In this study we investigate a well-known biometric model of infant mortality that has been used both to distinguish major causes of mortality before 1 year of age and to examine historical data to identify possible underregistration of infant mortality particularly that which occurred shortly after birth. We show that arguments for the under-registration of infant deaths based solely on the biometric model cannot be supported empirically and are likely due to mis-specifications of the model as well as inappropriate methods for estimating the parameters of interest. The study uses data for 21 parishes in nineteenth-century Sweden. (EXCERPT)