The first sea-ice forecast for waters north of the U.S.S.R. was issued in 1923, and since about 1935 more than a dozen Soviet scientists have been studying the question. Before any method of forecasting could be evolved, meteorological and oceanographical observations, and information on the state of the sea ice had to be available. This information was collected by polar stations, merchant ships, icebreakers, vessels on patrol duty at the ice edge and aircraft on ice reconnaissance. In the late 'thirties, however, forecasters were still complaining of lack of data; not necessarily because they thought the system of collecting information was inefficient, but because there were so few observations for the preceding period. Nevertheless, a number of correlations were advanced and tested. As time went on and the volume of observations grew, it became possible to elaborate principles of more general application. This was done for instance by V. Yu. Vize, one of the earliest investigators of the subject, in a monograph called Osnovy dolgosrochnykh ledovykh prognozov [Principles of long-term ice forecasting]. The work was awarded a Stalin prize in 1946, but apparently no copy has come out of the U.S.S.R. It is not possible, therefore, to examine the methods which have been developed by Soviet scientists on the basis of ten to fifteen years' observations, but the earlier ideas which are considered below no doubt form the basis of the later system.