Population-based studies in twins have been of insufficient size to explore the relationship between risk of cerebral palsy and intrauterine growth. Earlier studies in singletons have suggested an optimum size at birth for minimum cerebral palsy risk between the 75th and 90th percentiles of weight for gestational age. We aggregated data from nine European cerebral palsy registers for 1976 to 1990. Using sex-specific fetal growth standards for twins, a z score of weight-for-gestation was derived for each of the 373 twin cases. The rates of cerebral palsy in each z-score band were compared to the rate in the a priori reference band of 0.67 to less than 1.28 (equivalent to the 75th to less than 90th percentiles). In twins born at 32 weeks' gestation or more (92% of all twins), cerebral palsy rates were higher for both light and heavy-for-gestation babies compared to an optimum (i.e., minimum risk) in the reference band. However, the rate ratio for heavy babies (90th percentile or greater) did not reach conventional (95% confidence intervals [CI]) statistical significance (rate ratios = 1.76; 90% CI 1.02-3.03). For twins born at less than 32 weeks, the significantly higher risk for cerebral palsy was observed consistently in all z-score bands less than average compared to the reference band. This multi-center study demonstrates that for twins born at 32 weeks' gestation or more, an increased risk of cerebral palsy is associated with deviations from optimal intrauterine growth at about 1 standard deviation above mean weight, as was earlier reported for singletons. For twins born at less than 32 weeks' gestation, this pattern is only demonstrable for babies weighing below the optimum weight-for-gestation.