SUMMARYA survey of 100 species and 122 observations has shown an average reduction in stomatal density of 14-3% (sE + 2-2 %) with COj enrichment, with 74 % of the cases exhibiting a reduction in stomatal density. A sign test demonstrated that stomatal density decreases, in response to CO^, significantly more often than expected by chance. Repeated observations on the same species indicated a significant repeatability in the direction of the stomatal response. Analyses which removed tbe potential effect of taxonomy on this data set showed no significant patterns in the dependency of the degree of stomatal change on growth form (woodiness vs. non-woodiness; trees vs. shrubs), habitat (cool vs. warm) or stomata] distribution on the leaf (amphi-vs. hypostomatous).Forty-three of the observations had been made in controlled environments and under a typical range in COê nrichment of 350-700 //mol mor'. For these cases the average stomatal density declined by 9"o (SE ± 3-3 %) and 60 "o of the cases showed reductions in stomatal density. When analyses were restricted to these 43 observations, amphistomatous samples more frequently bad greater changes in stomatal density than did hypostomatous samples.The degree of reduction in stomatal density with increasing CO^ increases with initial stomatal density, after the influence of taxonomy is removed using analyses of independent contrasts. When tbe data were examined for patterns that might be due explicitly to the effects of relatedness, the subclasses of the Hamamelidae and the Rosidae showed highly significant reductions in stomatal density with CO^ (87 "o of the species studied in the Hamamelidae and 80",, of the species in tbe Rosidae showed reduction with CO^ enrichment) and correlations between initial stomatal density and degree of reduction in stomatal density. The species sampled in the Hamamelidae were dominantly trees, whereas herbs dominated tbe species in the Rosidae. There were insufficient species studied at lower taxonomic levels to warrant further statistical analyses. This problem results from experimental and observational data being most often restricted to one species per taxonomic level, typically up to the level of order, a feature which can severely limit tbe extraction of taxonomicaliy-related and ecologicallyrelated plant responses.