[bjourne]

Well, then you can go and prove that. :) Because the lack of evidence in either direction (wrt cognitive load and commit messages) points us towards the null hypothesis. In my own experience, fixing bugs in C code long abandoned by the original authors, it doesn't matter one bit if the commit message reads like a novel or literally just "fsjdlkfdjlsfl fskjl". All that matters is the change is self-contained and references bug tracker ids and mailing list discussions.

Also in my experience, passive-aggressive power trippers love anal retentive rules. Hence adding more than necessary carries a significant risk.