Another post from the QQML conference. Carol Tenopir (with a presentation coauthored with prject partnered) reported on an international study which is investigating researchers' behaviours with sharing articles etc. They have undertaken focus groups in the USA and the UK and administered a questionnaire survey in various countries. They have discovered that people are sharing both informally and informally, in all sorts of ways e.g. via email, reference management software, printing things out. 16.7 full text articles and 8.6 links/references were on average shared by respondents (asked what they had shared in their last project). The mean number of articles downloaded per project were 65 (with a range between zero and 500, so working out percentage of those shared based just on average items shared seems too simplistic). Just over half of the downlaods came from library collections. Out of the works shared (whether their own work or others) the large majority were of the published version of the article.
There were some differences based on age e.g. people under 30 and over 60 shared less. Also there were some disciplinary differences e.g. engineers shared more via email and through other informal methods, social scientists shared a lot and shared more on social media. The project website is http://libraryconnect.elsevier.com/beyond-downloads and Carol Tenopir's publications on Researchgate are here: https://www.researchgate.net/profile/Carol_Tenopir/publications
There is, for example, a paper about this research here (so here I am, as a social science researcher, sharing links informally....)