The Canadian Grain Commission decided to introduce segregation of the top grade of Canadian wheat into three subgrades aiming at guaranteed protein levels for export of 14.5, 13.5%, etc., or 14.0, 13.0%, etc. Primary samples taken on loading carlots at country elevators were mailed to central laboratories for single Kjeldahl determinations and results were made available at terminal elevators when carlots were unloaded. Monitoring was undertaken with official samples taken during unloading. Computer studies were made with paired primary and unload data for 9,800 carlots unloaded at six Pacific terminals over 12 wk. Sequential segregation from the top of the distribution showed that two subgrades at required levels could be obtained by adjusting minimum boundaries for each subgrade, and a third more variable residual subgrade also seemed satisfactory; but to obtain maximum uniformity of means for subgrades required that boundaries be adjusted (1) as the mean and distribution of the main grade (1 C.W.) changed from week to week, and (2) from terminal to terminal, irrespective of whether cars were pooled or distributed by company of origin. In accordance with regression theory, unload subgrade means were lower than corresponding primary means when segregation was undertaken above the 1 C.W. mean. Floating station means for earlier shipments could be substituted for missing primary samples with some loss of precision; complete substitution permitted segregation of only two subgrades. For controlling shipments from country points, there are 48 Blocks with 5–10 Train Runs in each. Both the 1 C.W. mean and the distribution for terminal unloads could be changed by directing shipments from selected Train Runs or Blocks, but this might cause segregation problems. As the 1 C.W. mean changes, segregation boundaries must be adjusted and amounts of subgrades changed substantially. For each system, segregation at even or half percentage levels, the boundaries and amounts for subgrades repeat themselves, one percentage unit lower for each one percentage unit drop in the 1 C.W. mean. These studies indicated that the proposed system would prove practicable. The Commission introduced it forthwith, using guaranteed levels of 15.0, 14.0%, etc. at Thunder Bay, and of 14.5, 13.5%, etc. at Vancouver.