This weekend I spent a lot of time writing letters of recommendation for people in various stages of their careers: undergrads and master’s students applying for PhD programs, PhD students applying for postdoctoral opportunities and faculty jobs, and postdocs applying for faculty positions. As a good procrastinator, I spent a lot of time reflecting on what my theory for letter writing should be—and even wondered what the heck letters of recommendation are anyway.
It turns out you can trace the history of recommendation letters back as far as you want, as Wikipedia puts it: “in the New Testament, Saint Paul discusses whether he needs a letter of recommendation to commend his ministry to the church in Corinth.” Apparently, Romans’ letters were a genre of their own, litterae commendaticiae, and they praised the writer as much as the candidate the letter supported. One key tension around letter writing dates back at least to Cicero, who feared that “because I am recommending so many people to you with special emphasis, I may seem, out of a kind of eagerness to please, to be putting all my recommendations on the same level.”
Some more recent historical letters of recommendation impress by their brevity. This is Charles Darwin’s recommendation to Thomas Huxley, for a position as a Chair of Natural History at UofT (a position Huxley ultimately didn’t get):
Dear Sir, I have much pleasure in expressing my opinion, from the high character of your published contributions to Science, and from the course of your studies during your long Voyage, that you are excellently qualified for a Professorship in Natural History.— You have my best wishes for success in your present application.
That’s it. Having this short letter of recommendation feels impressive somehow. Someone once told me this story about receiving a letter from a Russian institute that went something like this: “Ivan Ivanovich (the Russian John Doe) is a good student. I have given him an unsolved conjecture from the 50s. He solved it. It was not a very hard problem, though. I’m sure he will do well in your program.” According to my undisclosed friend, this was the strongest letter of recommendation he had seen that year. He may or may not have framed it on his wall (an anonymized version, of course!).
So now that you’ve noticed I spent way too much time on Wikipedia, let me give my best take on letter writing. Academic recommendation letters are an institutional ritual through which more senior academics lend their credibility and reputation to support the applications of more junior individuals. They complement a candidate’s CV by providing context, qualifying accomplishments, offering subjective assessments, and revealing qualities that are otherwise difficult to observe.
In practice, modern letters of recommendation are a cascading series of claims, each of which should hopefully be supported by convincing evidence. They typically begin with an overarching endorsement, followed by increasingly specific claims that justify it. Clichés include personal assessments of intellectual ability, contribution, and character. Oftentimes, a good writer will also compare the candidate with other relevant peers and predict their future success.
Evidence provided includes factual evidence (they did X, which is verifiable in their CV) and anecdotal evidence (they once did this very impressive thing in my lab). The latter is most relevant, since most factual evidence can be read from the candidates’ vitae. Confusingly, personal judgments function as both claims (“She is brilliant”) and evidence (“she’s the strongest student I’ve advised in ten years”, a report only the writer can make). The best evidence of all is “costly” to send (you are using a superlative you can only spend so often).
As a letter writer, two central tensions emerge. First, letter writing is a repeated game. People have an incentive to write strong letters that benefit candidates; it benefits the letter writer when your mentees succeed! But on the other hand, your reputation as a letter writer is on the line, and if you inflate your letters today, tomorrow’s letter will carry less weight. Second, letter writing is a genre full of clichés (my strongest possible recommendation!), which help people interpret and compare candidates. But at the same time, your goal as a letter-writer is to make your letters stand out.
Good letter writing takes tremendous effort, which peers and academics, especially outside the US, often ignore. It requires balancing three contextual factors. First, you must consider the candidate you are writing the letter for. This is your one opportunity to tell something about this candidate that isn’t captured in their CV --- what is the most relevant thing you can tell others about them? Second, it depends on the opportunity! The position the candidate is pursuing. Maybe there’s a broad job market context: everyone is hiring for AI, and you must convince them that the candidate can do that, too! Third, it depends on who you are, both because your expertise will shape the credibility of what you say and because your relationship with the candidate will shape how you write the letter.
So here’s my theory of letter-writing: write letters so the strength of your claims is proportional to the evidence supporting them and the credibility you are willing to stake. A good letter writer should therefore spend a lot of time gathering evidence about people you are trying to support. This, in my naïve assistant-professor viewpoint, offers a solution to both tensions. You can be a strong advocate because your superlatives are always backed by something you can report. But also, you can keep the conventions in the key claims and let the evidence do the differentiating. One small bit also encodes a part of the letter-writing that hasn’t been fully fleshed out yet: the one about costly signals
Rereading some of the strongest letters I’ve come across makes me somewhat confident in my “formula.” They tend to tell you a few good stories rather than pile on adjectives. Which brings me back to Ivan Ivanovich, the code golf of strong letters of recommendation, with one claim (he is a good student), one piece of evidence (he solved the conjecture), and nicely framed within an engaging story (that the advisor was a hard-ass).