Ninety-four distinct measures of condom use were identified in a review of 72 psychosocial studies of HIV-prevention. A coding scheme was developed which characterised each measure in terms of: (1) type of measure (e.g. frequency of use, use at last intercourse), (2) response alternatives (e.g. Likert scale), (3) recall period (e.g. lifetime, last month), (4) partner specification (e.g. 'new' partner), and (5) activity specification (e.g. vaginal or anal intercourse). Results show a great deal of inconsistency in measurements of condom use across studies. Single-item frequency or consistency measures were most common. Several difficulties with measures employed to date were identified. These included indeterminate reliability and validity of measures, lack of attention to the epidemiological significance of different types of use and terminological and self-presentational problems with items assessing use. Implications of extant measurement approaches for the accumulated knowledge base on predictors of 'safer' sex and suggestions for future research in this area are outlined.