The reliability of classification of cerebral palsy (CP) in low-birthweight children was assessed by using clinical and research study records sampled from population-based cohort studies in the USA, the Netherlands, Canada, and Germany. Records of neurological examination findings and functional motor assessments were submitted to up to five pediatricians with expertize in CP diagnosis, who grouped children into categories referred to as ‘disabling’ CP, ‘non-disabling’ CP, and no CP. Each study provided between 31 and 51 records of children assessed between 2 and 8 years of age, approximately equally divided among the three groupings. The discrimination between ‘any CP’ and ‘no CP’ was only fair (mean Kappa coefficients 0.37 to 0.69). However, when more detailed information describing motor function was used, children with ‘disabling’ CP could be distinguished, on the basis of records, from those without CP or with ‘non-disabling’ CP with good to excellent reliability (mean Kappa coefficients 0.69 to 0.88). Because of the substantially higher agreement observed when these functional distinctions are made, we recommend that reports or comparisons of rates of CP should include levels of motor function of children with CP, and not simply total CP, among the outcomes of interest.