In a way, the letter of recommendation is the good-old-boy-club side of a portfolio. Maybe a kinder way of stating this is that it is the human side of a person’s file. After all, the data that should be most crucial for the next step lie elsewhere. Essentially, the letter of recommendation tells evaluators when to ignore the data. By data I mean grades, test scores, publications, funding, and other sorts of verifiable things that belong in a CV. The letters are both a personal vouching for someone and an interpretation of the data.
Just by writing the letter, you validate the person and their application. The way it works in the US these days, if your letter cannot be positive, then you should not agree to write it. But that does not mean all letters are the same, by any means. I know at least one very famous person in my field that writes letters of one or two sentences. They are positive, without true content, and only fulfill the stamp-of-approval side of letters. This person writes enough of those, that even that is of little value. But I sympathize. I imagine this person is inundated with letter requests, making this a reasonable compromise.
Besides validation, one of the things people want is help interpreting the data, as when we are instructed to tell whether we think that a student’s grades are not reflective of their true abilities, perhaps because of an outside job, a sport, or an ill relative needing care. If the stipend offered by a graduate program lifts the need to work, the reasoning goes, the person will perform much better. If you have information on this, share it.
Another kind of data interpretation is that surrounding multiple author papers. It may be hard for evaluators to tell what a certain person did on a project. Tell them.
More importantly, evaluators want to know who this person is, and what they will be like should they be hired, or admitted. This is very difficult to convey. I think the best way to do this is by anecdote, since the data are elsewhere. If you can tell a few short stories that capture the person in a number of different dimensions, do so. It is good if these stories can be interesting and show the person in a variety of ways, particularly involving important things like coming up with new ideas, showing leadership by organizing something new, or helping others.
Be very careful that you don’t fall into the trap of describing women and men of equivalent abilities differently. Research says you probably do this, so be careful. I have seen things in letters that would dumbfound you. The study by Hebl, Martin, and Madera linked to above indicated that people rate letters lower if they have tons of communal characteristics. So make it clear your letters show strength and action for all your students.
I write three kinds of recommendation letters, those for undergraduates who may have taken a course or two from me, but did not work in my lab, those for high school students, undergraduates, grad students, and postdocs who did work in my lab, and those for colleagues whose work I know and whom I’m also likely to know personally. The three kinds of letters are different. I’ll try to give more detail on each kind in later entries. This is just an overview.
For the student that I know from class and did not work in my lab, I write a short letter that still tries to capture them. Since my classes all involve papers or projects, I can talk about their project, how they approached it, what its strengths were, how much teamwork they showed. That gives some fun information, and sets the student apart. It also gives a framework for the praise you will want to give the student.
The other two kinds of letters will have much more information. They should have several paragraphs about the actual work, giving what is most important, perhaps pointing to a key paper or two. Put the work in a broader context. Teach the reader why it is important, since they may well be from a different specialty. This is especially true of tenure letters that can go to committees that are very far from the specialty. I’ve often heard that a good letter explains why someone’s work is important and what it is all about better than the actual subject does.
I often ask people for whom I write letters to provide me with a few paragraphs on their best work, why it is important, and what special challenges they had in doing the work. I may or may not use the information they give me, but it can be very helpful.
In general, adjectives are not helpful in letters, except maybe in the first paragraph. Even though the letter is a stamp of approval, the readers also want a little meat. This is not the place for generic word-salad or kind-to-pets letters that could be written about anyone.
My standard letter formula begins with a short summary paragraph. The second paragraph tells exactly how I know the person. I try to put in the date of first meeting, or classes etc. This helps in future years more than relative times like “last year.” The next one to five paragraphs are the meat and consist of the anecdotes, interpretations of the research, and special information that helps interpret the data. The concluding paragraph is the endorsement: Hire this person; interview this person; admit this person; tenure this person, or something weaker, “enthusiastically”, “most highly”, and the like.
I do not put negative information in letters, because it is so hard to know how it will be read. Is someone who is stated to be “slightly shy” actually pathologically withdrawn? It is hard to know. But I do not put untruths. If someone is not at all creative, needing to be told exactly what to do, I will not say they are creative. I think a lot of other people write letters this way, so, if you are an evaluator, scan them for what people don’t say, as well as for what they do say. I say scan, because most letters are only scanned quickly. Only if a hard decision needs to be made, or someone is about to be interviewed or hired in a highly competitive search, are letters read closely. Letter-writing conventions vary with the country, so be sure when you read them you take this into account.
Whenever possible, I give the letters to my assistant and let her send them out. I forward the requests to her, and she handles them. I do not personalize letters in detail for each application, though if I have a student applying for research jobs and teaching jobs, I’ll have two different letters. If other specialization is needed, my assistant alerts me. If I know someone well at a place one of my people has applied, I might send a brief email to that person, alerting them of the opportunity.
Writing these letters can take too much time. I write letters for a lot of people: 28 in 2011, 53 in 2010, 50 in 2009. I try to do them quickly, in scraps of time between other tasks. I like to do them when asked, so they don’t hang over me. I estimate that a new letter for someone takes between 20 minutes and 2 hours, no more. I might spend longer reading papers if they are interesting enough.
Oh, another thing. I do not reiterate the data. I do not repeat grade points, numbers of publications, dates of degrees or any of that. I don’t see the point. Maybe I should do this.
There are a number of places you can get more specific advice on letter writing. Here’s one on getting a great letter. Here’s one from someone who reads a lot of letters quickly. Here is the standard kind of stuff they tell undergraduates. I would say that you should email your professors copies of any papers or projects you did for their classes, but the other stuff can also be useful.
What we want to hear is that this is a great person, full of brilliance and grit, hard working, but always taking time to help others. Let such people shine. But also help the ones that are not so stellar socially, but may be even more brilliant. OK, let’s be honest, even the non-brilliant, honest, fairly hard working, fairly smart people need those letters too.
Oh, I should say something about those horrible buttons where you have to say if someone is the best in 10 years, best ever, or mediocre in each of 20 different ways. I hate them. I never know what to put. They are inflated, but how much? When I look at results of these, I pay more attention to relative than absolute values. If someone is high in everything except writing, then that tells me something. Maybe. I just fill out quite high values on these things. They are asking us to quantify something that can’t be quantified, and I resent it. For this, look at the data. For opinions, read the letters.
As a final note, I’ll say that letters are much more important for people still living with parents who may squeeze performance out of uninterested children. There strong school support in the form of letters can be extremely informative.
For better or worse, these letters are a part of our system. I know plenty of people have taken the time to write letters for me, using their precious time. All I can say is thank you. Maybe I can return the favor, or at least pass it on.
– Madera, J., Hebl, M., & Martin, R. (2009). Gender and letters of recommendation for academics: Agentic and communal differences. Journal of Applied Psychology. 94, 1591-1599