Measurements of final height were made on more than 4000 children in the final grade of school in Gothenburg in 1992; at the same time, mid-parental heights were recorded. These data were combined with other information (sex, length of gestation, size at birth, estimated age at peak height velocity (PHV) and height measurements made before 8 years of age) and used in a multiple regression analysis to assess the strength of the linear relationship between attained final height and these other potentially predictive measures of adult height. The R2 value increased from 0.16 at birth to 0.64 when the child was 8 years old. The inclusion of mid-parental height in the regression analysis contributed significantly to the explained variation in final height, especially at the earlier ages; the further addition of size at birth and age at PHV provides a small increase in the explained variation. The probability that the final height of a child will be below -2 or above +2 standard deviation scores (SDS) was assessed, based on previous SDS values for height when younger and on mid-parental height SDS. As a result of the large sample size included in the analyses, considerable confidence can be placed on the accurate prediction of final height values in the range -2.5 to +2.5 SDS.