As Mark recently mentioned, we are deep in recommendation letter season. I've been in the biz long enough that I've probably written at least a hundred letters (estimating more than ten a year for more than a decade), and read far more than that. After you read enough letters, they can start blend together. But, in a big stack of applications, there are usually a few letters that stand out as risible, causing a good chuckle and round of comment from the committee. And they are almost always letters written on behalf of women. In a standard letter of recommendation at the postdoc/faculty level, there is frequently a comparison to other successful scientists. The letter usually reads something like "reminds me of person X, Y, or Z at a similar level of their career" or "shows the same persistence and insight as person Q, and stronger big picture thinking than person P". These comparisons are almost always favorable, saying that the applicant is in the same league as other people who are recognized as having had a significant scientific impact. But, for some reason, some fraction of letter writers insist upon doing these comparisons only within a single gender, when the applicant is a woman. In other words, "(woman) X shows a similar level of insight as (woman) Y and (woman) Z". I'm not saying that these comparisons are not favorable -- they're usually comparing a strong female applicant favorably with other successful female scientists. Their praise is genuine and well meant. However, one can't but help perceive that they see women as somehow swimming in a different pool than the rest of the guys. Now the good news is that most committees that I've been on have seen right through this. We note it, and have a small laugh at the letter writer's expense. In addition, it's not common -- usually only affecting a couple of letters in an applicant pool. So, if you're writing a letter for someone in an underrepresented group, please save yourself from mockery by examining exactly how you perceive the applicant's comparison sample.