The effectiveness of the use of objective data available at birth to predict and prevent child abuse was assessed in a group of 2802 non-Asian infants born in Bradford in 1979. Two-thirds of the actual abuse occurred in the 18% predicted to be at risk, but supportive measures did not prevent abuse. Those who needed and received the most attention from social workers and health visitors fared worst. Other indices suggested poor parenting in the at-risk families. Child abuse seems to be predictable but prevention is more difficult.