In academic medicine, bullying, harassment, and discrimination are not rare outliers but documented, patterned phenomena that intersect with mentorship and training structures [5][6][7]. Sponsorship, the active use of influence to advance a junior colleague, is also unevenly distributed, with women and underrepresented groups often receiving less advocacy even when mentored [8,14]. Despite this accumulated knowledge, most institutional responses remain strangely generic: workshops, policy statements, aspirational values on websites. Trainees still lack a practical map for what to do when mentorship goes wrong: how to name what is happening, how to decide whether to repair, exit, or escalate, and what they can reasonably expect from institutions that claim to value mentoring.I have been mentoring medical scientists, physicians, biotechnologists and biologists for over two decades, in several countries. In that time, I have celebrated students who flourished with generous guidance and honest feedback, but I have also watched talented people shrink, stall, or quietly leave because a mentoring relationship went wrong. Many of them did not complain formally; they came to my office, or caught me in a corridor, and told me stories that sounded uncomfortably similar, no matter the country, specialty, or institution. I have also had to confront my own blind spots as a mentor. Moments when, under pressure for grants, papers, and committees, I postponed meetings too often, answered emails too late, or failed to use my position to open doors for someone who trusted me. This piece grows out of that mix of privilege and discomfort: the privilege of seeing so many careers from close range, and the discomfort of recognizing how easily a system like ours can turn good intentions into harm. This Opinion article sits at that junction. It weaves together composite vignettes that many in academic medicine will recognize with an explicit decision pathway for trainees and a simple institutional framework, COMPASS, that links everyday mentoring practices to structural levers and concrete metrics. The goal is not to re-describe the problem, but to move from lament to a testable playbook for safer, fairer academic training.The damage caused by harmful mentorship is often quiet before it becomes dramatic. The literature captures this duality: on average, mentoring has modestly positive effects, but negative or neglectful relationships can reverse that benefit completely [1,3,5].Consider a junior infectious diseases fellow paired with a renowned tuberculosis researcher. On paper, it looks ideal. The lab is high profile; the mentor is generous with ideas but largely absent in practice. Meetings slide, then disappear. Deadlines are flexible for the mentor but rigid for the fellow. Authorship remains a moving target discussed in passing, never in writing. Feedback arrives as late-night line edits with no real conversation about direction. No single episode feels catastrophic. Yet, by the end of the year, the fellow has a thin CV, a narrow skill set tied to the mentor's pipeline, and an expanding sense of dread. This is not cinematic abuse. It is neglect, and neglect is one of the most common negative mentoring experiences described by protégés [3].The consequences extend beyond individual morale. When overwork, public shaming, or chronic unavailability substitute for guidance, team learning collapses. Psychological safety, the shared belief that it is safe to take interpersonal risks, is a prerequisite for the rapid error detection and adaptation that good science requires [4]. In unsafe teams, people withhold dissent, soften inconvenient data, and keep nascent ideas to themselves. Science becomes more fragile just as the rhetoric about excellence becomes louder. Reviews of academic bullying in medical settings show how mistreatment correlates with burnout, mental health problems, and attrition, with trainees in oppressive environments more likely to abandon particular specialties or academic careers altogether [5].The cost is also cumulative and material. It appears as stalled manuscripts, missed funding cycles, lost introductions, and delayed independence. Careers do not usually derail in one dramatic moment; they bend slowly under the weight of repeated, preventable failures of supervision, sponsorship, and safety.Negative mentoring is not a single phenomenon. Classic work by Eby and colleagues organized protégés' reports into categories such as mentor distancing (being unavailable), manipulative behavior, lack of expertise, and general dysfunction [3]. Subsequent studies have shown that these experiences are widespread; in some workplace samples, more than half of mentees report at least one negative mentoring event [3]. Academic medicine adds its own layers of steep hierarchies, intense apprenticeship, and dependence on a few key gatekeepers [2,[5][6][7].In practice, these patterns often appear as recognizable archetypes, summarized in Table 1, rather than as abstract categories:• The Absentee Sponsor. Prestige is abundant; presence is scarce. Meetings are perpetually postponed, feedback is sporadic, and the mentee slowly internalizes the message that their development is peripheral. When deadlines slip, the cost is borne almost entirely by the trainee. • The Extractor. Here, the mentee is treated less as a developing colleague and more as a convenient source of labor. Tasks multiply, the link to the mentee's goals becomes tenuous, and authorship remains vague until the moment of submission. Loyalty flows upward; credit does not. • The Gatekeeper. Access becomes currency. Introductions, collaborations, and data are controlled in ways that primarily preserve the mentor's power. The mentee's visibility is rationed: conference invitations, committees, and awards are deferred indefinitely because "the timing is not right." • The Unsafe Teacher. Feedback wounds rather than sharpens. Errors are tallied; embarrassment is used as a teaching method. In such climates, curiosity shrinks and people stop taking intellectual risks. The day-to-day atmosphere violates the core features of psychological safety described in team research [4].• The Boundary Crosser. Here, the problem is not misalignment but danger. Lines between professional and personal are blurred, and behavior crosses into harassment or discrimination. Surveys among surgical residents, for example, have documented high rates of gender discrimination and sexual harassment in training environments [7].These archetypes do not cover every scenario, but they help trainees name what they are living. They also interact with structural harms that must be called out clearly. Sexual harassment and gender-based discrimination in academic medicine are tightly linked to concentrated power, weak accountability, and under-resourced reporting systems [6,7]. Sponsorship is itself structured by inequity: women and underrepresented groups are less likely to receive the active, public advocacy that moves careers forward, even when they are formally "well mentored" [8,14]. A taxonomy that ignores these dynamics risks framing predictable design failures as individual fragility.Recognizing the pattern is not about pathologizing imperfect mentors. It is about creating a shared language for behaviors that, left unnamed, silently shape who advances, who stalls, and who leaves.If stories of "mentorship gone wrong" sound familiar across specialties and continents, it is because they are produced by structure as much as by personality. Academic medicine concentrates scarce rewards, grant funding, first and last authorship, protected time, prestige, inside supervisory relationships. At the same time, institutions often collapse three distinct roles into one person: the supervisor who controls data and funding, the evaluator who writes letters and signs off on progression, and the sponsor who opens doors.This role conflation creates conflicts of interest even for well-intentioned people. Under conditions of grant scarcity and tournament-like competition for positions, rational actors may under-invest in mentoring, hoard credit, or limit a mentee's visibility to protect their own program [5][6][7][8]14]. The system, viewed from a distance, can resemble a pyramid: many trainees at the base, a narrowing set of secure posts at the top, and a culture in which everyone is encouraged to believe that exceptional effort alone will secure a place.Three mechanisms recur in conversations with trainees and faculty. First, scarcity of time and funding converts mentoring into an optional luxury. Meetings are the first thing to go when deadlines loom, even though data show that structured mentoring contributes to productivity [1,2,9]. Second, opaque norms around authorship, data access, and contribution statements make it easy for credit to flow toward those already in power. Third, the absence of explicit, written expectations about goals, timelines, and sponsorship leaves both parties improvising, which usually favors the person with more experience and leverage.The result is a system in which negative mentoring is not an aberration but a predictable byproduct. Fixing this requires more than exhortations to "be a better mentor." It requires dyadic tools that help pairs function better and institutional safeguards that separate incompatible roles, protect reporting, reward good mentoring, and make invisible work visible.Trainees often sense trouble long before they have language for it. The early signs are usually mundane. Meetings are repeatedly rescheduled and then quietly abandoned. Feedback, when it arrives, is non-specific ("keep working") or purely editorial, with no discussion of direction or feasibility. Authorship plans remain verbal, resurface only at submission, and seem to shift with each iteration. Opportunities, talks, committees, collaborative calls, consistently flow toward the mentor, with little thought given to the mentee's visibility.Qualitative studies in academic health centers describe how mentees in such relationships gradually reinterpret structural problems as personal failure [2]. They start to believe they are "slow," "not ready," or "lucky to be here at all," even when objective outputs suggest otherwise.Research on psychological safety helps explain this pattern: when people perceive interpersonal risk as high and their status as fragile, they self-silence [4] Another signal is the body's own barometer. When the prospect of sending an email reliably produces dread, when you rehearse simple requests for days, when you carefully edit out any sentence that might sound like advocacy for your own development, the relationship is no longer a neutral container for learning. It has become a psychological stressor, and the literature on academic bullying suggests that prolonged exposure to such climates carries real risk [5].Once a trainee recognizes that something is wrong, the next question is brutally simple: what now? Over years of discussing these situations with mentees and colleagues, I have found it useful to frame the options in three verbs: repair, exit, or escalate. These are not rigid steps but distinct moves, each with its own risks and evidence base.Some misaligned relationships can be repaired, particularly when both parties are acting in good faith but swimming in institutional chaos. The most powerful tool here is specificity. Rather than a vague conversation about "needing more mentoring," it is often more productive to bring a written, time-bound proposal: clear goals, concrete deliverables, timelines, and an initial authorship and data-use plan.Randomized trials of mentor training in clinical and translational research demonstrate that focused curricula can improve communication, expectation-setting, and support for career development [9,11]. Instruments such as the Mentoring Competency Assessment (MCA) translate those skills into concrete domains, aligning expectations, maintaining effective communication, fostering independence, that can be discussed explicitly [10,12]. Mentees can adapt the same domains to frame their requests: regular meetings with agendas; agreed milestones; explicit discussion of authorship and sponsorship.When a mentor engages sincerely with this process, adjusts meeting cadence, clarifies credit, follows through on agreed actions, the relationship may not only stabilize but strengthen. Not every poorly structured dyad is doomed. Some simply need the kind of deliberate scaffolding that we routinely build into clinical trials and rarely build into mentoring.If repeated attempts at repair produce only temporary or cosmetic change, the problem may not be skills but incentives and bandwidth. In those cases, exit becomes an act of risk management rather than betrayal. Changing mentors is painful, particularly when trainees worry about being perceived as disloyal or ungrateful. Yet the cumulative opportunity cost of staying in a stagnant or exploitative relationship is often greater than the disruption of leaving.One intermediate step is to diversify rather than immediately abandon: adding a co-mentor with complementary expertise, seeking an external sponsor, or joining a collaborative network that is not fully controlled by the current mentor. Multi-mentor models are increasingly encouraged for exactly this reason. Over time, as alternative sources of guidance and sponsorship solidify, the mentee can renegotiate or step away from the original dyad.Empirical work reminds us that mentoring is, at best, one factor among many [1,2]. There is no virtue in persisting indefinitely in a relationship that consistently narrows rather than enlarges a trainee's trajectory.Escalation is different. When boundaries are crossed, harassment, discrimination, retaliation, plagiarism, the goal is not repair. It is protection and accountability. Reviews and commentaries on sexual harassment in academic medicine how these are and how often career risks when they these of who or did in which and with what help move reports from vague to Trainees not have to this with trusted the or institutional can options and for their must reporting that are and by that are is when safety is at or the of the failures the of those with the least to substitute for institutional academic systems reliably both science and chronic mentoring then the of the need The in one such domains Meetings and and credit, everyday practices with structural safeguards and simple with written mentorship for each active These are not documents but shared goals, meeting cadence, and a authorship and data-use A is what of active have a and how often are these recognizes that advocacy is can and sponsorship actions, introductions, to or mentors to for each a Surveys and of sponsorship in academic medicine support the link between active advocacy and career progression, as as the current in who such advocacy and sound but often whether work moves or can meetings with simple and shared and they can meeting rates and of as of mentoring and more structural reporting or of supervision, and sponsorship to an for and of among trainees are of whether these in rather than only on that mentoring is not treated as a but as of mentoring to protect can and and studies have shown that mentor training can improve such training with and explicit protected time for mentoring of the between and credit authorship and authorship contribution statements, and data-use and the and the of papers, visibility of contribution statements, can be at the is through regular that explicitly about mentoring, sponsorship, and in reporting Psychological safety, in and clinical can be in academic as with documented when problems are these levers an rather than an aspirational and can a an academic year, and whether reports mentee visibility and perceived safety in more links the archetypes of harmful mentorship to trainee actions, structural and is an Opinion article in evidence and not a systematic The vignettes are from conversations and to protect while The decision repair, exit, escalate, is as a for trainees rather than a on steps that and and be formally work might the of mentorship sponsorship and on objective such as time to and attrition, as as on psychological safety and perceived The that mentoring failures are patterned that are both and mentorship goes wrong, it is to tell the as a of the but the the The evidence a mentoring is often the of design who over how credit is what behaviors are or decision pathway here trainees a to move from to to repair with written to exit relationships that consistently their when safety is to through protected that how institutions can make good mentoring and mentoring that sponsorship that that feedback, that make reporting less that values mentoring as real authorship that credit, and a that psychological safety as as of us who own of mentors who even when it cost them shared authorship and us in ways that We also often just as the experiences that us The is to move beyond to and to build systems in which mentoring that is and is not an act of individual but the domains that practices with institutional safeguards and each active mentor and mentee on a written goals, meeting cadence, and a authorship and data-use A is the of with an and the that a written authorship to advocacy such as introductions, talks, or and credit in public A practical is a target of documented sponsorship mentee example, three to by gender and meetings at a regular cadence, with simple and in a shared can meeting of documented and of agreed as of mentoring reporting separate or supervision, and sponsorship roles where and to an of these among the of and for are of whether function in is into and mentoring and explicit of protected mentoring time as of the use of mentoring in of mentor and between and mentoring time authorship contribution statements example, and clear data-use are at the or program The of visibility of contribution statements in and institutional and with credit in as that explicitly about psychological safety, exposure to bullying or harassment, in reporting and perceived of authorship and in a safety with documented on problems, practical of whether the is over time