Ecologists have long assumed that a more diverse plant community is a more productive one, and field surveys seem to bear this out: species-rich plots yield more biomass than species-poor ones. A confound lurks in the correlation. Diverse plots are more likely, by chance alone, to contain at least one highly productive species, and that single species may drive the extra yield—an effect of sampling, not of diversity as such. Experiments that plant fixed numbers of species while varying which ones are present let researchers separate the two. Where they have done so, diversity retains an independent effect on productivity, but a smaller one than the raw surveys suggested.