Those interested in Wikipedia are discussing the comparison of errors appearing in a sample of articles, reported in by Nature, of 42 article. While I agree with Jakob Voss's comments on the limitations of the study, for this sample the amount of errors does seem roughly comparable with Wikipedia -- hopefully that Wikipedia outlier for Dmitri Mendeleyev will be fixed soon. I was further intrigued to note that the errors per topic correlate between the two:

This is a strong correlation (r=0.574) implying perhaps a similarity in the difficulty of writing on that topic, or perhaps a difference in scrutiny by the experts (e.g., the person reviewing the Cambrian explosion is picky!).