What do you include in your graduate student letters of recommendation?

[rabbitandfox23]

Hi everyone,
I'm a junior faculty member in a Humanities department that has a PhD program. I'm now on the committee of two students and both have recently approached me for letters for their applications for predoctoral fellowships. I've written letters for undergraduates applying for graduate study as well as letters of support for tenure files, but I have never written letters for grad fellowships.
What do you think are the crucial components to a good letter for fellowship competitions? Should I focus mainly on the student's PhD research?  Should I describe the research and explain how it will advance the field? Or should I also spend time talking about his/her performance in graduate seminars and general service and contributions to the department? Or would such topics be considered irrelevant to the opportunity at hand?  Do you usually compare the promise of this student to other graduate students in his/her cohort?
And how long are these letters? 1-2 pages? Closer to 1 page?
Thanks in advance for your thoughts!

[AJ_Katz]

By the end of a graduate student's career, I usually can't fit it all onto one page, although I try to make just one page, when possible.  The content of the letter depends on the focus / interests of the student and what kind of career or fellowship they're applying for.  If it's a teaching fellowship, I'll talk about their mentoring and teaching impacts first and foremost, while still including a brief mention of their impact of research to demonstrate that they are not one-dimensional. 

In these letters, I usually do not talk about how the student compares to others.  I will instead highlight the awards that they've received and describe how selective each award is (for example, given to just one graduate student per year?  If so, that's a big deal). 

So, usually the structure of my letters is:
Paragraph one: who am I and what is my relationship to the person and for how long have I known them
Paragraph two:  description of their impact in teaching or research (depending on the award / job they're applying for) that uses citations to their published papers, awards received, or other works.
Paragraph three (optional): description of secondary area of impact, such as teaching, mentoring, or outreach / service
Paragraph four: This is the closing paragraph, which allows me to add a personal anecdote about the person, how they work or think as a scientist that is noteworthy OR a description of their impact on their profession or lab group community that goes beyond just the science / teaching.  Of course, at the end, I emphasize in one sentence that all of these elements collectively tell me this person is an outstanding candidate for the fellowship / job and, if I did want to compare the student to others, I would add that information here. 

Once you get the structure for these letters down, it should be pretty easy to go back to a letter and revise it for each person.  If you haven't seen it before, there is also an online tool for checking gender bias in letters of recommendation.  I highly recommend trying it out when you are first figuring out how to craft a letter of recommendation:  https://slowe.github.io/genderbias/

[mleok]

Once you get the structure for these letters down, it should be pretty easy to go back to a letter and revise it for each person.  If you haven't seen it before, there is also an online tool for checking gender bias in letters of recommendation.  I highly recommend trying it out when you are first figuring out how to craft a letter of recommendation:  https://slowe.github.io/genderbias/

Interestingly, when I checked two letters I wrote for former postdocs, the one I wrote for the female postdoc was 48% male-biased, compared to 38% male-biased for the male postdoc. What should I infer from these results?

[Puget]

Once you get the structure for these letters down, it should be pretty easy to go back to a letter and revise it for each person.  If you haven't seen it before, there is also an online tool for checking gender bias in letters of recommendation.  I highly recommend trying it out when you are first figuring out how to craft a letter of recommendation:  https://slowe.github.io/genderbias/

Interestingly, when I checked two letters I wrote for former postdocs, the one I wrote for the female postdoc was 48% male-biased, compared to 38% male-biased for the male postdoc. What should I infer from these results?

Same-- just tried it on letters for a female and male grad student, and the one for the female grade student was 32% male biased whereas the one for the male grad student was 13% female biased. The reason? The letter for the female student was for a postdoc and focused on research, the letter for the male student was for a position at a SLAC and focused on teaching and mentoring. So, this may work when they are for the same jobs, but the terms it is picking up on basically seem to be the difference between emphasizing research vs. teaching and mentoring.

[nonsensical]

Is this a fellowship where the applicant proposes to conduct a specific project under your mentorship? If so, in addition to the more general things others have mentioned, it may be useful to write about the strengths of the project and how it would contribute to the field, and to discuss how you will mentor the applicant. The fellowship itself may also have some instructions about what they want recommenders to include in their letters.

[Descartes]

What does this gender bias calculator actually measure?  Does it only work for letters of recommendation and the like, or any writing?

I was curious about it, so I ran about 4 of my own e-mails that I had authored through it.  One was 100% female biased, one 100% male biased, and two neutral.  What does that mean?

[Puget]

What does this gender bias calculator actually measure?  Does it only work for letters of recommendation and the like, or any writing?

I was curious about it, so I ran about 4 of my own e-mails that I had authored through it.  One was 100% female biased, one 100% male biased, and two neutral.  What does that mean?

I don't think it is meant for anything other than letters of rec., and specifically those for research-focused positions.  It is meant to pick up on the fact that (at least in the sciences) there are known biases where letters for women tend to talk more about being a good mentor, teacher, lab citizen, team player etc. and less about research skills, accomplishments, and intellect. It is just looking for use of key words (which it shows you). It probably works pretty well as an alert to accidentally doing this if you just use it for letters for postdocs and research TT positions, but as I discovered it will flag letters for teaching-focused positions when they appropriately emphasize teaching and mentoring (mine happened to be for a male grad student).

[Kron3007]

I usually start with who I am and my relationship with the student (how/why I recruited them, how long we have worked together, etc.).  This paragraph helps to highlight their background before we met (ie. I hired them based on their exceptional grades, and an interview in which they demonstrated their enthusiasm for XXXXX (hopefully relevant for the fellowship)).

The next paragraph is usually about their performance since joining my lab.  In there, I would point to awards, number of publications, etc., but I dont get too specific since it is all listed on their CV.  On this note, I try to emphasize things that may not come out in their CV (helpful around the lab, help organize lab BBQs, helping with conference prep in an unofficial capacity, etc).  The specifics obviously depend on the student.

In the final paragraph I usually go a little more forward thinking and highlight why their background and interests make them a good fit for this specific fellowship.

Once I have this, I just edit it for different purposes.  I'm not saying this is the best way to write one, but that is what I do. 

[Kron3007]

What does this gender bias calculator actually measure?  Does it only work for letters of recommendation and the like, or any writing?

I was curious about it, so I ran about 4 of my own e-mails that I had authored through it.  One was 100% female biased, one 100% male biased, and two neutral.  What does that mean?

Regarding that link, I think it is complete bunk.  I ran a couple through and they were both male biased (one was a male student, the other was a female), but when you look at the "male associated" and "female-associated" words it is just silly.  Publication is male?  This is one of the main outputs and needs to be discussed.  Likewise, scholarship is a male biased word?  If I am writing a letter for a student and they have won a scholarship, that will be highlighted.

Nice try I suppose, but this link is bunk...

Just ran another that was neutral.  I think the trend that I am seeing is that my undergrad references tend to come out more female biased and it shifts more male as the student advances.  If anything, this calculator is gender biased.