At present, frailty is defined variably. Some consensus on a definition is likely to emerge, but the basis for a successful definition needs to be explored. Here, a classic approach to validation is proposed: a successful definition of frailty should be multifactorial but must also manage the many factors in a way that takes their interactions into account. It is likely to be correlated with disability, co-morbidity and self-rated health, and should identify a group that is vulnerable to adverse outcomes. Ideally, it should also be susceptible to animal modelling. In that frailty and age are so bound together, it is also likely that there will be some age at which virtually all people will be frail, by any definition. Apart from being valid, the success of any definition of frailty will depend on it being useful to researchers and clinicians. The need for progress on our understanding of frailty is evident, but for now, there is insufficient evidence to accept a single definition of frailty.