Estimation of premorbid intelligence is often required in the context of testing adults as well as children. However, no prior study has yet developed a method of estimating premorbid intelligence in children. The aim of the present study was to develop a method of estimating premorbid intelligence based on school records. With this aim, we administered KEDI-WISC to 130 children. There was a significant correlation (.725) between FIQ and school grade. Based on this correlation, we developed two methods of estimating children's premorbid intelligence. First, we developed a regression formula in which school grade is the predictor and FIQ is the predicted variable. With this regression method, the proportion of the sample in which the estimation error was ≤±5, ≤±10, and ≤±15 was 42%, 66%, and 86%, respectively. Second, we developed an estimation method using the average FIQ of children whose school grades are in the comparable range. With this categorization method, the proportion of the sample in which the estimation error was ≤±5, ≤±10, and ≤±15 was 40%, 71%, and 88%, respectively. Discussion includes how much difference between predicted premorbid FIQ and postmorbid FIQ is required to diagnose 'intelligence decline'.