De vervaardiging van een systeem functie-ponskaarten

While making a series of 10,000 "function" -punched cards with the data of the well-known Barlow's tables and with the logarithms of the integers from the tables of Schrön, we took the opportunity to examine the accuracy with which our personnel and machines are working under normal conditions. We examined for instance in how many cases the first punching operator did not notice an error she made and how many of those errors remained after control-punching. These last-mentioned errors were traced by control-calculations with the punched cards, e.g. by calculating and checking n × n²= n ³ or by the calculation of differences. In this way it appeared that the first punching operator did not notice 270 errors in the part coming from Barlow's tables, being 0.055% of the number of punched columns or 2,7% of the number of cards. Further it was found, that only one error of the first operator was not noticed (or not corrected) by the controling operator.
Besides this a number of printing-errors, or anyway errors which were the result of the bad legibility of the digits, were found in Barlow's tables, where the punching operator and the controling operator had come to the same interpretation.
The article gives in some detail what precautions were taken by wiring the machines to attain that an error would be noticed as such.