I just came across this info in a book that I am reading (Irene Hames, Peer Review and Manuscript Management in Scientific Journals, Blackwell, Oxford 2007): In 2002 some 3000 or so scientists funded by US taxpayers through the US National Institutes of Health were interviewed (anonymously) about professional misconduct related to their scientific research. They were asked to report misbehavior during the preceding 3 years, and the only scientists interviewed were those in early or mid-career. One out of every three such scientists admitted some form of misbehavior 'in the top-10 most serious categories during the previous 3 years.' (pg 175) 6% of those interviewed failed to represent data that contradicted their hypothesis, 5% had published the same data already elsewhere (redundant publication), 10% assigned authorship inappropriately (eg to people who w ere not involved in the research), etc.
Based on my experience these problems start well at undergraduate level, because universities don't bother hard enough to try to stamp out academic misconduct. No dramatic surprise then that there's a substantial minority of scientists who think that such conduct is, if not acceptable, but certainly no big deal. What the data I have just reported also reveal is that folks usually get away with such conduct and that only a small minority get caught ... bad news all round.