§1. A recent paper described the records obtained at Kew and Eskdalemuir Observatories of a world-wide magnetic storm which occurred on December 16-17, 1917. The outstanding feature was the much larger size of the disturbance at the more northern station. This seems a general fact, of which the two magnetic storms discussed in the present paper afford further evidence. They were two of the principal storms of 1918. Their joint discussion affords an opportunity of emphasising the variety in the phenomena exhibited by storms of the same class. The results differ so much from some which Dr. S. chapman has given in a recent paper as representative of world-wide magnetic storms, that it may not be amiss to explain that the two storms were selected before the appearance of Dr. Chapman’s paper. The storm of March 7-8 was suggested by the kindness of the Director of the Meteorological Service of Canada in sending me, unasked, copies of the D (declination), H (horizontal force), and V (vertical force) curves from Agincourt (near Toronto). Eskdalemuir curves for that date and also for the second storm, which occurred on August 15-16, were kindly supplied by Dr. Crichton Mitchell. The Eskdalemuir magnetographs record N (north component) and W (west component), instead of H and D as at Kew and Agincourt. The simplest way to a comparison was to calculate H and D changes at Eskdalemuir from the observed N and W changes. The two storms are undoubtedly of the kind discussed by Dr. Chapman, according to whom “apparently all great world-wide magnetic storms commence simultaneously to within a few seconds, over the whole earth, although small local fluctuations may sometimes mask the commencement at particular stations.” All I think we can really say is that the particular form of disturbance known as a “sudden commencement,” or Sc, which precedes a considerable number of world-wide storms, appears simultaneously at all stations, to the degree of accuracy with which time can be measured on ordinary curves. This conclusion has been reached independently by several recent investigators, including Dr. Chapman and myself.§ Accuracy to half-a-minute in time measurements is a high claim, and tlie instant at which a movement becomes visible is not wholly independent of the sensitiveness of the instrument. If Dr. Chapman is correct in stating that all world-wide storms commence simultaneously, we must I think suppose that all have Sc s. If so, these must be “masked” more often than not. The existence of an Sc is important in relation to the question whether storms recur after a definite interval. When Mr. W. Maunder* claimed to have established a 27.3-day recurrence period, this aspect of the case was dealt with in a review of his paper which I contributed to “Terrestrial Magnetism.” Of the 276 storms in Mr. Maunder’s list for the period 1882 to 1903 only 77, or 28 per cent., were credited by him with an Sc. In many cases of world-wide storms disturbance seems to originate gradually. The storm of December 16-17, 1917, is a case in point. On that occasion the growth of disturbance was so rapid that I concluded “ its commencement may be accepted without hesitation as occurring between 8 h. and 9 h.” Often, however, differences of several hours occur in the estimate of the time of commencement made at different stations. Also, an Sc is only sometimes immediately followed by large disturbance. Often a comparatively quiet time intervenes between the Sc—which possesses at most stations a characteristic form—and the large movements, which in the absence of an Sc would be recognised as a storm. In some cases there is room for doubt whether there is a real connection between the Sc and the subsequent storm. There is all the more reason for hesitation on this point because, in a considerable number of cases, the Sc itself is the principal movement. It is followed in some instances by what I have described as a “crest.” In low or mean latitudes the Sc is mainly an H movement and is seldom visibly oscillatory, H undergoing a considerable rise in the course of a few minutes. The “crest” consists essentially in the maintenance of the enhanced value of H, the curve presenting a nearly level ridge for a considerable time, which may amount to several hours. A rapid fall, somewhat like an inversion of the original Sc, then ensues, and there may be no further disturbance worth mentioning.