Carl Zimmer is up on the web today at The New York Times, writing about a study that challenges the "comforting assumption" that scientific retractions are mostly due to honest error.
The study's authors analyzed 2,047 retracted papers in biomedicine and the life sciences and concluded that
fraud and suspected fraud were misconduct was behind three-quarters of the retractions for which they could determine the cause.
This is indeed discomfiting news. You might think that science preferentially draws people who are honest and curious, and who are smart enough to recognize that cheating in science is a bad bet. Not every case of scientific cheating or fraud is uncovered, but plenty of them are. This study and Zimmer's account remind us that while the truth may be out there, we should trust no one.
[Update: I borrowed the description of the study from Zimmer's story. He emailed me to say that the study found that "misconduct," rather than "fraud and suspected fraud," were behind three-quarters of the retractions for which the researchers could determine a cause. Hence the change. The abstract of the study is here.]