To make the carnival and Saturnalia coincide, Mr. Frazer points out that 'if the Saturnalia, like many other seasons of licence, was always observed at the end of the old year or the beginning of the new one, it must, like the carnival, have been originally held in February or March, at the time when March was the first month of the Roman year.'[65] Thus, in conservative rural districts, the Saturnalia would continue to be held in February, not, as at Rome, in December, though Roman writers do not tell us so, and though non-Roman pagan peoples held festival at the winter solstice. The soldiers who killed poor Dasius were ultra-conservative, but they killed him in November, when their month of Saturnalia began, not in February, when, as they held by old usage, their Saturnalia should have been kept. The hypothesis may be stated thus: 1. In rural districts 'the older and sterner practice' of murder may long have survived.[66] 2. In rural districts the Saturnalia continued to be held in February-March, not in December.[67] 3. Therefore the soldiers, who kept up 'the older and sterner practice' of remote districts where the Saturnalia fell in February-March, killed Dasius--in November! 4. Meanwhile, so wedded were the rural districts to Saturnalia in February-March, that the feast continued in these months under the Church and became our carnival. 5. The eclectic soldiers in Lower Mœsia kept up the _old_ killing and full month of revelry (though we never hear of a full month in older or later Rome), but they accepted the _new_ date, November (not kept in Rome) and December; though in their remote rural homes the Saturnalia were in February-March. Doubtless their officers insisted on the new official date, while permitting the old month of revel and the human sacrifice. Yet, apparently, of old there was but one day of revel.