In food-poisoning outbreaks, the ingestion time of suspect foods has been estimated using the Sartwell method (Hirayama method), the graphical method and maximum likelihood estimation. These methods are based upon the assumption that the incubation period is log-normally distributed. However, these methods may not always be valid because the bias and variability of the estimates are large. We constructed a database from 341 food-poisoning cases in the literature. Using this database, we developed a method to estimate the actual ingestion time based on the period from the ingestion time to the first outbreak, using a multiple linear regression equation. Since only 9.09% of cases showed log-normal distribution, the above three methods would not be valid in most cases. The new method was applicable in all cases. In addition, the values estimated by the new method showed higher correspondence and accuracy than the values estimated by the other methods. Therefore, we consider that the new method is superior to those methods.