Nature News features a write-up on patterns of article retraction by high- and low-impact journals. Murat Cokol and his colleagues at Columbia (M. Cokol et al. EMBO Rep. 5, 422–423; 2007) identified 596 retracted articles (out of 9.4 million articles published from 1950-2004 per PubMed) and noted that journals with high impact factors were more likely to retract papers than low-impact journals. The authors suggest that high- and low-impact journals have the same batting average in detecting flawed articles before they are published.
The Nature piece goes on to provide the sort of informal peer review not given to the EMBO correspondence. Drummond Rennie, deputy editor of JAMA, and Sandra Titus, director of intramural research at ORI, point out a number of flaws, such as the introduction of scientific misconduct policies only within the last 20 years (while the Cokol sample goes back to 1950) and the need to retract other papers by an author once one instance of misconduct is identified. They also comment on the vagaries of impact factors and legal barriers to retraction.