A commentary in Nature reports a survey of the actual behaviour of scientists. Several thousand NIH funded scientists were surveyed anonymously and asked to report which of a list of questionable practices they had themselves engaged in. The results are presented in the commentary - practices reported included falsifying data, ignoring the requirements of human subjects, and not disclosing the involvement of commercial concerns.
From a library point of view, interesting sins included publishing the same data in more than one publication, assigning authorship credit where none was due and using someone else's ideas without credit.
There is an interesting letter in Nature for 11th August, which argues that acceptable practices differ between basic and clinical research, and that some of this reported misbehaviour might not actually be misbehaviour. The original study surveyed basic and clinical scientists.