Lawrence’s study using articles in computer science, comparing articles from the same venue (mostly refereed conference proceedings, which is the main form of publication in computer science) and showing that the ones that were self-archived had 300% more citations than the ones in the same venue that were not self-archived.