We are entering the season of recommendation letters. If the applicant makes it to the final stage, the letters are scrutinized and every word said or unsaid gets interpreted. How does one go about writing a recommendation letter? I am not very good at it, but here is my usual template:
In the first sentence, I say where I know the person from and how long I have known them. Then I summarize our joint work, if there is any.
If it is about a student who took my course, in the second paragraph I recall his or her academic performance, highlighting if there was anything special, any interesting remark that came up in class for example.
If it is about someone who I did research with, in the second and third paragraphs I talk about the technical content of our work, giving my own assessment of how deep or interesting the results are, and mentioning any particular role that might have been played by the collaborator. I make sure to insert some reservation of some kind, because no one is perfect, and a letter is not believable if it does not contain a negative. The compliments are more compelling if there is something to balance them. If necessary, I add a comparison to specific people who have applied for similar positions.
If it is an application for a position that has some teaching, I add a sentence or two about teaching.
Then I insert a few words about the person's personality.
In my final sentence, I give my overall opinion, which I try to make as multidimensional as I can, so that each person is best possible in his or her own dimension and will be sought after by the places that are looking for exactly this kind of person.
After that initial draft, depending on how much time I have, I polish the letter: I reconsider the choice of words in the final sentence. I look up some research papers or documents to refresh my mind on the most interesting features of the work, and I develop the technical paragraphs a little bit. I search for adjectives that are an evaluation, and look for errors in emphasis. I remove expressions such as "it seems that", "I believe that", etc. Writing "this work is deep" is much more forceful than "I believe that this work is deep". Obviously, if I write it, it's because I believe it, so why add "I believe"? Pointless. I look at the relative balance between the various parts. For example, it's not good to be too effusive about somebody's personality, because then in comparison it makes the rest of the letter look bad. It's better to put the reservation, whatever it may be, in the middle of the technical evaluation rather than first or last. I may ask a colleague to take a look and tell me the impression conveyed by my letter. Sometimes it is quite different from what I had intended! Sometimes I may add some comment about the place where the applicant is applying, and the suitability of the place for the person or vice versa (but that has to be tailored to each place, so mostly I do not bother). The one thing I never do is polish the style of the English to remove ambiguous antecedents, to correct mismatches in number, or to fix faulty co-references.
The most important thing about writing a good recommendation letter has nothing to do with the applicant: it is that I have to be in a good mood. If I am in a good mood, good words will come to my mind, upbeat expressions, accurate but positive turns of phrase.
The next most important thing also has nothing to do with the applicant: it is that good letters take time to write. (At least an hour, but easily half a day for more senior applicants.) Usually, the first time I write a letter for someone, I am pressed for time and it is not that great; by the second or third iteration I have had some time to polish it, and it is a much better description of the person's strengths.