Now the root error of all the earlier commentators who attempted the elucidation of the passage under consideration consisted in treating ‘eykt’ not as a solar bearing, but as a definite clock time. Three o’clock clearly would not do, for sunset at 3 p.m. on the shortest day in winter indicates a latitude too far north to correspond in any way with the climate indicated. Torfaeus, the earliest writer on the subject, accordingly interpreted ‘the south-west airt’ as the whole quarter between south and west, and dividing the _time_ between noon and 6 p.m. (equinoctial west) into thirds he arrived at 4 p.m. as the time of sunset, which with 8 a.m. for Dagmál gave an eight hours day, or a latitude of approximately 49° N. Of course, for the reasons already given, the bearing corresponding to such a division of the horizon (S. 60° W.), assuming the latter to be justifiable, would not unalterably represent 4 p.m. even in Iceland, and the clock time for which the bearing stood in Iceland would be indicated by a wholly different position of the sun in another latitude. [Illustration: DIAGRAM OF SUN’S BEARINGS AT THREE-HOUR INTERVALS LATITUDE 65. A.D. 1000] Next came what may be called the school of Rafn, who claimed to have located the Wineland of the sagas with certainty in the neighbourhood of Rhode Island. For them an interpretation which resulted in a latitude of 49° was unsatisfactory. They accordingly prayed in aid a passage from Snorri’s Edda, in which the winter is said to begin at the point where the sun sets in ‘Eyktarstad’. It was known that winter, according to the Icelandic calendar, began in the week preceding the 18th of October, and observation in the latitude of Snorri’s home showed that the sun set there on the 17th of October at 4.30 p.m. As the passage is drawing a distinction between autumn and winter it could hardly refer to the Icelandic winter beginning about the 18th of October, for as Vigfusson has pointed out with regard to this division of the calendar, which persists in modern Iceland, it is a division of the year into summer and winter only, and leaves spring and autumn out of account.[95] But it led Rafn and his followers to assert, in the teeth of all the other evidence, that ‘eykt’ was not a point but a period of time, and that ‘eyktarstad’ was a point which could be interpreted as 4.30 p.m. apparently in any latitude! This, with ‘dagmálastad’ at 7.30 a.m. gave a day of nine hours, from which Rafn claimed to deduce the latitude—to a second of arc—as 41° 24′ 10″, an observation which, accepting Rafn’s theory as to the locality visited, would be beyond the accuracy of a modern sextant. Unfortunately for this surprising result, the method of calculation was hardly so correct, for, apart from the fallacy of treating the local _time_ as transferable, no correction was made for the effects of refraction, &c., and the declination assumed was not that of the eleventh but of the nineteenth century.