It is argued here that the effects of variable site occupation span on frequency seriation have not yet been clearly elucidated. The necessary clarification is provided by graphic models used to generate simulated data which are then seriated. After an analysis of the seriation results, it is concluded that site duration variability poses a more serious problem for frequency seriations than was once thought. Possible solutions are then discussed. It is shown that no solution is in the offing for typological seriations, although it is possible to estimate the magnitude of the errors produced. Something approaching a solution does exist, however, for attribute microseriations.