For the most part, it is a pretty big deal for a journal to retract a scientific article after it has been published. In doing so, either the authors of the manuscript or the editors of the journal are admitting significant breakdown in some portion of the scientific method and/or publication process. Sometimes, the retraction is because a manuscript is sent to and accepted by more than one journal at the same time. This is never supposed to be done, but some do it anyway, either as a mistake or as an attempt to get more exposure for the same amount of work. Sometimes, someone realizes that a significant error was made in data collection or processing, rendering the conclusions invalid. The worst though, are the instances where true wrongdoing or fraud has occurred.
In an ideal world, peer review would prevent this. Peer review is a process whereby journal editors provide manuscripts to the author's peers, scientists in the same fields of study, who provide the editors with an unbiased review of the quality and scientific accuracy of the manuscript. Having reviewed manuscripts for about a dozen or so journals, I can say that I have certainly read my fair share of manuscripts that I consider to be flawed in design or interpretation that need to be seriously reconsidered for publication. Unfortunately, however, that opinion only gets shared with the editors of the journal the article was submitted to and with the authors themselves. If the article is not accepted, the authors are free to resubmit somewhere else. Hopefully reviewers for other journals would draw the same conclusions, but they may not have the same experiences as me, and clearly I lack many experiences that others have. Furthermore, journals depend on reviewers doing work gratis, and therefore are limited by those willing to do the work. You can imagine, then, how some substandard manuscripts manage to make it through peer review from time to time.
I digress, but the point of this post is to highlight a blog that covers this topic by scanning journals for retractions and posting them for all to see. RetractionWatch tracks retractions but also provides background info on some of the more egregious offenders. It has received some excellent reviews since its inception only one year ago, and I have added it to my RSS feeds for regular updates.