What does a great mentor actually give you? (Not their topic.)

Everyone gives the same advice to a new PhD student. Find a great mentor, and do what they do. The first half is right. The second half, according to a large study of how scientific careers actually unfold, is close to backwards.

The paper is Yifang Ma, Satyam Mukherjee and Brian Uzzi, “Mentorship and protégé success in STEM fields”, in PNAS (2020). They assembled genealogical records on nearly 40,000 scientists who published 1,167,518 papers in biomedicine, chemistry, math and physics between 1960 and 2017, and asked a question most of us only answer with anecdotes: does having a great mentor actually change where you end up, and if so, what is it that gets transmitted?

Their headline: mentorship is associated with a 2x to 4x higher chance that a protégé wins a major prize, gets into the National Academy of Sciences, or becomes what they call a superstar, compared with matched protégés who had similar talent and similar starting odds. Not a small nudge.

What the mentor is actually transmitting

Here is the clever part of the design. They found clusters of mentors with similar records and reputations, who attracted protégés of similar talent. Within each cluster, one mentor had something extra: they went on to win a prize. Crucially, the skill that made them prizeworthy existed before the prize was awarded, which means their protégés were exposed to it early, before anyone knew to call it special.

And the authors are specific about what that hidden skill is. In their words, it is “skill in creating and communicating prizewinning research.” Read that twice. It is not only about having good ideas. It is about creating and communicating them. The thing that separates the future laureate from an equally credentialed peer is partly the ability to make important work legible to other people, and that ability rubs off on the students standing nearby. I run a company built on the second half of that sentence, so take my enthusiasm with salt. But it is their sentence, not mine.

The counterintuitive part: do not copy your mentor

Now the finding that should change behavior. Protégés did not succeed most by working on their mentor’s topics. They succeeded by studying original topics, and by coauthoring only a small fraction of their papers with the mentor. The students who broke away and staked out their own ground did better than the ones who stayed in the mentor’s lane. Mentorship also predicted something related: protégés who pioneered their own research topics, and who were more likely to be midcareer late bloomers than instant stars.

So the transmission is real, but it is not imitation. You absorb the mentor’s tacit skill for making and communicating important work, and then you point it at a question that’s yours. Copying the topic is the mistake. Copying the craft is the point.

What the study can’t rule out

I like this paper, and I don’t want to oversell it.

It is observational. Nobody randomly assigned students to great mentors. The authors work hard with matched comparisons to rule out the boring explanation, that great mentors simply attract great students, and that matching is the backbone of the whole argument. But matching narrows a gap, it does not close it. Some of the 2x to 4x could still be selection they could not see.

The models explain 34 to 44 percent of the variance in protégé success. That is a lot for social data, and it also means most of what decides a career still sits outside the model. Mentorship matters. It is not destiny.

And it is STEM, mostly the elite tip of STEM, measured through prizes and academy membership and “superstardom.” Those are the outcomes at the top of a very steep pyramid. Whether the same pattern holds for a solid, non-prizewinning career, or in the humanities, this paper cannot say.

Why you should care, on either side of the desk

If you’re early, the practical read is unusually clean. Choose a mentor for the tacit skill you can only learn by proximity: how they frame a problem, how they decide what’s worth doing, how they make it land with other people. Then don’t spend your twenties as their coauthor. Take the craft and leave the topic.

If you’re the mentor, the uncomfortable implication is that the most valuable thing you pass on isn’t the project you hand a student, and not the papers you write together. It’s a way of working that you might not even know you’ve, including how you make your work visible. You can’t teach that by keeping students on your own topics. You teach it by letting them leave.

I keep coming back to that one word in the abstract: communicating. The study says the hidden, mentorable skill behind a celebrated career includes the ability to make the work known. If that is right, then “just do good science and it will speak for itself” is not merely incomplete advice. It is the one thing the laureates apparently did not believe. Did yours?