The common disease asthma is probably not a single disease, but rather a complex of multiple, separate syndromes that overlap. Although clinicians have recognised these different phenotypes for many years, they have remained poorly characterised, with little known about the underlying pathobiology contributing to them. Development of targeted therapies for asthma, and phenotype-specific clinical trials have raised interest in these phenotypes. Improved understanding of these phenotypes in complex diseases such as asthma will also improve our ability to link specific genotypes to their associated disease, which should help development of biomarkers. However, there is no standardised method to define asthma phenotypes. This Review analyses some of the methods that have been used to define asthma phenotypes and proposes an integrated method of classification to improve our understanding of these phenotypes.