I was somewhat disappointed in the 2003 edition of this collection, especially after enjoying so much the last three editions. Perhaps Oliver Sacks and I disagree on what makes science writing good. The two best articles were 1491 by Charles Mann about what the Americas were like before Columbus landed and the effect of the European arrival:
In North America, Indian torches had their biggest impact on the Midwestern prairie, much or most of which was created and maintained by fire. Millennia of exuberant burning shaped the plains into vast buffalo farms. When Indian societies disintegrated, forest invaded savannah in Wisconsin, Illinois, Kansas, Nebraska, and the Texas Hill Country. Is it possible that the Indians changed the Americas more than the invading Europeans did? "The answer is probably yes for most regions for the next 250 years or so" after Columbus, William Denevan wrote, "and for some regions right up to the present time."
and Atul Gawande's The Learning Curve, an article on how doctors need to learn on the job (while potentially making costly mistakes) in order to become more effective overall:
In medicine, there has long been a conflict betwenn the imperative to give patients the best possible care and the need to provide novices with expericne. Residencies attempt to mitigate potential harm through supervision and graduated responsibility. And there is reason to think that patients actually benefit from teaching. But there is no avoiding those first few unsteady times a young physician tries to put in a central line, removes a breast cancer, or sew together two segments of colon. No matter how many protections are in place, on average these cases go less well with the novice than with someone experienced.