The investigations discussed in this paper have been made for the purpose of examining the errors which may occur in a mortality table based on the experience of insurance offices when the Census Method of construction is used for a period which has been subject to extreme fluctuations. Hitherto, although the theory, with practical but empirical illustrations, has received considerable attention, no actual experience has been investigated. The present investigation is possible because, owing to the abnormality of the 1931-35 material for annuitants, further information was furnished by the offices.